Cray X1 assembly language Reference Manual
The Cray X1 assembly language X1 is a symbolic language for Cray X1 systems. The Cray X1 assembler generates object code from CAL source code for execution on Cray X1 systems.
Advertisement
Advertisement
Cray Assembly Language (CAL) for Cray X1™ Systems Reference
Manual
S–2314–50
© 2001, 2003 Cray Inc. All Rights Reserved. This manual or parts thereof may not be reproduced in any form unless permitted by contract or by written permission of Cray Inc.
U.S. GOVERNMENT RESTRICTED RIGHTS NOTICE
The Computer Software is delivered as "Commercial Computer Software" as defined in DFARS 48 CFR 252.227-7014.
All Computer Software and Computer Software Documentation acquired by or for the U.S. Government is provided with Restricted
Rights. Use, duplication or disclosure by the U.S. Government is subject to the restrictions described in FAR 48 CFR 52.227-14 or DFARS 48 CFR 252.227-7014, as applicable.
Technical Data acquired by or for the U.S. Government, if any, is provided with Limited Rights. Use, duplication or disclosure by the
U.S. Government is subject to the restrictions described in FAR 48 CFR 52.227-14 or DFARS 48 CFR 252.227-7013, as applicable.
Autotasking, CF77, Cray, Cray Ada, Cray Channels, Cray Chips, CraySoft, Cray Y-MP, Cray-1, CRInform, CRI/TurboKiva,
HSX, LibSci, MPP Apprentice, SSD, SuperCluster, UNICOS, UNICOS/mk, and X-MP EA are federally registered trademarks and Because no workstation is an island, CCI, CCMT, CF90, CFT, CFT2, CFT77, ConCurrent Maintenance Tools, COS,
Cray Animation Theater, Cray APP, Cray C90, Cray C90D, Cray CF90, Cray C++ Compiling System, CrayDoc, Cray EL, Cray
Fortran Compiler, Cray J90, Cray J90se, Cray J916, Cray J932, CrayLink, Cray MTA, Cray MTA-2, Cray MTX, Cray NQS,
Cray/REELlibrarian, Cray S-MP, Cray SSD-T90, Cray SV1, Cray SV1ex, Cray SV2, Cray SX-5, Cray SX-6, Cray T90,
Cray T94, Cray T916, Cray T932, Cray T3D, Cray T3D MC, Cray T3D MCA, Cray T3D SC, Cray T3E, CrayTutor, Cray X1,
Cray X-MP, Cray XMS, Cray-2, CSIM, CVT, Delivering the power . . ., DGauss, Docview, EMDS, GigaRing, HEXAR, IOS,
ND Series Network Disk Array, Network Queuing Environment, Network Queuing Tools, OLNET, RQS, SEGLDR, SMARTE,
SUPERLINK, System Maintenance and Remote Testing Environment, Trusted UNICOS, UNICOS MAX, and UNICOS/mp are trademarks of Cray Inc.
AIX and IBM are trademarks and RISC System/6000 is a product of International Business Machines Corporation. AXP and DEC are trademarks of Digital Equipment Corporation. DynaText, DynaVerse, DynaWeb, and EBT are trademarks of Electronic Book
Technologies, Inc. HP and HP-UX are trademarks and HP 9000 is a product of Hewlett-Packard Company. IRIX, Mindshare, SGI, and Silicon Graphics are trademarks of Silicon Graphics, Inc. Netscape and Netscape Navigator are trademarks of Netscape
Communications Corporation. NFS, Solaris, SPARC, Sun, and SunOS are trademarks of Sun Microsystems, Inc. Open Software
Foundation, OSF, and OSF/1 are trademarks of Open Software Foundation, Inc. UNIX is a registered trademark in the United States and other countries, licensed exclusively through X/Open Company Limited.
The UNICOS, UNICOS/mk, and UNICOS/mp operating systems are derived from UNIX System V. These operating systems are also based in part on the Fourth Berkeley Software Distribution (BSD) under license from The Regents of the University of California.
Record of Revision
Version
1.0
1.1
Description
August 19, 2002
Draft version to support the Programming Environment 4.1 release.
June 2003
Supports the Cray Assember 1.1 and Programming Environment 5.0 releases.
S–2314–50 i
Contents
Preface
Accessing Cray Documentation
Error Message Explanations
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Typographical Conventions .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Ordering Documentation
Reader Comments .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Introduction [1]
Capabilities .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Related Publications .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Cray X1 Instruction Set Overview [2]
Data Types .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Integer .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Floating-point .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Addresses .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Registers .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Address and Scalar Registers .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Vector Registers .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Vector Length Register .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Mask Registers .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Vector Carry Register .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Bit Matrix Multiply Register
Control Registers .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Program Counter
Performance Counters
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
CAL Source Statement Format .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
S–2314–50
Page
iii
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual
Page
Scalar Instructions .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Immediate Loads .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Scalar Memory References .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Branches and Jumps
Scalar Integer Functions
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Scalar Floating-point Functions
Bit Matrix Multiplication .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Byte and Halfword Access .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Scalar Atomic Memory Operations .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Other Scalar Instructions .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Vector Instructions .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Elemental Vector Operations .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Vector Memory References .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Elemental Vector Functions
Mask Operations .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Other Vector Instructions .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Memory Ordering
Summary of Rules
Special Syntax Forms
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
CAL Pseudo Instruction Overview [3]
Pseudo Instruction Format .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Program Control .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Loader Linkage .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Mode Control
Section Control .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Message Control .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Listing Control .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Symbol Definition
Data Definition
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Conditional Assembly .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
iv
S–2314–50
Micro Definition .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Defined Sequences .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
File Control .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
CAL Program Organization [4]
Program Modules .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Global Definitions and Local Definitions .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Program Segments .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Program
Sections .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Examples .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Example 1: Global and Local Definitions .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Example 2: Sections and Qualifiers .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
CAL Assembler Invocation [5]
Assembler Command Line .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Environment Variables .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
ASDEF Environment Variable .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
LPP Environment Variable .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
TMPDIR Environment Variable .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
TARGET Shell Variable
Assembler Execution
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Reading Source Files .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Using Binary Definition Files .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
CPU Compatibility Checking .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Multiple References to a Definition .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Included Files .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Source Statement Listing File .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Source Statement Listing .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Cross-reference Listing .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Diagnostic Messages .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
S–2314–50
Contents
Page
v
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual
Page
Diagnostic Message Listing File .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Object File .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Creating a Binary Definition File .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Symbols .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Macros .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Opdefs .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Opsyns .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Micros
Linking
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
CAL Source Statements [6]
Source Statement Format .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Label Field
Result Field
Operand Field
Comment Field
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Old Format
Case Sensitivity
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Symbols .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Symbol Qualification .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Unqualified Symbol .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Qualified Symbols .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Symbol Definition
Symbol Attributes .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Type Attribute .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Relative Attributes .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Redefinable Attributes
Symbol Reference .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Tags .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Constants .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Floating Constant .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
vi
S–2314–50
Integer Constant .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Character Constants .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Data Items .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Floating Data Item
Integer Data Item
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Character Data Item .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Literals .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Micros .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Location Elements .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Location Counter .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Origin Counter .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Longword-bit-position Counter .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Force Longword Boundary .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Expressions .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Operators .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Operator Precedence .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Restrictions
Statement Editing
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Micro Substitution .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Concatenate
Append
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Continuation .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Comment .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Actual Statements and Edited Statements .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
CAL Pseudo Instruction Descriptions [7]
Equate Symbol (=) .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
(Deferred implementation) ALIGN .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
BASE
BITW
BSS
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
S–2314–50
Contents
Page
vii
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual
BSSZ .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
CMICRO .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
COMMENT .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
CON
DATA
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
DBSM
DECMIC
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
DMSG
DUP
ECHO
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
EDIT .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
EJECT .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
ELSE .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
END
ENDDUP
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
ENDIF .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
ENDM .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
ENDTEXT .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
ENTRY .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
ERRIF .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
ERROR .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
EXITM .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
EXT
FORMAT
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
IDENT .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
IFA
IFC
IFE
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
IFM .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
INCLUDE .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
viii S–2314–50
Page
LIST .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
LOC .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
LOCAL .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
MACRO .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
MICRO .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
MICSIZE .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
MLEVEL .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
MSG .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
NEXTDUP .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
NOMSG .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
OCTMIC .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
HEXMIC .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
OPDEF .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
OPSYN .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
ORG .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
OSINFO .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
QUAL .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
SECTION .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Local Sections .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Main Sections .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Literals Section .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Sections Defined by the SECTION Pseudo Instruction .
.
.
.
.
.
.
.
.
.
.
.
.
Common Sections .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Section Stack Buffer .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Generated Code Position Counters .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Origin Counter .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Location Counter .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Word-bit-position Counter .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
SET
Force Word Boundary .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
S–2314–50
Contents
Page
ix
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual
Page
SKIP .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
SPACE .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
STACK .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
START .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
STOPDUP .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
SUBTITLE
TEXT
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
TITLE .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
VWD .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Cray X1 Instruction Set and Encoding [8]
System and Memory Ordering Instructions .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
System Instructions .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Memory Ordering Instructions .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Register Move and Jump Instructions .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Register Move Instructions .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Jump Instructions .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
A Register Instructions .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
A Register Integer Instructions
A Register Logical Instructions .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
A Register Shift Instructions .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
A Register Immediate Instructions .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
A Register Integer Compare Instructions .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
A Register Byte and Halfword Instructions .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Other A Register Instructions .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
A Register Branch Instructions .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
A Register Memory Access Instructions .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Load and Store Instructions .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Prefetch Instructions .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Atomic Memory Instructions .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
S Register Instructions .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
x
S–2314–50
S Register Integer Instructions .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
S Register Logical Instructions .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
S Register Shift Instructions .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
S Register Immediate Instructions .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
S Register Integer Compare Instructions .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Other S Register Instructions .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
S Register Branch Instructions .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
S Register Floating Point Instructions .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
S Register Floating Point Compare Instructions .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
S Register Conversion Instructions .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
S Register Memory Instructions .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Vector Register Instructions .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Vector Register Integer Instructions .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Vector Register Logical Instructions .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Vector Register Shift Instructions .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Vector Register Integer Compare Instructions .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Vector Register Floating Point Instructions .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Vector Register Floating Point Compare Instructions .
.
.
.
.
.
.
.
.
.
.
.
.
Vector Register Conversion Instructions .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Other Vector Register Instructions
Vector Mask Instructions .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Vector Memory Instructions .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Privileged Instructions .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
CAL Defined Sequences [9]
Similarities Among Defined Sequences .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Editing .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Definition Format .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Formal Parameters .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Instruction Calls .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Interaction with the INCLUDE Pseudo Instruction .
.
.
.
.
.
.
.
.
.
.
.
.
.
S–2314–50
Contents
Page
xi
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual
Page
Macros (MACRO) .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Macro Definition .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Macro Calls .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Operation Definitions (OPDEF) .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Opdef Definition .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Opdef Calls
Duplication (DUP)
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Duplicate with Varying Argument (ECHO) .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Ending a Macro or Operation Definition (ENDM) .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Premature Exit from a Macro Expansion (EXITM) .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Ending Duplicated Code (ENDDUP) .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Premature Exit of the Current Iteration of Duplication Expansion (NEXTDUP) .
.
.
.
.
.
Stopping Duplication (STOPDUP) .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Specifying Local Unique Character String Replacements (LOCAL)
Synonymous Operations (OPSYN) .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Appendix A Calling Conventions
Appendix B ASDEF Macros and Opdefs
Appendix C Character Set
Glossary
Index
Figures
Figure 1.
CAL Program Organization .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Figure 2.
Page Header Format .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Figure 3.
Source Statement Listing Format .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Figure 4.
Cross-reference Listing Format .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Figure 5.
ASCII Character with Left-justification and Blank-fill
Figure 6.
ASCII Character with Left-justification and Zero-fill .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
xii
S–2314–50
Figure 7.
ASCII Character with Right-justification and Zero-fill .
.
.
.
.
.
.
.
.
.
Figure 8.
ASCII Character with Right-justification in 8 bits
Figure 9.
Storage of Unlabeled Data Items .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Figure 10.
Storage of Labeled and Unlabeled Data Items .
.
.
.
.
.
.
.
.
.
.
.
Tables
Table 1.
Cray X1 Registers .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Table 2.
Processor State Information .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Table 3.
Quadrant Modifications
Table 4.
Scalar Memory References
Table 5.
Branches and Jumps
Table 6.
Scalar Integer Functions
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Table 7.
Integer Relations .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Table 8.
Scalar Floating-point Functions .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Table 9.
Bit Matrix Multiply Instructions .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Table 10.
Byte and Halfword Access .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Table 11.
Scalar Atomic Memory Operations
Table 12.
Other Scalar Instructions .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Table 13.
Vector Memory References .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Table 14.
Elemental Vector Functions .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Table 15.
Mask Operations .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Table 16.
Other Vector Instructions .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Table 17.
Explicit Memory Ordering Instructions .
.
.
.
.
.
.
.
.
.
.
.
.
.
Table 18.
Special Syntax Forms .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Table 19.
Character Set .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Contents
Page
S–2314–50 xiii
Preface
This manual supports the Cray Assembler 1.1 release running on Cray X1 systems. This preface describes how to access Cray documentation and error message explanations, interpret our typographical conventions, order Cray documentation, and contact us about this document.
Accessing Cray Documentation
Each software release package includes the CrayDoc documentation system, a collection of open-source software components that gives you fast, easy access to and the ability to search all Cray manuals, man pages, and glossary in HTML and/or PDF format from a web browser at the following locations:
• Locally, using the network path defined by your system administrator
• On the Cray public web site at: http://www.cray.com/craydoc/
All software release packages include a software release overview that provides information for users, user services, and system administrators about that release.
An installation guide is also provided with each software release package.
Release overviews and installation guides are supplied in HTML and PDF formats as well as in printed form. Most software release packages contain additional reference and task-oriented documentation, like this document, in
HTML and/or PDF formats.
Man pages provide system and programming reference information. Each man page is referred to by its name followed by a number in parentheses: manpagename( n) where n is the man page section identifier:
3
4
1
2
5
7
8
User commands
System calls
Library routines
Devices (special files) and Protocols
File formats
Miscellaneous information
Administrator commands
S–2314–50 xv
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual
Access man pages in any of these ways:
• Enter the man command to view individual man pages in ASCII format; for example: man ftn
To print individual man pages in ASCII format, enter, for example: man ftn | col -b | lpr
• Use a web browser with the CrayDoc system to view, search, and print individual man pages in HTML format.
• Use Adobe Acrobat Reader with the CrayDoc system to view, search, and print from collections of formatted man pages provided in PDF format.
If more than one topic appears on a page, the man page has one primary name
(grep, for example) and one or more secondary names (egrep, for example).
Access the ASCII or HTML man page using either name; for example:
• Enter the command man grep or man egrep
• Search in the CrayDoc system for grep or egrep
Error Message Explanations
Access explanations of error messages by entering the explain msgid command, where msgid is the message ID string in the error message. For more information, see the explain(1) man page.
Typographical Conventions
The following conventions are used throughout this document:
Convention command variable
Meaning
This fixed-space font denotes literal items, such as file names, pathnames, man page names, command names, and programming language elements.
Italic typeface indicates an element that you will replace with a specific value. For instance, you may replace filename with the name datafile in xvi S–2314–50
Preface user input
[ ]
...
your program. It also denotes a word or concept being defined.
This bold, fixed-space font denotes literal items that the user enters in interactive sessions. Output is shown in nonbold, fixed-space font.
Brackets enclose optional portions of a syntax representation for a command, library routine, system call, and so on.
Ellipses indicate that a preceding element can be repeated.
Ordering Documentation
To order software documentation, contact the Cray Software Distribution Center in any of the following ways:
E-mail: [email protected]
Web: http://www.cray.com/craydoc/
Click on the Cray Publications Order Form link.
Telephone (inside U.S., Canada):
1–800–284–2729 (BUG CRAY), then 605–9100
Telephone (outside U.S., Canada):
Contact your Cray representative, or call +1–651–605–9100
Fax:
+1–651–605–9001
Mail:
Software Distribution Center
Cray Inc.
1340 Mendota Heights Road
Mendota Heights, MN 55120–1128
USA
S–2314–50 xvii
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual
Reader Comments
Contact us with any comments that will help us to improve the accuracy and usability of this document. Be sure to include the title and number of the document with your comments. We value your comments and will respond to them promptly. Contact us in any of the following ways:
E-mail: [email protected]
Telephone (inside U.S., Canada):
1–800–950–2729 (Cray Customer Support Center)
Telephone (outside U.S., Canada):
Contact your Cray representative, or call +1–715–726–4993 (Cray Customer
Support Center)
Mail:
Software Publications
Cray Inc.
1340 Mendota Heights Road
Mendota Heights, MN 55120–1128
USA xviii S–2314–50
Introduction [1]
The Cray X1 Assembly Language (CAL) is a symbolic language for Cray X1 systems. The Cray X1 assembler generates object code from CAL source code for execution on Cray X1 systems.
1.1 Capabilities
The primary capability of CAL is to express with symbolic machine instructions the full Cray X1 instruction set. This manual does not describe the machine instructions in detail but does give an overview of the instructions and lists the
instructions (Chapter 8, page 187). The Cray X1 instruction set is described in
detail in the System Programmer Reference for Cray X1 Systems.
In addition, CAL provides these capabilities:
• The free-field source statement format of CAL lets you control the size and location of source statement fields.
• With some exceptions, you can enter source statements in uppercase, lowercase, or mixed-case letters.
• You can assign code or data segments to specific areas to control local and common sections.
• You can use preloaded data by defining data areas during assembly and loading them with the program.
• You can designate data in integer, floating-point, and character code notation.
• You can control the content of the assembler listing.
• You can define a character string in a program and substitute the string for each occurrence of its micro name in the program by using micro coding.
• You can define sequences of code in a program or in a library and substitute the sequence for each occurrence of its macro name in the program by using macro coding.
S–2314–50 1
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual
1.2 Related Publications
The following documents contain additional information that may be helpful:
• System Programmer Reference for Cray X1 Systems
• Cray Fortran Compiler Commands and Directives Reference Manual
• Fortran Language Reference Manual, Volume 1
• Fortran Language Reference Manual, Volume 2
• Fortran Language Reference Manual, Volume 3
• Cray C and C++ Reference Manual
• Loader man page ld(1)
2 S–2314–50
Cray X1 Instruction Set Overview [2]
2.1 Data Types
2.1.1 Integer
S–2314–50
This chapter gives an overview of the Cray X1 system instruction set. The
Cray X1 system is an implementation of the Cray NV-1 instruction set
architecture. The full instruction set is listed in Chapter 8, page 187 and is
described in detail in the System Programmer Reference for Cray X1 Systems.
The Cray X1 system has four fundamental data types:
• W: 32-bit integer (Word)
• L: 64-bit integer (Longword)
• S: 32-bit IEEE-754 floating-point (Single)
• D: 64-bit IEEE-754 floating-point (Double)
The default for the data type of an instruction is L. The other data types are usually specified in assembly language notation with a suffix on the result register.
An instruction’s type is usually the type of its result. For comparisons, however, the instruction’s type is the type of its operands.
The result of a scalar comparison is a Boolean truth value (0 or 1) of type L, and the result of a vector comparison is a vector of Boolean truth values in a mask register.
In addition, some instructions reference 8-bit quantities (Bytes) and 16-bit quantities (Halfwords).
The convention used for numbering bits in a data type is that bit 2 n as bit n. In a range of bits within a data type from 2 n to 2 m where 2 n is referred to is a higher order bit than 2 m
, the bits are specified as n:m.
The Word type W comprises 32-bit twos complement signed integers in the range
-2147483648 ... 2147483647
-0x80000000 ... 0x7fffffff
3
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual
2.1.2 Floating-point
The Longword type L comprises 64-bit twos complement signed integers in the range
-9223372036854775808 ... 9223372036854775807
-0x8000000000000000 ... 0x7fffffffffffffff
Interpretations of these types as C unsigned integers mod 2
32 and 2
64 are supported in comparisons and implicitly in the hi(j*k) multiplication function.
The assembly language syntax uses the notations UW and UL for unsigned comparisons.
Special instructions accelerate loads and stores of 8-bit and 16-bit integers from
64-bit Longword containers.
The Cray X1 system conforms to the IEEE-754 standard for floating-point arithmetic with the exception of subnormal numbers.
The 32-bit single precision floating-point type S has this format:
Bit
2
31
2
30
2
32
:2
23
:2
0
Definition
Sign
Exponent
Mantissa (implicit normalization)
The 64-bit double precision floating-point type D has this format:
Bit
2
63
2
62
2
51
:2
52
:2
0
Definition
Sign
Exponent
Mantissa (implicit normalization)
A 128-bit floating-point format is implemented by software. It comprises a standard 64-bit D value with 64 additional mantissa bits.
2.1.3 Addresses
4
Memory is addressed for instruction fetches, address and scalar register loads and stores, vector loads and stores, and atomic memory operations. All addresses have the same format and interpretation. They are 64-bit values that identify a byte offset in a virtual address space.
S–2314–50
2.2 Registers
Cray X1 Instruction Set Overview [2]
Addresses must be strongly aligned except for the special unaligned (,ua) load and store instructions. This means that instruction fetches and other 32-bit references must have zero bits in the lowest order two bit positions and that
64-bit references must have three lowest order zero bits.
The highest order bit positions (63:48) of a user virtual address must be zero.
Distributed memory applications use bits 47:38 of the virtual address for a logical node number.
Data caches are always coherent, but the instruction caches are not.
Self-modifying code must synchronize the instruction cache with an explicit lsync,i instruction.
The Cray X1 system addresses memory in big-endian order. If a 64-bit Longword is stored at aligned address x, a 32-bit load from x+0 will retrieve its upper order
32 bits and a 32-bit load from x+4 will retrieve its lower order 32 bits.
The Cray X1 processor has the following set of registers: a0,...,a63 s0,...,s63 v0,...,v31 vl m0,...,m7 vc bmm c0,...,c63
Table 1. Cray X1 Registers
64-bit address registers
64-bit scalar registers
32-bit or 64-bit vector registers
Vector length register
Mask registers
Vector carry register
64x64 bit matrix multiply register
Control registers
S–2314–50 5
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual
In addition to these registers, processor state information is contained in:
(no syntax)
(no syntax)
Table 2. Processor State Information
Program counter
Performance counters
In this manual, the term register refers to a specific register, such as a23, the term
register type refers to the letter or letters that identify the specific set of registers, such as a for address registers, and register designator refers to the number that specifies a register in that set, such as 23. CAL accepts either upper case or lower case for the register type, so a23 and A23 are treated the same.
2.2.1 Address and Scalar Registers
There are 64 address and 64 scalar registers, each 64 bits in width. They constitute computational way stations between memory and functional units in the processor for serial regions of the program. They also serve vectorized code with addresses, strides, and scalar values.
Registers a0 and s0 are unmodifiable zero values. Instructions that would write to them discard their results and may ignore exceptions.
Both the address and scalar registers are general-purpose and support the same memory reference instructions, immediate loads, integer functions, and conditional branches. However, address registers must be used for:
• Memory base addresses, offsets, and strides
• Indirect jump destinations and return addresses
• Vector element indexing
• Vector Length computation with cvl()
• Reading and writing vl and control registers
• Receiving the results of the mask analysis instructions first(), last(), and pop()
• Supplying the stride for a vector scan() and cidx()
• 8-bit and 16-bit accesses
And scalar registers must be used for:
6 S–2314–50
Cray X1 Instruction Set Overview [2]
• Scalar bit matrix multiplications
• Floating-point operations
• Scalar operands to vector operations
Unlike some other instruction set architectures, the Cray X1 system does not mandate a strict data typing rule. Integer operations may be performed on floating-point data and vice versa.
When 32-bit data enter a 64-bit address or scalar register, they are always right-justified and sign-filled, even if the value is notionally unsigned. 32-bit operations on 64-bit registers always ignore the upper bits and are guaranteed to return a sign-extended result.
There is a restriction that prohibits the use of both register designators n and
n+32 or n and n-32 as operands for the same instruction if they are the same register types.
2.2.2 Vector Registers
S–2314–50
There are 32 vector registers, each capable of holding 64 32-bit elements or
64 64-bit elements. They constitute computational way stations between the memory and the functional units of the processor for parallel regions of the program.
32-bit and 64-bit data are stored differently in vector registers. In particular, 32-bit data are not sign-extended as they are in the address and scalar registers.
A vector register holding the result of a 32-bit load or operation is defined for use only as an operand of a 32-bit store or operation. Similarly, a vector of 64-bit data can be used only in 64-bit instructions. Explicit conversion operations must be used to change the width of the data in the elements of vector registers.
This rule permits a hardware implementation to pack 32-bit vector elements in non-obvious ways that can vary from one generation to the next depending on pipe width. Packing permits 32-bit vector operations to execute at double speed.
When a vector register is written by a memory load or vector instruction, only the elements that are actually written are well-defined in the result vector register.
The other elements, which may be those past the limiting value vl or which correspond to zero bits in the controlling mask register, become undefined.
Programs must not assume that these other elements are preserved.
There is a restriction that prohibits the use of the same vector register as both the operand and the result of a type conversion operation.
7
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual
2.2.3 Vector Length Register
The maximum number of elements that a vector register can hold is not actually specified by the architecture. It is only guaranteed to be a power of 2 and at least
64. It may vary between hardware implementations.
A vector’s length is always a count of its elements, not its Bytes, Words, or
Longwords. A vector of 32-bit data cannot hold any more elements than a vector of 64-bit data can.
The vector length register vl specifies the number of elements to be processed by vector register operations. Once set, it is an implicit operand to every vector register operation that follows.
Programs should use the cvl() function to compute legal and well-balanced
Vector Length values.
Vector Length can be set to zero.
2.2.4 Mask Registers
The vrip instruction declares that the contents of the vector registers need no longer be maintained. It is used at the end of a vector sequence to avoid expensive context switch times.
Each of the eight mask registers contains a bit corresponding to each vector register element position. Since this may be larger than 64 bits, the instruction set contains instructions that manipulate mask registers directly.
Masks are set with the results of vector comparison operations. They can then be used to generate vector values with the scan() and cidx() functions. Masks are also used to control vector instructions on a per-element basis. Only the first four masks, m0:m3, can be used to control elemental vector operations. Values in m4:m7 must be moved to m0:m3 for use in vector instructions.
By software convention, mask register m0 can be assumed to have every bit set.
2.2.5 Vector Carry Register
There is a single Vector Carry (vc) Register that is both an operand to and a result of the 64-bit vector add with carry and subtract with borrow instructions. Like the mask registers, it holds one bit for every vector register element position.
8 S–2314–50
Cray X1 Instruction Set Overview [2]
2.2.6 Bit Matrix Multiply Register
The Cray X1 system implements bit matrix multiplication (see Section 2.4.6, page
18). There is a 64 by 64 bit matrix multiply register that must be loaded from a
vector register.
2.2.7 Control Registers
Access to most of the control registers is privileged to code running in kernel mode. These privileged control registers provide the means for programming the processor’s address translation units and other critical functions.
Some control registers are available to user mode code and are therefore part of the instruction set architecture: c0 : Floating-point control (rounding mode and interrupt/exception masks) c1 : Floating-point status (interrupt/exception flags) c2 : Read-only MSP configuration information c3 : Read-only user time clock c4 : Read-only additional MSP information c28 : 32 performance counter enable flags c29 : 32 2-bit performance counter event selections c30 : Performance counter access control c31 : Performance counter value
2.2.8 Program Counter
The 64-bit virtual Program Counter holds the byte address of the next instruction to fetch. This counter is not visible to the user but its content is referenced in the description of some instructions, with the notation pc.
2.2.9 Performance Counters
The processor has 32 64-bit performance counters that are accessed indirectly through control register c31. Each counter can be programmed with one of four event codes that determine when it increments.
S–2314–50 9
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual
2.3 CAL Source Statement Format
The format of the Cray Assembly Language (CAL) source statement is briefly described here before we summarize the Cray X1 system machine instructions.
For details, see Chapter 8, page 187. Detail on the format of the listing files is
Each machine instruction is denoted by a source statement with three fields separated by white space and optionally followed by a comment.
10
Label Result Operand Comment
The Label field is optional for machine instructions. It can contain an identifier that must begin in column 1. The label is given the value of the instruction’s byte address.
The Result field is always present. It typically contains the name of the instruction’s result register and the type when necessary.
The Operand field is usually present. It typically contains a brief expression in
C-like syntax with registers and constant expressions.
A Comment can follow the instruction. Comments begin with a semicolon, which is not part of the syntax for any instruction. A comment can also begin with an asterisk (*) in column 1.
Example:
Label Result Operand Comment top a1 a2+a3 ; a1 gets the sum of a2 and a3
In the instruction descriptions that follow, similar instructions are often combined with a notation like: a i,W or L a j+ak or imm instead of fully elaborating the four distinct instructions: a i,W a i,L a i,W a i,L a j+ak a j+ak a j+imm a j+imm
Operators in CAL statements very closely follow the syntax, meaning, and
operator precedence rules of the C language. For details, see Section 6.9, page 94.
S–2314–50
Cray X1 Instruction Set Overview [2]
2.4 Scalar Instructions
This section gives an overview of the Cray X1 system scalar instruction set.
2.4.1 Immediate Loads
S–2314–50
These CAL statements load the value of expression into the 64-bit a or s register.
a i s i expression expression
; Immediate load a i
; Immediate load s i
However, because all Cray X1 system machine instructions are 32 bits wide, a
64-bit expression, depending on the bit pattern of the expression to be loaded, cannot necessarily be loaded into a register using a single machine instruction.
To aid the programmer in loading a value of up to 64 bits into a register, the assembler supports these immediate load statements that the assembler expands into a sequence of one or more of the primitive 16-bit immediate load instructions described below.
The hardware supports a set of primitive immediate load instructions, each of which has a 16-bit expression (expr in the rest of this section) as part of the instruction and each of which modifies 1, 2, 3 or 4 16-bit quadrants (or
Halfwords) of the result register.
For one set of primitives, one quadrant of the result register receives the 16-bit value (expr) specified in the machine instruction, and 0, 1, 2, or 3 of the quadrants containing the higher order bits may be filled with copies of the immediate value’s uppermost bit for sign-extension; the remaining quadrants are set to zero.
In other words, no part of the old value in the register remains unchanged.
For another set of primitives, one quadrant of the result register receives the
16-bit value (expr) specified in the machine instruction, and 0, 1, 2, or 3 of the quadrants containing the higher order bits may be filled with copies of the immediate value’s uppermost bit for sign-extension; the remaining quadrants are left unchanged. In other words, a new value is merged with part of the old value in the register.
Which quadrants get modified and how they are modified depends on the quadrant (or Halfword) specifiers a (for bits 63:48), b (for bits 47:32), c (for bits
31:16) and d (for bits 15:0). The assembler uses two factors:
• The combination of specifiers (ab, abc, or abcd)
• The field in which the specifier is located (result field or operand field) to determine which of the other three quadrants are modified and how.
11
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual
Result a i or si a i or si a i or si a i or si a i or si a i or si a i or si a i or si:a a i or si:b a i or si:c a i or si:d a i or si:ab a i or si:abc
Operand
expr:a
expr:b
expr:c
expr:d
expr:ab
expr:abc
expr:abcd expr expr expr expr expr expr
The following table details the syntax and quadrant modifications.
0
-
-
sign sign sign sign sign expr a (63:48) expr
0
0
Table 3. Quadrant Modifications
0 expr
-
expr sign expr sign sign
b (47:32)
0 expr
0 c (31:16)
0
0 expr
0
0 expr
-
expr
-
expr sign d (15:0)
0
0
0 expr
0
0
-
expr
expr
-
-
2.4.2 Scalar Memory References
Any of the data types W, L, S, or D can be used as type in the following instructions but only the size of type matters:
Table 4. Scalar Memory References a i or si,type [aj+expr] a a i
i or si,type
[aj+expr]
[aj+ak]
[aj+ak]
[aj],ua
; Load from aj plus scaled offset expr
; Load from aj plus scaled index ak
; Load Longword containing unaligned address a j a i or si,type ; Store to aj plus scaled offset expr a i or si,type ; Store to aj plus scaled index ak
12 S–2314–50
Cray X1 Instruction Set Overview [2]
[aj] pref pref a i,ua
[aj+expr]
[aj+ak]
; Store to Longword containing unaligned address aj
; Prefetch Longword from aj plus scaled offset expr
; Prefetch Longword from aj plus scaled index a k
Offset expressions (expr) and index values (ak) are always multiplied by the data size before being added to the aj base address. A 32-bit memory reference multiplies the offset by 4; a 64-bit reference multiplies the offset by 8. This means that alignment can be checked solely from the low order 2 or 3 bits of the base address aj. These bits are ignored for the unaligned (,ua) references and must be zero for the others.
Like all other instructions that enter 32-bit data into the 64-bit address and scalar registers, 32-bit loads sign-extend their results.
As a consequence of the Cray X1 system register numbering restriction mentioned above, indexed references must not use aj and ak registers whose register designators differ by exactly 32.
The constant offset values (expr) are signed twos complement 14-bit values prior to being scaled (-8192...8191). The byte offset ranges are therefore
-32768...32767
for 32-bit references and -65536...65535 for 64-bit references.
The data cache prefetch (pref) instructions cannot raise any memory exceptions.
Prefetches from bad addresses are therefore silently ignored.
2.4.3 Branches and Jumps
The Cray X1 system has six position-independent conditional branch instructions and one indirect jump instruction: bz bn blt ble a i or si,displ a i or si,displ a i or si,displ a i or si,displ
Table 5. Branches and Jumps
; Branch if == 0
; Branch if != 0
; Branch if < 0
; Branch if <= 0
S–2314–50 13
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual bgt bge
j,ai a a a
i or si,displ
i or si,displ
j,SR or RT
; Branch if > 0
; Branch if >= 0
; Jump to aj and set ai to pc+4 with optional subroutine call or return hint
The conditional branches test all 64 bits of ai or si. 32-bit values will be correctly interpreted if the usual sign extension convention is observed.
The conditional branch instructions identify their targets with a signed twos complement 20-bit instruction displacement relative to the next instruction at pc +4. The offset range for conditional branches is therefore:
-524287 ... 524288 instructions
-2097151 ... 2097152 bytes relative to the branch instruction.
The single indirect jump instruction can be annotated with hints that distinguish subroutine calls and returns from other jumps. These hints are used by the hardware jump target prediction mechanism.
Jump destination addresses (aj) must be properly aligned so that their lowest two bits are zero.
2.4.4 Scalar Integer Functions
These are the integer operations available to address and scalar registers. To avoid clutter, only the address register forms are shown here. The scalar register forms are analogous, except for the lack of the cvl() function. Type L is the default type. In many instructions, the ak register can be replaced with an unsigned 8-bit immediate constant value (imm).
a i,W or L a i,W or L a i,W or L a i,W or L a i a j+ak or imm a j-ak or imm a j*ak a k/aj hi (aj*ak)
Table 6. Scalar Integer Functions
; Addition
; Subtraction
; Multiplication
; Signed division
; Upper 64 bits of full 128-bit multiplication
14 S–2314–50
S–2314–50
Cray X1 Instruction Set Overview [2] a i a i a i a i a i a i,W or L a i,W or L a i a i,L or UL a i,W or L a i,W or L a i a i a i a j&ak or imm a j|ak or imm
; Bitwise AND
; Bitwise OR a j^ak or imm
~a j&ak
; Bitwise XOR
; Bitwise AND with complement
~a j^ak a j<<ak or imm a j>>ak or imm
; Bitwise Equivalence
; Left shift
; Logical right shift
+a j>>ak or imm ; Arithmetic right shift a j<ak or imm ; Comparison lz (aj) pop (aj) a j?ak:ai a j?ai:ak
; Leading zero count
; Population count
; Conditional move, ai gets ak if aj != 0
; Conditional move, ai gets ak if aj == 0 cvl (ak or imm) ; Compute Vector Length
The rules and guidelines below apply to integer functions:
• A properly encoded integer function can raise no exception.
• Division by zero is undefined, as is the signed division of the most negative twos complement integer by –1:
-9223372036854775808 / -1
The result cannot be represented in 64 bits.
• Shift count values are interpreted as unsigned integers of the same size as the shifted value. This means that the upper 32 bits of the shift count are ignored for 32-bit shifts, and that negative shift counts are out of range.
• Shift count values that are out of range yield zero results for logical left and right shifts and a word full of sign bits for arithmetic right shifts.
• The hi(aj*ak) function is useful for multiprecision integer arithmetic and for implementing division by constant values that are not exact powers of 2.
Compute a normalized fixed-point reciprocal of the denominator accurate to
64 bits at compilation time, then implement division with hi(aj*ak) and a right shift.
15
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual
16
• The Vector Length computation function cvl() computes a balanced and legal Vector Length value for a given count of remaining loop iterations.
Specifically,
Value of cvl(ak)
0 a k a k/2
MAXVL
If a k < 0
0 <= ak <= MAXVL
MAXVL<a k < 2 × MAXVL a k > 2 × MAXVL where MAXVL is the actual maximum vector length at run time.
• As long as 32-bit data are maintained in the canonical sign extended format in the 64-bit address and scalar registers, an explicit integer type conversion is necessary only when one is:
– converting unsigned 32-bit integer data to a 64-bit integer type, or
– converting a 64-bit integer type to a 32-bit integer type
The first case is best implemented with an immediate load of zero into the upper 32 bits of the register (ai:ab 0). The second case is best implemented with a 32-bit integer addition to zero (ai,W aj+0). And this latter conversion is not required when the result is used only as a operand to 32-bit operations.
• Here are two useful idioms: a i,UL a i,UL a j< 1 a0<a k
; C logical negation (a i !aj)
; C logical normalization (a i !!ak)
• Only a less-than comparison function is necessary. It yields 1 for true and 0 for false. The full set of integer relations can be constructed thus: a a i i a i,UL a i
Table 7. Integer Relations a j<ak ; a j < ak a k<aj a i<1 a k<aj
; a j <= ak
; a j > ak
S–2314–50
Cray X1 Instruction Set Overview [2] s i,S or D s i,S or D s i,S or D s i,S or D s i,S or D s i,S or D s i,S or D s i,S or D s i,S or D s i,S or D s i,S or D s i,S or D s i,t1 s i,W or L s i,W or L s i,W or L s i,W or L a a i
i,UL a j<ak a i<1 ; a j >= ak
• When a comparison is used as the predicate of a conditional branch, the second instruction can be removed by reversing the branch condition.
2.4.5 Scalar Floating-point Functions
Unlike the integer functions, scalar floating-point instructions are available only to the scalar (sj) registers. There is no default data type, nor are there immediate operand forms.
s j+sk s j-k s j*sk s k/sj abs (sj) sqrt (sj) cpys (sj,sk) s j==sk s j!=sk s j<sk s j<=sk s j?sk s j,t2 round (sj),S or D trunc (sj),S or D ceil (sj),S or D floor (sj),S or D
Table 8. Scalar Floating-point Functions
; Addition
; Subtraction
; Multiplication
; Division
; Arithmetic absolute value
; Square root
; Copy value of sj with sign of sk
; Equality comparison (unordered yields 0)
; Inequality comparison (unordered yields 1)
; Less than comparison (unordered yields trap)
; Less than or equal comparison (unordered yields trap)
; Test ordered (ordered yields 1, unordered yields 0)
; Convert from type t2 to distinct type t1 (not both integer)
; Convert floating-point to integer, rounding to nearest
; Convert floating-point to integer, rounding to zero
; Convert floating-point to integer, rounding up
; Convert floating-point to integer, rounding down
S–2314–50 17
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual
The rules and guidelines below apply to floating-point functions:
• Comparisons yield the 64-bit integer values 0 and 1 as results.
• To ensure that IEEE-754 signed zero values are properly recognized, programs should use a floating-point comparison against s0 before a conditional branch instead of using bz and bn directly on floating-point data.
• IEEE-754 defines the absolute value function as a non-arithmetic operation, or one that should not trap on an invalid operand. Implement a non-arithmetic absolute value function with cpys(sj,s0).
• Similarly, a non-arithmetic negation should also be implemented with cpys() from a logical complement of the operand.
• Arithmetic floating-point operations signal floating-point exceptions and cause traps depending on the mask fields in the floating-point control register c0. Floating-point traps are not precise. The lsync fp instruction should be used when it is necessary to force pending traps to be taken before proceeding.
2.4.6 Bit Matrix Multiplication
There are three bit matrix multiply instructions:
18 bmm s i v i v k bmm bmm
Table 9. Bit Matrix Multiply Instructions
(sk)
(vk),mm
; Bit matrix load
; Scalar bit matrix multiply
; Vector bit matrix multiply
The bmm vk instruction loads bits from the vector register vk into the bmm register.
The si bmm(sk) instruction performs the basic bit matrix multiplication operation.
Each bit j of the 64-bit integer result si, counting from the highest order bit position down to the lowest, is computed thus: s ij = pop(sk & bmmj) (mod 2) where bmmj is the j the row of the bmm register.
In other words, bit j of the result si is set if the bitwise AND of sk and element j of the vector register that was loaded into the bmm register has an odd number of
S–2314–50
Cray X1 Instruction Set Overview [2] bits set. When the bmm register is loaded such that exactly one bit is set in each row and each column, it performs a bit permutation function such as a centrifuge.
The bit matrix multiplication can also be viewed as the matrix product (mod 2) of a 1-by-64 row vector in sk and the 64-by-64 transpose of bmm. Within this perspective, each 64-bit word loaded into bmm constitutes a column.
The vi bmm(vk),mm instruction is described in the Vector instructions section
2.4.7 Byte and Halfword Access
The Cray X1 system instruction set supports direct memory references to 32-bit and 64-bit data. The functions described in this section accelerate access to
8-bit (Byte) and 16-bit (Halfword) data. They manipulate only address register operands.
S–2314–50 a i a i a i a i a i a i extb (aj,ak) exth (aj,ak) mskb (aj,ak) mskh (aj,ak) insb (aj,ak) insh (aj,ak)
Table 10. Byte and Halfword Access
; Extract unsigned Byte from aj at ak
; Extract unsigned Halfword from aj at ak
; Clear Byte in aj at ak
; Clear Halfword in aj at ak
; Shift Byte aj into position ak
; Shift Halfword aj into position ak
All of these functions use the lowest order 2 or 3 bits of ak to identify a 16-bit
Halfword or 8-bit Byte or position in a 64-bit field in big-endian order. The 64-bit field is assumed to hold the Longword containing the byte addressed by ak.
The functions listed above perform the following operations:
• The ext() functions extract the indexed field, right-justified and zero-extended to 64 bits.
• The msk() functions clear the indexed field.
• The ins() functions truncate the value of aj and shift it into the indexed field.
To load an unsigned Byte from memory address a2, use:
19
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual a1 a1
[a2],ua extb(a1,a2)
To store a Byte in a1 to memory address a2, use: a3 a3 a4 a3
[a2]
[a2],ua mskb(a3,a2) insb(a1,a2) a3|a4 a3,ua
Note: This store sequence is not atomic with respect to stores into the same
64-bit Longword from other processors. Use the atomic aax operation (Section
2.4.8, page 20) when such conflicts are possible.
The 16-bit Halfword access functions are defined only for properly aligned
(even) ak addresses.
2.4.8 Scalar Atomic Memory Operations
There are three basic atomic memory operations. Two of them have variant forms to be used when no result is necessary. The result of an atomic memory operation is always the old Longword value from memory.
[a j] a i,aadd a i [a j],afadd,ak
[a j] a i,aax,ak a a i i
[a
[a
Table 11. Scalar Atomic Memory Operations
j],afax,ak
j],acswap,ak
; Add ai to Longword addressed by aj
; Atomic add, returning old value
; AND ai to Longword addressed by aj, then
XOR with ak
; Atomic logical, returning old value
; Store ak to Longword addressed by aj if its old value equals ai
Notes on atomic scalar memory operations:
• Only atomic operations on aligned 64-bit Longwords are provided.
Analogous atomic operations on shorter fields are synthesized in software.
• Register ai is used as an operand to afax and acswap and, as a result, is also overwritten with the old value from memory.
20 S–2314–50
Cray X1 Instruction Set Overview [2]
• The general atomic logical operation can be used to synthesize more specific useful operations:
AND x
OR x
XOR x
Equivalence x ai value x
x complement
0 complement
0 complement
Set bits x and clear y (x OR y) complement
Mask with x and store y x x x ak value
0
x complement x y
• Atomic memory operations can be suffixed with ,NA to hint to the caches that the Longword should not be allocated, or with ,EX to hint that the
Longword should be exclusively allocated. The default hint is ,NA. Cache hints are ignored on atomic memory operations.
• Atomic memory operations are not ordered with respect to other memory references in the same processor, apart from those required by register-to-register data dependences. They must be surrounded by gsync instructions to ensure proper ordering.
2.4.9 Other Scalar Instructions
These nonprivileged scalar instructions do not fit neatly into any of the categories above, and so they are described here for completeness: syscall break a i,W or L s i,W or L a i c i expr expr s k a k ck a k
Table 12. Other Scalar Instructions
; System call
; Debugging breakpoint
; Transfer sk to ai
; Transfer ak to si
; Read Control register
; Write Control register
S–2314–50 21
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual
2.5 Vector Instructions
This section gives an overview of the Cray X1 system vector instruction set.
2.5.1 Elemental Vector Operations
The Cray X1 system vector instruction set contains vector versions of most scalar integer functions, floating-point functions, and memory references. These elemental vector operations process each element independently. They execute under control of a mask register (m0,...,m3) and the vector length register
(vl). Though there are 8 mask registers, only the first 4 can be used in the vector instructions.
The assembler’s default controlling mask register is m0. By software convention, every bit in m0 is always set.
In the vector result of an elemental vector operation, those elements that are past vl or which correspond to a zero bit in the controlling mask register become undefined (merges are the exception).
Elemental vector operations therefore have these semantics: for (I = 0; I < MAXVL; I++) if (I >= vl || !mm[I]) else vi[I] = undefined; vi[I] = vj[I] op (sk or vk[I]);
2.5.2 Vector Memory References
The Cray X1 system loads and stores vector registers from a sequence of properly aligned byte addresses. The address sequence is computed from an aligned base address aj and either a scalar stride value or a vector of 64-bit offset values.
Strided references add multiples of the stride value to the base address, while gather/scatter references add the corresponding offset vector element value. Both scalar strides and vector offset values are scaled by the data size.
v i,type v i,type
[a j,ak or imm]
Table 13. Vector Memory References
[a j,ak or imm],mm
[a j,vk],mm vi,type,mm
; Strided load
; Gather
; Strided store
22 S–2314–50
Cray X1 Instruction Set Overview [2] v i,type v i,type v i,vc v i,vc v i,type v i v i,S or D v i,W or L v i,W or L v i,W or L v i,W or L
S–2314–50
[a j,vk]
[a j,vk] v i,type,mm v i,type,mm ,ord
; Scatter with distinct offsets
; Scatter with arbitrary offsets
The rules and guidelines below apply to vector memory references:
• Immediate stride values are 6-bit signed twos complement integers in the range -32...31
• A poorly aligned aj base address may or may not cause a trap if vl is zero or the controlling mask mm is clear.
• All vector memory reference instructions can be annotated with an optional cache hint code:
– ,NA Data should not allocate space in cache if not present
– ,SH New allocations should be in shared state
– ,EX New allocations should be in exclusive state (default)
2.5.3 Elemental Vector Functions
Most elemental vector functions with two operands permit the use of sk as a fixed scalar operand in place of vk.
v j+sk or vk,mm s k or vk-vj,mm v j+sk or vk,mm s k or vk-vj,mm v j*sk or vk,mm hi(v j*sk or vk),mm s k or vk/vk,mm v j&sk or vk,mm v j|sk or vk,mm v j^sk or vk,mm
~v j&sk or vk,mm
Table 14. Elemental Vector Functions
; Addition
; Subtraction
; 64-bit addition with carry
; 64-bit subtraction with borrow
; Multiplication
; Upper 64 bits of full 128-bit multiplication
; Floating-point division
; Bitwise AND
; Bitwise OR
; Bitwise XOR
; Bitwise AND with complement
23
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual v i,W or L v i,W or L v i,W or L v i,W or L v i,W or L v i,W or L v i v i,S or D v i,S or D v i,S or D v i,t1 v i,W or L v i,W or L v i,W or L v i,W or L v i,W or L m i,type m i,type m i,type m i,type m i,type m i,type m i,S or D
24
~v j^sk or vk,mm v j<<sk or vk,mm v j>>sk or vk,mm
+v j>>sk or vk,mm lz(v j),mm pop(v j),mm bmm(v k),mm abs(v j),mm sqrt(v j),mm cpys(v j,vk),mm v j,t2,mm round(v j),S or D,mm trunc(v j),S or D,mm ceil(v j),S or D,mm floor (v j),S or D,mm m m?sk or vk:vj v j==sk or vk,mm v j!=sk or vk,mm v j<sk or vk,mm v j<=sk or vk,mm v j>sk,mm v j>=sk or vk,mm v j?sk or vk,mm
; Bitwise Equivalence
; Logical left shift
; Logical right shift
; Arithmetic right shift
; Leading zero count
; Population count
; Bit matrix multiplication
; Arithmetic absolute value
; Square root
; Copy of vj with signs of vk
; Convert type t2 to distinct type t1
; Convert floating-point to integer, rounding to nearest
; Convert floating-point to integer, rounding to zero
; Convert floating-point to integer, rounding up
; Convert floating-point to integer, rounding down
; Merge
; Compare for equal
; Compare for not equal
; Compare for less than
; Compare for less than or equal
; Compare for greater than
; Compare for greater than or equal
; Compare for unordered
The rules and guidelines below apply to elemental vector functions:
• The comparisons <, <=, >, and >= can also use types UW and UL for unsigned integer comparisons.
• There is no vector form of integer division.
• There is no vector/scalar division.
S–2314–50
Cray X1 Instruction Set Overview [2]
• As in scalar floating-point functions, non-arithmetic absolute value and negation should be implemented with cpys().
• Subtraction with borrow will clear bits in vc when a borrow occurs.
• Unselected elements of vc are cleared in an add-with-carry or subtract-with-borrow function.
• Vectors of 32-bit data must not be used as operands to 64-bit functions and vice versa.
• Comparisons are elemental operations, but every bit of the resulting mask mi is well defined. Result mask bits that correspond to element positions past vl or to zero bits in the controlling mask mm are set to zero. The result mask mi is therefore a subset of the controlling mask mm.
• Most commutative functions with a vj op sk syntax have a reversed sk op vj
synonym supported by the assembler. For details, see Section 2.8, page 31.
• The bmm() function is the elemental vector analog to the scalar si bmm(sk) instruction.
2.5.4 Mask Operations m i m i m i m i m i m i a i a i a i
The Cray X1 system provides operations that operate directly on mask registers.
This feature improves decoupled execution and permits the maximum vector length to exceed the width of the Address and scalar registers.
m j&mk m j|mk m j^mk
~m j&mk
~m j^mk fill(a k) first(m k) last(m k) pop(m k)
Table 15. Mask Operations
; Bitwise AND
; Bitwise OR
; Bitwise XOR
; Bitwise AND with complement
; Bitwise equivalence
; Set leading ak bits, clear remainder
; Find lowest set bit index (MAXVL if none)
; Find highest set bit index (MAXVL if none)
; Count number of bits set
S–2314–50 25
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual
26 v v v i i i scan(a k,mj) cidx(a k,mj)
; Running sums of mask bits (times ak)
; Indices of set mask bits (times ak) cmprss(v k,mj) ; Compress vector
The rules and guidelines below apply to mask operations:
• Any of the mask registers can be used in the mask operation instructions, but only m0–m3 can be used in the vector instructions.
• The scan(), cidx(), and cmprss() functions use vl. The other mask operations manipulate all MAXVL bits of the masks.
• The scan() function generates a vector from a mask register, a stride, and the vector length register. For each element position 0 <= n < vl, v i[n] = ak × 6
0
≤t<n m j[t]
The remaining elements of the result are undefined.
Equivalently, scan() can be defined with this algorithm: for (I = sum = 0; I < MAXVL; I++) if (I >= vl) vi[I] = undefined; else { vi[I] = sum; if (mj[I]) sum += ak;
}
The scan() function is particularly useful for vectorizing conditionally incremented loop induction variables and for constructing gather/scatter offset vectors for compression and expansion.
• The cidx() function defines the leading elements of the result vector register v i and undefines the remainder. The number of elements written to vi equals the number of bits set in the mask mj.
for (I = sum = 0; I < vl; I++) if (mj[I]) vi[sum++] = ak * I; for (; sum < MAXVL; sum++) vi[sum] = undefined;
S–2314–50
Cray X1 Instruction Set Overview [2]
• The cmprss() function is similar to cidx() except that it stores values from another vector register.
for (I = sum = 0; I < vl; I++) if (mj[I]) vi[sum++] = vk[I]; for (; sum < MAXVL; sum++) vi[sum] = undefined;
2.5.5 Other Vector Instructions
These vector instructions do not fit neatly into any of the categories above, so they are described here for completeness.
vl a i a i or si,W or L v i,aj or imm bmm vrip
Table 16. Other Vector Instructions a k vl v k,aj or imm a k or sk,W or L v k
; Set Vector Length
; Retrieve Vector Length
; Read element
; Write element
; Load bit matrix
; Declare vector state dead
The rules and guidelines below apply to these vector instructions:
• Immediate vector element indices (imm) are unsigned 6-bit integer values, so only the first 64 elements of a vector can be directly indexed.
• Vector element indices (aj) that are outside the range 0 <= aj <MAXVL are undefined and may cause a trap.
• Writing a value outside the range 0 <= ak <= MAXVL to vl will cause a trap. Use the cvl() function to compute balanced and legal vector lengths before writing vl.
• The bmm vk instruction copies the leading 64-bit elements of vk into the corresponding rows of the 64 by 64 bit matrix. Only min(vl,64) elements are written. The remaining rows are cleared, so every bit in the matrix becomes well defined.
S–2314–50 27
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual
This operation is not elemental because it does not operate under control of a mask register. The vk register must therefore be completely defined up to vl with 64-bit values.
• The vrip instruction undefines all of the vector registers, the vector carry register vc, and the mask registers, apart from m0. If the contents of the bmm register are defined, they remain defined. Mask register m0 remains defined if it still has all of its bits set; otherwise, it too becomes undefined.
The vrip instruction is used at the end of a sequence of vector instructions to reduce the size of the processor state that must be saved and restored when switching processor contexts, and also to release physical registers in implementations that rename the vector registers.
a i m i,aj or imm m a
k,aj or imm k
; Read mask bits
; Write mask bits
The aj or immediate value is multiplied by 64 to index the first mask bit to be transferred to or from the highest-order bit position in the address register. If it is out of range, the results are undefined and a trap may ensue.
2.6 Memory Ordering
The Cray X1 system provides minimal implicit memory ordering guarantees between different references from a processor to the same word of memory.
Scalar references are mutually well ordered, but scalar/vector and vector/vector interactions generally are not. While atomic memory operations on the same
Longword are mutually ordered, they are not ordered relative to scalar and vector references.
Explicit memory ordering instructions must appear in the program to ensure that references to overlapping words will execute in the right order.
Suppose that A and B are different references to memory from a given processor.
They can be scalar loads, scalar stores, atomic memory operations, or distinct elements of vector load or store instructions. Suppose that A precedes B in the naive program order, and that they reference overlapping words of memory. If both A and B are loads, no ordering question pertains. Otherwise, the Cray X1 system guarantees that B will reference memory after A only if:
• A and B are both scalar references
28 S–2314–50
Cray X1 Instruction Set Overview [2] gsync gsync gsync gsync msync msync msync lsync lsync lsync lsync lsync lsync lsync lsync a j a j,cpu a j,a a j,r a j a j,p a j,v s,v v,s vr,s v,v v j,v v j,v,el fp i
• B is a store whose address or data cannot be computed without the result of the load A
• A and B are elements of the same ordered vector scatter or zero-stride vector store
• A and B are elements of the same unordered vector scatter and are storing the same value
All other intraprocessor ordering cases require the presence of explicit memory ordering instructions.
Table 17. Explicit Memory Ordering Instructions
; Global ordering of all prior references before all later references
; Global ordering, with all prior instructions complete
; Global ordering of prior scalar loads before all later references
; Global ordering of all prior references before later scalar stores
; MSP ordering of all prior references before all later references
; MSP ordering of all prior stores before all later loads
; MSP ordering of all prior vector references before all later vector references
; Local ordering of prior scalar references before later vector references
; Local ordering of prior vector references before later scalar references
; Local ordering of prior vector loads before later scalar references
; Local ordering of prior vector references before later vector references
; Local ordering of latest vj reference before later vector references
; Local elemental ordering of latest vj reference before later vector references
; Local ordering of prior floating-point traps before later instructions
; Local ordering of prior stores to instruction memory before later instructions
S–2314–50 29
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual
Both the gsync and msync instructions require an aj processor mask operand that specifies a set of processors within the local MSP group of four. They perform a local barrier operation that must be satisfied by all processors in the aj mask with identical mask values before execution can proceed. The aj processor mask is right-justified and the lowest-order bit position corresponds to processor
0 in the local group.
The conjunction of the barrier semantics with the memory ordering effects is unfortunate. When no barrier is needed, a processor mask must be used that has just the bit set that corresponds to the issuing processor.
Interprocessor memory reference ordering is also weakly guaranteed. The architecture specifies only that the references of one processor will be seen in the same order by others, and that once a store is visible to another processor, it will be visible to all. (References to I/O space in system code by a single processor are implicitly mutually ordered.)
2.7 Summary of Rules
To summarize, these rules apply to the Cray X1 system instruction set:
• When a 32-bit datum enters an address or scalar register, it is always right-justified and sign extended to 64 bits.
• When an address or scalar register is used as a 32-bit operand, its upper
32 bits are ignored.
• On memory reference instructions, all immediate offsets, ak indices, and vk gather/scatter offset values are scaled by the data size.
• Whenj and k are distinct register operands of the same kind to the same instruction, then j != k (mod 32).
• A vector register cannot be both operand and result of the same vector type conversion instruction.
• 32-bit vector results must not be used as operands to 64-bit operations and vice versa.
• Memory addresses must be strongly aligned, apart from ,ua references.
• Result vector register elements that are not written become undefined and cannot be assumed to be preserved.
30 S–2314–50
Cray X1 Instruction Set Overview [2]
• Explicit memory ordering instructions are necessary for proper ordering of overlapping references that are not both scalar and are not both loads, unless there is an implicit ordering due to register data dependences.
• gsync instructions are required around atomic memory operations.
• MAXVL is a power of 2 and at least 64 but cannot be assumed to be fixed.
2.8 Special Syntax Forms
For programmer convenience, the assembler supports the following special syntax forms for machine instructions:
Special Form a s v m a a i i i i i i a.SP
a.FP
a.EA
a s v m j j j j expr symbol
Table 18. Special Syntax Forms a i:d a i a i:b a i:c a i:d a63
3
Translated Into a i,L s i,L v i,L m i a i a i:b a i:c a62 a61 a s v
j|0
j|0
j|s0,m0 m j|mj
expr1:a
1 expr2 expr3 expr4
expr1:a
2 expr2 expr3 expr4
1
2
3
The assembler analyzes expr and tries to generate the minimum number of load immediate instructions.
If the symbol is relocatable, the assembler always generates four instructions and the loader fills in the values.
S–2314–50 31
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual a.RA
a.CI
a.CSP
a.CFP
a60 a59 a58 a57
32 S–2314–50
S–2314–50
CAL Pseudo Instruction Overview [3]
Pseudo instructions direct the assembler in its task of interpreting source statements and generating an object file. This chapter briefly describes all the pseudo instructions. For detailed descriptions of the pseudo instructions, see
Chapter 7, page 103. Chapter 9, page 215 further describes the use of the pseudo
instructions that are used for defined sequences; that is, for defining macros and opdefs.
Pseudo instructions are classified and described according to their applications, as follows:
Class
Program control
Loader linkage
Mode control
Section control
Message control
Listing control
Symbol definition
Data definition
Conditional assembly
Micro definition
File control
Defined sequences
Pseudo instructions
IDENT , END, COMMENT
ENTRY , EXT, START,OSINFO
BASE , QUAL, EDIT, FORMAT
SECTION , STACK, ORG, LOC, BITW, BSS, ALIGN
ERROR , ERRIF, MLEVEL, DMSG, MSG, NOMSG
LIST , SPACE, EJECT, TITLE, SUBTITLE, TEXT,
ENDTEXT
= , SET, MICSIZE, DBSM
CON , BSSZ, DATA, VWD
IFA , IFC, IFE, IFM, SKIP, ENDIF, ELSE
CMICRO , MICRO, OCTMIC, DECMIC,HEXMIC
INCLUDE
MACRO , OPDEF, DUP, ECHO, ENDM, ENDDUP,
STOPDUP , LOCAL, OPSYN, EXITM, NEXTDUP
Note: You can specify pseudo instructions in uppercase or lowercase, but not in mixed case.
33
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual
3.1 Pseudo Instruction Format
The source statement format for the pseudo instructions has the same format as used for symbolic machine instructions, but the contents of the fields differ.
Label Result Operand Comment label name arguments ; comment
The fields are separated by white space. Depending on the pseudo instruction, the label field may be required, optional, or ignored. The name in the Result field is the name of the pseudo instruction. The name may be all uppercase or all lowercase but not mixed case. No other subfield besides the name is in the
Result field. The arguments in the Operand field may be required, optional, or ignored. Comments are optional.
3.2 Program Control
The program control pseudo instructions define the limits of a program module and are as follows:
Pseudo
IDENT
END
COMMENT
Description
Marks the beginning of a program module.
Marks the end of a program module.
Enters comment (for example, a copyright) into the generated binary load module.
3.3 Loader Linkage
The loader linkage pseudo instructions provide for the linking of multiple object program modules into one executable program:
Pseudo
ENTRY
EXT
START
Description
Specifies symbols defined as objects, function entry points, or values so that they can be used by other program modules linked by a loader.
Specifies linkage to objects, function entry points, or values defined as entry symbols in other program modules.
Specifies symbolic address at which execution begins.
34 S–2314–50
CAL Pseudo Instruction Overview [3]
OSINFO Specifies information to be passed to the loader about how the object file is to be linked.
3.4 Mode Control
Mode control pseudo instructions define the characteristics of an assembly:
Pseudo
BASE
QUAL
EDIT
FORMAT
Description
Specifies data as being octal, decimal, hex, or a mixture of octal and decimal.
Designates a sequence of code where symbols may be defined with a qualifier, such as a common routine with its own labels.
Turns editing on or off.
Changes the format to old or new.
3.5 Section Control
Section control pseudo instructions control the use of sections and counters in a CAL program:
Pseudo
SECTION
STACK
ORG
LOC
BITW
BSS
ALIGN
Description
Defines specific program sections.
Increments the size of the stack.
Resets location and origin counters.
Resets location counter.
Sets the current bit position relative to the current Longword.
Reserves memory. The reserved memory is uninitialized.
(Deferred implementation) Aligns code.
3.6 Message Control
S–2314–50
The message control pseudo instructions are as follows:
Pseudo
ERROR
ERRIF
Description
Sets an assembly error flag
Sets an assembly error flag according to the conditions being tested
35
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual
MSG and
NOMSG
DMSG
Specifies that certain messages should be issued or not issued
Issues a comment-level message containing the string found in the operand field
3.7 Listing Control
The listing control pseudo instructions are as follows:
Pseudo
LIST
TEXT
ENDTEXT
Description
Controls listing by specifying particular listing features that will be enabled or disabled
SPACE
EJECT
Inserts blank lines in listing
Begins new page
TITLE Prints main title on each page of listing
SUBTITLE Prints subtitle on each page of listing
Declares beginning of global text source
Terminates global text source
3.8 Symbol Definition
The symbol definition pseudo instructions are as follows:
Pseudo
=
SET
MICSIZE
DBSM
Description
Equates a symbol to a value; not redefinable.
Sets a symbol to a value; redefinable.
Equates a symbol to a value equal to the number of characters in micro string; redefinable.
Generates a named label entry in the debug symbol tables with a specific type specified.
3.9 Data Definition
36
Data definition pseudo instructions are the only pseudo instructions that generate object binary. The only other instructions that are translated into object binary are the symbolic machine instructions. An instruction that generates
S–2314–50
CAL Pseudo Instruction Overview [3] binary cannot be used within a section that does not allow instructions, data, or both.
The data definition pseudo instructions are as follows:
Pseudo
CON
BSSZ
DATA
VWD
Description
Places an expression value into one or more Longwords
Generates bytes that have been initialized to 0
Generates one or more Longwords of numeric or character data
Generates a variable-width field, 0-64 bits wide, of data
3.10 Conditional Assembly
The conditional assembly pseudo instructions permit the optional assembly or skipping of source code.
The conditional assembly pseudo instructions are as follows:
Pseudo
IFA
IFE
IFC
IFM
SKIP
ENDIF
ELSE
Description
Tests expression attributes; address and relative attributes.
Tests two expressions for some assembly condition; less than, greater than, and equal to.
Tests two character strings for assembly condition; less than, greater than, and equal to.
Test for machine characteristics.
Unconditionally skip subsequent statements.
Terminates conditional code sequence.
Reverses assembly condition.
3.11 Micro Definition
Through the use of micros, programmers can assign a name to a character string and subsequently refer to the character string by its name. A reference to a micro results in the character string being substituted for the name before assembly of the source statement containing the reference.
The following pseudo instructions specify micro definition:
S–2314–50 37
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual
Pseudo
CMICRO
MICRO
OCTMIC
DECMIC
HEXMIC
Description
Constant , non-redefinable micro; assigns a name to a character string.
Redefinable micro; assigns a name to a character string.
Converts the octal value of an expression to a character string and assigns it a redefinable name.
Converts the decimal value of an expression to a character string and assigns it a redefinable micro name.
Converts the hex value of an expression to a character string and assigns it a redefinable micro name.
In addition to the micros previously listed, the assembler provides predefined micros. They can be specified in all uppercase or all lowercase, but not mixed case. CAL provides the following predefined micros:
Micro
$DATE
$JDATE
$TIME
$MIC
$CNC
$QUAL
$CPU
$CMNT
$APP
Description
Current date: ‘ mm / dd / yy ’
Julian date: ‘ yyddd ’
Time of day - ‘ hh : mm : ss ’
Micro character: double quotation mark (")
Concatenation character: underscore (_).
Name of qualifier in effect; if none, the string is null.
Target machine: ‘CRAY X1’.
Comment character used with the new format: semicolon (;).
Append character used with the new format: backslash (must be immediately followed by a newline).
The following example illustrates the use of a predefined micro ($DATE):
DATA ’THE DATE IS "$DATE"’
DATA ’THE DATE IS 01/01/00’ <edited line>
You can reference micro definitions anywhere in a source statement, except in a comment, by enclosing the micro name in quotation marks. If column 72 of a line is exceeded because of a micro substitution, the assembler creates additional continuation lines. No replacement occurs if the micro name is unknown or if one of the micro marks was omitted.
38 S–2314–50
CAL Pseudo Instruction Overview [3]
In the following example, a micro called PFX is defined as the character string
ID . A reference to PFX is in the label field of a line.
"PFX"TAG A1 S1 ; Left-shifted three spaces when edited.
In the following example, before the line is interpreted, the assembler substitutes the definition for PFX producing the following line:
IDTAG A1 S1 ; Left-shifted three spaces when edited.
The following example shows the use of the predefined micros, AREGSIZE and PREGSIZE:
A = "AREGSIZE"
CON A
B = "PREGSIZE"
CON B
; Size of the A registers.
; Store value in memory.
; Size of the Program register.
; Store value in memory.
3.12 Defined Sequences
The defined sequences pseudo instructions are as follows:
Pseudo
MACRO
OPDEF
DUP
ECHO
ENDM
ENDDUP
STOPDUP
LOCAL
Description
A sequence of source program instructions saved by the assembler for inclusion in a program when called for by the macro name. The macro call resembles a pseudo instruction.
A sequence of source program instructions saved by the assembler for inclusion in a program called for by the OPDEF pseudo instruction. The opdef resembles a symbolic machine instruction.
Introduces a sequence of code that is assembled repetitively a specified number of times; the duplicated code immediately follows the DUP pseudo instruction.
Introduces a sequence of code that is assembled repetitively until an argument list is exhausted.
Ends a macro or opdef definition.
Terminates a DUP or ECHO sequence of code.
Stops the duplication of a code sequence by overriding the repetition condition.
Specifies unique strings that are usually used as symbols within a
MACRO , OPDEF, DUP, or ECHO pseudo instruction.
S–2314–50 39
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual
OPSYN
EXITM
Defines a label field name that is the same as a specified operation in the operand field name.
Terminates the innermost nested MACRO or OPDEF expansion.
3.13 File Control
The file control pseudo instruction, INCLUDE, inserts a file at the current source position. The INCLUDE pseudo instruction always prepares the file for reading by opening it and positioning the pointer at the beginning.
You can use this pseudo instruction to include the same file more than once within a particular file.
You can also nest INCLUDE instructions. Because you cannot use INCLUDE recursively, you should review nested INCLUDE instructions for recursive calls to a file that you have already opened.
40 S–2314–50
CAL Program Organization [4]
The CAL program consists of sequences of source statements organized into program modules, program segments, and sections. This chapter describes the organization of a CAL program. Source statements are described in detail in
The end of the chapter includes some small sample programs that provide the basic structure for getting started in writing CAL programs.
4.1 Program Modules
A program module is the main body of source statements and resides between the IDENT and END pseudo instructions. (For more information on pseudo
instructions, see Chapter 3, page 33 and Chapter 7, page 103.) The IDENT
pseudo instruction marks the beginning of a program module. The END pseudo instruction marks the end of a module. Any definitions between these two pseudo instructions apply only to the program module in which the definition resides.
4.2 Global Definitions and Local Definitions
CAL definitions are pseudo instruction source statements that do not generate code; they define and assign values to symbols, macros, opdefs, and micros. (For
information on symbols, see Section 5.3.9.1, page 64. For information on macros,
opdefs, and micros, see Section 5.3.2.2.3, page 56.)
CAL definitions that occur before the first IDENT pseudo instruction or between the END pseudo instruction that terminates one program module and the IDENT that begins the next program module, and are not the exceptions noted below, are
global definitions. A global definition can be referenced without redefinition from within any of the program segments that follow the definition. Symbols defined within the global definitions area cannot be qualified.
CAL definitions that occur between an IDENT pseudo instruction and the following END pseudo instruction are local definitions. A local definition can be referenced only from within the program module in which it was defined.
The exceptions are redefinable micros, redefinable symbols, and symbols of the form %%x where x is 0 or more identifier characters. These definitions are local to the program segment in which they were defined, even if they were defined
S–2314–50 41
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual
4.3 Program Segments
A program segment consists of global definitions, a program module, or a combination of global definitions and a program module. The END statement that terminates a module always terminates the segment containing it.
4.4 Program within the global definition area. They are not known to the assembler after the
END pseudo instruction that terminates that program segment, and they are not included in the cross-reference listing.
The CAL program consists of one or more segments. Figure 1, page 43 illustrates
the organization of the CAL program.
42 S–2314–50
S–2314–50
CAL Program Organization [4]
Program
Segment - 1
Global definitions - 1
Program module - 1
Segment - 2
Global definitions - 2
Program module - 2
Segment - n
Global definitions - n
Program module - n
Figure 1. CAL Program Organization a12249
43
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual
4.5 Sections
A CAL program module can be divided into blocks of memory called sections.
By dividing a module into sections, you can conveniently separate sequences of code from data. As the assembly of a program progresses, you can explicitly or implicitly assign code to specific sections or reserve areas of a section.
The assembler assigns locations in a section consecutively as it encounters instructions or data destined for that particular memory section.
Use the main and literals sections for implicitly assigned code. The assembler maintains a stack of section names assigned by the SECTION pseudo instruction.
All sections except stack sections are passed directly to the loader.
Sections can be local or common. A local section is available to the CAL program module in which it resides. A common section is available to another CAL program module.
To assign code explicitly to a section, use the SECTION pseudo instruction. For
more details, see Section 7.53, page 165.
4.6 Examples
This section includes examples that show the basic structure of CAL programs.
4.6.1 Example 1: Global and Local Definitions
The following example illustrates a CAL program and the definition and the use of symbols in the context of segments and modules.
SYM1 = 1
SYM2 SET 2
%%SYM3 = 3
%%SYM4 SET 4
; Begin segment 1 global
; SYM1 cannot be redefined
; SYM2 equals 2 for this module
; Gone at the end of the module
; Gone at the end of the module
IDENT TEST1
S1 SYM1
S2 SYM2
S3 %%SYM3
S4 %%SYM4
END
SYM2 SET 3
; Beginning of module 1
; Register S1 gets 1
; Register S2 gets 2
; Register S3 gets 3
; Register S4 gets 4
; End of segment 1 and module TEST 1
; Beginning of segment 2
44 S–2314–50
CAL Program Organization [4]
%%SYM3 = 5
IDENT TEST2
S1 SYM1
S2 SYM2
S3 %%SYM3
S4 %%SYM4
END
IDENT TEST3
S1 SYM1
S2 SYM2
S3 %%SYM3
END
; Global definitions
; Beginning of module TEST 2
; Register S1 gets 1
; Register S2 gets 3
; Register S3 gets 5
; Error: not defined
; End of segment 2 and module TEST 2
; Beginning of segment 3 and module TEST 3
; Register S1 gets 1
; Error: not defined
; Error: not defined
; End of segment 3 and module TEST 3
4.6.2 Example 2: Sections and Qualifiers
The following example illustrates section specification and deletion and indicates the current section. The example includes the QUAL pseudo instruction. For a
description of the QUAL pseudo instruction, see Section 7.52, page 163.
IDENT STACK
SYM1 = 1
QUAL QNAME1
SYM2 = 2
SNAME SECTION MIXED
MLEVEL ERROR
SYM3 = *
MLEVEL *
SECTION *
SYM4 = 4
QUAL QNAME2
; The IDENT statement puts the first entry
; on the list of qualifiers. This entry
; starts the symbol table for unqualified
; symbols.
; SYM1 is relative to the main section.
; Second entry on the list of qualifiers.
; SYM2 is the first entry in the symbol
; table for QNAME1.
; SNAME is the second entry on the list of
; sections
; Reset message level to error eliminate
; warning level messages.
; SYM3 is the second entry in the symbol
; table for QNAME1 and is relative to the
; SNAME section.
; Reset message level to default in effect
; before the MLEVEL specification.
; SNAME is deleted from the list of
; sections.
; SYM4 is the third entry in the symbol
; table for QNAMEl and is relative to the
; main section.
; Third entry on the list of qualifiers.
S–2314–50 45
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual
SYM5 = 5 ; SYM5 is the first entry in the symbol
; table for QNAME2.
SYM6 = /QNAME1/SYM2 ; SYM6 gets SYM2 from the symbol table for
; QNAMEl even though QNAME1 is not the
SYM7
QUAL
= 6
*
; current qualifier in effect.
; QNAME2 is removed as the current
; qualifier name.
; SYM7 is the fourth entry in the symbol
SYM8 = 7
; table for QNAME1.
; Second entry in the symbol table for
; unqualified symbols.
46 S–2314–50
CAL Assembler Invocation [5]
This chapter describes the as command and its options for invoking the CAL assembler.
5.1 Assembler Command Line
The as(1) command invokes the CAL assembler. The format of the as(1) command is: as
[-b bdflist]
[-B]
[-c bdfile]
[-C cpu]
[-D micdef]
[-f]
[-F]
[-g symfile]
[-G]
[-h]
[-H]
[-i nlist]
[-I options]
[-j]
[-J]
[-l lstfile]
[-L msgfile]
[-m mlevel]
[-M [no]message= n]
[-n number]
[-o objfile]
[-U]
[-v]
[-V]
[-x] file
S–2314–50 47
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual
The as(1) command assembles the specified file. The following options, each a separate argument, can appear in any order, but they must precede the file argument:
-b bdflist
Reads the binary definition files specified by bdflist. bdflist can be in one of the following forms:
• List of files separated by a comma
• List of files enclosed in double quotation marks and separated by a comma and/or one or more spaces
The assembler first reads the default binary definition file, unless suppressed by the -B option, and then reads the files in the order specified in bdflist. For details on binary
definition files, see Section 5.3.2, page 54.
-B
Suppresses reading of the default binary definition file. The default is to read the default binary definition file.
-c bdfile
Creates binary definition file bdfile. By default, the assembler
does not create a binary definition file. See Section 5.3.9, page 63.
-C cpu
Generates code for the specified CPU. By default, code is generated for the CPU specified by the TARGET environment variable. The -C option overrides the TARGET environment variable. If neither the TARGET environment variable nor the
-C option are specified, the assembler generates code for a generic Cray X1 system.
cpu can be in one of the following forms: primary charc
[,charc] where primary is cray-x1 and charc is a comma-separated list of CPU characteristics (currently, no CPU characteristics are supported).
48 S–2314–50
S–2314–50
CAL Assembler Invocation [5]
-D micdef
-f
-F
-g symfile
-G
-h
-H
-i nlist
Defines a globally defined constant micro mname, as follows:
micdef is mname[=[string]]
mname must be a valid identifier. If the = character is specified, it must immediately follow mname. The string that immediately follows the = character, if any, is associated with mname. If you do not specify string, then mname will be associated with an empty string.
If mname was defined as a micro by use of a binary definition file, the mname specified on the command line overrides the
mname defined within the binary definition file; in that case, the assembler issues a note-level diagnostic message.
Enables new statement format (default).
Enables old statement format.
Creates assembly output symbol file symfile. By default, the assembler does not create an output symbol file.
Forces all symbols to symfile. By default, unreferenced symbols are excluded.
Enables all list pseudo instructions, regardless of location field name.
Disables all list pseudos, regardless of location field name.
Restricts list pseudo processing to pseudos with location field names that are in nlist.
49
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual
50
-I options m
M d
D e
E h
H l
L s
S t
T
Lists options; overrides the list pseudo instruction. Specifying conflicting options is prohibited. Can be one or more options in any order without intervening spaces.
b
B c
C n
N p
P
Enables macro/opdef/dup/echo expansion binary only
Disables macro/opdef/dup/echo expansion binary only (default)
Enables macro/opdef/dup/echo expansion conditionals
Disables macro/opdef/dup/echo expansion conditionals (default)
Enables dup/echo expansions
Disables dup/echo expansions (default)
Enables edited statement listing (default)
Disables edited statement listing
Enables page header in listings
Disables page header in listings (default)
Enables listing control pseudo instructions
Disables listing control pseudo instructions
(default)
Enables macro/opdef expansions
Disables macro/opdef expansions (default)
Enables inclusion of nonreferenced local symbols in the cross-reference listing (default)
Disables inclusion of nonreferenced local symbols in the cross-reference listing
Enables macro/opdef/dup/echo expansion of pre-edited lines
Disables macro/opdef/dup/echo expansion of pre-edited lines (default)
Enables source statement listing (default)
Disables source statement listing
Enables text source statement listing
Disables text source statement listing (default)
S–2314–50
S–2314–50
CAL Assembler Invocation [5] x
X
Enables cross-reference listing (default)
Disables cross-reference listing
-j
Enables editing. Editing is enabled by default.
-J
Disables editing. Editing is enabled by default.
-l lstfile
Writes source statement listing to lstfile. By default, the assembler does not generate a diagnostic listing.
-L msgfile
Writes diagnostic message listing to msgfile. By default, the assembler does not generate a diagnostic message listing.
-m mlevel
Sets the message level for diagnostic listing, diagnostic message listing, and the standard error file. mlevel must be one of the following: comment, note, caution, warning, or error
Listing messages of the specified mlevel and higher are issued.
The default message level is warning.
-M [no]message=n[:n]
Enables or disables specified diagnostic messages. n is the number of the message to be enabled or disabled. You can specify more than one message number; multiple numbers must be separated by a colon with no intervening spaces. For example, to disable messages as-17 and as-74, specify:
-M nomessage=17:74
This option overrides the -m mlevel option for the specified messages. If n is not a valid message number, it is ignored. Only messages of level comment, note, caution, and warning are affected by this option. Messages of level error or internal are not affected.
51
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual
-n number
-o objfile
-U
-V
-x file
Sets the maximum number of diagnostic messages inserted into the listing files. Files used are the source listing, diagnostic message listing, and the standard error file. number must be 0 or greater; the default is 100.
The relocatable object file is objfile. By default, the relocatable object file name is formed by removing the directory name, if any, and the .s suffix, if any, from the input file and by appending the .o suffix.
Forces source code to be converted to uppercase. Quoted strings are protected, as are embedded micros. Both new and old format statement types are supported.
Writes the assembler version to standard error.
Assemble instructions in interactive input/output mode. In this mode, the assembler takes as input from standard input a symbolic machine instruction and writes to standard output the
32-bit hex form. This can be used as input to the debugger to modify code being debugged.
Binary definition files are not processed. No instruction labels, macros, opdefs or micros are allowed. The interactive session is terminated by a ctrl-D.
Assembles the named file; all options must precede the file argument.
5.2 Environment Variables
The following subsections describe environment variables used by the assembler.
How the environment variables are set depends on the type of shell being used.
52 S–2314–50
CAL Assembler Invocation [5]
5.2.1 ASDEF Environment Variable
The environment variable ASDEF specifies the full pathname of the default binary definition file. This environment variable is normally set by the cal modulefile in the programming environment but the user may redefine the variable. If
ASDEF is not set, the assembler uses the pathname relative to the assembler executable (../lib/asdef), normally installed with the assembler, as the default path to binary definition file.
5.2.2
LPP Environment Variable
The LPP environment variable sets the number of lines per page for the source statement listing file and diagnostic message file. The valid range is 4-999. If the specified value is outside this range, the value is set to the default value. By default, the number of lines per page is 55.
5.2.3
TMPDIR Environment Variable
The TMPDIR shell variable specifies a directory that the assembler uses for its temporary file. If the directory is not specified or is specified incorrectly, the assembler uses the system default. The default is site specific.
5.2.4
TARGET Shell Variable
The TARGET environment variable determines the characteristics of the machine for which the code is generated.
The value of TARGET can be in one of the following forms:
primary [,charc]
, charc where primary is cray-x1 and charc is a comma separated list of CPU characteristics (currently, no cpu characteristics are supported).
5.3 Assembler Execution
When the assembler executes, it reads the specified source file and optionally reads one or more binary definition files. The source file may have one or more
INCLUDE pseudo instructions which will cause the assembler to also read those source files.
S–2314–50 53
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual
The assembler makes two passes for each program segment. During the first pass, specified binary definition files are read, each source statement in the source file and included files is read, defined sequences (such as macros) are expanded, and memory addresses are assigned. During the second pass, values are substituted for the symbolic operands and addresses, any appropriate diagnostic messages are written, listing files, if specified, are written, and if there are no errors of ERROR or INTERNAL severity, an object file is generated. If specified, a binary definition file is also generated.
5.3.1 Reading Source Files
The assembler takes one source file as the primary input for assembly. The source
file may contain more than one program segments. See Chapter 4, page 41.
5.3.2 Using Binary Definition Files
If one or more binary definition files is specified, the assembler reads the binary definition files before reading the source file. The global definitions (symbols, micros, macros, opdefs, and opsyns) in the binary definition file are then visible to the assembler as it processes the source statements in the source file.
Note: System- and user-defined binary definition files are identical in all respects. Both types of files are created and used in exactly the same manner.
In this manual, they are treated as separate entities to encourage you to define binary definition files that meet your particular programming requirements.
You can create user-defined binary definition files by using either of the following methods:
• Copying the system-defined binary definition files and then modifying the new file either by adding new definitions or by redefining existing definitions.
• Disabling the recognition of system-defined binary definition files and accumulating the defined sequences entirely from an assembler source program.
You can specify more than one binary definition file with each assembly. If more than one binary definition file is specified, the files are processed from left to right in the order specified by the -b option.
Lines or sequences of code assembled and stored in a binary definition file, can be accessed without reassembly. This means accessing a binary definition file directly saves assembler time.
54 S–2314–50
CAL Assembler Invocation [5]
When binary definition files are read, they are checked for the following:
• CPU compatibility
• Multiple references to the same definition
5.3.2.1 CPU Compatibility Checking
The assembler permits access to any previously defined file with one restriction.
Binary definition files are marked with the CPU type for which they were created. Binary definition files created on one Cray system is not necessarily compatible with all Cray systems. If a binary definition file is not compatible with the system you are using, the binary definition file is not accepted and the following message is issued:
Incompatible version of binary definition file ‘ file’
This check ensures that the machine on which the binary definition file was created is compatible with the program trying to use it. Some CAL pseudo instructions have restricted use based on hardware and software requirements.
The binary definition file compatibility check prevents the mixing of binary definition files and ensures that hardware and software restrictions are not violated.
5.3.2.2 Multiple References to a Definition
The assembler checks for multiple references to definition names for macros and opsyns, location field names for symbols and micros, and syntax for opdefs.
The following subsections describe how multiple references to a definition are resolved.
5.3.2.2.1 Symbols
If a symbol is defined in more than one binary definition file, the definitions are compared. If the definitions are identical, the assembler disregards the duplicates and makes one entry for the symbol from the binary definition files. If a symbol is defined more than once and the definitions are not identical, the assembler uses the last definition associated with the location field name and issues the following diagnostic message:
Symbol ‘name’ is redefined in file ‘file’
S–2314–50 55
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual
5.3.2.2.2 Macros
5.3.2.2.3 Opdefs
5.3.2.2.4 Opsyn
56
If a macro with the same name is defined in more than one binary definition file, the definitions are compared. If the definitions associated with the macro’s name are identical character by character, the assembler disregards the duplicate definition and makes one entry for the macro from the binary definition files. If the name of the macro is used more than once, and the definitions associated with the name are not identical character by character, the assembler uses the definition associated with the last reference to the name and issues the following diagnostic message:
Macro name in filefile replaces previous definition
If a macro is defined with the same name as a pseudo instruction, the macro replaces the pseudo instruction and the assembler issues the same message as shown above.
If an opdef with the same syntax is defined in more than one binary definition file, the definitions of the opdefs are compared. If the definitions of the two opdefs are exactly the same, the assembler disregards the duplicate definition and makes one entry for the opdef from the binary definition files. If the same syntax appears more than once and the definitions are not exactly the same, the syntax associated with the last reference to the opdef is used as its definition and the assembler issues the following diagnostic message:
Opdef name in file file replaces previous definition
If an opdef is defined with the same syntax as a machine instruction, the opdef replaces the machine instruction and the assembler issues the message shown above.
If an opsyn with the same name is defined in more than one binary definition file, the definitions are compared. If the definitions are identical, the assembler disregards the duplicate definition and makes one entry for the opsyn from the binary definition files. If the name for an opsyn is used more than once and the definitions are not identical, the assembler uses the definition associated with the last reference to the opsyn name and issues the following diagnostic message:
Opsyn name in file file replaces previous definition
S–2314–50
CAL Assembler Invocation [5]
If an opsyn is defined with the same name as a pseudo instruction, the opsyn replaces the pseudo instruction and the assembler issues the message as shown above. Pseudo instructions have an internal code that permits the assembler to identify them when they are encountered. When an opsyn is used to redefine an existing pseudo instruction, the assembler copies the predefined internal code of that pseudo instruction and uses it for identification in the binary definition file.
5.3.2.2.5 Micros
If a micro with the same location field name is defined in more than one binary definition file, the micro strings associated with the location field names are compared. If the strings are identical, the assembler disregards the duplicate definition and makes one entry for the micro from the binary definition files. If the micro is used more than once and the strings associated with the micro names are not exactly identical, the assembler uses the string associated with the last reference to the micro name and issues the following diagnostic message:
Micro ‘name’ in file ‘file’ replaces previous definition
5.3.3 Included Files
The INCLUDE pseudo instruction inserts a file at the current source position. The
INCLUDE pseudo instruction always prepares the file for reading by opening it and positioning the pointer at the beginning.
5.3.4 Source Statement Listing File
The assembler generates a source statement listing and a cross-reference listing.
You can control the format of these listings by using the listing control pseudo instructions (see Section 5.6, page 101) or by using the -i, -I, -l, -L, -a, -n, -h,
and -H options on the as(1) command line (see Section 5.1, page 47).
Each page of listing output produced by the assembler contains three header lines. Figure 1 shows the format of the page header.
S–2314–50 57
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual
Assembler version #
Date and time
Section and qualifier
Title
Subtitle
Scale (1-72 characters wide)
Global page #
Local page #
Cray Inc. System a12245
Figure 2. Page Header Format
The three lines of the page header are described as follows:
• The first line contains, from left to right, the assembler version number, the title of the program, and a page number that is global over all programs assembled by the current assembly. If you do not specify a title by using a
TITLE pseudo instruction, the title is taken from the operand field of the
IDENT pseudo instruction.
• The second line contains, from left to right, the date and time of assembly, a subtitle if specified with a SUBTITLE pseudo instruction, and a page number that is local for this listing.
• On the third line, the left-most entry is a local section name if specified in a SECTION pseudo instruction. To the right of the local section name is a symbol qualifier name if specified by a QUAL pseudo instruction. The next field is a horizontal scale that is 72 characters wide, numbered from 1 through 72. This scale appears directly over your source code and helps you to differentiate the four fields of your source statements. On the far right of the third line is the name of the Cray Inc. system for which the code was generated.
5.3.4.1 Source Statement Listing
The format of the source statement listing, as shown in Figure 3, appears directly
under the page header and contains five columns of information, as follows:
58 S–2314–50
CAL Assembler Invocation [5]
Location address Hex code Line number
Source statement or
Diagnostic message
Sequence field a12246
Figure 3. Source Statement Listing Format
The five columns of the source statement listing are described as follows:
• The location address column contains the hex representation of the byte address of the current statement.
• The hex code column contains the hex representation of the current instruction or numeric value.
If the numeric value is an address, the hex code has one of the following suffixes:
– + (Relocation in program block)
– C (Common section)
– D (Dynamic section)
– S (Stack section)
– T (Task common)
– Z (Zero common)
– : (Immobile attribute)
– X (External symbol)
– None (Absolute address)
• The results of several pseudo instructions can also appear in the hex code column:
– The hex value of symbols defined by the SET, MICSIZE, or = pseudo instruction
– The hex value of the number of bytes reserved by the BSS or BSSZ pseudo instruction
S–2314–50 59
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual
– The hex value of the number of bytes skipped as a result of the ALIGN pseudo instruction
– The hex value of the number of characters in a micro string defined by a
MICRO , OCTMIC, or DECMIC pseudo instruction
• The line number column contains the line number of the source code.
• The source statement column is 72 characters wide and holds columns 1 through 72 of each source statement.
• The diagnostic column contains a diagnostic message immediately following a statement that contains an error.
• The sequence field column contains either an identifier or information taken from columns 73 through 90 of the source line image (which may be empty).
It contains an identifier if the line is an expansion of a macro or opdef, or if the line was edited.
5.3.5 Cross-reference Listing
The assembler generates a cross-reference listing in the format shown in Figure
4. The assembler lists symbols alphabetically and groups them by qualifier if the QUAL pseudo instruction has declared qualifiers. If qualifiers were declared, each new qualifier appears on a fresh page. The qualifier name appears on the third line of the page header.
The cross-reference listing does not include unreferenced symbols that are defined between TEXT and ENDTEXT pseudo instructions and it does not include symbols of the form %%xxxxxx; x is zero or more identifier characters.
Note: The cross reference page header is nearly identical to the page header of the assembler listing; the difference is that the string Symbol cross reference is printed out in the middle field of the third line of the cross-reference listing.
60 S–2314–50
CAL Assembler Invocation [5]
Assembler version #
Date and time
Section and qualifier
Symbol Value .
Title
Subtitle
"Symbol cross reference"
Symbol references
Global page #
Local page #
Cray Inc. system a12247
Figure 4. Cross-reference Listing Format
The information in each column is described as follows:
• The symbol column contains the symbol name.
• The value column contains the hex value of the symbol.
The hex value of the symbol may have one of the following suffixes:
– + (Relocation in program block)
– C (Common section)
– D (Dynamic section)
– S (Stack section)
– T (Task common)
– Z (Zero common)
– : (Immobile attribute)
– X (External symbol)
– None (Absolute address)
• The period (.) column separates the value field from the symbol reference fields and is called the separator.
• The symbol references column contains one or more references to the symbol.
The assembler references symbols using the following format: page:line x
S–2314–50 61
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual where line is the decimal representation of the line number that contains the reference.
x represents the type of reference as follows:
– A blank in this column means the symbol value is used at the specified line.
– D means the symbol is defined in the location field of an instruction or else by a SET, =, or EXT pseudo instruction.
– E means the symbol is an entry name.
– F means the symbol is used in an expression on an IFE, IFA, or ERRIF conditional pseudo instruction.
– R means the symbol is used in an address expression in a memory read instruction.
– S means the symbol is used in an address expression in a memory store instruction.
If a symbol is defined in text between TEXT and ENDTEXT pseudo instructions, the cross-reference listing reports the associated TEXT name on the line below the symbol reference.
If a symbol is defined in a binary definition file, the cross-reference listing reports the associated file name on the line below the symbol reference.
5.3.6 Diagnostic Messages
The assembler generates diagnostic messages that provide user information about the assembly. Diagnostic messages are classified by level of severity from low to high, as follows:
• Comment (statistical information)
• Note (possible assembly problems)
• Caution (definite user errors during assembly)
• Warning (possible error such as truncation of a value)
• Error (fatal assembly error, you should check the message and source code carefully for possible mistakes)
62 S–2314–50
CAL Assembler Invocation [5]
5.3.7 Diagnostic Message Listing File
If the -L option is specified, the diagnostic messages are also written to the diagnostic message listing file.
5.3.8 Object File
Note: To print comment-, note-, and caution-level diagnostic messages to the standard error file, you must specify the -m option on the as(1) command line.
• Internal (The assembler detected an error in the assembler itself. Assembly cannot continue.)
The diagnostic messages are written to the standard error file. If the -l option is specified, the diagnostic messages are also written to the source statement listing file immediately following the source statement that caused the diagnostic message.
The object file that is generated consists of a sequence of ELF tables. The loader processes the object file. For details on ELF files, see the elf(3e) man page.
5.3.9 Creating a Binary Definition File
To create binary definition files, include the -b and -c options on the as(1) command line. The -b option accepts a list of files separated by commas or a list of files enclosed in double quotation marks and separated by spaces or commas as arguments.
Only certain types of lines or sequences of code are permitted in a binary definition file. Binary definition files are always created from the global part of program segments and from any currently accessed binary definition files.
Typically, binary definition files are created from source programs that include
one segment that contains a global part, but has no program module (see Section
Additions can be made to binary definition files from assembler source programs that include program modules, however, not all lines or sequences of code in the global part are added.
Note: Under no circumstance is any line or sequence of code added to a binary definition file from an assembler program module. All additions to binary definition files come from the global part of the segment.
S–2314–50 63
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual
5.3.9.1 Symbols
5.3.9.2 Macros
Binary definition files are composed of lines or sequences of code classified as follows:
• Symbols
• Macros
• Opdefs
• Opsyns
• Micros
Each line or sequence of code added to a binary definition file must be in one of these classes and must satisfy the requirements for that particular class.
The assembler accumulates the symbols to be included in a new binary definition file from all currently accessed binary definition files and from all of the global parts of program segments that fit the following requirements:
• Symbols cannot be redefinable.
To be included in a binary definition file, a symbol must be defined with the
= (equates) pseudo instruction. Symbols defined with the SET or MICSIZE pseudo instruction are redefinable; therefore, they are not included in a binary definition file.
• Symbols cannot be preceded by %%.
This exclusion applies to symbols that are created by the LOCAL and = pseudo instructions.
The assembler identifies all of the symbols in the global part of program segments that meet the preceding requirements and includes them in the creation of a binary definition file.
The assembler accumulates the macros to be included in a new binary definition file from all currently accessed binary definition files and from all of the global parts of segments within a source program.
64 S–2314–50
CAL Assembler Invocation [5]
5.3.9.3 Opdefs
5.3.9.4 Opsyns
5.3.9.5 Micros
5.4 Linking
The assembler accumulates opdefs (operation definitions) to be included in a new binary definition file from all currently accessed binary definition files and from all of the global parts of segments within a source program.
The assembler accumulates opsyns (operation synonyms) to be included in a new binary definition file from all currently accessed binary definition files and from all of the global parts of segments within a source program.
The assembler accumulates micros to be included in a new binary definition file from all currently accessed binary definition files and from all of the global parts of segments within source program. Only micros that cannot be redefined are included in a binary definition file. A micro must be defined using the CMICRO pseudo instruction to be included in a binary definition file.
To create an executable program containing one or more assembler-generated object files, invoke the loader to link the object files with compiler- generated object files and/or object files contained in libraries. The loader may be invoked with commands ld(), cc(), or ftn().
S–2314–50 65
S–2314–50
CAL Source Statements [6]
This chapter describes in detail the format and syntax for source statements, the components of source statements, and the attributes of symbols and expressions.
The assembly language capabilities described here and in Chapter 9, page 215
provide very powerful programming capabilities beyond simply specifying symbolic machine instructions.
A CAL program is expressed in the form of source statements. CAL source statements fall into five categories:
• A symbolic machine instruction describes symbolically a Cray X1 machine instruction.
• A pseudo instruction controls the assembly process or results in generated data but does not result in generated machine instructions.
• A macro instruction represents a user-defined sequence of symbolic machine instructions and/or pseudo instructions.
• An opdef instruction represents a user-defined sequence of one or more symbolic machine instructions and has the same syntax as a symbolic machine instruction.
• A comment allows for annotation of a program but does not result in any generated machine instructions nor does it affect the assembly process.
The components of a statement are one or more of:
• A symbol is an identifier that has a value and several attributes.
• A tag is an identifier used in the label field of macros and other defined sequences but does not have a value or attributes.
• A mnemonic is an identifier that is a fixed part of a symbolic machine instruction or pseudo instruction.
• A constant is a character string that does not contain any operators and can be evaluated to a fixed value at assembly time.
• A data item specifies a numerical or character value in the operand field of the
DATA pseudo instruction and in literals.
• A micro is an identifier that names a character string.
• A location element represents the current value of the location counter.
67
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual
• An expression is a combination of one or more operands (constants, symbols, location elements) and zero or more operators that evaluates to a value with attributes. In CAL, registers are not a part of an expression, but registers and expressions with operators can be part of a symbolic machine instruction.
6.1 Source Statement Format
Each CAL source statement has three fields separated by white space and optionally followed by a comment.
Label Result Operand Comment
6.1.1 Label Field
The content of the label field depends on the requirements of the result and/or operand fields of each particular source statement. The label field of all machine instructions can optionally contain a symbol. If the label field of a machine instruction contains a symbol, the symbol value is set equal to the current value
of the location counter. (See Chapter 4, page 41.)
When an instruction uses the label field, it begins in column 1 and is terminated by white space. The label field also can contain an asterisk (*) in column 1 to identify a comment line.
6.1.2 Result Field
The content of the result field depends on the particular instruction. The result field of pseudo instructions and macro instructions must match existing pseudo instructions or macros. Machine or opdef instructions can contain 1–5 subfields.
The subfield can be empty, contain expressions, or consist of registers or operators. The result field begins with the first non-whitespace character following a label field that is not empty and usually ends with white space or a semicolon. If column l is empty, the result field can begin in column 2 or subsequent columns. A blank result field following a label field produces a diagnostic message.
6.1.3 Operand Field
Before the operand field can be specified, it must be preceded by a result field.
68 S–2314–50
CAL Source Statements [6]
If the statement is a symbolic machine instruction, the operand field contains the operation being performed. However, it can contain other information, depending on the particular instruction. The syntax of the operand field is identical to that of the result field. Machine or opdef instructions can contain one through five subfields.
Usually, the operand field begins with the first white space character following a result field that is not empty and ends with white space or a semicolon.
6.1.4 Comment Field
The comment field is optional and can be specified with an asterisk or a semicolon. A semicolon comment can be in any column, including column
1. If an asterisk is used to indicate a comment, it must appear in column 1.
Generally, a comment that begins in column 1 is specified by using an asterisk and a comment that begins in any other column is specified by using a semicolon.
Usually, comment fields are not edited. For more information about editing
comment fields, see Section 6.10, page 97.
The following example illustrates the use of the comment field: ident test1
*Asterisk in column 1 denotes comment line
; Semicolon begins comment end test1
6.1.5 Old Format
S–2314–50
For compatibility with a prior generation of assemblers, an old source statement format is supported. The old format has the following differences:
• Only the space character is treated as a field separator.
• The semicolon character does not start a comment.
• The label field can start in either column 1 or column 2.
• A comma in column 1 indicates a continuation of the statement on the previous line.
• If the label field terminates before column 33, the result field must begin before column 35; otherwise, the field is considered empty.
• If the result field terminates before column 33, the operand field must begin before column 35; otherwise, the field is considered empty. If the result field
69
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual extends beyond column 32, however, the operand field must begin after not more than one blank separator and can begin after column 35.
• The comment field begins with the first nonblank character following the operand field or, if the operand field is empty, does not begin before column
35. If the result field extends beyond column 32 and no operand entry is provided, two or more blanks must precede the comment field.
• If editing is enabled, comments are edited.
The old format is specified by either the FORMAT pseudo instruction or the -F parameter of the assembler invocation statement.
6.1.6 Case Sensitivity
Formal parameters, symbols, names, and macro names are case sensitive.
To be recognized, subsequent references to a previously defined formal parameter, symbol, name, or macro name must match the original definition character-for-character and case-for-case (uppercase or lowercase).
The following rules govern the use of uppercase and lowercase characters in CAL statements:
• Pseudo instructions and mnemonics are case sensitive; they can be uppercase or lowercase, but not mixed case.
• Register types are case insensitive; they can be uppercase, lowercase, or mixed case.
• Macro names, opdef mnemonics, symbol names, and other names are case sensitive; they are interpreted as coded.
6.2 Symbols
70
A symbol is an identifier that can be from 1–255 characters long and has an associated value and attributes. You can use symbols in expressions and in the following ways:
• In the label field of a source statement to define the symbol for use in the program and to assign it a value and certain characteristics called attributes.
• In the operand or result field of a source statement to reference the symbol.
• In loader linkage.
S–2314–50
CAL Source Statements [6]
A symbol can be local or global depending on where the symbol is defined; that is, a symbol used within a single program module is local, and a symbol used by
a number of program segments is global (see Section 4.2, page 41). A symbol also
can be unique to a code sequence (see Section 6.2.1.2, page 72).
In addition to symbols specified in source statements, the assembler may generate symbols for internal use of the following form (where n is a decimal digit):
%% nnnnnn
Symbols that begin with the character sequence %% are discarded at the end of a program segment regardless of whether they are redefinable or defined in the global definitions part, and regardless of whether they are user-defined or generated by the assembler.
For more detailed information about symbols generated by the assembler, see the
description of the LOCAL pseudo instruction in Section 7.38, page 151.
If a symbol name is the same as a Cray X1 register name, a warning message is issued.
6.2.1 Symbol Qualification
Symbols defined within a program module (between IDENT and END pseudo instructions) can be unqualified or qualified. They are unqualified unless
preceded by the QUAL pseudo instruction (see Section 7.52, page 163, for more
information).
6.2.1.1 Unqualified Symbol
The following statements describe ways in which unqualified symbols can be referenced:
• Unqualified symbols defined in an unqualified code sequence can be referenced without qualification from within that sequence.
• If the symbol has not been redefined within the current qualifier, unqualified symbols can be referenced without qualification from within the current qualifier.
• Unqualified symbols can be referenced from within the current qualifier by using the form // symbol.
S–2314–50 71
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual
Unqualified symbols are defined as follows:
symbol = n ; symbol is equal to n
The following example illustrates unqualified symbol definition:
EDIT OFF
IDENT TEST
SYM_1 = *
A1 SYM_1
SYM_2 SET 2
SYM_3 =
END
3
; SYM_1 has a value equal to the location
; counter.
; Register Al gets SYM_1’s value.
; SYM_2 is redefinable
; SYM_3 is not redefinable.
6.2.1.2 Qualified Symbols
72
You can make a symbol that is not a global symbol unique to a code sequence by specifying a symbol qualifier that will be appended to all symbols defined within the sequence.
Qualified symbols must be defined with respect to the following rules:
• A qualified symbol cannot be defined with a name that is reserved for registers.
• Symbols can be qualified only in a program module.
Qualified symbols can be referenced as follows:
• If a qualified symbol defined in a code sequence is referenced from within that sequence, it can be referenced without qualification.
• If a qualified symbol is referenced outside of the code sequence in which it was defined, it must be referenced in the form / qualifier / symbol. The
qualifier variable is a 1- to 8-character identifier defined by the QUAL pseudo instruction and the symbol variable is a 1- to 255-character identifier.
Qualified symbols are defined as follows: qualified_symbol = /[ identifier]/symbol
The following example illustrates the use of qualified symbols:
SYM1
IDENT
= 1
TEST
QUAL NAME1
; Assignment
; Declare qualifier name
S–2314–50
CAL Source Statements [6]
SYM1 =
S1
S1
S1
2
SYM1
//SYM1
/NAME1/SYM1
QUAL *
S1 SYM1
S1
S1
END
//SYM1
/NAME1/SYM1
; Qualified symbol SYM1
; Register S1 gets 2 (qualified SYM1)
; Register S1 gets 1 (unqualified SYM1)
; Register S1 gets 2 (qualified SYM1)
; Pop the top of the qualifier stack
; Register S1 gets 1
; Register S1 gets 1
; Register S1 gets 2
6.2.2 Symbol Definition
S–2314–50
A symbol is defined by assigning it a value and attributes. The value and attributes of a symbol depend on how the program uses the symbol. The assignment can occur in the following three ways:
• When a symbol is used in the label field of a symbolic machine instruction or certain pseudo instructions, it is defined as follows:
– It has the address of the current value of the location counter (for a
description of counters, see Section 6.8.1, page 93).
– It is an object or function entry point.
– It has a byte-address attribute.
– It is absolute, immobile, or relocatable.
– It is not redefinable.
• A symbol used in the label field of a symbol-defining pseudo instruction is defined as having the value and attributes derived from an expression in the operand field of the instruction. Some symbol-defining pseudo instructions cause the symbol to have a redefinable attribute. When a symbol is redefinable, a redefinable pseudo instruction must be used to define the symbol the second time. Redefinition of the symbol causes it to be assigned a new value and attributes.
• A symbol can be defined as external to the current program module. A symbol is external if it is defined in a program module other than the module currently being assembled. The true value of an external symbol is not known within the current program module.
Here are examples of a symbol:
START = * ; The symbol START has the current value of
73
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual
PARAM SET D’18
EXT SECOND
; the location counter and cannot be
; redefined.
; The symbol PARAM is equal to the decimal
; value 18 and can be redefined.
; Identifies SECOND as an external symbol.
6.2.3 Symbol Attributes
When a symbol is defined, it assumes two or more attributes. These attributes are in three categories:
• Type
• Relative
• Redefinable
Every symbol is assigned one attribute from each of the first two categories.
Whether a symbol is assigned the redefinable attribute depends on how the symbol is used. Each symbol has a value of up to 64 bits associated with it.
6.2.3.1 Type Attribute
The following type attributes are supported:
• Object. This symbol represents the memory address at which data is stored.
• Function. This symbol represents the memory address of the beginning of a sequence of executable instructions.
• Value. This symbol represents a value that is not a memory address.
All memory addresses are byte addresses.
6.2.3.2 Relative Attributes
74
Each symbol is assigned one of the following relative attributes:
• Absolute
A symbol is assigned the relative attribute of absolute when the current location counter is absolute and it appears in the label field of a machine instruction, BSS pseudo instruction, or data generation pseudo instruction such as BSSZ or CON or if it is equated to an expression that is absolute. All globally defined symbols have a relative attribute of absolute. The symbol is known only at assembly time.
S–2314–50
S–2314–50
CAL Source Statements [6]
• Immobile
A symbol is assigned the relative attribute of immobile when the current location counter is immobile and it appears in the label field of a machine instruction, BSS pseudo instruction, or data generation pseudo instruction, such as BSSZ or CON, or if it is equated to an expression that is immobile. The symbol is known only at assembly time.
• Relocatable
A symbol is assigned the relative attribute of relocatable when the current location counter is relocatable and it appears in the label field of a machine instruction, BSS pseudo instruction, or data generation pseudo instruction such as BSSZ or CON. A symbol also is relocatable if it is equated to an expression that is relocatable.
• External
A symbol is assigned the relative attribute of external when it is defined by an
EXT pseudo instruction. An external symbol defined in this manner is entered in the symbol table with a value of 0. The address attribute of an external symbol is specified as value (V), byte ( B), or word ( W); the default is value.
A symbol is also assigned the relative attribute of external if it is equated to an expression that is external. Such a symbol assumes the value of the expression and can have an attribute of byte address, word address, or value.
Note: The assignment of an unknown variable with a register at assembly time is made by using a symbol with a relative attribute of external.
In the following example, register s1 is loaded with variable ext1 at assembly time: ident test1 ext test1 ; Variable extl is defined as an external
; variable
; ext1 transmits value to register s1 s1 end ext1 ident test2 entry ext1 ext1 = 3 ; When the two modules are linked, register
; S1 gets 3.
end
75
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual
6.2.3.3 Redefinable Attributes
In addition to its other attributes, a symbol is assigned the attribute of redefinable if it is defined by the SET or MICSIZE pseudo instructions. A redefinable symbol can be defined more than once in a program segment and can have different values and attributes at various times during an assembly. When such a symbol is referenced, its most recent definition is used by the assembler. All redefinable symbols are discarded at the end of a program segment without regard to whether they were defined in the global definitions.
The following example illustrates the redefinable attribute:
SYM1
SYM2
SYM1
SYM2
IDENT
=
SET
SET
SET
END
2
2
TEST
1
3
; Not redefinable
; Redefinable
; Error: SYM1 previously defined as 1
; Redefinable
6.2.4 Symbol Reference
When a symbol is in a field other than the label field, the symbol is being referenced. Reference to a symbol within an expression causes the value and attributes of the symbol to be used in place of the symbol. Symbols can be found in the operand fields of pseudo instructions.
An expression containing a symbol may have a different value and attributes depending on operators and other operands in the expression.
The following example illustrates a symbol reference:
S1 SYM1+1
IFA DEF,SYM1
; Register Sl gets the value of SYMl+1.
; SYMl+1 is an example of a symbol in an
; operand field used in an expression.
; Symbols can also be used outside of an
; expression. In this instance, SYMl is
; not used within an expression; it is a
; symbol.
6.3 Tags
76
A tag is an identifier that does not have an associated value or attribute and cannot be used in expressions. Tags that are 1–32 characters in length are used to identify the following types of information:
S–2314–50
S–2314–50
CAL Source Statements [6]
• Macro instructions
• Micro character strings
• Conditional sequences
• Duplicated sequences
The first character must be one of the following:
• Alphabetic character (A through Z or a through z)
• Dollar sign ($)
• Percent sign (%)
• At sign (@)
• Underscore character (_)
Characters 2–32 can also be decimal digits (0 through 9).
Tags that are 1–255 characters in length can be used to identify the following types of information:
• Program modules
• Sections
The first character must be one of the valid tag characters or the underscore (_) character. Characters 2–255 can also be decimal digits (0 through 9).
Different types of tags do not conflict with each other or with symbols. For example, a micro can have the same tag as a macro, and a program module can have the same tag as a section.
Examples of valid and not valid tags:
Valid count
@ADD
_SUBTRACT
ABCDE465
Comment
Lowercase is permitted
@ legal beginning character
_ beginning character and 9 characters are legal
Combinations of letters and digits are legal if the first character is legal
77
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual
Not valid
9knt
Y+Z3
+YZ3
Comment
Begins with a numeric character
Contains an illegal character
Begins with +
Note: If you plan to use a source manager to store your CAL program, avoid using special character sequences such as the three-character string %U%. A source manager may replace these strings throughout your source program with other text. Because this type of string is allowed within identifiers and long-identifiers, avoid using it in names, long names, and symbols.
The underscore character (_) also is used as the concatenation character (see
Section 6.10, page 97). Usually the assembler edits this character out of a source
line. To insert this character into a long name, either disable editing or use the predefined concatenation micro ($CNC). To disable editing, use either the invocation statement or the EDIT pseudo instruction.
6.4 Constants
Constants can be defined as floating, integer, or character.
6.4.1 Floating Constant
78
A floating constant is evaluated as a 32-bit, 64-bit, or 128-bit quantity, depending on the precision specified.
The floating constant is defined as follows:
[decimal-prefix] floating-decimal [floating-suffix] [binary-scale decimal-integer]
In the preceding definition, variables are defined as follows:
• decimal-prefix
This variable specifies the numeric base for the floating-decimal and/or the
decimal-integer variables. D’ or d’ specifies a decimal-prefix and is the only prefix available for a floating constant.
• floating-decimal
The floating-decimal variable can include the decimal-integer, decimal-fraction and/or decimal-exponent variables. A decimal-integer is a nonempty string of
S–2314–50
S–2314–50
CAL Source Statements [6] decimal digits. A decimal-integer or a decimal-fraction is a nonempty string of decimal digits representing a whole number, a mixed number, or a fraction.
A floating-decimal can be defined as follows:
– A decimal-integer followed by a decimal-fraction with an optional
decimal-exponent and decimal-integer. For example: n .
n or n .
n E n or n .
n E + n or n .
n D n or n.
D +n
• A decimal-integer followed by a period (.) with a decimal-exponent and
decimal-integer. For example: n .
or n .E
or n. +n or n .
nn or nn D + n
• A decimal-integer followed by a decimal-exponent and decimal-integer. For example: n E n or n E or n D n or n D n
• A decimal-fraction followed by an optional decimal-exponent and decimal-integer.
For example:
.
n or .
n E n or .
n E + n or .n D n or .
n D n
• decimal-exponent
The power of 10 by which the integer and/or fraction will be multiplied; indicates whether the constant will be single precision (E or e; one 64-bit word) or double precision (D or d; two 64-bit words). n is an integer in the base specified by prefix.
If no decimal-exponent is provided, the constant occupies one word.
decimal-exponents are defined as follows:
– E n (Positive decimal exponent, 64-bit double precision)
– E+ n (Positive decimal exponent, 64-bit double precision)
– E- n (Negative decimal exponent, 64-bit double precision)
– D n (Positive decimal exponent, 128-bit long double precision)
– D+ n (Positive decimal exponent, 128-bit long double precision)
– D- n (Negative decimal exponent, 128-bit long double precision)
• floating-suffix (one of f or FlL)
• binary-scale decimal-integer
79
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual
The integer and/or fraction will be multiplied by a power of 2. Binary scale is specified with S or s and an optional add-operator (+ or -). n is an integer in the base specified by the decimal-prefix. For example:
Sn or S+n sn or s+n
S-n or s-n
Positive binary exponent
Positive binary exponent
Negative binary exponent
Note: Double-precision floating-point numbers are truncated to single-precision floating-point numbers if pseudo instructions, which can reserve only one memory word (such as the CON pseudo instruction) are used.
The following examples illustrate floating constants:
CON
CON
CON
CON
D’1.5
4.5E+10
4.D+15
D’1.0E-6
; Mixed decimal of the form n.n.
; Single-precision floating constant of
; the form n.nE+n.
; Double-precision floating constant of
; the form n.D+n.
; Negative floating constant of the form
; n.nE-n.
SYM
CON
=
1000e2 ; Single precision floating constant of
; the form nD+n.
1777752d+l0 ; Double-precision floating constant of
; the form nD+n.
6.4.2 Integer Constant
80
An integer constant is evaluated as a 64-bit twos complement integer. The integer constant is defined as follows:
base-integer [binary-scalebase-integer]
octal-prefix octal-integer [binary-scale octal-integer]
decimal-prefix decimal-integer [binary-scale decimal-integer]
hex-prefix hex-integer [binary-scale hex-integer]
In the preceding definition, variables are defined as follows:
• base-integer
S–2314–50
S–2314–50
CAL Source Statements [6]
A string of decimal digits (0, 1, 2, 3, 4, 5, 6, 7, 8, 9) of any length
• binary-scale
The integer and/or fraction that will be multiplied by a power of 2.
binary-scale is specified with S or s and an optional add-operator (+ or -). n is an integer in the base specified by the decimal-prefix. For example:
Sn or S+n sn or s+n
(positive binary exponent)
(positive binary exponent) s-n or S-n (negative binary exponent)
• base-integer, octal-prefix, decimal-prefix, or hex-prefix
Numeric base used for the integer. If no prefix is used, base-integer is determined by the default mode of the assembler or by the BASE pseudo instruction. A prefix can be one of the following character combinations:
D’ or d’
O’ or o’
X’ , x’, 0X, or 0x
• octal-integer
Decimal (default mode)
Octal
Hexadecimal
A string of octal integers (0, 1, 2, 3, 4, 5, 6, 7) of any length
• decimal-integer
A string of decimal integers (0, 1, 2, 3, 4, 5, 6, 7, 8, 9) of any length
• hex-integer
A string of hexadecimal integers (0, 1, 2, 3, 4, 5, 6, 7, 8, 9, A or a, B or b, C or c , D or d, E or e, F or f) of any length
The following examples illustrate integer constants:
S1 O’1234567
A4 D’50
SYM =
SYM2 = x’ffffffa
0Xbeef
; Octal-prefix followed by octal-integer.
; Integer-constant of the form
; decimal-prefix followed by
; decimal-integer.
; Integer-constant of the form hex-prefix
; followed by hex-integer.
;
81
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual
6.4.3 Character Constants
The character constant is defined as follows:
character-string [character-suffix]
In the preceding definition, variables are defined as follows:
• character-string
The default is a string of zero or more characters (enclosed in apostrophes) from the ASCII character set. Two consecutive apostrophes (excluding the delimiting apostrophes) indicate one apostrophe.
• character-suffix
The justification and fill of a character string:
H or h
L or l
R or r
Z or z
Left-justified, blank-filled (default)
Left-justified, zero-filled
Right-justified, zero-filled
Left-justified, zero-filled, at least one trailing binary zero character guaranteed
The following examples illustrate character constants:
S3 ‘*’R
CON ’ABC’L
S1 ’XYZ’H
CON ’OUT’
; ASCII character, right justified, zero filled.
; ASCII characters, left justified, zero filled.
; ASCII characters, left justified, blank filled.
; ASCII characters, left justified, blank
; filled (default).
VWD 32/’EFG’ ; ASCII characters, left justified, blank
; filled within a 32-bit field (all default).
6.5 Data Items
A character or data item can be used in the operand field of the DATA pseudo instruction and in literals. The length of the data field occupied by a data item is determined by its type and size. Data items can be floating, integer, or character.
The sections that follow describe these types of data items.
82 S–2314–50
CAL Source Statements [6]
6.5.1 Floating Data Item
Single-precision floating data items occupy one word and double-precision floating data items occupy two words. A floating data item is defined as follows:
[sign] floating-constant
In the preceding definition, the sign variable is defined as follows:
• sign
The sign variable determines how the floating data item will be stored. The sign variable can be specified as follows:
+ or omitted
-
~
Not complemented
Negated (twos complemented)
Ones complemented
Note: Although syntactically correct, ~ is not permitted; a semantic error is generated with floating data.
• floating-constant
The syntax for a floating data item is the same as the syntax for floating
constants. Floating constants are described in Section 6.4.1, page 78.
The following example illustrates floating constants for data items:
DATA D’1345.567
DATA 1345.E+1
DATA 4.5E+1O
DATA 4.D+15
DATA D’1.0E-6
DATA 1000e2
DATA 1.5S2
; Decimal floating data item
; of the form n.n.
; Decimal floating data item
; of the form n.E+n.
; Single-precision floating
; constant of the form n.nE+n.
; Double-precision floating
; constant of the form n.D+n.
; Negative floating constant
; of the form n.nE-n.
; Single-precision floating
; constant of the form nen.
; Floating binary scale data
; item of the form n.nSn.
S–2314–50 83
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual
6.5.2 Integer Data Item
An integer data item occupies one 64-bit word and is defined as follows:
[ sign ] integer-constant
In the preceding definition the sign variable defines the form of a data item to be stored. The sign variable can be replaced in the integer data item definition with any of the following:
+ or omitted
-
~
Not complemented
Negated (twos complemented)
Ones complemented
The syntax for integer-constant is described in Section 6.4.2, page 80.
The following example illustrates integer constants for data:
DATA +o’20
VWD 40/0,24/O’200
; Octal integer
6.5.3 Character Data Item
The character data item is as follows:
[character-prefix] character-string [character-count] [character suffix]
In the preceding definition, variables are defined as follows:
• character-prefix
This variable specifies the character set used for the stored constant. It is specified as follows:
A or a
• character-string
ASCII character set (default)
The default is a string of zero or more characters (enclosed in apostrophes) from the ASCII character set. Two consecutive apostrophes (excluding the delimiting apostrophes) indicate one apostrophe.
• character-count
The length of the field, in number of characters, into which the data item will be placed. If count is not supplied, the length is the number of words
84 S–2314–50
6.6 Literals
S–2314–50
CAL Source Statements [6] needed to hold the character string. If a count field is present, the length is the character count times the character width; therefore, length is not necessarily an integral number of words. The character width is 8 bits for ASCII or
EBCDIC, and 6 bits for control data display code.
If an asterisk is in the count field, the actual number of characters in the string is used as the count. Two apostrophes that are used to represent one apostrophe are counted as one character.
If the base is mixed, the assembler assumes that the count is decimal
• character-suffix
This variable specifies justification and fill of the character string as follows:
H or h
L or l
R or r
Z or z
Left-justified, blank-filled (default)
Left-justified, zero-filled
Right-justified, zero-filled
Left-justified, zero-filled, at least one trailing zero character guaranteed
The following example illustrates character data items:
DATA A’ERROR IN DSN’ ; ASCII character set left justified and
; blank fill by default; two words
DATA A’error in dsn’R ; ASCII character set right justified,
; zero filled; stored in two words.
DATA ‘Error’ ; Default ASCII character set left
; justified and blank filled by default
; stored in one word.
Literals are read-only data items whose storage is controlled by the assembler.
Specifying a literal lets you implicitly insert a constant value into memory. The actual storage of the literal value is the responsibility of the assembler. Literals can be used only in expressions because the address of a literal, rather than its value, is used.
The first use of a literal value in an expression causes the assembler to store the data item in one or more words in a special local block of memory known as the literals section. Subsequent references to a literal value do not produce multiple copies of the same literal.
85
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual
Because literals can map into the same location in the literals section, the assembler checks for the presence of matching literals before new entries are added. This check is made bit by bit. If the current string is identical to any string currently stored in the literals section, the assembler maps that string to the location of the matching string. If the current string is not identical to any of the strings currently stored, the current string is considered to be unique, and is assigned a location in the literals section.
The following special syntaxes are in effect for literals:
• Literals always have the following attributes:
– Relocatable (relative) to a constant section
– Byte (address)
• Literals cannot be specified as character strings of zero bits. The actual constant within a literal must have a bit length greater than 0. In actual use, you must specify at least one 8-bit character for the ASCII character set.
• By default, literals always fall on full-word boundaries. Trailing blanks are added to fill the word to the next word boundary.
When used as an element of an expression, a literal is defined as follows:
= data-item
A data item for literals is the same as data items for constants. Data items for
constants are described in Section 6.5, page 82.
Single-precision literals are stored in one 64-bit word (default). Double-precision literals are stored in two 64-bit words. The following example shows how literals can be specified with single or double precision:
CON =1.5
CON =1.sD1
; Single-precision literal
; Double-precision literal
Figure 5 illustrates how the ASCII character a is stored by either of the following
instructions (^ represents a blank character):
CON =‘a’H
CON =‘a’
86 S–2314–50
CAL Source Statements [6]
01100001 ^^^^^^^^ ^^^^^^^^ ^^^^^^^^ ^^^^^^^^ ^^^^^^^^ ^^^^^^^^ ^^^^^^^^ a12250
Figure 5. ASCII Character with Left-justification and Blank-fill
Figure 6 illustrates how the ASCII character a is stored by any of the following
instructions:
CON =‘a’L
CON ‘a’R
CON -‘a’S
01100001 00000000 00000000 00000000 00000000 00000000 00000000 00000000 a12251
Figure 6. ASCII Character with Left-justification and Zero-fill
Figure 7 illustrates how the ASCII character a is stored by the following
instruction:
CON =‘a’R
This example illustrates how the ASCII character a is stored when =‘a’R is specified.
00000000 00000000 00000000 00000000 00000000 00000000 00000000 01100001 a12252
Figure 7. ASCII Character with Right-justification and Zero-fill
Figure 8 illustrates how the ASCII character a is stored by the following
instruction:
CON =‘a’*R
S–2314–50 87
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual
01100001 00000000 00000000 00000000 00000000 00000000 00000000 00000000 a12253
Figure 8. ASCII Character with Right-justification in 8 bits
The character set available to CAL is declared as follows:
CON =‘A’ ; 8-bit ASCII character.
The following example illustrates the use of the H, L, R, or Z options when specifying literals:
CON =‘AB’3
CON =‘AB’3H
CON =‘AB’6R
CON =‘AB’6Z
; Left-justified with one blank-padded on the
; right (default).
; Left-justified with one blank-padded on the
; right (default).
; Right-justified, filled with four leading
; zeros.
; Left-justified, padded with four trailing
; zeros
6.7 Micros
88
Through the use of micros, you can assign a name to a character string and subsequently refer to the character string by its name. A reference to a micro results in the character string being substituted for the name before assembly of the source statement containing the reference. The CMICRO, MICRO, OCTMIC,
and DECMIC pseudo instructions (described in Chapter 7, page 103) assign the
name to the character string.
Refer to a micro by enclosing the micro name in double quotation marks (" ") anywhere in a source statement other than within a comment. If column 72 of a line is exceeded because of a micro substitution, the assembler creates additional continuation lines. No replacement occurs if the micro name is unknown or if one of the quotation marks is omitted.
When a micro is edited, the source statement that contains the micro is changed.
Each substitution produces one of the following cases:
• The length of the micro name and the pair of quotation marks is the same as the predefined substitute string. When the micro is edited, the length of the source statement is unchanged.
S–2314–50
S–2314–50
CAL Source Statements [6]
• The length of the micro name and the double quotation marks is greater than the predefined substitute string. When the string is edited, all characters to the right of the edited string shift left the number of spaces equal to the difference between the length of the micro name including the double quotation marks and the predefined substitute string.
• The length of the micro name and the double quotation marks is less than the predefined substitute string. If column 72 of a line is exceeded because of a micro substitution, the assembler creates additional continuation lines.
Resulting lines are processed as if they were one statement.
In the following example, the length of the micro name (including quotation marks) is equal to the length of the predefined substitute string. A micro named
PFX is defined as EQUAL. A reference to PFX is in the label field of the statement, as follows:
"PFX"TAG S0 S1 ; The location of S0 and S1 on the
; source statement is unchanged
When the line is interpreted, the assembler substitutes EQUAL for "PFX", producing the following line:
EQUALTAG S0 S1 ; The location of S0 and S1 on the
; source statement is unchanged
In the following example, the length of the micro name (including quotation marks) is greater than the length of the predefined substitute string. A micro named PFX is defined as LESS. A reference to PFX is in the label field of the statement, as follows:
"PFX"TAG S0 S1 ; Because LESS is one character shorter
; than the micro string name "PFX", the
; values in the result and operand
; fields are shifted one space to the
; left.
Before the line is interpreted, the assembler substitutes LESS for "PFX", producing the following line:
LESSTAG S0 S1 ; Because LESS is one character shorter
; than the micro string name "PFX", the
; values in the result and operand
; fields are shifted one space to the
; left.
89
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual index micro null micro
\
\\
In the following example, the length of the micro name (including quotation marks) is less than the length of the predefined substitute string. A micro named pfx is defined as greater. A reference to pfx is in the label field of the following statement:
"pfx" tag S0 S1 ; Because greater is two characters
; longer than micro string name "pfx",
; the values in the result and operand
; fields are shifted two spaces to the
; right.
Before the line is interpreted, the assembler substitutes the predefined string greater for "pfx". Because the predefined substitute string is 2 characters longer than micro name, the fields to the right of the substitution are shifted 2 characters to the right, producing the following statement: greatertag S0 S1 ; Because greater is two characters
; longer than the micro string name
; "pfx", the values in the result and
; operand fields are shifted
One or more micro substitutions can occur between the beginning and ending quotation marks of a micro. These substitutions create a micro name that is substituted, along with the surrounding quotation marks, for the corresponding micro string. Substitutions of this type are embedded micros. An embedded micro consists of a micro name included between a left ({) and a right brace ( }) and is specified as follows:
{ microname}
When a micro that contains one or more embedded micros is encountered, the assembler edits all embedded micros within the micro until a micro name is recognized or until the micro name is determined to be illegal (undefined or exceeding the maximum allowable string length of 8 characters). When an illegal micro is encountered, the assembler issues an appropriate message and terminates the editing of the micro. An embedded micro also can contain one or more embedded micros.
The following example includes valid and not valid defined embedded micros
; Assigns literal value to index
; Assigns literal value to null array "index" micro \Some string\ array1 micro \Some string\
90 S–2314–50
CAL Source Statements [6]
* "array1" - an explicit reference
* Some string - an explicit reference; edited by the assembler
* "array" "index" - not valid, because "array" was not defined
* "array"1 - not valid, because "array" was not defined; edited by the assembler
* "array{index}" - This is an example of an embedded micro
* Some string - This is an example of an embedded micro; edited by the assembler
* "{null}array{index}" - This is an example of two embedded micros
* Some string - This is an example of two embedded micros; edited by the assembler
The assembler places no restrictions on the number of recursions that are necessary to identify a micro name. The following example demonstrates the unlimited recursive editing capability of the assembler on embedded micros: index null micro micro
\1\ ; Assigns literal value to index
\\ ; Assigns literal value to null array"index" micro \Some string\ arrayl micro \Some string\; edited by the assembler
* "{nu{n{null}u{null}ll}ll}ar{null{null}}ray{ind{null}ex}" - Micro
* Some string - Micro; edited by the assembler
The assembler issues a warning- or error-level listing message when an invalid micro name is specified. If a micro name is recognized as invalid before editing begins, a warning-level message is issued. If an embedded micro has been edited and the resulting string is not a valid micro name, an error-level listing message is issued.
The following examples demonstrate how the assembler assigns levels to messages when a micro that is not valid is encountered: identity micro null micro
\The substitute string for this example\;
\ ; Assigns literal value to null
* "identity{null}" - This is a valid micro
* The substitute string for this example - This is a valid
* micro; edited by the assembler
* The following micro is invalid, because the maximum micro
* name length of eight characters is exceeded.
When a micro
* name is identified as being invalid before editing occurs,
* a warning-level listing message is issued:
* "identity9{null}" - This is a not valid micro; edited by the assembler
* "identity9 - This is a not valid micro
* The following micro is not valid, because the maximum
* micro name length of eight characters is exceeded.
When a
* micro name is identified as being not valid after editing
S–2314–50 91
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual
* occurs, an error-level listing message is issued:
* "id{null}entity9{null}" - This is a not valid micro
* "identity9" - This is a not valid micro; edited by the assembler
6.8 Location Elements
Location elements are used to obtain the current value of the location counter, the origin counter, and the Longword-bit-position counter. Location elements can occur as elements of expressions. For a description of expression elements, see
Section 6.9, page 94. The origin, location, word-bit-position, and byte-bit-position
counters are described in Chapter 7, page 103.
The following elements have special meanings to the assembler:
• *. Location counter.
The asterisk (*) denotes a value equal to the current value of the location counter with byte-address attribute and absolute, immobile, or relocatable attributes. The location counter is absolute if the LOC pseudo instruction modified it by using an expression that has a relative attribute of absolute.
The location counter is immobile if it is relative to either a STACK or TASKCOM section. The location counter is relocatable in all other cases.
• *A or *a. Absolute location counter.
The *A or *a denotes a value equal to the current value of the location counter with byte-address and absolute attributes.
• *B or *b. Absolute origin counter.
The *B or *b denotes a value equal to the current value of the origin counter relative to the beginning of the section with byte-address and absolute attributes.
• *O or *o. Origin counter.
The *O or *o denotes a value equal to the current value of the origin counter relative to the beginning of the current section. The origin counter has an address attribute of byte. If the current section is a section with a type of
STACK or TASKCOM, it has an immobile attribute. In all other cases, it has a relative attribute of relocatable.
• *W or *w. Longword pointer.
The *W or *w denotes a value equal to the current value of the
Longword-bit-position counter with absolute and value attributes. *W is
92 S–2314–50
CAL Source Statements [6] relative to the word and the word-bit-position counter is almost always equal to 0, 16, 32, or 48. The assembler issues a warning message when the word-bit-position counter has a value other than 0 (not pointing at a word boundary) and is used in an expression.
6.8.1 Location Counter
Usually, the location counter is the same value as the origin counter and the assembler uses it to define symbolic addresses within a section. The counter is incremented when the origin counter is incremented. Use the LOC pseudo instruction to adjust the location counter so that it differs in value from the origin counter or so that it refers to the address relative to a section other than the one currently in use. When the location element * is used in an expression, the assembler replaces it with the current byte-address value of the location counter for the section in use. To obtain the 64-bit word-address value of the location counter, use W.*.
6.8.2 Origin Counter
The origin counter controls the relative location of the next word that will be assembled or reserved in the section. You can reserve blank memory areas by using either the ORG or BSS pseudo instructions to advance the origin counter.
When the location element *O is used in an expression, the assembler replaces it with the current byte-address value of the origin counter for the section in use. To obtain the word-address value of the origin counter, use W.*O. (In this context, a word is a 64-bit long word.).
6.8.3 Longword-bit-position Counter
As instructions and data are assembled and placed into a 64-bit long word, the assembler maintains a pointer that indicates the next available bit within the word currently being assembled. This pointer is known as the word-bit-position
counter. It is 0 at the beginning of a new word and is incremented by 1 for each completed bit in the word. Its maximum value is 63 for the rightmost bit in the word. When a word is completed, the origin and location counters are incremented by 1, and the word-bit-position counter is reset to 0 for the next word.
When the location element *W is used in an expression, the assembler replaces it with the current value of the word-bit-position counter. The normal advancement of the word-bit-position counter is in increments of 32 as instructions are
S–2314–50 93
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual generated. You can alter this normal advancement by using the BITW, BITP,
DATA , and VWD pseudo instructions.
6.8.4 Force Longword Boundary
If either of the following conditions are true, the assembler completes a partial word and sets the word-bit-position counter to 0:
• The current instruction is an ALIGN, BSS, BSSZ, CON, LOC, or ORG pseudo instruction.
• The current instruction is a DATA or VWD pseudo instruction and the instruction has an entry in the label field.
6.9 Expressions
The result and operand fields for many source statements contain expressions.
An expression consists of one or more constants (see Section 6.4, page 78), literals
(see Section 6.6, page 85), location elements (see Section 6.8, page 92) or symbols
(see Section 6.2, page 70) and zero or more operators (see Section 6.9.1, page 95).
There are some restrictions on the combinations of operands and operators that are allowed in the different contexts, as specified below.
The assembler will evaluate the expression, applying appropriate restrictions and if no restrictions are violated, insert the value of the expression into the immediate field in the machine instruction or the operand field of the pseudo instruction.
Registers are not included as part of a CAL expression. A register may be an operand in an instruction but the assembler cannot determine the value of the register. For example, in the instruction: s5 s3+2*3 the assembler parses the instruction such that 2*3 is an expression that it evaluates and then inserts the value of 6 into the immediate field of the machine instruction. It does not treat s3 as part of the expression.
Each CAL expression has a type and a value. The type is one of:
• absolute
• object
94 S–2314–50
6.9.1 Operators
S–2314–50
CAL Source Statements [6]
• function
If the type of the expression is either object or function, it has a relative attribute that is one of:
• relocatable
• external
• immobile and has an address attribute that is one of:
• byte address
• stack offset
The CAL operators are:
• unary operators:
– + (unary plus)
– – (unary minus)
– ~ (bitwise complement)
– ! (logical NOT)
– B. (byte prefix)
– b. (byte prefix)
– W. (word prefix)
– w. (word prefix)
– L. (longword prefix)
– l. (longword prefix)
– < (mask right)
– > (mask left)
• binary operators:
– + (add)
95
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual
6.9.2 Operator Precedence
CAL expressions are evaluated according to the following operator precedence:
&
^
|
&&
||
! ~ +(unary) -(unary) B. W. L. b. w. l. < > Right to left
* / Left to right
+ -
<< >>
Left to right
Left to right
Left to right
Left to right
Left to right
Left to right
Left to right
Parentheses can be used to group an expression or subexpressions.
6.9.3 Restrictions
– – ( subtract)
– * (multiply)
– / (divide)
– & (bitwise AND)
– | (bitwise OR)
– ^ (bitwise XOR)
– && (logical AND)
– || (logical OR)
– << (bitwise left shift)
– >> (bitwise right shift)
96
If the machine instruction has an immediate field (imm6, imm8, imm14, imm16, or imm20 ), then an expression is allowed in that subfield of the CAL instruction. If the value of the expression exceeds the size of the immediate field, the value will be truncated and a diagnostic message will be issued.
If a pseudo instruction has an expression field or subfield, the restrictions are
described in Chapter 7, page 103. Only integer expressions are allowed in pseudo
instructions unless explicitly stated otherwise.
S–2314–50
CAL Source Statements [6]
If the expression includes one or more relocatable or external symbols that is known only at link time, the assembler will evaluate the expression including the symbol name and pass to the loader the symbol name and an offset value to allow the loader to complete the evaluation and insert the final value in the code of the executable.
A symbolic register designator may be an expression but may not include any relocatable symbols and the entire expression must be enclosed in parentheses.
Floating constants are allowed only in the CON pseudo instruction.
Two relocatable symbols can occur in the same expression only if the difference of the two symbols is taken.
6.10 Statement Editing
The assembler processes source statements sequentially from the source file.
Statement editing is a form of preprocessing in which the assembler deletes or replaces characters before processing the statement as source code.
The assembler performs the following types of statement editing:
• Concatenation
The assembler recursively deletes all underscore characters and combines the character that preceded the underscore with the character following the underscore.
• Micro substitution
The assembler replaces a micro name with a predefined character string. The character string replacement is not edited a second time.
A macro or opdef definition is not immediately interpreted but is saved and interpreted each time it is called. Before interpreting a statement, the assembler performs editing operations. The assembler does not perform micro substitution or concatenate lines when editing is disabled. (Editing is disabled by using the
EDIT pseudo instruction or by including the -J option in the invocation line of the assembler.)
The edit invocation statement option does not affect appending, continuation, and the processing of comments.
S–2314–50 97
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual
The following special characters signal micro substitution, concatenation, append, continuation, and comments:
;
,
Character
" name "
_
*
Edit
Yes
Yes
No
No
No
Description
Micro; affected by the EDIT pseudo instruction on the invocation statement option.
Concatenate; (underscore) affected by the EDIT pseudo instruction on the invocation statement option.
Comment line; (asterisk) unaffected by the EDIT pseudo instruction on the invocation statement option.
Comment line; (semicolon) unaffected by the
EDIT pseudo instruction on the invocation statement option.
Continuation line; (comma) unaffected by the
EDIT pseudo instruction on the invocation statement option (old format only).
Note: When the assembler edits "$CMNT", "$MIC", "$CNC", or "$APP", the string name and the pair of double quotation marks (" ") are replaced by a previously defined string. For example, when the assembler edits "$CMNT", a semicolon is substituted for the micro name $CMNT and the double quotation marks (" "). After the substitution occurs, the semicolon is not edited again and editing continues on the line. Using the predefined "$CMNT" micro permits a comment to be edited.
For example,
"$CMNT" Cray Inc "$DATE" - "$TIME" is edited as follows:
; Cray Inc 12/31/02 - 8:15:45
The characters to the right of the substituted character are shifted six positions to the left after editing, because the character string substituted for "$CMNT"
(;) is six characters shorter than the micro name.
98 S–2314–50
CAL Source Statements [6]
6.10.1 Micro Substitution
You can assign a micro name to a character string. You can refer to that character string in subsequent statements by its micro name. The assembler searches for quotation marks (") that delimit micro names. The first quotation mark indicates the beginning of a micro name; the second quotation mark identifies the end of a micro name. Before a statement is interpreted, the assembler replaces the micro name with the character string that comprises the micro. For more information
on micros, see Section 6.7, page 88.
6.10.2 Concatenate
If statement editing is on, the concatenate feature combines characters connected by the underscore (_) character. the assembler examines each line for the underscore character and deletes each occurrence of the underscore. The two adjoining columns are linked before the statement is interpreted. The concatenate symbol can be in any column and tells the assembler to concatenate the characters following the last underscore to the character preceding the first underscore.
6.10.3 Append
The append feature combines source statements that continue for more than one line. It is available only when the new format is specified. The exact number of lines that the assembler can append depends on memory limitations.
The append symbol is a backslash character immediately followed by a newline character and appends one line to another. It can be used in any column on any line.
When the current line contains a blackslash character immediately followed by a newline character, the assembler appends the first non-whitespace character and all characters that follow from the next line to the current line. The characters are appended at the position in the current line that contains the backslash, the backslash and newline are replaced.
6.10.4 Continuation
A comma in column 1 indicates a continuation of the previous line. Columns 2 through 72 become a continuation of the previous line. Continuation is permitted only when the old format is specified.
S–2314–50 99
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual
6.10.5 Comment
A semicolon (;) in any column (new format only) or an asterisk (*) in column 1 indicates a comment line. The assembler lists comment lines, but they have no effect on the program. When a semicolon or an asterisk has an editing symbol after it, the symbol is treated as part of the comment and is not used. In the new format, comment statements with semicolons or asterisks are not appended.
Note: Asterisk comment statements are not included in macro definitions. To include a comment line in a macro definition, enter an underscore in column 1 of the comment line followed by an asterisk and then the comment. Because editing is disabled at definition time, the statement is inserted. If editing is enabled at expansion time, the underscore is edited out and the statement is treated as a comment.
The following example illustrates the use of comment statements in a macro:
MACRO
EXAMPLE
* This comment is not included in the definition.
_* This comment is included in the definition.
SYM = 1
EXAMPLE ENDM
The macro in the preceding example is expanded as follows:
LIST
EXAMPLE
LIS,MAC
;Macro call
* This comment is included in the definition.
SYM = 1
6.10.6 Actual Statements and Edited Statements
CAL statements can be divided into two categories: actual and edited. An actual statement is the unedited version of a statement that includes any appending of lines. It contains all of the editing symbols rather than the results of the editing. If an actual statement has a corresponding edited statement, further processing is done on the edited statement. The following examples show actual and edited statements.
100 S–2314–50
S–2314–50
CAL Source Statements [6]
This following example shows an actual statement:
LOC MCALL ARG1,\
ARG2,\
ARG3,\
ARG4,\
ARG5
An actual statement can have a corresponding edited statement. The edited statement displays the statement without any editing symbols. The following example shows the edited version of the actual statement in the preceding example:
LOC MCALL ARG1,ARG2,ARG3,ARG4,ARG5
In the following example, the actual statement has no corresponding edited statement:
ENTER ARG1,ARG2,ARG3 ; Comments
101
CAL Pseudo Instruction Descriptions [7]
This chapter lists the pseudo instructions presented throughout Chapter 3, page
33, in alphabetical order for easy reference.
Note: You can specify pseudo instructions in uppercase or lowercase, but not in mixed case.
Throughout this chapter, pseudo instructions with ignored fields (location or operand) are defined as follows: ignored pseudox ignored pseudox
The assembler ignores the label field of this statement. If the field is not empty, then all of the characters in the field are skipped until a blank character is encountered and a caution-level message is issued. The first nonblank character following the blank character is assumed to be the beginning of the result field.
Name of the Pseudo instruction with a blank label field.
pseudoy ignored pseudoy ignored
Name of the Pseudo instruction with a blank operand field.
The assembler ignores the operand field of this statement. If the field is not empty, then all of the characters in the field are skipped until a blank character is encountered and a caution-level message is issued. The first nonblank character following the blank character is assumed to be the beginning of the comment field.
7.1 Equate Symbol (
=
)
The equate symbol (=) when used as a pseudo instruction defines a symbol with the value and attributes determined by the expression. The symbol is not redefinable.
You can specify the = pseudo instruction anywhere within a program segment.
If the = pseudo instruction is found within a definition, it is defined and is not recognized as a pseudo instruction. If the = pseudo instruction is found within a skipping sequence, it is skipped and is not recognized as a pseudo instruction.
S–2314–50 103
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual
The format of the = pseudo instruction is as follows:
[symbol] = expression[,[attribute]]
The symbol variable represents an optional unqualified symbol. The symbol is implicitly qualified by the current qualifier. The symbol must not be defined already. The label field can be blank. symbol must satisfy the requirements for
symbols as described in Section 6.2, page 70.
All symbols found within expression must have been previously defined. The
expression operand must meet the requirements for an expression as described in
The attribute variable specifies a byte, word (W), or value ( V) attribute. If present, it is used instead of the expression’s attribute. If a byte-address attribute is specified, an expression with a word-address attribute is multiplied by 8; if a word-address attribute is specified, an expression with byte-address attribute is divided by 8. You cannot specify a relocatable expression as having value attribute.
In the following example, the symbol SYMB is assigned the value of
(A*B+100)/4 . The following illustrates the use of the = pseudo instruction:
SYMB = (A*B+100)/4
7.2 (Deferred implementation)
ALIGN
7.3
BASE
104
The BASE pseudo instruction specifies the base of numeric data as octal, decimal, or mixed when the base is not explicitly specified by an O’, D’, or X’ prefix.
The default is decimal.
You can specify the BASE pseudo instruction anywhere in a program segment.
However, if the BASE pseudo instruction is located within a definition or skipping section, it is not recognized as a pseudo instruction.
The format of the BASE pseudo instruction is as follows: ignored BASE option/*
S–2314–50
S–2314–50
CAL Pseudo Instruction Descriptions [7]
The option variable specifies the numeric base of numeric data. It is a required single character specified as follows:
• O or o (Octal)
• D or d (Decimal)
• X or x (Hex)
• M or m (Mixed)
Numeric data is assumed to be octal, except for numeric data used for the following (assumed to be decimal):
• Statement counts in DUP and conditional statements
• Line count in the SPACE pseudo instruction
• Bit position or count in the BITW, BITB, or VWD pseudo instructions
• Character counts as in CMICRO, MICRO, OCTMIC, and DECMIC pseudo instructions
• Character count in data items (see Section 6.5.3, page 84).
When the asterisk (*) is used with the BASE pseudo instruction, the numeric base reverts to the base that was in effect prior to the specification of the current prefix within the current program segment. Each occurrence of a BASE pseudo instruction other than BASE * can modify the current prefix. Each
BASE * releases the most current prefix and reactivates the prefix that preceded the current prefix. If all BASE pseudo instructions specified are released, a caution-level message is issued, and the default mode (decimal) is used.
The following example illustrates the use of the BASE pseudo instruction:
.
.
BASE 0 ; Change base from default to octal.
VWD 50/12 ; Field size and constant value both octal.
.
BASE D
VWD
.
49/19
; Change base from octal to decimal.
; Field size and constant value both decimal.
.
.
BASE M
VWD 39/12
; Change from decimal to mixed base.
; Field size decimal, constant value octal.
105
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual
.
.
.
BASE *
BASE *
BASE *
; Resume decimal base.
; Resume octal base.
; Stack empty - resume decimal base (default)
7.4
BITW
106
The BITW pseudo instruction resets the current bit position to the value specified, relative to bit 0 of the current word. If the current bit position is not bit 0, a value of 64 (decimal) forces the following instruction to be assembled at the beginning of the next word (force word boundary). If the current bit position is bit 0, the
BITW pseudo instruction with a value of 64 does not force a word boundary, and the instruction following BITW is assembled at bit 0 of the current word.
If the origin and location counters are set lower than the current value, any code previously generated in the overlapping part of the word is ORed with any new code.
The BITW pseudo instruction is restricted to sections that allow data or instructions and data. If the BITW pseudo instruction is found within a definition, it is defined and is not recognized as a pseudo instruction. If the BITW pseudo instruction is found within a skipping sequence, it is skipped and is not recognized as a pseudo instruction.
The format of the BITW pseudo instruction is as follows: ignored BITW [expression]
The expression variable is optional. If expression is not specified, the default is the absolute value of 0. If expression is specified, it must have an address attribute of value, a relative attribute of absolute, and be a positive value in the range from 0 to 64 (decimal). All symbols within expression (if any) must have been defined previously. If the current base is mixed, decimal is used.
The expression operand must meet the requirements for expressions as described
The value generated in the code field of the listing is equal to the value of the expression.
The following example illustrates the use of the BITW pseudo instruction:
S–2314–50
7.5
BSS
S–2314–50
CAL Pseudo Instruction Descriptions [7]
BITW D’39
The BSS pseudo instruction reserves a block of memory in a section. A forced byte boundary occurs, and the number of bytes specified by the operand field expression is reserved. This pseudo instruction does not generate data. To reserve the block of memory, the location and origin counters are increased.
You must specify the BSS pseudo instruction from within a program module. If the BSS pseudo instruction is found within a definition, it is defined and is not recognized as a pseudo instruction. If the BSS pseudo instruction is found within a skipping sequence, it is skipped and is not recognized as a pseudo instruction.
The format of the BSS pseudo instruction is as follows:
[symbol] BSS [expression]
The symbol variable is optional. It is assigned the byte address of the location counter after the force byte boundary occurs. symbol must meet the requirement
for symbols as described in Section 6.2, page 70.
The expression variable is an optional absolute expression with a byte address or value attribute and with all symbols, if any, previously defined. The value of the expression must be positive. A force byte boundary occurs before the expression is evaluated.
The expression operand must meet the requirements for an expression as
described in Section 6.9, page 94.
The left margin of the listing shows the hex byte count.
The following example illustrates the use of the BSS pseudo instruction:
BSS 4
A CON ’NAME’
CON
CON
BSS
1
2
16+A-W.* ; Reserve more words so that the total
; starting at A is 16.
107
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual
7.6
BSSZ
The BSSZ pseudo instruction generates a block of bytes that contain 0’s. When
BSSZ is specified, a forced byte boundary occurs, and the number of zeroed bytes specified by the operand field expression is generated.
The BSSZ pseudo instruction is restricted to sections that have a type of data or instructions and data. If the BSSZ pseudo instruction is found within a definition, it is defined and is not recognized as a pseudo instruction. If the BSSZ pseudo instruction is found within a skipping sequence, it is skipped and is not recognized as a pseudo instruction.
The format of the BSSZ pseudo instruction is as follows:
[symbol] BSSZ [expression]
The symbol variable represents an optional symbol. It is assigned the byte-address value of the location counter after the force byte boundary occurs. symbol must
meet the requirements for a symbol as described in Section 6.2, page 70.
The expression variable represents an optional absolute expression with an attribute of byte address or value and with all symbols previously defined. The expression value must be positive and specifies the number of bytes containing
0’s that will be generated. A blank operand field results in no data generation.
The expression operand must meet the requirement for an expression as described
The octal word count of a BSSZ is shown in the left margin of the listing.
7.7
CMICRO
108
The CMICRO pseudo instruction assigns a name to a character string. After the name is defined, it cannot be redefined. If the CMICRO pseudo instruction is defined within the global definitions part of a program segment, it can be referenced at any time after its definition by any of the segments that follow. If the CMICRO pseudo instruction is defined within a program module, it can be referenced at any time after its definition within the module. However, a constant micro defined within a program module is discarded at the end of the module and cannot be referenced by any segments that follow.
If the CMICRO pseudo instruction is found within a definition, it is defined and is not recognized as a pseudo instruction. If the CMICRO pseudo instruction
S–2314–50
S–2314–50
CAL Pseudo Instruction Descriptions [7] is found within a skipping sequence, it is skipped and is not recognized as a pseudo instruction.
The format of the CMICRO pseudo instruction is as follows: name CMICRO [string[,[exp][,[exp],[case]]]]
The name variable is required and is assigned to the character string in the operand field. It has attributes that cannot be redefined. If name was previously defined and the string represented by the previous definition is not the same string, an error message is issued and definition occurs. If the strings match, no error message is issued and no definition occurs. name must meet the
requirements for identifiers as described in Section 6.3, page 76.
The string variable represents an optional character string that can include previously defined micros. If string is not specified, an empty string is used. A character string can be delimited by any character other than a space. Two consecutive occurrences of the delimiting character indicate a single such character (for example, a micro consisting of the single character * can be specified as ‘*’ or ****).
The exp variable represents optional expressions. The first expression must be an absolute expression that indicates the number of characters in the micro character string. All symbols, if any, must be previously defined. If the current base is mixed, decimal is used for the expression. The expressions must meet the
requirements for expressions as described in Section 6.9, page 94.
The micro character string is terminated by the value of the first expression or the final apostrophe of the character string, whichever occurs first. If the first expression has a 0 or negative value, the string is considered empty. If the first expression is not specified, the full value of the character string is used. In this case, the string is terminated by the final apostrophe.
The second expression must be an absolute expression indicating the micro string’s starting character. All symbols, if any, must be defined previously. If the current base is mixed, decimal is used for the expression.
The starting character of the micro string begins with the character that is equal to the value of the second expression, or with the first character in the character string if the second expression is null or has a value of 1 or less.
The optional case variable denotes the way uppercase and lowercase characters are interpreted when they are read from string. Character conversion is restricted
109
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual to the letter characters (A - Z and a - z) specified in string. You can specify case in uppercase, lowercase, or mixed case, and it must be one of the following values:
• MIXED or mixed
string is interpreted as you entered it and no case conversion occurs. This is the default.
• UPPER or upper
All lowercase alphabetic characters in string are converted to their uppercase equivalents.
• LOWER or lower
All uppercase alphabetic characters in string are converted to their lowercase equivalents.
7.8
COMMENT
The COMMENT pseudo instruction defines a character string of up to 256 characters that will be entered as an informational comment in the generated binary load module.
If the operand field is empty, the comment field is cleared and no comment is generated. If a comment is specified more than once, the most recent one is used. If the last comment differs from the previous comment, a caution-level message is issued.
If a subprogram contains more than one COMMENT pseudo instruction, the character string from the last COMMENT pseudo instruction goes into the binary load module.
You must specify the COMMENT pseudo instruction from within a program module. If the COMMENT pseudo instruction is found within a definition, it is defined and is not recognized as a pseudo instruction. If the COMMENT pseudo instruction is found within a skipping sequence, it is skipped and is not recognized as a pseudo instruction.
The format of the COMMENT pseudo instruction is as follows: ignored COMMENT [del-char[string-of-ASCII]del-char]
110 S–2314–50
7.9
CON
S–2314–50
CAL Pseudo Instruction Descriptions [7]
The del-char variable designates the delimiter character. It must be a single matching character on both ends of the ASCII character string. A character string can be delimited by a character other than an apostrophe. Any ASCII character other than a space can be used. Two consecutive occurrences of the delimiting character indicate that a single such character will be included in the character string.
The string-of ASCII variable is an optional ASCII character string of any length.
The following example illustrates the use of the COMMENT pseudo instruction:
IDENT CAL
COMMENT ’COPYRIGHT CRAY INC.
2002’
COMMENT -CRAY X1 computer system-
COMMENT @ABCDEF@@FEDCBA@
END
The CON pseudo instruction generates one or more full 64-bit words of binary data. This pseudo instruction always causes a forced 64-bit word boundary.
The CON pseudo instruction is restricted to sections that have a type of data or instructions and data. If the CON pseudo instruction is found within a definition, it is defined and is not recognized as a pseudo instruction. If the CON pseudo instruction is found within a skipping sequence, it is skipped and is not recognized as a pseudo instruction.
The format of the CON pseudo instruction is as follows:
[symbol] CON [expression]{,[expression]}
The symbol variable is an optional symbol. It is assigned the byte address value of the location counter after the force 64–bit word boundary occurs. symbol must
meet the requirements for a symbol as described in Section 6.2, page 70.
The expression variable is an expression whose value will be inserted into one
64-bit word. If an expression is null, a single zero word is generated. A force word boundary occurs before any operand field expressions are evaluated. A double-precision, floating-point constant is not allowed. expression must meet the
requirements for an expression as described in Section 6.9, page 94.
The following example illustrates the use of the CON pseudo instruction:
111
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual
A CON
CON
O’7777017
A ; Generates the
; address of A.
7.10
DATA
112
The DATA pseudo instruction generates zero or more bits of code for each data item parameter found in the operand field. If a label exists in the label field, a forced word boundary occurs and the symbol is assigned an address attribute and the value of the current location counter.
If a label is not included in the label field, a forced word boundary does not occur.
The DATA pseudo instruction is restricted to sections that have a type of data, constants, or instructions and data. If the DATA pseudo instruction is found within a definition, it is defined and is not recognized as a pseudo instruction. If the DATA pseudo instruction is found within a skipping sequence, it is skipped and is not recognized as a pseudo instruction.
The length of the field generated for each data item depends on the type of constant involved. Data items produce zero or more bits of absolute value binary code, as follows:
Data item
Floating
Integer
Character
Description
One or two binary words, depending on whether the data item is a single- or double-precision data item
One binary word
Zero or more bits of binary code depending on the following:
• Character set specified
• Number of characters in the string
• Character count (optional)
• Character suffix (optional)
A word boundary is not forced between data items.
The format of the DATA pseudo instruction is as follows:
S–2314–50
CAL Pseudo Instruction Descriptions [7]
[symbol] DATA [data_item][,[data_item]]
The symbol variable represents an optional symbol that is assigned the word address of the location counter after a force byte boundary. If no symbol is present, a force word boundary does not occur. symbol must meet the
requirements for a symbol as described in Section 6.2, page 70.
The data_item variable represents numeric or character data. data_item must meet
the requirements for a data item as described in Section 6.5, page 82.
The DATA pseudo instruction works with the actual number of bits given in the data item.
In the following example, unlabeled data items are stored in the next available
IDENT
DATA
DATA
END
EXDAT
’abcd’*
’efgh’
; Unlabeled data item 1.
; Unlabeled data item 2.
01100001 01100010 01100011 01100100 01100101 01100110 01100111 01101000
Unlabeled data item number 1 Unlabeled data item number 2
a12271
Figure 9. Storage of Unlabeled Data Items
In the following example, labeled data items cause a forced word boundary
IDENT EXDAT
DATA ’abcd’* ; Unlabeled data item 1.
ALPHA DATA ’efgh’* ; Labeled data item 1.
BETA DATA ’ijkl’* ; Labeled data item 2.
DATA ’mnop’ ; Unlabeled data item 2.
S–2314–50 113
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual
01100001 01100010 01100100 01100100 00000000 00000000
Unlabeled data item number 1
00000000 00000000
01100101 01100110 01100111 01101000
Labeled data item number 1
00000000 00000000 00000000 00000000
114
01101001 01101010 01101011 01101100
Labeled data item number 2
01101101 01101110 011001111 01110000
Unlabeled data item number 2
a12272
Figure 10. Storage of Labeled and Unlabeled Data Items
In the following example, if no forced word boundary occurs, data is stored bit by bit in consecutive words.
The following example shows the code generated by each source statement:
LL2
IDENT EXAMPLE
DATA 0’5252,A’ABC’R ; 0000000000000000005252
DATA
DATA ’EFGH’
DATA
DATA ’EFGH’*
DATA
’ABCD’
’ABCD’*
’ABCD’12R
; 0000000000000020241103
; 0405022064204010020040
; 0425062164404010020040
; 040502206420
; 10521443510
; 0000000000000000000000
DATA
DATA
’EFGHIJ’*
’ABCD’
; 040502206420
;
; 044512
10521443510
; 0405022064204010020040
S–2314–50
7.11
DBSM
S–2314–50
CAL Pseudo Instruction Descriptions [7]
DATA 100
DATA 1.25E-9
; 0000000000000000000144
; 0377435274616704302142
DATA ’THIS IS A MESSAGE’*L
; 0521102225144022251440
VWD
END
8/0
; O4044O232425232464O5O7
; 0424
; 000
The DBSM pseudo instruction generates a named label entry in the debug symbol tables with the type specified.
The format of the DBSM pseudo instruction is as follows:
[ignored] DBSM TYPE=symbol
TYPE is specified as either ATP or BOE (after the prologue or beginning of epilogue). symbol is user defined and marks these two points in the code. The
symbol can appear anywhere in the code, but the address that is entered into the debug symbol table is the address of where the pseudo instruction appears in the code. This pseudo instruction is ignored unless you specify the debug option on the command line.
The following example illustrates the use of the DBSM pseudo instruction:
IDENT TEST
ENTRY FRED
FRED =
BSSZ
*
16
S4
CHK =
DBSM
; Fake prolog.
S4
*
ATP=FRED ; Should be the same as CHK address.
A1
A1
A1
DBSM
S1
J
S1
S1
S1
BOE=FRED
5
B00
; Address should be the same as the next
; instruction
115
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual
From the debugger, you can do a stop in FRED to generate a breakpoint at CHK. A call to this routine from a program executing in the debugger stops the execution.
7.12
DECMIC
The DECMIC pseudo instruction converts the positive or negative value of an expression into a positive or negative decimal character string that is assigned a redefinable micro name. The final length of the micro string is inserted into the code field of the listing.
You can specify DECMIC with zero, one, or two expressions. DECMIC converts the value of the first expression into a character string with a character length indicated by the second expression. If the second expression is not specified, the minimum number of characters needed to represent the decimal value of the first expression is used.
If the second expression is specified, the string is equal to the length specified by the second expression. If the number of characters in the micro string is less than the value of the second expression, and the value of the first expression is positive, the character value is right-justified with the specified fill characters
(zeros or blanks) preceding the value.
If the number of characters in the string is less than the value of the second expression, and the value of the first expression is negative, a minus sign precedes the value. If zero fill is indicated, zeros are used as fill between the minus sign and the value. If blank fill is indicated, blanks are used as fill before the minus sign.
If the number of characters in the string is greater than the value of the second expression, the characters at the beginning of the string are truncated and a warning message is issued.
You can specify the DECMIC pseudo instruction anywhere within a program segment. If the DECMIC pseudo instruction is found within a definition, it is defined and is not recognized as a pseudo instruction. If the DECMIC pseudo instruction is found within a skipping sequence, it is skipped and is not recognized as a pseudo instruction.
The format of the DECMIC pseudo instruction is as follows: name DECMIC [[[expression
1
],expression
2
[,[option]]]]
116 S–2314–50
1
Generated by the assembler
S–2314–50
CAL Pseudo Instruction Descriptions [7]
name is assigned to the character string that represents the decimal value of expression
1 and has redefinable attributes. name must meet the requirements for
identifiers as described in Section 6.3, page 76.
expression1 is optional and represents the micro string equal to the value of the expression. If specified, expression
1 must have an address attribute of value and a relative attribute of absolute with all symbols, if any, previously defined.
If the first expression is not specified, the absolute value of 0 is used. If the current base is mixed, a default of octal is used. If the first expression is not specified, the absolute value of 0 is used when creating the micro string. The expression
1 operand must meet the requirements for expressions as described in
expression2 is optional and provides a positive character count less than or equal to decimal 20. If this parameter is present, the necessary leading zeros or blanks (depending on option) are supplied to provide the requested number of characters. If specified, expression
2 must have an address attribute of value and a relative attribute of absolute with all symbols, if any, previously defined. If the current base is mixed, a default of decimal is used. expression
2 must meet the
requirements for expressions as described in Section 6.9, page 94.
If expression
2 is not specified, the micro string is represented in the minimum number of characters needed to represent the decimal value of the first expression.
option represents the type of fill characters (ZERO for zeros or BLANK for spaces) to be used if the second expression is present and fill is needed. The default is
ZERO . You can enter option in mixed case.
The following example illustrates the use of the DECMIC and MICSIZE pseudo instructions:
MIC MICRO ‘ABCD’
V MICSIZE MIC ; The value of V is the number of
; characters in the micro string
DECT DECMIC V,2
; represented by MIC.
; DECT is a micro name.
_*There are "DECT" characters in MIC.
* There are 19 characters in MIC.1
117
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual
The following example demonstrates the ZERO and BLANK options with positive and negative strings:
BASE D
ONE DECMIC 1,2
_*
*
"ONE"
01
; The base is decimal
TWO DECMIC 5*8+60+900,3
_* "TWO"
* 000
THREE DECMIC -256000,l0,ZERO
_*
*
"THREE"
-000256000
; Returns 1 in 2 digits.
; Returns 1 in 2 digits.
; Decimal 1000.
; Returns 1000 as 3 digits (000).
; Returns 1000 as 3 digits (000).
; Decimal string with zero fill.
; Minus sign, zero fill, value.
; Minus sign, zero fill, value.
; Decimal string with blank fill.
; Blank fill, minus sign, value.
; Blank fill, minus sign, value.
FOUR DECMIC -256000,l0,BLANK
_* "FOUR"
* ^^^-256000
FIVE DECMIC 256000,10,ZERO
_*
*
"FIVE"
0000256000
SIX DECMIC 256000,10,BLANK
_* "SIX"
* ^^^^256000
END
; Zero fill on the left.
; Zero fill on the left.
; Blank fill (^) on the left.
; Blank fill (^) on the left.
SEVEN DECMIC 256000,5
_* "SEVEN"
* 56000
EIGHT DECMIC 777777777,3
_*
*
"EIGHT"
777
; Truncation warning issued.
; Truncation warning issued.
; Truncation warning issued.
; Truncation warning issued.
7.13
DMSG
The DMSG pseudo instruction issues a comment level diagnostic message that contains the string found in the operand field, if a string exists. If the string consists of more than 80 characters, a warning message is issued and the string is truncated.
Comment level diagnostic messages might not be issued by default on the operating system in which the assembler is executing. For more information, see
118 S–2314–50
7.14
DUP
7.15
ECHO
S–2314–50
CAL Pseudo Instruction Descriptions [7]
The assembler recognizes up to 80 characters within the string, but the string may be truncated further when the diagnostic message is issued (depending on the operating system in which the assembler is executing).
You can specify the DMSG pseudo instruction anywhere within a program segment. If the DMSG pseudo instruction is found within a definition, it is defined and is not recognized as a pseudo instruction. If the DMSG pseudo instruction is found within a skipping sequence, it is skipped and is not recognized as a pseudo instruction.
The format of the DMSG pseudo instruction is as follows: ignored DMSG [del-char[string-of-ASCII]del-char]
The del-char variable represents the delimiting character. It must be a single matching character on both ends of the ASCII character string. Apostrophes and spaces are not legal delimiters; all other ASCII characters are allowed.
Two consecutive occurrences of the delimiting character indicate a single such character will be included in the character string.
The string-of ASCII variable represents the ASCII character string that will be printed to the diagnostic file. A maximum of 80 characters is allowed.
Note: Using the DMSG pseudo instruction for assembly timings can be deceiving. For example, if the DMSG pseudo instruction is inserted near the beginning of an assembler segment, more time could elapse (from the time that the assembler begins assembling the segment to the time the message is issued) than you might have expected.
The DUP pseudo instruction introduces a sequence of code that is assembled repetitively a specified number of times The duplicated code immediately follows the DUP pseudo instruction.
The DUP pseudo instruction is described in detail in Section 9.4, page 250.
The ECHO pseudo instruction defines a sequence of code that is assembled zero or more times immediately following the definition.
119
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual
The ECHO pseudo instruction is described in detail in Section 9.5, page 252.
7.16
EDIT
The EDIT pseudo instruction toggles the editing function on and off within a program segment. Appending (\ in the new format) and continuation (, in the old format) are not affected by the EDIT pseudo instruction. The current editing status is reset at the beginning of each segment to the editing option specified on the assembler invocation statement. For a description of statement editing, see
You can specify the EDIT pseudo instruction anywhere within a program segment. If the EDIT pseudo instruction is found within a definition, it is defined and is not recognized as a pseudo instruction. If the EDIT pseudo instruction is found within a skipping sequence, it is skipped and is not recognized as a pseudo instruction.
The format of the EDIT pseudo instruction is as follows: ignored EDIT */ option
The option variable turns editing on and off. option can be specified in uppercase, lowercase, or mixed case, and it can be one of the following:
• ON (enable editing)
• OFF (disable editing)
• No entry (reverts to the format specified on the assembler invocation statement)
An asterisk (*) resumes use of the edit option in effect before the most recent edit option within the current program segment. Each occurrence of an EDIT other than an EDIT * initiates a new edit option. Each EDIT * removes the current edit option and reactivates the edit option that preceded the current edit option.
If the EDIT * statement is encountered and all specified edit options were released, a caution-level message is issued and the default is used.
7.17
EJECT
120
The EJECT pseudo instruction causes the beginning of a new page in the output listing. EJECT is a list control pseudo instruction and by default, is not listed. To
S–2314–50
7.18
ELSE
S–2314–50
CAL Pseudo Instruction Descriptions [7] include the EJECT pseudo instruction on the listing, specify the LIS option on the LIST pseudo instruction.
You can specify the EJECT pseudo instruction anywhere within a program segment. If the EJECT pseudo instruction is found within a definition, it is defined and is not recognized as a pseudo instruction. If the EJECT pseudo instruction is found within a skipping sequence, it is skipped and is not recognized as a pseudo instruction.
The format of the EJECT pseudo instruction is as follows: ignored EJECT ignored
The ELSE pseudo instruction terminates skipping initiated by the IFA, IFC,
IFE , ELSE, or SKIP pseudo instructions with the same label field name. If statements are currently being skipped under control of a statement count,
ELSE has no effect.
You can specify the ELSE pseudo instruction anywhere within a program segment. If the assembler is not currently skipping statements, ELSE initiates skipping. Skipping is terminated by an ELSE pseudo instruction with a matching label field name. If the ELSE pseudo instruction is found within a definition, it is defined and is not recognized as a pseudo instruction.
The format of the ELSE pseudo instruction is as follows: name ELSE ignored
The name variable specifies a required name for a conditional sequence of
code. name must meet the requirements for identifiers as described in Section
The following example illustrates the use of the ELSE pseudo instruction:
SYM
L
DEF
BUF
A
=
=
=
MICRO
IFA
=
1
’LESS THAN’
1000
100
#DEF,A,1
10
121
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual
X
X
Y
BTEST
WARNING
IFA
ERROR
EXT,SYM
; Generate warning message if SYM is
; absolute.
BTEST ELSE
A1 SYM ; Assemble if SYM is not absolute.
BTEST ENDIF
*Assemble BSSZ instruction if W.* is less than BUF, otherwise
*assemble ORG
Z
Z
IFE
BSSZ
SKIP
ORG
IFC
ERROR ERROR
IFC
S1
S2
ENDIF
IFC
S3
S4
IFC
S5
S6
ENDIF
W.*,LT,BUF,2
BUF-W.*
1
BUF
; Generate words of zero to address BUF.
; Skip next statement.
’ "L" ’,EQ,,2
; Error message if micro string defined
; by L is empty.
‘ABCD’,GT,‘ABC’
; If ABCD is greater than ABC,
DEF ; Statement is included.
BUF ; Statement is included.
5
6
’ ’,GT,,2
; If single space is greater than null
; string,
DEF ; Statement is included.
BUF ; Statement is included.
’’’’,EQ,*’*,2
; If single apostrophe equals single
; apostrophe.
; Statement is included.
; Statement is included.
7.19
END
122
The END pseudo instruction terminates a program segment (module initiated with an IDENT pseudo instruction) under the following conditions:
• If the assembler is not in definition mode
• If the assembler is not in skipping mode
S–2314–50
S–2314–50
CAL Pseudo Instruction Descriptions [7]
• If the END pseudo instruction does not occur within an expansion
The format of the END pseudo instruction is as follows: ignored END ignored
If the END pseudo instruction is found within a definition, a skip sequence, or an expansion, a message is issued indicating that the pseudo instruction is not allowed within these modes and the statement is treated as follows:
• Defined if in definition mode
• Skipped if in skipping mode
• Do-nothing instruction if in an expansion
You can specify the END pseudo instruction only from within a program module.
If the END pseudo instruction is valid and terminates a program module, it causes the assembler to take the following actions:
• Generates a cross-reference for symbols if the cross-reference list option is enabled and the listing is enabled
• Clears and resets the format option
• Clears and resets the edit option
• Clears and resets the message level
• Clears and resets all list control options
• Clears and resets the default numeric base
• Discards all qualified, redefinable, nonglobal, and %% symbols
• Discards all qualifiers
• Discards all redefinable and nonglobal micros
• Discards all local macros, opdefs, and local pseudo instructions (defined with an OPSYN pseudo instruction)
• Discards all sections
123
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual
7.20
ENDDUP
The ENDDUP pseudo instruction ends the definition of the code sequence to be repeated. An ENDDUP pseudo instruction terminates a dup or echo definition with the same name.
The ENDDUP pseudo instruction is described in detail in Section 9.8, page 256.
7.21
ENDIF
The ENDIF pseudo instruction terminates skipping initiated by an IFA, IFE, IFC,
ELSE , or SKIP pseudo instruction with the same label field name; otherwise,
ENDIF acts as a do-nothing pseudo instruction. ENDIF does not affect skipping, which is controlled by a statement count.
You can specify the ENDIF pseudo instruction anywhere within a program segment. Skipping is terminated by an ENDIF pseudo instruction with a matching label field name. If the ENDIF pseudo instruction is found within a definition, it is defined and is not recognized as a pseudo instruction.
The format of the ENDIF pseudo instruction is as follows: name ENDIF ignored
The name variable specifies a required name for a conditional sequence of
code. name must meet the requirements for identifiers as described in Section
Note: If an END pseudo instruction is encountered in a skipping sequence, an error message is issued and skipping is continued. You should not use the END pseudo instruction within a skipping sequence.
7.22
ENDM
An ENDM pseudo instruction terminates the body of a macro or opdef definition.
The ENDM pseudo instruction is described in detail in Section 9.6, page 254.
124 S–2314–50
CAL Pseudo Instruction Descriptions [7]
7.23
ENDTEXT
S–2314–50
The ENDTEXT pseudo instruction terminates text source initiated by a TEXT instruction. An IDENT or END pseudo instruction also terminates text source.
The ENDTEXT is a list control pseudo instruction and by default, is not listed unless the TXT option is enabled. If the LIS option is enabled, the ENDTEXT instruction is listed regardless of other listing options.
You can specify the ENDTEXT pseudo instruction anywhere within a program segment. If the ENDTEXT pseudo instruction is found within a definition, it is defined and is not recognized as a pseudo instruction. If the ENDTEXT pseudo instruction is found within a skipping sequence, it is skipped and is not recognized as a pseudo instruction.
The format of the ENDTEXT pseudo instruction is as follows: ignored ENDTEXT ignored
The following example illustrates the use of the ENDTEXT pseudo instruction
(with the TXT option off).
The following represents the source listing:
B
C
A
IDENT
=
TXTNAME TEXT
=
=
ENDTEXT
A1
A2
END
TEXT
2
’An example.’
3
4
A
B
The following represents the output listing:
A
IDENT TEXT
= 2
TXTNAME TEXT ’An example.’
A1
A2
END
A
B
125
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual
7.24
ENTRY
The ENTRY pseudo instruction specifies symbolic addresses or values that can be referred to by other program modules linked by the loader. Each entry symbol must be an absolute, immobile, or relocatable symbol defined within the program module.
The ENTRY pseudo instruction is restricted to sections that allow instructions or data or both. If the ENTRY pseudo instruction is found within a definition or skipping sequence, it is defined and not recognized as a pseudo instruction.
The format of the ENTRY pseudo instruction is as follows: ignored ENTRY [symbol:[attr]],[symbol:[attr]]
The symbol variable specifies the name of zero, one, or more symbols. Each of the names must be defined as an unqualified symbol within the same program module. The corresponding symbol must not be redefinable, external, or relocatable relative to either a stack or a task common section.
The length of the symbol is restricted depending on the type of loader table that the assembler is currently generating. If the symbol is too long, the assembler will issue an error message.
The symbol operand must meet the requirements for symbols as described in
The attr attribute can be either OBJ or FUNC, specifying whether the symbol is an object or a function entry point. If attr is not specified and the ENTRY pseudo instruction occurs in a DATA, CONST, or COMMON section, the default is OBJ.
The following example illustrates the use of the ENTRY pseudo instruction:
EPTNME =
TREG =
.
.
ENTRY EPTNME,TREG
.
*
O’17
7.25
ERRIF
126
The ERRIF pseudo instruction conditionally issues a listing message. If the condition is satisfied (true), the appropriate user-defined message is issued. If
S–2314–50
S–2314–50
CAL Pseudo Instruction Descriptions [7] the level is not specified, the ERRIF pseudo instruction issues an error-level message. If the condition is not satisfied (false), no message is issued. If any errors are encountered while evaluating the operand field, the resulting condition is handled as if true and the appropriate user-defined message is issued.
You can specify the ERRIF pseudo instruction anywhere within a program segment. If the ERRIF pseudo instruction is found within a definition, it is defined and is not recognized as a pseudo instruction. If the ERRIF pseudo instruction is found within a skipping sequence, it is skipped and is not recognized as a pseudo instruction.
The format of the ERRIF pseudo instruction is as follows:
[option] ERRIF [expression],condition,[expression]
The option variable used in the ERRIF pseudo instruction is the same as in the
ERROR pseudo instruction. See the ERROR pseudo instruction for information.
Zero, one, or two expressions to be compared by condition. If one or both of the expressions are missing, a value of absolute 0 is substituted for every expression that is not specified. Symbols found in either of the expressions can be defined later in a segment.
The expression operand must meet the requirements for expressions as described
The condition variable specifies the relationship between two expressions that causes the generation of an error. For LT, LE, GT, and GE, only the values of the expressions are examined. You can enter condition in uppercase, lowercase, or mixed case, and it can be one of the following:
• LT (less than)
The value of the first expression must be less than the value of the second expression.
• LE (less than or equal)
The value of the first expression must be less than or equal to the value of the second expression.
• GT (greater than)
The value of the first expression must be greater than the value of the second expression.
127
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual
• GE (greater than or equal)
The value of the first expression must be greater than or equal to the value of the second expression.
• EQ (equal)
The value of the first expression must be equal to the value of the second expression. Both expressions must be one of the following:
– Absolute
– Immobile relative to the same section
– Relocatable in the program section or the same common section
– External relative to the same external symbol.
The word-address, byte-address, or value attributes must be the same.
• NE (not equal)
The first expression must not equal the second expression. Both expressions cannot be absolute, or external relative to the same external symbol, or relocatable in the program section or the same common section. The word-address, byte-address, or value attributes are not the same.
The ERRIF pseudo instruction does not compare the address and relative attributes. A CAUTION level message is issued.
The following example illustrates the use of the ERRIF pseudo instruction:
P ERRIF ABC,LT,DEF
7.26
ERROR
128
The ERROR pseudo instruction unconditionally issues a listing message. If the level is not specified, the ERROR pseudo instruction issues an error level message.
If the condition is not satisfied (FALSE), no message is issued.
You can specify the ERROR pseudo instruction anywhere within a program segment. If the ERROR pseudo instruction is found within a definition, it is defined and is not recognized as a pseudo instruction. If the ERROR pseudo instruction is found within a skipping sequence, it is skipped and is not recognized as a pseudo instruction.
The format of the ERROR pseudo instruction is as follows:
S–2314–50
7.27
EXITM
7.28
EXT
S–2314–50
CAL Pseudo Instruction Descriptions [7]
[option] ERROR ignored
The option variable specifies the error level. It can be entered in upper, lower, or mixed case. The following error levels are mapped directly into a user-defined message of the corresponding level:
COMMENT, NOTE, CAUTION, WARNING, or ERROR
The following levels are mapped into an error-level message:
C, D, E, F, I, L, N, O, P, R, S, T, U, V, or X
The following levels are mapped into warning-level messages:
W, W1, W2, W3, W4, W5, W6, W7, W8, W9, Y1, or Y2
Messages C through Y2 provide compatibility with Cray Assembly Language, version 1 (CAL1).
The assembler can produce five similar messages with differing levels (error, warning, caution, note, or comment). The ERROR pseudo instruction can be used to check for valid input and to assign an appropriate message.
In the following example, a user-defined error level message is specified:
ERROR ERROR ; ***ERROR*** Input is not valid
The EXITM pseudo instruction immediately terminates the innermost nested macro or opdef expansion, if any, caused by either a macro or an opdef call.
The EXITM pseudo instruction is described in detail in Section 9.7, page 255.
The EXT pseudo instruction specifies linkage to symbols that are defined as entry symbols in other program modules. They can be referred to from within the program module, but must not be defined as unqualified symbols elsewhere within the program module. Symbols specified in the EXT instruction are defined as unqualified symbols that have relative attributes of external and specified address.
129
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual
130
You can specify the EXT pseudo instruction anywhere within a program module.
If the EXT pseudo instruction is found within a definition or skipping sequence, it is defined and not recognized as a pseudo instruction.
The format of the EXT pseudo instruction is as follows: ignored EXT [symbol:[attribute]],[symbol:[attribute]]
The variables associated with the EXT pseudo instruction are described as follows:
• symbol
The symbol variable specifies the name of zero, one, or more external symbols.
Each of the names must be an unqualified symbol that has a relative attribute of external and the corresponding address attribute.
The length of the symbol is restricted depending on the type of loader table that the assembler is currently generating. If the symbol is too long, the assembler will issue an error message.
The symbol operand must meet the requirements for symbols as described in
• attribute
The attribute variable specifies the attribute symbol-type as follows:
– The symbol-type will be assigned to the external symbol; it can be one of the following:
V or v
OBJ
Value (default)
Object
FUNC Function
– The linkage-attribute type is the linkage attribute that will be assigned to the external symbol. Linkage attributes can be specified in uppercase, lowercase, or mixed case, and they can be one of the following:
HARD (default)
SOFT
If the linkage-attribute is not specified on the EXT pseudo instruction, the default is HARD. All hard external references are resolved at load time.
S–2314–50
7.29
FORMAT
S–2314–50
CAL Pseudo Instruction Descriptions [7]
A soft reference for a particular external name is resolved at load time only when at least one other module has referenced that same external name as a hard reference.
You conditionally reference a soft external name at execution time. If a soft external name was not included at load time and is referenced at execution time, an appropriate message is issued.
If the operating system for which the assembler is generating code does not support soft externals, a caution-level message is issued and soft externals are treated as hard externals.
Note: Typically, a soft external is used for references to large software packages (such as graphics packages) that may not be required in a particular load. When such code is required, load time is shorter and the absolute module is smaller in size. For most uses, however, hard externals are recommended.
The following example illustrates the use of the EXT pseudo instruction:
.
.
IDENT A
.
ENTRY VALUE
VALUE = 2.0
.
.
.
END
IDENT B
EXT VALUE
CON VALUE ; The 64-bit external. External value 2.0 is
; stored here by the loader.
END
The assembler supports both the CAL version 1 (CAL1) statement format and a new statement format. The FORMAT pseudo instruction lets you switch between statement formats within a program segment. The current statement format is reset at the beginning of each section to the format option specified on the assembler invocation statement. For a description of the recommended
formatting conventions for the new format, see Section 6.2.1.2, page 72.
131
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual
You can specify the FORMAT pseudo instruction anywhere within a program segment. If the FORMAT pseudo instruction is found within a definition, it is defined and is not recognized as a pseudo instruction. If the FORMAT pseudo instruction is found within a skipping sequence, it is skipped and is not recognized as a pseudo instruction.
The format of the FORMAT pseudo instruction is as follows: ignored FORMAT */ option
The option variable specifies old or new format. option can be specified in uppercase, lowercase, or mixed case, and it can be one of the following:
• OLD (old format)
• NEW (new format)
• No entry (reverts to the EDIT option specified on the assembler invocation statement)
An asterisk (*) resumes use of the format option in effect before the most recent format option within the current program segment. Each occurrence of a FORMAT other than a FORMAT * initiates a new format option. Each FORMAT * removes the current format option and reactivates the format that preceded the current format. If the FORMAT * statement is encountered and all specified format options were released, a caution-level message is issued and the default is used.
7.30
IDENT
132
The IDENT pseudo instruction identifies a program module and marks its beginning. The module name appears in the heading of the listing produced by the assembler (if the title pseudo instruction has not been used) and in the generated binary load module.
You must specify the IDENT pseudo instruction in the global part of a CAL program. If the IDENT pseudo instruction is found within a definition, it is defined and is not recognized as a pseudo instruction. If the IDENT pseudo instruction is found within a skipping sequence, it is skipped and is not recognized as a pseudo instruction.
The format of the IDENT pseudo instruction is as follows: ignored IDENT lname
S–2314–50
7.31
IFA
S–2314–50
CAL Pseudo Instruction Descriptions [7]
The lname variable is the long name of the program module. lname must meet the
requirements for long names as described in Section 6.3, page 76.
The length of the long name is restricted depending on the type of loader table the assembler is currently generating. If the name is too long, the assembler issues an error message.
The following example illustrates the use of the IDENT pseudo instruction:
; Beginning of the EXAMPLE program module IDENT EXAMPLE
.
.
.
END
; Other code goes here
; End of the EXAMPLE program module
The IFA pseudo instruction tests an attribute of an expression. If the expression has the specified attribute, assembly continues with the next statement. If the result of the attribute test is false, subsequent statements are skipped. If a label field name is present, skipping stops when an ENDIF or ELSE pseudo instruction with the same name is encountered; otherwise, skipping stops when the statement count is exhausted.
If any errors are encountered while evaluating the attribute-condition, the resulting condition is handled as if true and the appropriate listing message is issued.
You can specify the IFA pseudo instruction anywhere within a program segment.
If the IFA pseudo instruction is found within a definition, it is defined and is not recognized as a pseudo instruction. If the IFA pseudo instruction is found within a skipping sequence, it is skipped and is not recognized as a pseudo instruction.
The format of the IFA pseudo instruction is as follows:
[name]
[name]
[name]
[name]
IFA
IFA
IFA
IFA
[# or !]exp-attribute,expression[,[count]]
[# or !]redef-attribute,symbol[,[count]]
[# or !]reg-attribute,reg-arg_value[,[count]]
[# or !]micro-attribute,mname[,[count]]
133
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual
134
The name variable specifies an optional name of a conditional sequence of code.
A conditional sequence of code that is controlled by a name is ended by an
ENDIF pseudo instruction with a matching name. To reverse the condition of a conditional sequence of code controlled by a name, use an ELSE pseudo instruction with a matching name. If both name and count are present, name takes precedence. name must meet the requirements for names as described in
Either the pound sign (#) or the exclamation mark (!) is optional and negates the condition. If errors occur in the attribute condition, the condition is evaluated as if it were true. Although # or ! does not change the condition, it does specify the if not condition.
The exp-attribute variable is a mnemonic that signifies an attribute of expression.
expression must meet the requirement for an expression as described in Section
An expression has only one address attribute (VAL, PA, or WA) and relative attribute (ABS, IMM, REL, or EXT). An attribute also can be any of the following mnemonics preceded by a complement sign (#), indicating that the second subfield does not satisfy the corresponding condition. You can specify all of the following mnemonics in mixed case:
Mnemonic
VAL
PA
WA
ABS
IMM
REL
EXT
CODE
Attribute
Value; requires all symbols within the expression to be defined previously.
byte address; requires all symbols, if any, within the expression to be defined previously.
Word address; requires all symbols, if any, within the expression to be defined previously.
Absolute; requires all symbols, if any, within the expression to be defined previously.
Immobile; requires all symbols, if any, within the expression to be defined previously.
Relocatable; requires all symbols, if any, within the expression to be defined previously.
External; requires all symbols, if any, within the expression to be defined previously.
Immobile or relocatable; relative to a code section.
CODE requires all symbols, if any, within the expression to be defined previously.
S–2314–50
S–2314–50
COMMON
TASKCOM
ZEROCOM
DYNAMIC
STACK
CM
DATA
ZERODATA
CONST
MIXED
COM
EM
CAL Pseudo Instruction Descriptions [7]
Immobile or relocatable; relative to a data section.
DATA requires all symbols, if any, within the expression to be defined previously.
Immobile or relocatable; relative to a zero data section. ZERODATA requires all symbols, if any, within the expression to be defined previously.
Immobile or relocatable; relative to a constant section. CONST requires all symbols, if any, within the expression to be defined previously.
Immobile or relocatable; relative to a common section. MIXED requires all symbols, if any, within the expression to be defined previously.
Immobile or relocatable; relative to a common section. COM requires all symbols, if any, within the expression to be defined previously.
Immobile or relocatable; relative to a common section. COMMON requires all symbols, if any, within the expression to be defined previously.
Immobile or relocatable; relative to a task common section. TASKCOM requires all symbols, if any, within the expression to be defined previously.
Immobile or relocatable; relative to a zero common section. ZEROCOM requires all symbols, if any, within the expression to be defined previously.
Immobile or relocatable; relative to a dynamic section. DYNAMIC requires all symbols, if any, within the expression to be defined previously.
Immobile or relocatable; relative to a stack section.
STACK requires all symbols, if any, within the expression to be defined previously.
Immobile or relocatable; relative to a section that is placed into common memory. CM requires all symbols, if any, within the expression to be defined previously.
Immobile or relocatable; relative to a section that is placed into extended memory. EM requires all symbols, if any, within the expression to be defined previously. If EM is specified, the condition always fails.
135
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual
136
LM
DEF
Immobile or relocatable; relative to a section that is placed into local memory. LM requires all symbols, if any, within the expression to be defined previously. If LM is specified for a Cray system, the condition always fails.
True if all symbols in the expression were defined previously; otherwise, the condition is false.
The redef-attribute variable specifies a redefinable attribute. The condition is true if the symbol following redef-attribute is redefinable; otherwise, the condition is false. Redefinable attribute is defined as follows:
Mnemonic
SET
Attribute
The symbol in the second subfield is a redefinable symbol. symbol must meet the requirements for a
symbol as described in Section 6.2, page 70.
The reg-attribute variable specifies a register attribute. reg-arg-value is any
ASCII character up to but not including a legal terminator (blank character or semicolon; new format) and element separator character (,). If you specify REG, the condition is true if the following string is a valid complex-register; otherwise, the condition is false. Register-attribute is defined as follows:
Mnemonic
REG
Attribute
The second subfield contains a valid A, S, or C register designator.
The micro-attribute variable specifies an attribute of the micro specified by mname.
is an existing micro name; otherwise, the condition is false. micro-attribute is defined as follows:
Mnemonic
MIC
MICRO
CMICRO
Attribute
The name in the second subfield is a micro name.
The name in the second subfield is a micro name and the corresponding micro can be redefined.
The name in the second subfield is a micro name and the corresponding micro is constant.
The count variable specifies the statement count. It must be an absolute expression with positive value. All symbols in the expression, if any, must be
S–2314–50
7.32
IFC
S–2314–50
CAL Pseudo Instruction Descriptions [7] previously defined. A missing or null count subfield gives a zero count. count is used only when the label field is not specified. If name is not present and count is present in the operand field, skipping stops when count is exhausted. If neither
name nor count is present, no skipping occurs.
The following example illustrates the use of the IFA pseudo instruction:
SYM1 SET 1
SYM2 = 2
IFA SET,SYM1,2
S1 SYM1
S2 SYM2
SYM2 = 1
IFA SET,SYM2,1
S3 SYM2
; If the condition is true,
; include this statement
; include this statement
; If the condition is false,
; skip this statement.
The IFC pseudo instruction tests a pair of character strings for a condition under which code will be assembled if the relation specified by conditionis satisfied
(true). If the relationship is not satisfied (false), subsequent statements are skipped. If a label field name is present, skipping stops when an ENDIF or ELSE pseudo instruction with the same name is encountered; otherwise, skipping stops when the statement count is exhausted.
If any errors are encountered during evaluation of the string condition, the resulting condition is handled as if true and an appropriate listing message is issued.
You can specify the IFC pseudo instruction anywhere within a program segment.
If the IFC pseudo instruction is found within a definition, it is defined and is not recognized as a pseudo instruction. If the IFC pseudo instruction is found within a skipping sequence, it is skipped and is not recognized as a pseudo instruction.
The format of the IFC pseudo instruction is as follows:
[name] IFC [string],condition,[string] [,[count]]
The name variable specifies an optional name of a conditional sequence of code.
A conditional sequence of code that is controlled by a name is ended by an
ENDIF pseudo instruction with a matching name. To reverse the condition of a conditional sequence of code controlled by a name, use an ELSE pseudo instruction with a matching name. If both name and count are present, name
137
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual
138 takes precedence. name must meet the requirements for names as described in
The string variable specifies the character string that will be compared. The first and third subfields can be null (empty) indicating a null character string. The
ASCII character code value of each character in the first string is compared with the value of each character in the second string. The comparison is from left to right and continues until an inequality is found or until the longer string is exhausted. A value of 0 is substituted for missing characters in the shorter string.
Micros and formal parameters can be contained in the character strings.
The string operand is an optional ASCII character string that must be specified with one matching character on both ends. A character string can be delimited by any ASCII character other than a comma or space. Two consecutive occurrences of the delimiting character indicate a single such character will be included in the character string.
The following example compares the character strings O’100 and ABCD*:
AIF IFC =O’l00=,EQ,*ABCD***
The condition variable specifies the relation that will be satisfied by the two strings. You can enter condition in mixed case, and it must be one of the following:
• LT (less than)
The value of the first string must be less than the value of the second string.
• LE (less than or equal)
The value of the first string must be less than or equal to the value of the second string.
• GT (greater than)
The value of the first string must be greater than the value of the second string.
• GE (greater than or equal)
The value of the first string must be greater than or equal to the value of the second string.
• EQ (equal)
The value of the first string must be equal to the value of the second string.
• NE (not equal)
S–2314–50
7.33
IFE
S–2314–50
CAL Pseudo Instruction Descriptions [7]
The value of the first string must not equal the value of the second string.
The count variable specifies the statement count. It must be an absolute expression with positive value. All symbols in the expression, if any, must be previously defined. A missing or null count subfield gives a zero count. The
count operand is used only when the label field is not specified. If name is not present and count is present in the operand field, skipping stops when count is exhausted. If neither name nor count is present, no skipping occurs.
The following examples illustrates the use of the IFC pseudo instruction. The first string is delimited by the at sign (@), and the second string is delimited by the percent sign (%). The first string is equal to the second string.
IDENT TEST
EX1 IFC @ABC@@D@,EQ,%ABC@D%
EX1
EX1
S1
S2
ELSE
S3
ENDIF
END
1
2
3
; The condition is true.
; Skipping does not occur.
; Statement is included.
; Statement is included.
; Statements within the ELSE sequence
; are included only if the condition
; fails.
; Statement is skipped.
; End of skip sequence.
In the next example, the first string is not equal to the second string, the two statements following the IFC are skipped.
IDENT TEST
EX1 IFC @ABBCD@,EQ,@ABCD@ ; The condition is false.
EX1
S1 1
S2 2
ENDIF
S3 3
; Skipping occurs.
; Statement is skipped.
; Statement is skipped.
; End of skip sequence
; This statement is included regardless
; of whether the condition is true or
; false.
END
The IFE pseudo instruction tests a pair of expressions for a condition. If the relation (condition)specified by the operation is satisfied, code is assembled. If
139
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual
140
conditionis true, assembly resumes with the next statement; if condition is false, subsequent statements are skipped. If a label field name is present, skipping stops when an ENDIF or ELSE pseudo instruction with the same name is encountered; otherwise, skipping stops when the statement count is exhausted.
If any errors are encountered during the evaluation of the expression-condition, the resulting condition is handled as if true and an appropriate listing message is issued.
If an assembly error is detected, assembly continues with the next statement.
You can specify the IFE pseudo instruction anywhere within a program segment.
If the IFE pseudo instruction is found within a definition, it is defined and is not recognized as a pseudo instruction. If the IFE pseudo instruction is found within a skipping sequence, it is skipped and is not recognized as a pseudo instruction.
The format of the IFE pseudo instruction is as follows:
[name] IFE [expression],condition,[expression] [,[count]]]
The name variable specifies an optional name of a conditional sequence of code.
A conditional sequence of code that is controlled by a name is ended by an
ENDIF pseudo instruction with a matching name. To reverse the condition of a conditional sequence of code controlled by a name, use an ELSE pseudo instruction with a matching name. If both name and count are present, name takes precedence. name must meet the requirements for names as described in name
must meet the requirements for names as described in Section 6.3, page 76.
The expression variables specify the expressions to be compared. All symbols in the expression must be defined previously. If an expression is not specified, the absolute value of 0 is used. expressions must meet the requirements for
expressions as described in Section 6.9, page 94.
The condition variable specifies the relation to be satisfied by the two strings. You can enter condition in mixed case, and it must be one of the following:
• LT (less than)
The value of the first expression must be less than the value of the second expression. The attributes are not checked.
• LE (less than or equal)
The value of the first expression must be less than or equal to the value of the second expression. The attributes are not checked.
S–2314–50
S–2314–50
CAL Pseudo Instruction Descriptions [7]
• GT (greater than)
The value of the first expression must be greater than the value of the second expression. The attributes are not checked.
• GE (greater than or equal)
The value of the first expression must be greater than or equal to the value of the second expression. The attributes are not checked.
• EQ (equal)
The value of the first expression must be equal to the value of the second expression. Both expressions must be one of the following:
– Attributes must be the same
– Immobile relative to the same section
– Relocatable relative to the same section
– External relative to the same external symbol.
– The word-address, byte-address, or value
• NE (not equal)
The first expression and the second expression do not satisfy the conditions required for EQ described above.
The count variable specifies the statement count. It must be an absolute expression with a positive value. All symbols in the expression, if any, must be previously defined. A missing or null count subfield gives a zero count. count is used only when the label field is not specified. If name is not present and count is present in the operand field, skipping stops when count is exhausted. If neither
name nor count is present, no skipping occurs.
The following example illustrates the use of the IFE pseudo instruction:
SYM1
SYM2
SYM3
SYM4
IDENT TEST
=
=
SET
0
*
1000
SET 500
NOTEQ IFE SYM1,EQ,SYM2
S1 SYM1
S2 SYM2
; Condition fails, values are the same,
; but the attributes are different.
; The ELSE sequence is assembled.
141
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual
NOTEQ ELSE
S1 SYM3
S2 SYM4
NOTEQ ENDIF
END
; Statement is included.
; Statement is included.
; End of conditional sequence.
7.34
IFM
142
The IFM pseudo instruction tests characteristics of the current target machine.
If the result of the machine condition is true, assembly continues with the next statement. If the result of the machine condition is false, subsequent statements are skipped. If a label field name is present, skipping stops when an ENDIF or
ELSE pseudo instruction with the same name is encountered; otherwise, skipping stops when the statement count is exhausted.
If any errors are encountered during the evaluation of the string condition, the resulting condition is handled as if true and an appropriate listing message is issued.
You can specify the IFM pseudo instruction anywhere within a program segment.
If the IFM pseudo instruction is found within a definition, it is defined and is not recognized as a pseudo instruction. If the IFM pseudo instruction is found within a skipping sequence, it is skipped and is not recognized as a pseudo instruction.
The format of the IFM pseudo instruction is as follows:
[name]
[name]
IFM
IFM
[# or !]logical-name[,[count]]
numeric-name,condition,[expression] [,[count]]
The name variable specifies an optional name of a conditional sequence of code.
A conditional sequence of code that is controlled by a name is ended by an
ENDIF pseudo instruction with a matching name. To reverse the condition of a conditional sequence of code controlled by a name, use an ELSE pseudo instruction with a matching name. If both name and count are present, name takes precedence. name must meet the requirements for names as described in
The logical-name variable specifies the mnemonic that signifies a logical condition of the machine for which the assembler is currently targeting code. If the logical name is preceded by a pound sign (#) or an exclamation mark (!), its resulting condition is complemented. For a detailed list of the mnemonics, see the logical
S–2314–50
S–2314–50
CAL Pseudo Instruction Descriptions [7]
traits of the CPU option for the appropriate operating system in Chapter 5, page 47.
The numeric-name variable specifies the mnemonic that signifies a numeric condition of the machine for which the assembler is currently targeting code.
For a detailed list of the mnemonics, see the numeric traits of the CPU option for
the appropriate operating system in Chapter 5, page 47. You can specify these
mnemonics in mixed case.
The condition variable specifies the relation to be satisfied between the numeric name and the expression, if any. You can enter condition in mixed case, and it must be one of the following:
• LT (less than)
The value of the numeric name must be less than the value of the expression.
• LE (less than or equal)
The value of the numeric name must be less than or equal to the value of the expression.
• GT (greater than)
The value of the numeric name must be greater than the value of the expression.
• GE (greater than or equal)
The value of the numeric name must be greater than or equal to the value of the expression.
• EQ (equal)
The value of the numeric name must be equal to the value of the expression.
• NE (not equal)
The value of the numeric name must not equal the value of the expression.
The expression variable specifies the expression to be compared to the numeric name. All symbols in the expression must be defined previously and must have an address attribute of value and a relative attribute of absolute. If the current base is mixed, a default of decimal is used. If an expression is not specified, the absolute value of 0 is used. expression must meet the requirements for expressions
as described in Section 6.9, page 94.
143
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual
The count variable specifies the statement count. It must be an absolute expression with a positive value. All symbols in the expression, if any, must be previously defined. A missing or null count subfield gives a zero count. count is used only when the label field is not specified. If name is not present and count is present in the operand field, skipping stops when count is exhausted. If neither
name nor count is present, no skipping occurs.
The following example illustrates the use of the IFM pseudo instruction: ident test ex1 ifm vpop
.
.
.
ex1 ifm numcpus,eq,4
.
.
.
ex2 else
.
.
.
ex2 endif end
; Assuming the condition is true,
; skipping does occur within the IFM
; part.
; Assuming the condition is false,
; skipping occurs.
; Toggles the condition so that the else
; part is not skipped.
7.35
INCLUDE
144
The INCLUDE pseudo instruction inserts a file at the current source position. The
INCLUDE pseudo instruction always prepares the file for reading by opening it and positioning the pointer at the beginning.
You can use this pseudo instruction to include the same file more than once within a particular file.
You can also nest INCLUDE instructions. Because you cannot use INCLUDE recursively, you should review nested INCLUDE instructions for recursive calls to a file that you have already opened.
You can specify the INCLUDE pseudo instruction anywhere within a program segment. If the INCLUDE pseudo instruction occurs within a definition, it is recognized as a pseudo instruction and the specified file is included in the definition. If the INCLUDE pseudo instruction occurs within a skipping sequence, it is recognized as a pseudo instruction and the specified file is included in the
S–2314–50
S–2314–50
CAL Pseudo Instruction Descriptions [7] skipping sequence. The INCLUDE pseudo instruction statement itself is not inserted into a defined sequence of code.
Note: The INCLUDE pseudo instruction can be forced into a definition or skipped sequence of code. When editing is enabled, INCLUDE is expanded during execution and the file is read in at that point. This method is not recommended because formal parameters are not substituted correctly into statements when the INCLUDE macro is expanded during execution. If using this method, insert an underscore (_) anywhere within the pseudo instruction, as follows: IN_CLUDE. If editing is disabled during execution, INCLUDE is not expanded.
The format of the INCLUDE pseudo instruction is as follows: ignored INCLUDE filename
The filename variable is an ASCII character string that identifies the file to be included. The ASCII character string must be a valid file name depending on the operating system under which the assembler is executing. If the ASCII character string is not a valid file name or the assembler cannot open the file, a listing message is issued.
filename must be specified with one matching character on each end. Any ASCII character other than a comma or space can be used. Two consecutive occurrences of the delimiting character indicate a single such character will be included in the character string.
In the following examples, the module named INCTEST contains an INCLUDE pseudo instruction. The file to be included is named DOG and the CAT file is included within the DOG file.
The INCTEST module is as follows:
IDENT INCTEST
INCLUDE *DOG*
END
; Call file DOG with INCLUDE.
The file DOG contains the following:
S1 1
INCLUDE ’CAT’
S2 2
; Register S1 gets 1.
; Call file CAT with INCLUDE.
; Register S2 gets 2.
145
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual
146
The file CAT contains the following:
S3 3 ; Register S3 gets 3.
The expansion of the INCTEST module is as follows:
IDENT INCTEST
INCLUDE *DOG*
S1 1
INCLUDE ’CAT’
S3
S2
END
3
2
; Call file DOG with INCLUDE.
; Register S1 gets 1.
; Call file CAT with INCLUDE.
; Register S3 gets 3.
; Register S2 gets 2.
The following example demonstrates that it is illegal to include a file recursively within nested INCLUDE instructions.
The INCTEST module is as follows:
IDENT INCTEST
INCLUDE *DOG*
END
; Call file DOG with INCLUDE.
The file DOG contains the following:
S1 1
INCLUDE ’CAT’
S2 2
; Register S1 gets 1.
; Call file CAT with INCLUDE.
; Register S2 gets 2.
The file CAT includes the following:
S3 3 ; Register S3 gets 3.
INCLUDE -DOG; Illegal. If file B was included by
; file A, it cannot include file A.
The following example demonstrates that it is legal to include a file more than once if it is not currently being included.
The INCTEST module is as follows:
IDENT INCTEST
INCLUDE *DOG*
INCLUDE *DOG*
END
; Call file DOG with INCLUDE.
; Call file DOG with INCLUDE.
The file DOG contains the following:
S–2314–50
7.36
LIST
S–2314–50
CAL Pseudo Instruction Descriptions [7]
S1
S2
1
2
; Register S1 gets 1.
; Register S2 gets 2.
The expansion of the INCTEST module is as follows:
IDENT INCTEST
INCLUDE *DOG* ; Call file DOG with INCLUDE.
S1
S2
1
2
; Register S1 gets 1.
; Register S2 gets 2.
INCLUDE *DOG* ; Call file DOG with INCLUDE.
S1 1 ; Register S1 gets 1.
S2
END
2 ; Register S2 gets 2.
The LIST pseudo instruction controls the listing. LIST is a list control pseudo instruction and by default, is not listed. To include the LIST pseudo instruction on the listing, specify the LIS option on this instruction. An END pseudo instruction resets options to the default values.
You can specify the LIST pseudo instruction anywhere within a program segment. If the LIST pseudo instruction is found within a definition, it is defined and is not recognized as a pseudo instruction. If the LIST pseudo instruction is found within a skipping sequence, it is skipped and is not recognized as a pseudo instruction.
The format of the LIST pseudo instruction is as follows:
[name] LIST [option]{,[option]}*
The name variable specifies the optional list name. name must meet the
requirements for identifiers as described in Section 6.3, page 76.
If name is present, the instruction is ignored unless a matching name is specified on the list parameter on the assembler invocation statement. LIST pseudo instructions with a matching name are not ignored. LIST pseudo instructions with a blank label field are always processed.
The option variable specifies that a particular listing feature be enabled or disabled. All option names can be specified in some form as assembler invocation statement parameters. The selection of an option on the assembler invocation statement overrides the enabling or disabling of the corresponding feature by
147
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual
148 a LIST pseudo instruction. If you use the no-list option on the assembler invocation statement, all LIST pseudo instructions in the program are ignored.
There can be zero, one, or more options specified or an *. If no options are specified, OFF is assumed. The allowed options are described as follows:
• ON (enables source statement listing)
Source statements and code generated are listed (default).
• OFF (disables source statement listing)
While this option is selected, only statements with errors are listed. If the LIS option is enabled, listing control pseudo instructions are also listed.
• ED (enables listing of edited statements)
Edited statements are included in the listing file (default).
• NED (disables listing of edited statements)
Edited statements are not included in the listing file.
• XRF (enables cross-reference)
Symbol references are accumulated and a cross-reference listing is produced
(default).
• NXRF (disables cross-reference)
Symbol references are not accumulated. If this option is selected when the
END pseudo instruction is encountered, no cross-reference is produced.
• XNS (includes nonreferenced local symbols in the reference)
Local symbols that were not referenced in the listing output are included in the cross-reference listing (default).
• NXNS (excludes nonreferenced local symbols from the cross-reference)
If this option is selected when the END pseudo instruction is encountered, local symbols that were not referenced in the listing output are not included in the cross-reference.
• LIS (enables listing of the listing pseudo instructions)
The LIST, SPACE, EJECT, TITLE, SUBTITLE, TEXT, and ENDTEXT pseudo instructions are included in the listing.
• NLIS (disables listing of the listing pseudo instructions)
S–2314–50
S–2314–50
CAL Pseudo Instruction Descriptions [7]
The LIST, SPACE, EJECT, TITLE, SUBTITLE, TEXT, and ENDTEXT pseudo instructions are not included in the listing (default).
• TXT (enables global text source listing)
Each statement following a TEXT pseudo instruction is listed through the
ENDTEXT instruction if the listing is otherwise enabled.
• NTXT (disables global text source listing)
Statements that follow a TEXT pseudo instruction through the following
ENDTEXT instruction are not listed (default).
• MAC (enables listing of macro and opdef expansions)
Statements generated by macro and opdef calls are listed. Conditional statements and skipped statements generated by macro and opdef calls are not listed unless the macro conditional list feature is enabled (MIF).
• NMAC (disables listing of macro and opdef expansions)
Statements generated by macro and opdef calls are not listed (default).
• MBO (enables listing of generated statements before editing)
Only statements that produce generated code are listed. The listing of macro expansions (MAC) or the listing of duplicated statements (DUP) must also be enabled.
• NMBO (disables listing of statements that produce generated code)
Statements generated by a macro or opdef call (MAC), or by a DUP or ECHO
(DUP) pseudo instruction, are not listed before editing (default).
Note: Source statements containing a micro reference (see MIC and NMIC options) or a concatenation character are listed before editing regardless of whether this option is enabled or disabled.
• MIC (enables listing of generated statements before editing)
Statements that are generated by a macro or opdef call, or by a DUP or ECHO pseudo instruction, and that contain a micro reference or concatenation character are listed before and after editing. The listing of macro expansions or the listing of duplicated statements must also be enabled.
• NMIC (disables listing of generated statements before editing)
Statements generated by a macro or opdef call, or by a DUP or ECHO pseudo instruction, are not listed before editing (default).
149
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual
Note: Conditional statements (see NIF and NMIF options) and skipped statements in source code are listed regardless of whether this option is enabled or disabled.
• MIF (enables macro conditional listing)
Conditional statements and skipped statements generated by a macro or opdef call, or by a DUP or ECHO pseudo instruction, are listed. The listing of macro expansions or the listing of duplicated statements must also be enabled.
• NMIF (disables macro conditional listing)
Conditional statements and skipped statements generated by a macro or opdef call, or by a DUP or ECHO pseudo instruction, are not listed (default).
• DUP (enables listing of duplicated statements)
Statements generated by DUP and ECHO expansions are listed. Conditional statements and skipped statements generated by DUP and ECHO are not listed unless the macro conditional list feature is enabled (MIF).
• NDUP (disables listing of duplicated statements)
Statements generated by DUP and ECHO are not listed (default).
The asterisk (*) reactivates the LIST pseudo instruction in effect before the current LIST pseudo instruction was specified within the current program segment. Each occurrence of a LIST pseudo instruction other than LIST initiates a new listing control. Each LIST releases the current listing control and reactivates the listing control that preceded the current list control. If all specified listing controls were released when a LIST * is encountered, the assembler issues a caution-level message and uses the defaults for listing control.
7.37
LOC
150
The LOC pseudo instruction sets the location counter to the first byte of the word address specified. The location counter is used for assigning address values to label field symbols. Changing the location counter allows code to be assembled and loaded at one location, controlled by the origin counter, then moved and executed at another address controlled by the location counter. The LOC pseudo instruction forces a word boundary within the current section before the location counter is modified.
S–2314–50
7.38
LOCAL
S–2314–50
CAL Pseudo Instruction Descriptions [7]
The LOC pseudo instruction is restricted to sections that allow instructions or data, or both. If the LOC pseudo instruction is found within a definition, it is defined and is not recognized as a pseudo instruction. If the LOC pseudo instruction is found within a skipping sequence, it is skipped and is not recognized as a pseudo instruction.
The format of the LOC pseudo instruction is as follows: ignored LOC [expression]
The expression variable is optional and represents the new value of the location counter. If the expression does not exist, the counter is reset to the absolute value of 0. If the expression does exist, all symbols (if any) must be defined previously.
If the current base is mixed, octal is used as the base.
The expression operand cannot have an address attribute of byte, a relative attribute of external, or a negative value. A force word boundary occurs before the expression is evaluated. The expression operand must meet the requirements
for an expression as described in Section 6.9, page 94.
The following example illustrates the use of the LOC pseudo instruction:
LBL
ORG W.*+1000
LOC 200
A1
.
0
.
.
J LBL
Note: In the preceding example, the code is generated and loaded at location
W.*+1000 and the user must move it to absolute location 200 before execution.
The LOCAL pseudo instruction specifies unique character string replacements within a program segment that are defined only within the macro, opdef, dup, or echo definition. These character string replacements are known only in the macro, opdef, dup, or echo at expansion time. The most common usage of the
LOCAL pseudo instruction is for defining symbols, but it is not restricted to the definition of symbols.
The LOCAL pseudo instruction is described in detail in Section 9.11, page 260.
151
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual
7.39
MACRO
The MACRO pseudo instruction marks the beginning of a sequence of source program instructions saved by the assembler for inclusion in a program when called for by the macro name.
Macros are described in detail in Section 9.2, page 221.
7.40
MICRO
152
The MICRO pseudo instruction assigns a name to a character string. The assigned name can be redefined. You can reference and redefine a redefinable micro after its initial definition within a program segment. A micro defined with the MICRO pseudo instruction is discarded at the end of a module and cannot be referenced by any of the segments that follow.
You can specify the MICRO pseudo instruction anywhere within a program segment. If the MICRO pseudo instruction is found within a definition, it is defined and is not recognized as a pseudo instruction. If the MICRO pseudo instruction is found within a skipping sequence, it is skipped and is not recognized as a pseudo instruction.
The format of the MICRO pseudo instruction is as follows: name MICRO [string[,[exp][,[exp],[case]]]]
The name variable is required and is assigned to the character string in the operand field. It has redefinable attributes. If name was previously defined, the previous micro definition is lost. name must meet the requirements for identifiers
as described in Section 6.3, page 76.
The string variable represents an optional character string that can include previously defined micros. If string is not specified, an empty string is used. A character string can be delimited by any character other than a space. Two consecutive occurrences of the delimiting character indicate a single such character (for example, a micro consisting of the single character * can be specified as ‘*’ or ****).
The exp variable represents optional expressions. The first expression must be an absolute expression that indicates the number of characters in the micro character string. All symbols, if any, must be previously defined. If the current base is mixed, decimal is used for the expression. The expressions must meet the
requirements for expressions as described in Section 6.9, page 94.
S–2314–50
CAL Pseudo Instruction Descriptions [7]
The micro character string is terminated by the value of the first expression or the final apostrophe of the character string, whichever occurs first. If the first expression has a 0 or negative value, the string is considered empty. If the first expression is not specified, the full value of the character string is used. In this case, the string is terminated by the final apostrophe.
The second expression must be an absolute expression that indicates the micro string’s starting character. All symbols, if any, must be defined previously. If the current base is mixed, decimal is used for the expression.
The starting character of the micro string begins with the character that is equal to the value of the second expression, or with the first character in the character string if the second expression is null or has a value of 1 or less.
The optional case variable denotes the way uppercase and lowercase characters are interpreted when they are read from string. Character conversion is restricted to the letter characters (A - Z and a - z) specified in string. You can specify case in uppercase, lowercase, or mixed case, and it must be one of the following:
• MIXED or mixed
string is interpreted as entered and no case conversion occurs. This is the default.
• UPPER or upper
All lowercase alphabetic characters in string are converted to their uppercase equivalents.
• LOWER or lower
All uppercase alphabetic characters in string are converted to their lowercase equivalents.
The following example illustrates the use of the MICRO pseudo instruction:
MIC
MIC2
MICRO
MICRO
MIC2; CAL has edited these lines
MIC3 MICRO
’THIS IS A MICRO STRING’
’"MIC"’,1
MICRO ’THIS IS A MICRO STRING’,1
’"MIC2"’
MIC3; CAL has edited these lines
MIC4 MICRO ’"MIC"’,10
MICRO ’T’
; CALL TO MICRO MIC2.
’THIS IS A MICRO STRING’,10 MIC4; CAL has edited these lines MICRO
MIC5 MICRO
MICRO
’"MIC4"’ MIC5
’THIS IS A ’
MIC6 MICRO ’"MIC" ‘,5,11
MIC6; CAL has edited these lines MICRO ’THIS IS A MICRO STRING’,5,11
S–2314–50 153
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual
MIC7 MICRO ’"MIC6"’
MIC7; CAL has edited these lines MICRO
MIC8 MICRO ’"MIC"’,11,5
MIC8; CAL has edited these lines MICRO
MIC9 MICRO ’"MIC8"’
MIC9; CAL has edited these lines MICRO
7.41
MICSIZE
’MICRO’
’THIS IS A MICRO STRING’,11,5
’ IS A MICRO’
The MICSIZE pseudo instruction defines the symbol in the label field as a symbol with an address attribute of value, a relative attribute of absolute, and a value equal to the number of characters in the micro string whose name is in the operand field. Another SET or MICSIZE instruction with the same symbol redefines the symbol to a new value.
You can specify the MICSIZE pseudo instruction anywhere within a program segment. If the MICSIZE pseudo instruction is found within a definition, it is defined and is not recognized as a pseudo instruction. If the MICSIZE pseudo instruction is found within a skipping sequence, it is skipped and is not recognized as a pseudo instruction.
The format of the MICSIZE pseudo instruction is as follows:
[symbol] MICSIZE name
The symbol variable specifies an optional unqualified symbol. symbol is implicitly qualified by the current qualifier. The label field can be blank. symbol must meet
the requirement for a symbol as described in Section 6.2, page 70.
The name variable represents the name of a micro string that has been previously
defined. name must meet the requirements for identifiers as described in Section
7.42
MLEVEL
The MLEVEL pseudo instruction sets the message level for the output listing, the message listing, and the standard error file.
If the option accompanying the MLEVEL pseudo instruction is not valid, a diagnostic message is generated and MLEVEL is ignored.
154 S–2314–50
7.43
MSG
S–2314–50
CAL Pseudo Instruction Descriptions [7]
You can specify the MLEVEL pseudo instruction anywhere within a program segment. If the MLEVEL pseudo instruction is found within a definition, it is defined and is not recognized as a pseudo instruction. If the MLEVEL pseudo instruction is found within a skipping sequence, it is skipped and is not recognized as a pseudo instruction.
The format of the MLEVEL pseudo instruction is as follows: ignored MLEVEL [option]/*
The option variable specifies an optional message level. It can be entered in uppercase, lowercase, or mixed case, and it must be one of the following levels:
• ERROR (enables ERROR level messages only)
• WARNING (enables WARNING and ERROR level messages)
• CAUTION (enables CAUTION, WARNING, and ERROR level messages)
• NOTE (enables NOTE, CAUTION, WARNING, and ERROR level messages)
• COMMENT (enables COMMENT, NOTE, CAUTION, WARNING, and ERROR level messages)
• The asterisk (*) reactivates the message level in effect before the current message level was specified within the current program segment. Each occurrence of an MLEVEL pseudo instruction other than MLEVEL * initiates a new message level. Each MLEVEL * releases the current message level and reactivates the message level that preceded the current message level.
If all specified message levels have been released when an MLEVEL * is encountered, the assembler issues a CAUTION message to alert you to the situation and then reverts to the default level, WARNING.
• If you do not specify a level, the level is reset to the default, which is
WARNING .
The MSG pseudo instruction enables specified assembler messages to be printed.
It essentially increases the severity of the specified messages so that when encountered, the message will be printed. The format is as follows:
MSG msgnum[:msgnum ... ]
155
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual msgnum Number of the assembler message to print.
Multiple message numbers are separated by colons.
If the value of the -M message option on the as statement differs from the option on the MSG pseudo instruction, the as statement value is used.
If the option accompanying the MSG pseudo instruction is not valid, a diagnostic message is generated, and MSG is ignored.
If MSG is placed in global scope (outside of any IDENT/END pair), it applies to the file. If MSG is placed in module scope (inside an IDENT/END pair), it applies to
just that module. If there are multiple MSG and NOMSG (see Section 7.45, page 156)
pseudo instructions for the same message number, the last applies throughout.
The two pseudo instructions cannot be toggled.
7.44
NEXTDUP
The NEXTDUP pseudo instruction stops the current iteration of a duplication sequence indicated by a DUP or an ECHO pseudo instruction. Assembly of the current repetition of the dup sequence is terminated immediately and the next repetition, if any, is begun.
The NEXTDUP pseudo instruction is described in detail in Section 9.9, page 256.
7.45
NOMSG
156
The NOMSG pseudo instruction disables specified assembler messages to prevent them from being printed. It essentially decreases the severity of the messages so that when encountered, the message will not be printed. The format is as follows:
NOMSG msgnum[:msgnum ... ] msgnum Number of the assembler message to disable.
Multiple message numbers are separated by colons.
If the value of the -M message option on the as statement differs from the option on the NOMSG pseudo instruction, the as statement value is used.
If the option accompanying the NOMSG pseudo instruction is not valid, a diagnostic message is generated, and NOMSG is ignored.
S–2314–50
7.46
OCTMIC
S–2314–50
CAL Pseudo Instruction Descriptions [7]
If NOMSG is placed in global scope (outside of any IDENT/END pair), it applies to the file. If NOMSG is placed in module scope (inside an IDENT/END pair), it
applies to just that module. If there are multiple NOMSG and MSG (see Section
7.43, page 155) pseudo instructions for the same message number, the last applies
throughout. The two pseudo instructions cannot be toggled.
The OCTMIC pseudo instruction converts the value of an expression to a character string that is assigned a redefinable micro name. The character string that the pseudo instruction generates is represented as an octal number. The final length of the micro string is inserted into the code field of the listing.
You can specify OCTMIC with zero, one, or two expressions. The value of the first expression is converted to a micro string with a character length equal to the second expression. If the second expression is not specified, the minimum number of characters needed to represent the octal value of the first expression is used.
If the second expression is specified, the string is equal to the length specified by the second expression. If the number of characters in the micro string is less than the value of the second expression, the character value is right justified with the specified fill characters (zeros or blanks) preceding the value. If the number of characters in the string is greater than the value of the second expression, the beginning characters of the string are truncated and a warning message is issued.
You can specify the OCTMIC pseudo instruction anywhere within a program segment. If the OCTMIC pseudo instruction is found within a definition, it is defined and is not recognized as a pseudo instruction. If the OCTMIC pseudo instruction is found within a skipping sequence, it is skipped and is not recognized as a pseudo instruction.
The format of the OCTMIC pseudo instruction is as follows: name OCTMIC [expression
1
] [","[expression
2
[","[option]]]]
The name variable is required and specifies the name of the micro. name must
meet the requirements for identifiers as described in Section 6.3, page 76.
The expression
1 variable is an optional expression and is equal to name. If specified, expression
1 must have an address attribute of value and a relative attribute of absolute. All symbols used must be previously defined. If the current
157
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual base is mixed, a default of octal is used. If the first expression is not specified, the absolute value of 0 is used in creating the micro string. The expression
1 operand
must meet the requirements for expressions as described in Section 6.9, page 94.
expression
2 provides a positive character count less than or equal to decimal 22.
If this parameter is present, leading zeros or blanks (depending on option) are supplied, if necessary, to provide the requested number of characters. If specified, expression
2 must have an address attribute of value and a relative attribute of absolute with all symbols, if any, previously defined. If the current base is mixed, a default of decimal is used. If expression
2 is not specified, the micro string is represented in the minimum number of characters needed to represent the octal value of the first expression. The expression
2 operand must meet the requirements
for expressions as described in Section 6.9, page 94.
option represents the type of fill characters (ZERO for zeros or BLANK for spaces) that will be used if the second expression is present and fill is needed. The default is ZERO. You can enter option in mixed case.
The following example illustrates the use of the OCTMIC pseudo instruction:
IDENT EXOCT
BASE 0
ONE OCTMIC 1,2
_* "ONE"
* 01
TWO OCTMIC 5*7+60+700,3
*
*
_* "TWO"
*
023
; The base is octal.
; Returns 1 in 2 digits.
; Returns 1 in 2 digits.
; Returns 1023 in 3 digits.
; Returns 1023 in 3 digits.
*
*
THREE OCTMIC 256000,10,ZERO
_* "THREE"
* 00256000
FOUR OCTMIC 256000,l0,BLANK
; Zero fill on the left.
; Zero fill on the left.
_*
*
" FOUR" ; Blank fill (^) on the left.
^^256000 ; Blank fill (^) on the left.
END
7.47
HEXMIC
158
The HEXMIC pseudo instruction converts the value of an expression to a character string that is assigned a redefinable micro name. The character string that the
S–2314–50
S–2314–50
CAL Pseudo Instruction Descriptions [7] pseudo instruction generates is represented as an hexadecimal number. The final length of the micro string is inserted into the code field of the listing.
You can specify HEXMIC with zero, one, or two expressions. The value of the first expression is converted to a micro string with a character length equal to the second expression. If the second expression is not specified, the minimum number of characters needed to represent the hexadecimal value of the first expression is used.
If the second expression is specified, the string is equal to the length specified by the second expression. If the number of characters in the micro string is less than the value of the second expression, the character value is right justified with the specified fill characters (zeros or blanks) preceding the value. If the number of characters in the string is greater than the value of the second expression, the beginning characters of the string are truncated and a warning message is issued.
You can specify the HEXMIC pseudo instruction anywhere within a program segment. If the HEXMIC pseudo instruction is found within a definition, it is defined and is not recognized as a pseudo instruction. If the HEXMIC pseudo instruction is found within a skipping sequence, it is skipped and is not recognized as a pseudo instruction.
The format of the HEXMIC pseudo instruction is as follows: name HEXMIC [expression
1
] [","[expression
2
[","[option]]]]
The name variable is required and specifies the name of the micro. name must
meet the requirements for identifiers as described in Section 6.3, page 76.
The expression
1 variable is an optional expression and is equal to name. If specified, expression
1 must have an address attribute of value and a relative attribute of absolute. All symbols used must be previously defined. If the current base is mixed, a default of hexadecimal is used. If the first expression is not specified, the absolute value of 0 is used in creating the micro string. The expression
1 operand must meet the requirements for expressions as described in
expression
2 provides a positive character count less than or equal to decimal 22.
If this parameter is present, leading zeros or blanks (depending on option) are supplied, if necessary, to provide the requested number of characters. If specified, expression
2 must have an address attribute of value and a relative attribute of absolute with all symbols, if any, previously defined. If the current base is mixed, a default of decimal is used. If expression
2 is not specified, the micro string is represented in the minimum number of characters needed to represent the
159
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual hexadecimal value of the first expression. The expression
2 operand must meet the
requirements for expressions as described in Section 6.9, page 94.
option represents the type of fill characters (ZERO for zeros or BLANK for spaces) that will be used if the second expression is present and fill is needed. The default is ZERO. You can enter option in mixed case.
The following example illustrates the use of the HEXMIC pseudo instruction:
IDENT EXOCT
BASE 0
ONE HEXMIC 1,2
_*
*
"ONE"
01
TWO HEXMIC 5*7+60+700,3
_* "TWO"
; The base is hexidecimal.
; Returns 1 in 2 digits.
; Returns 1 in 2 digits.
; Returns 1023 in 3 digits.
*
* 023
*
THREE HEXMIC 256000,10,ZERO
; Returns 1023 in 3 digits.
*
*
*
_* "THREE"
* 00256000
; Zero fill on the left.
; Zero fill on the left.
FOUR HEXMIC 256000,l0,BLANK
_* " FOUR" ; Blank fill (^) on the left.
^^256000 ; Blank fill (^) on the left.
END
7.48
OPDEF
The OPDEF pseudo instruction marks the beginning of an operation definition
(opdef). The opdef identifies a sequence of statements to be called later in the source program by an opdef call. Each time the opdef call occurs, the definition sequence is placed into the source program.
The OPDEF pseudo instruction is described in detail in Section 9.3, page 237.
7.49
OPSYN
160
The OPSYN pseudo instruction defines an operation that is synonymous with another macro or pseudo instruction operation.
S–2314–50
7.50
ORG
S–2314–50
CAL Pseudo Instruction Descriptions [7]
The OPSYN pseudo instruction is described in detail in Section 9.12, page 262.
The ORG pseudo instruction resets the location and origin counters to the byte address specified. ORG resets the location and origin counters to the same address relative to the same section.
The ORG pseudo instruction forces a 64-bit longword boundary within the current section and also within the new section specified by the expression. These forced longword boundaries occur before the counter is reset. ORG can change the current working section without modifying the section stack.
The ORG pseudo instruction is restricted to sections that allow instructions or data, or instructions and data. If the ORG pseudo instruction is found within a definition, it is defined and not recognized as a pseudo instruction. If the ORG pseudo instruction is found within a skipping sequence, it is skipped and not recognized as a pseudo instruction.
The format of the ORG pseudo instruction is as follows: ignored ORG [expression]
The expression variable is an optional immobile or relocatable expression with positive relocation within the section currently in use. If the expression is blank, the longword address of the next available longword in the section is used. A force 64-bit longword boundary occurs before the expression is evaluated.
The expression must have a value or word-address attribute. If the expression has a value attribute, it is assumed to be a longword address. If the expression exists, all symbols (if any) must be defined previously. If the current base is mixed, hexadecimal is used as the base.
The expression cannot have a relative attribute of absolute or external, or a negative value. The expression operand must meet the requirements for an
expression as described in Section 6.9, page 94.
The following example illustrates the use of the ORG pseudo instruction:
ORG W.*+0’200
161
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual
7.51
OSINFO
162
The OSINFO pseudo instruction allows the assembler to pass specific information to the linker. The format is as follows:
OSINFO keyword=value
The keywords and the valid values for each are as follows:
Keyword
TEXTPAGESIZE
DATAPAGESIZ
MINMSPWIDTH
MAXMSPWIDTH
MINMAXVL
MAXMAXVL
VADDRWIDTH
NUMBARRIERS
SVXMODEL
PROGMODEL
HANDLECPF
HANDLEUVL
Value
0
1
2
0
System default
System minimum
System maximum
System default
1
2
System minimum
System maximum
Minimum acceptable MSP width. The valid values are 1 through 4.
Maximum acceptable MSP width. The valid values are 1 through 4.
Minimum acceptable max-VL. Valid values are between 1 and 64.
Maximum acceptable max-VL. Valid values are between 1 and 64.
Virtual address width in bits. The value must be 48.
Number of barriers required. The value must be 0.
The value must be 2.
The value must be 0.
0 Does not install a handler for continuous page fault signals.
Anything else
0
Does install a handler for continuous page fault signals.
Does not install a handler for unaligned vector load signals.
S–2314–50
7.52
QUAL
CAL Pseudo Instruction Descriptions [7]
HANDLEMTO
Anything else
0
Anything else
Does install a handler for unaligned vector load signals.
Does not install a handler for MSYNC timeout signals.
Does install a handler for MSYNC timeout signals.
The value is a constant or a symbol that resolves to a constant.
A QUAL pseudo instruction begins or ends a code sequence in which all symbols defined are qualified by a qualifier specified by the QUAL pseudo instruction or are unqualified. Until the first use of a QUAL pseudo instruction, symbols are defined as unqualified for each program segment. Global symbols cannot be qualified. The QUAL pseudo instruction must not occur before an IDENT pseudo instruction.
A qualifier applies only to symbols. Names used for sections, conditional sequences, duplicated sequences, macros, micros, externals, formal parameters, and so on, are not affected.
You must specify the QUAL pseudo instruction from within a program module. If the QUAL pseudo instruction is found within a definition or skipping sequence, it is defined and is not recognized as a pseudo instruction.
At the end of each program segment, all qualified symbols are discarded.
The format of the QUAL pseudo instruction is as follows: ignored QUAL */ [name]
The name variable is optional and indicates whether symbols will be qualified or unqualified and, if qualified, indicates the qualifier to be used. The name operand
must meet the requirements for names as described Section 6.3, page 76.
S–2314–50 163
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual
164
The name operand causes all symbols defined until the next QUAL pseudo instruction to be qualified. A qualified symbol can be referenced with or without the qualifier that is currently active. If the symbol is referenced while some other qualifier is active, the reference must be in the following form:
/ qualifier/symbol
When a symbol is referenced without a qualifier, the assembler tries to find it in the currently active qualifier. If the qualified symbol is not defined within the current qualifier, the assembler tries to find it in the list of unqualified symbols. If both of these searches fail, the symbol is undefined.
An unqualified symbol can be referenced explicitly using the following form:
// symbol
If the operand field of the QUAL is empty, symbols are unqualified until the next occurrence of a QUAL pseudo instruction. An unqualified symbol can be referenced without qualification from any place in the program module, or in the case of global symbols, from any program segment assembled after the symbol definition.
An asterisk (*) resumes use of the qualifier in effect before the most recent qualification within the current program segment. Each occurrence of a QUAL other than a QUAL * causes the initiation of a new qualifier. Each QUAL * removes the current qualifier and activates the most recent prior qualification. If the QUAL * statement is encountered and all specified qualifiers are released, a caution-level message is issued and succeeding symbols are defined as being unqualified.
The following example illustrates the use of the QUAL pseudo instruction:
* Assembler default for symbols is unqualified.
ABC = 1
; ABC is unqualified.
QUAL QNAME1
; Symbol qualifier QNAME1
ABC = 2
; ABC is qualified by QNAME1.
J XYZ
XYZ S1 A2
; XYZ is qualified by QNAME1.
.
.
.
QUAL QNAME2
S–2314–50
7.53
SECTION
S–2314–50
CAL Pseudo Instruction Descriptions [7]
; Symbol qualifier QNAME2.
ABC = 3
J /QNAME1/XYZ
.
.
.
QUAL *
; Resume the use of symbols qualified with
; qualifier QNAME1.
.
.
.
QUAL *
; Resume the use of unqualified symbols
.
.
A
.
IFA DEF,ABC
; Test whether ABC is defined.
B IFA DEF,/QNAME1/ABC
; Test if ABC is defined within qualifier
; QNAME1
C IFA DEF,/QNAME2/ABC
; Test if /QNAME2/ABC is defined within
; qualifier QNAME2.
.
.
.
The SECTION pseudo instruction establishes or resumes a section of code. The section can be common or local, depending on the options found in the operand field. Each section has its own location, origin, and bit-position counters.
You must specify the SECTION pseudo instruction from within a program module. If the SECTION pseudo instruction is found within a definition, it is defined and is not recognized as a pseudo instruction. If the SECTION
165
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual
166 pseudo instruction is found within a skipping sequence, it is skipped and is not recognized as a pseudo instruction.
The format of the SECTION pseudo instruction is as follows:
[lname]
[lname]
SECTION
SECTION
[type][","[location]][","[ENTRY]]
[location]","[type][","[ENTRY]]
[lname] SECTION [type][","[ENTRY]][","[location]]
[lname]
[lname]
[lname] ignored
SECTION
SECTION
SECTION
SECTION
[location]","[[ENTRY]][","[type]]
[ENTRY][","[location]][","[type]]
[ENTRY][","[type]][","[location]]
*
The variables associated with the SECTION pseudo instruction are described as follows:
• lname
The lname variable is optional and names the section. lname must meet the
requirements for tags as described in Section 6.3, page 76.
• type
The type variable specifies the type of section. It can be specified in uppercase, lowercase, or mixed case. type can be one of the following (for a
description of local sections, see Section 7.53.1, page 172):
– MIXED
Defines a section that permits both instructions and data. MIXED is the default type for the main section initiated by the IDENT pseudo instruction. If type is not specified, MIXED is the default. The loader treats a MIXED section as a local section.
S–2314–50
S–2314–50
CAL Pseudo Instruction Descriptions [7]
– CODE
Restricts a section to instructions only; data is not permitted. The loader treats a CODE section as a local section.
– DATA
Restricts a section to data only (CON, DATA, BSSZ, and so on); instructions are not permitted. The loader treats the DATA section as a local section.
– ZERODATA
Neither instructions nor data are allowed within this section. The loader treats a ZERODATA section as a local section. At load time, all space within a
ZERODATA section is set to 0.
– CONST
Restricts a section to constants only (CON, DATA, BSSZ, and so on); instructions are not permitted. The loader treats the CONST section as a local section.
– STACK
Sets up a stack frame (designated memory area). Neither data nor instructions are allowed. All symbols that are defined using the location or origin counter and are relative to a section that has a type of STACK are assigned a relative attribute of immobile.
These symbols may be used as offsets into the STACK section itself. These sections are treated like other section types except relocation does not occur after assembly. Because relocation does not occur, sections with a type of stack are not passed to the loader.
Sections with a type of STACK conveniently indicate that symbols are relative to an execution-time stack frame and that their values correspond to an absolute location within the stack frame relative to the base of the stack frame. Symbols with stack attributes are indicated as such in the debug tables that the assembler produces.
– COMMON
Defines a common section that can be referenced by another program module.
Instructions are not allowed.
Data cannot be defined in a COMMON section without a name (no name in label field); only storage reservation can be defined in an unnamed COMMON section. The label field that names a COMMON section cannot match the label
167
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual
168 field name of a previously defined section with a type of COMMON, DYNAMIC,
ZEROCOM , or TASKCOM. If duplicate label field names are specified, an error level message is issued.
For a description of unnamed (blank) COMMON, see Section 7.53.5, page 174.
– DYNAMIC
Allocates an expandable common section at load time. DYNAMIC is a common section. Neither instructions nor data are permitted within a DYNAMIC section; only storage reservation can be defined in an unnamed DYNAMIC section. The label field that names a DYNAMIC section cannot match the label field name of a previously defined section with a type of COMMON,
DYNAMIC , ZEROCOM, or TASKCOM. If duplicate label field names are specified, an error-level message is issued.
For a description of blank DYNAMIC, see Section 7.53.5, page 174.
– ZEROCOM
Defines a common section that can be referenced by another program module.
Neither instructions nor data are permitted within a ZEROCOM section; only storage reservation can be defined.
At load time, all uninitialized space within a ZEROCOM section is set to 0 . If a COMMON section with the same name contains the initialized text that was referenced by another module that will be loaded, portions of a ZEROCOM section can be explicitly initialized to values other than 0.
ZEROCOM must always be named. The label field that names a ZEROCOM section cannot match the label field name of a previously defined section with a type of COMMON, DYNAMIC, ZEROCOM, or TASKCOM. If duplicate label field names are specified, an error level message is issued.
– TASKCOM
Defines a task common section. Neither instructions nor data are allowed at assembly time. At execution time, TASKCOM is set up and can be referenced by all subroutines that are local to a task. Data also can be inserted at execution time into a TASKCOM section by any subroutine that is executed within a single task.
When a section is defined with a type of TASKCOM, the assembler creates a symbol that is assigned the name in the label field of the SECTION pseudo instruction that defines the section. This symbol is not redefinable, has a value of 0, an address attribute of word, and a relative attribute that is relocatable relative to the section. The loader relocates this symbol, and it
S–2314–50
S–2314–50
CAL Pseudo Instruction Descriptions [7] is used as an offset into an execution time task common table. The word at which it points within this table contains the address of the base of the task common section in memory.
All symbols defined using the location or origin counter within a task common section are assigned a relative attribute of immobile. These symbols are treated like other symbols, but relocation does not occur after assembly.
These symbols can be used as offsets into the task common section itself.
Sections with a type of TASKCOM indicate that their symbols are relative to an execution-time task common section, and their values correspond to an absolute location within the task common section relative to the beginning of the task common section. These values are indicated as such in the debug tables that the assembler produces. For a description of local sections, see
TASKCOM must always be named. The label field that names a TASKCOM section cannot match the label field name of a previously defined section with a type of COMMON, DYNAMIC, ZEROCOM, or TASKCOM . If duplicate label field names are specified, an error level message is issued.
Note: Accessing data from a task common section is not as straightforward as accessing data directly from memory. For more information about task common, see the Fortran Language Reference Manual, Volume 3.
• ENTRY
Sets a bit in the Program header to direct the loader to create an entry point at the same address as the first word of the section.
• *
The name, type, and location of the section in control reverts to the name, type, and location of the section in effect before the current section was specified within the current program module. Each occurrence of a SECTION pseudo instruction other than SECTION * causes a section with the name, type, and
location specified to be allocated. Each SECTION * releases the currently active section and reactivates the section that preceded the current section. If all specified sections were released when a SECTION * is encountered, the assembler issues a caution-level message and uses the main section.
When type is not specified, MIXED is used by default.
If type is not specified, the default is MIXED for type. Because a module within a program segment is initialized without a name, these defaults, when acting together, force this initial section entry to become the current working section.
169
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual
170
If the section name and attributes are previously defined, the SECTION pseudo instruction makes the previously defined section entry the current working section. If the section name and attributes are not defined, the SECTION pseudo instruction tries to create a new section with the name and attributes. The following restrictions apply when a new section is created:
• A section of the type TASKCOM, COMMON, ZEROCOM, and a section with a specified entry must always have a label field name.
• If a section with a type of COMMON, DYNAMIC, ZEROCOM, or TASKCOM is being created for the first time, it must never have a name that matches a section that was created previously with a type of COMMON, DYNAMIC, ZEROCOM, or TASKCOM.
The following example illustrates the use of the SECTION pseudo instruction: ident exsect
; The Main section has by default a type of mixed.
.
.
con 1
; Data and instructions are permitted in
S1 1
; the Main section.
.
.
dsect section data
; This section is defined with a name of
.
; dsect and a type of data.
.
con 3
; Data is permitted in dsect.
bszz 2
; Data is permitted in dsect.
.
.
.
S2 S3
; The assembler generates an error-level
.
; message because instructions are illegal in a
.
; section with a type of data.
S–2314–50
S–2314–50
CAL Pseudo Instruction Descriptions [7]
.
csect section common
; This section is defined with a name of
.
; csect and a type of common.
.
data ’12345678’
; Data is permitted in a named common section.
S2 A1
; The assembler generates an error-level
.
; message, because instructions are not permitted
.
; in a common section.
.
section
; This section is unnamed and is assigned
; by default a type of mixed. When a section is
; specified without a name and a type, the main
; section becomes the current section.
section *
; The current section reverts to the previous section
; in the stack buffer csect.
section *
; The current section reverts to the previous section in the
; stack buffer dsect.
con 2
; A memory location with a value of 2 is inserted into dsect.
.
.
section *
; The current section reverts to the main
.
; section.
.
.
dsect section code
; The assembler considers this section specification unique
; and different from the previously defined section named
; dsect. Sections with types of mixed, code, data, and stack
; are treated as local sections that are specified with the
171
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual
; same name therefore, are, considered unique if they
.
; are specified with different types.
.
s1 s2 ; Instructions are permitted in dsect.
csect section common ; The current section reverts to the
; section defined previously as csect.
; When a section is specified with the
.
.
.
section *
; name, type, and location of a previously
; defined section, the previously defined
; section becomes the current section.
.
.
.
con 2
; The current section reverts to the main
; section
.
.
.
section *
; The assembler generates an error-level
; message because data is not permitted in a
; section with a type of code.
.
.
; This current section reverts to the main
; section.
.
csect section dynamic ; The assembler generates an error-level
; message, because the loader does not treat
; sections with types of common, dynamic,
; and taskcom as local sections Specifying
; a section with a previously defined name
; is illegal when the accompanying type
; does not define a local section.
end
7.53.1 Local Sections
172
A local section is a block of code that is usable only by the program module in which it resides. CAL uses three types of local sections:
• Main section
• Literals section
• Sections defined by the SECTION pseudo instruction
S–2314–50
CAL Pseudo Instruction Descriptions [7]
When a SECTION pseudo instruction is used, every SECTION type except
COMMON , DYNAMIC, TASKCOM, and ZEROCOM is local. For more information about
SECTION
types, see the SECTION pseudo instruction in Section 7.53, page 165.
7.53.2 Main Sections
The main section is initiated by the IDENT pseudo instruction and is always the first section in a program module. This section is used for all local code other than that generated by the occurrence of a literal reference or code between two SECTION pseudo instructions.
Generally, sections may not have names but must be assigned types and locations. The default name of the main section is always empty. The defaults for type and location are MIXED and CM, respectively. For more information about the MIXED and CM section names, see the SECTION pseudo instruction in
7.53.3 Literals Section
The first use of a literal value in an expression causes the assembler to store the data item in a literals section. Data is generated in the literals section implicitly by the occurrence of a literal. Explicit data generation or memory reservation is not allowed in the literals section. The assembler supports the literals section as a
constant section. For more information about literals, see Section 6.6, page 85.
7.53.4 Sections Defined by the SECTION Pseudo Instruction
When a SECTION pseudo instruction is used, all code generated or memory reserved (other than literals) between occurrences of SECTION pseudo instructions is assigned to the designated section.
Until the first SECTION pseudo instruction is specified, the main section is used.
If you specify the ORG pseudo instruction, an exception to these conditions can occur. Specifying the ORG pseudo instruction may cause the placement of code or memory reservations to be different from the currently specified working section.
The SECTION pseudo instruction is recommended for use because it has all of the same capabilities as the BLOCK and COMMON pseudo instructions.
When a section is released, the type and location of the previous section is used. When the number of sections released is equal to or greater than the number specified, CAL uses the defaults of the main section for type (MIXED) and location (CM).
S–2314–50 173
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual
A section with the same name, type, and location used in different areas of a program is recognized as the same section. For more information, see the
SECTION
pseudo instruction in Section 7.53, page 165.
7.53.5 Common Sections
When a SECTION pseudo instruction is used with a type of COMMON, DYNAMIC,
ZEROCOM , or TASKCOM, all code generated (other than literals) or memory reserved between occurrences of SECTION pseudo instructions is assigned to the designated common, dynamic, zero common, or task common section.
At program end, each common section is identified to the loader by its SECTION name and is available for reference by another program module. If you specify the ORG pseudo instruction, an exception to these conditions can occur.
Specifying the ORG pseudo instruction may cause the placement of code or memory reservations to be different from the currently specified working section.
If a common section is specified, the identifier in the label field that names the section must be unique within the module in which it is defined. When a section is assigned a type (COMMON, DYNAMIC, ZEROCOM, or TASKCOM) that differs from the type of a previously defined section, it cannot be assigned the name of a previously defined section within the same module.
7.53.6 Section Stack Buffer
The assembler maintains a stack buffer that contains a list of the sections specified. Each time a SECTION pseudo instruction names a new section, the assembler adds the name of the section to the list and identifies the new section as the current section. You also can use the BLOCK and COMMON pseudo instructions to name sections.
The assembler remembers the order in which sections are specified. An entry is deleted from the list each time a SECTION pseudo instruction contains an asterisk
(*) . When an entry is deleted, the name, location, and type of the previously specified section is enabled.
The first section on the list is the last section that will be deleted from the list. If the program contains more SECTION * instructions than there are entries, the assembler uses the main section. (The BLOCK * and COMMON * instructions replace the current section with the most recent previous section that was specified by the BLOCK and COMMON pseudo instructions.)
174 S–2314–50
S–2314–50
CAL Pseudo Instruction Descriptions [7]
For each section used in a program, the assembler maintains an origin counter, a location counter, and a bit position counter. When a section is first established or its use is resumed, the assembler uses the counters for that section.
The following example illustrates section specification and deletion and indicates the current section. The example includes the QUAL pseudo instruction. For a
description of the QUAL pseudo instruction, see Section 7.52, page 163.
IDENT
SYM1
SYM2
=
STACK
QUAL
=
SNAME SECTION
MLEVEL
SYM3 =
1
2
QNAME1
*
MIXED
ERROR
; The IDENT statement puts the first entry
; on the list of qualifiers.
This entry
; starts the symbol table for unqualified
; symbols.
; SYM1 is relative to the main section.
; Second entry on the list of qualifiers.
; SYM2 is the first entry in the symbol
; table for QNAME1.
; SNAME is the second entry on the list of
; sections
; Reset message level to error eliminate
; warning level messages.
; SYM3 is the second entry in the symbol
SYM4
SYM7
SYM8
MLEVEL
SECTION *
=
=
=
6
7
*
4
; table for QNAME1 and is relative to the
; SNAME section.
; Reset message level to default in effect
; before the MLEVEL specification.
; SNAME is deleted from the list of
; sections.
; SYM4 is the third entry in the symbol
; table for QNAMEl and is relative to the
SYM5
QUAL
=
QNAME2
5
; main section.
; Third entry on the list of qualifiers.
; SYM5 is the first entry in the symbol
; table for QNAME2.
SYM6 = /QNAME1/SYM2 ; SYM6 gets SYM2 from the symbol table for
; QNAMEl even though QNAME1 is not the
QUAL *
; current qualifier in effect.
; QNAME2 is removed as the current
; qualifier name.
; SYM7 is the fourth entry in the symbol
; table for QNAME1.
; Second entry in the symbol table for
; unqualified symbols.
175
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual
7.53.7 Generated Code Position Counters
This subsection describes the generated code position counters.
7.53.7.1 Origin Counter
The origin counter controls the relative location of the next word that will be assembled or reserved in the section. You can reserve blank memory areas by using either the ORG or BSS pseudo instructions to advance the origin counter.
When the special element *O is used in an expression, the assembler replaces it with the current byte-address value of the origin counter for the section in use. To obtain the word-address value of the origin counter, use W.*O. (In this context, a word is a 64-bit long word.) For more information about the special elements and
W.
prefix, see Section 6.8, page 92.
7.53.7.2 Location Counter
Usually, the location counter is the same value as the origin counter and the assembler uses it to define symbolic addresses within a section. The counter is incremented when the origin counter is incremented. Use the LOC pseudo instruction to adjust the location counter so that it differs in value from the origin counter or so that it refers to the address relative to a section other than the one currently in use. When the special element * is used in an expression, the assembler replaces it with the current byte-address value of the location counter for the section in use. To obtain the 64-bit word-address value of the location
counter, Use W.* (see Section 6.8, page 92).
7.53.7.3 Word-bit-position Counter
As instructions and data are assembled and placed into a 64-bit long word, the assembler maintains a pointer that indicates the next available bit within the word currently being assembled. This pointer is known as the word-bit-position
counter. It is 0 at the beginning of a new word and is incremented by 1 for each completed bit in the word. Its maximum value is 63 for the rightmost bit in the word. When a word is completed, the origin and location counters are incremented by 1, and the word-bit-position counter is reset to 0 for the next word.
When the special element *W is used in an expression, the assembler replaces it with the current value of the word-bit-position counter. The normal advancement of the word-bit-position counter is in increments of 32 as instructions are generated. You can alter this normal advancement by using the BITW, BITP,
DATA , and VWD pseudo instructions.
176 S–2314–50
CAL Pseudo Instruction Descriptions [7]
7.53.7.4 Force Word Boundary
If either of the following conditions are true, the assembler completes a partial word and sets the word-bit-position counter to 0:
• The current instruction is an ALIGN, BSS, BSSZ, CON, LOC, or ORG pseudo instruction.
• The current instruction is a DATA or VWD pseudo instruction and the instruction has an entry in the label field.
7.54
SET
The SET pseudo instruction resembles the = pseudo instruction; however, a symbol defined by SET is redefinable.
You can specify the SET pseudo instruction anywhere within a program segment.
If the SET pseudo instruction is found within a definition, it is defined and is not recognized as a pseudo instruction. If the SET pseudo instruction is found within a skipping sequence, it is skipped and is not recognized as a pseudo instruction.
The format of the SET pseudo instruction is as follows:
[symbol] SET expression[,[attribute]]
The symbol variable specifies an optional unqualified symbol. The symbol is implicitly qualified by the current qualifier. A symbol defined with the SET pseudo instruction can be redefined with another SET pseudo instruction, but the symbol must not be defined prior to the first SET pseudo instruction. The label field can be blank. symbol must meet the requirements for symbols as
described in Section 6.2, page 70.
All symbols found within expression must have been previously defined. The
expression operand must meet the requirements for an expression as described in
The attribute variable specifies a byte (B), word (W), or value (V) attribute.
Attribute, if present, is used rather than the expression’s attribute. If a byte-address attribute is specified, an expression with word-address attribute is multiplied by four; if word-address attribute is specified, an expression with byte-address attribute is divided by four. An immobile or relocatable expression cannot be specified as having a value attribute.
S–2314–50 177
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual
The following example illustrates the use of the SET pseudo instruction:
SIZE = o’100
PARAM SET D’18
WORD
BYTE
SET
SET
*W
*B
SIZE = SIZE+1
PARAM SET PARAM+2
; Illegal
; Legal
7.55
SKIP
178
The SKIP pseudo instruction unconditionally skips subsequent statements.
If a label field name is present, skipping stops when an ENDIF or ELSE with the same name is encountered; otherwise, skipping stops when the statement count is exhausted.
You can specify the SKIP pseudo instruction anywhere within a program segment. If the SKIP pseudo instruction is found within a definition, it is defined and is not recognized as a pseudo instruction. If the SKIP pseudo instruction is found within a skipping sequence, it is skipped and is not recognized as a pseudo instruction.
The format of the SKIP pseudo instruction is as follows:
[name] SKIP [count]
The name variable specifies an optional name for a conditional sequence of code.
If both name and count are present, name takes precedence. name must meet the
requirements for identifiers as described in Section 6.3, page 76.
The count variable specifies a statement count. It must be an absolute expression with a positive value. All symbols in the expression, if any, must be previously defined. A missing or null count subfield gives a zero count. count is used only when the label field is not specified. If name is not present and count is present in the operand field, skipping stops when count is exhausted. If neither name nor
count is present, no skipping occurs.
The following example illustrates the use of the SKIP pseudo instruction:
SNAME1
.
.
SKIP
SKIP
; No skipping occurs.
; Statements are skipped if an ENDIF or
; ELSE with a matching label field
; label is found.
S–2314–50
7.56
SPACE
S–2314–50
CAL Pseudo Instruction Descriptions [7]
SNAME1
SNAME2
.
.
.
ENDIF
.
SKIP 10
SNAME2
.
.
ENDIF
.
.
.
SKIP 4
; Statements are skipped until an ENDIF
; or ELSE with a matching label field
; label is found.
; Four statements are skipped.
The SPACE pseudo instruction inserts the number of blank lines specified into the output listing. SPACE is a list control pseudo instruction and by default, is not listed. To include the SPACE pseudo instruction on the listing, specify the LIS option on the LIST pseudo instruction.
You can specify the SPACE pseudo instruction anywhere within a program segment. If the SPACE pseudo instruction is found within a definition, it is defined and is not recognized as a pseudo instruction. If the SPACE pseudo instruction is found within a skipping sequence, it is skipped and is not recognized as a pseudo instruction.
The format of the SPACE pseudo instruction is as follows: ignored SPACE [expression]
The expression variable specifies an optional absolute expression that specifies the number of blank lines to insert in the listing. expression must have an address attribute of value, a relative attribute of absolute, and a value of 0 or greater.
If expression is not specified, the absolute value of 1 is used and one blank line is inserted into the output listing. If the current base is mixed, a default of decimal is used for the expression.
The expression operand must meet the requirement for an expression as described
179
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual
7.57
STACK
The STACK pseudo instruction increases the size of the stack. Increments made by the STACK pseudo instruction are cumulative. Each time the STACK pseudo instruction is used within a module, the current stack size is incremented by the number of words specified by the expression in the operand field of the STACK pseudo instruction.
The STACK pseudo instruction is used in conjunction with sections that have a type of STACK. If either a STACK section or the STACK pseudo instruction is specified within a module, the loader tables that the assembler produces indicate that the module uses one or more stacks. The stack size indicated in the loader tables is the combined sizes of all STACK sections, if any, added to the total value of all STACK pseudo instructions, if any, specified within a module.
You must specify the STACK pseudo instruction from within a program module.
If the STACK pseudo instruction is found within a definition, it is defined and is not recognized as a pseudo instruction. If the STACK pseudo instruction is found within a skipping sequence, it is skipped and not recognized as a pseudo instruction.
The format of the STACK pseudo instruction is as follows: ignored STACK [expression]
The expression variable is optional. If specified, it must have an address attribute of word or value, a relative attribute of absolute, a positive value, and all symbols within it (if any) must be defined previously.
If STACK is specified without expression, the stack is not incremented. The
expression operand must meet the requirements for an expression as described in
7.58
START
180
The START pseudo instruction specifies the main program entry. The program uses the START pseudo instruction to specify the symbolic address at which execution begins following the loading of the program. The named symbol can optionally be an entry symbol specified in an ENTRY pseudo instruction.
You must specify the START pseudo instruction from within a program module.
If the START pseudo instruction is found within a definition or skipping sequence, it is defined and is not recognized as a pseudo instruction.
S–2314–50
CAL Pseudo Instruction Descriptions [7]
The format of the START pseudo instruction is as follows: ignored START symbol
The symbol variable must be the name of a symbol that is defined as an unqualified symbol within the same program module. symbol must not be redefinable, must have a relative attribute of relocatable, and cannot be relocatable relative to any section other than a section that allows instructions or a section that allows instructions and data. The START pseudo instruction cannot be specified in a section with a type of data only.
The length of the symbol is restricted depending on the type of loader table that the assembler is currently generating. If the symbol is too long, an error message results.
The symbol operand must meet the requirements for symbols as described
The following example illustrates the use of the START pseudo instruction:
HERE =
.
IDENT EXAMPLE
START HERE
*
.
.
END
7.59
STOPDUP
7.60
SUBTITLE
S–2314–50
The STOPDUP pseudo instruction stops duplication of a code sequence indicated by a DUP or ECHO pseudo instruction.
The STOPDUP pseudo instruction is described in detail in Section 9.10, page 257.
The SUBTITLE pseudo instruction specifies the subtitle that will be printed on the listing. The instruction also causes a page eject. SUBTITLE is a list control pseudo instruction and is, by default, not listed. To include the SUBTITLE pseudo instruction on the listing, specify the LIS option on the LIST pseudo instruction.
181
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual
You can specify the SUBTITLE pseudo instruction anywhere within a program segment. If the SUBTITLE pseudo instruction is found within a definition, it is defined and is not recognized as a pseudo instruction. If the SUBTITLE pseudo instruction is found within a skipping sequence, it is skipped and is not recognized as a pseudo instruction.
The format of the SUBTITLE pseudo instruction is as follows: ignored SUBTITLE [del-char[string-of-ASCII]del-char]
The del-char variable is the delimiting character. It must be a single matching character on both ends of the ASCII character string. Apostrophes and spaces are not legal delimiters; all other ASCII characters are allowed. Two consecutive occurrences of the delimiting character indicate a single such character will be included in the character string.
The string-of-ASCII variable is an ASCII character string that will be printed as the subtitle on subsequent pages of the listing. This string replaces any previous string found within the subtitle field.
7.61
TEXT
182
Source lines that follow the TEXT pseudo instruction through the next ENDTEXT pseudo instruction are treated as text source statements. These statements are listed only when the TXT listing option is enabled. A symbol defined in text source is treated as a text symbol for cross-reference purposes; that is, such a symbol is not listed in the cross-reference unless a reference to the symbol from a listed statement exists. The text name part of the cross-reference listing contains the text name.
If the text appears in the global part of a program segment, Symbols defined in
text source are global. If the text appears within a program module, symbols in
text source are local.
TEXT is a list control pseudo instruction and is, by default, not listed. The TEXT pseudo instruction is listed if the listing is on or if the LIS listing option is enabled regardless of other listing options.
The TEXT and ENDTEXT pseudo instructions have no effect on a binary definition file.
You can specify the TEXT pseudo instruction anywhere within a program segment. If the TEXT pseudo instruction is found within a definition, it is defined
S–2314–50
7.62
TITLE
S–2314–50
CAL Pseudo Instruction Descriptions [7] and is not recognized as a pseudo instruction. If the TEXT pseudo instruction is found within a skipping sequence, it is skipped and is not recognized as a pseudo instruction.
The format of the TEXT pseudo instruction is as follows:
[name] TEXT [del-char[string-of-ASCII]del-char]
The name variable is optional. It is used as the name of the following source until the next ENDTEXT pseudo instruction. The name found in the label field is the text name for all defined symbols in the section, and it is listed in the text name part of the cross-reference listing.
The name location must meet the requirements for names as described in Section
The del-char variable is the delimiting character. It must be a single matching character on both ends of the ASCII character string. Apostrophes and spaces are not legal delimiters; all other ASCII characters are allowed. Two consecutive occurrences of the delimiting character indicate a single such character will be included in the character string.
The string-of-ASCII variable is an ASCII character string that will be printed as the subtitle on subsequent pages of the listing. A maximum of 72 characters is allowed. This string replaces any previous string found within the subtitle field.
The TITLE pseudo instruction specifies the main title that will be printed on the listing. TITLE is a list control pseudo instruction and is, by default, not listed. To include the TITLE pseudo instruction on the listing, specify the LIS option on the LIST pseudo instruction.
You can specify the TITLE pseudo instruction anywhere within a program segment. If the TITLE pseudo instruction is found within a definition, it is defined and is not recognized as a pseudo instruction. If the TITLE pseudo instruction is found within a skipping sequence, it is skipped and is not recognized as a pseudo instruction.
The format of the TITLE pseudo instruction is as follows: ignored TITLE [del-char[string-of-ASCII]del-char]
183
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual
The del-char variable is the delimiting character. It must be a single matching character on both ends of the ASCII character string. Apostrophes and spaces are not legal delimiters; all other ASCII characters are allowed. Two consecutive occurrences of the delimiting character indicate a single such character will be included in the character string.
The string-of -ASCII variable is an ASCII character string that will be printed to the diagnostic file. A maximum of 72 characters is allowed.
7.63
VWD
184
The VWD pseudo instruction allows data to be generated in fields that are from
0 to 64 bits wide. Fields can cross Longword boundaries. Data begins at the current bit position unless a symbol is used in the label field. If a symbol is present within the label field, a forced Longword boundary occurs, and the data begins at the new current bit position.
Code for each subfield is packed tightly with no unused bits inserted.
The VWD pseudo instruction is restricted to sections that have a type of instructions, data, or both. If the VWD pseudo instruction is found within a definition, it is defined and is not recognized as a pseudo instruction. If the VWD pseudo instruction is found within a skipping sequence, it is skipped and is not recognized as a pseudo instruction.
The format of the VWD pseudo instruction is as follows:
[symbol] VWD [count/[expression]][,[count/[expression]]]
The symbol variable represents an optional symbol. If symbol is present, a force word boundary occurs. The symbol is defined with the value of the location counter after the force word boundary and has an address attribute of word.
symbol must meet the requirements for symbols as described in Section 6.2, page 70.
The count variable specifies the number of bits in the field. It can be a numeric constant or symbol with absolute and value attributes. count must be positive and less than or equal to 64. If a symbol is specified for count, it must have been previously defined. If one or more count entries are not valid, no code is generated for the entire set of subfields in the operand field; however, each subfield is still evaluated.
S–2314–50
S–2314–50
CAL Pseudo Instruction Descriptions [7]
The expression variable represents the expression whose value will be inserted in the field. If expression is missing, the absolute value of 0 is used. If count is not equal to 0, the count is the number of bits reserved to store the following expression, if any. expression must meet the requirement for expressions as
described in Section 6.9, page 94.
The following example illustrates the use of the VWD pseudo instruction:
BASE M
PDT BSS 0
VWD 1/SIGN,3/0,60A’"NAM"’R
; 1000000000000023440515
; 10000000653
REMDR = 64-*W
VWD REMDR/DSN
; 41
; 00011044516
In the preceding example, the value of SIGN is 1, the value of FC is 0, the value of
ADD is 653 (octal), and the value of DSN is $IN in ASCII code.
185
S–2314–50
Cray X1 Instruction Set and Encoding [8]
This chapter lists the symbolic machine instructions for Cray X1 systems. Some
of the values in these tables are in binary; for instance, the j column in Section
The notations used in the instructions have the following meanings: ai
{ } ck
{d ... } hnt hw
Location i in the A registers.
Encloses optional elements in the syntax descriptions.
Location k in the control registers.
Double precision floating point value, 64 bits.
A hint in regards to making the best use of cache.
For information on the meaning of the values when using atomic memory instructions, see
Section 8.3.9.3, page 196. For the meanings of the
values with vector memory instructions, see
Halfword value, 16 bits. The hw values for
{ab:abc:abcd} are:
{ab:abc:abcd} hw imm6
{ ... l} mm
PC
{s ... } ab abc abcd
01
10
11
Immediate; how wide (in bits) the field containing a constant value is. In the case of imm6, the field is 6 bits wide.
Long value, 64 bits.
Register number m in the vector mask.
Program counter. The program counter holds the address of the instruction currently executing.
Single precision floating point value, 32 bits.
187
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual sd sj vi
{ ... w} wl
Single precision (32 bits) or double precision
(64 bits) floating point value. The sd values for
{s:d} are:
{s:d} sd s d
00
01
Location j in the S registers.
Location i in the vector registers.
Word value, 32 bits.
Word (32 bits) or longword (64 bits). The wl values for {,w:,l:} are:
{,w:,l:} wl w l
10
11
This appendix is organized as follows:
• System and memory ordering instructions, see Section 8.1, page 189.
• Register moves and jump instructions, see Section 8.2, page 190.
• A register instructions, see Section 8.3, page 192.
• S register instructions, see Section 8.4, page 197.
• Vector register instructions, see Section 8.5, page 203.
188 S–2314–50
Cray X1 Instruction Set and Encoding [8]
8.1 System and Memory Ordering Instructions
8.1.1 System Instructions g
000000 i xxxxxx j xxxxxx k xxxxxx t xx f
000000
Syntax break {n}
000001 000000 000000 000000 00 000000 syscall
Description
Break. n is an integer of up to 20 bit that is placed in bits 25..6 of the instruction.
System call
8.1.2 Memory Ordering Instructions g
000010
000010
000010
000010
000010
000010
000010 i
000000
000000
000000
000000
000000
000000
000000 aj aj j aj aj
000000
000000
000000 k
000000
000000
000000
000000
000000
000000
000000
00
00
00
00
00 t
00
00
000010
000010
000000
000000
000000 vj
000000
000000
00
00 f
000000
000001
000010
000011
001000
001001
001010
001011
001100
Syntax gsync aj gsync aj,cpu Global ordering and processor quiet gsync aj,a Global ordering with respect to scalar loads gsync aj,r
Description
Global ordering lsync s,v
Global ordering with respect to scalar stores
Local ordering of prior scalar with later vector lsync v,s Local ordering of prior vector with later scalar lsync vr,s lsync v,v lsync vj,v
Local ordering of prior vector reads with later scalar
Local ordering of vector references
Local ordering of latest vj references with later vector
S–2314–50 189
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual g
000010 i
000000
000010
000010
000010
000010
000010
000010
000000
000000
000000
000000
000000
000000 j vj k
000000 t
00
000000
000000 aj aj aj
000000
000000
000000
000000
000000
000000
000000
00
00
00
00
00
00
001110
001111
010000
010001
010010
010100 f
001101
Syntax Description lsync vj,v,el Local elem ordering of latest vj references with later vector lsync i lsync fp msync aj msync aj,p msync aj,v
Local ordering of stores and instruction fetches
Local ordering of floating-point operations
Meta CPU synch
Meta CPU producer synch
Meta CPU vector synch mint Meta CPU interrupt
8.2 Register Move and Jump Instructions
The following instructions are described in this section:
• Register Move Instructions (see Section 8.2.1, page 190).
• Jump Instructions (see Section 8.2.2, page 191).
8.2.1 Register Move Instructions g
000011
000011
000011
000011
000011
000011
000011
000011 i ai ai j k
000000 sk
000000 sk
000000 ak si si ai ai
000000 ak
000000 ck ci 000000 ak
000000 000000 ak
01
01
000000 000000 00
01
00
01 t
01
00 f Syntax
000000 ai{,d:,l:} sk
000000 ai,{s:w} sk
000001 si{,d:,l:} ak
000001 si,{s:w} ak
111101 ai ck
000011 ci ak
000000 vl ak
000001 ai vl
Description
S to A move 64-bit
S to A move 32-bit
A to S move 64-bit
A to S move 32-bit
C to A move 64-bit
A to C move 64-bit
A to VL move
VL to A move
190 S–2314–50
Cray X1 Instruction Set and Encoding [8] vi vi vi ai si si vi vi vi vi vi si si ai i ai ai g
000011
000011
000011
000011
000011
000011
000011
000011
000011
000011
000011
000011
000011
000011
000011
000011 j aj k vk aj aj aj vk vk vk imm6 vk imm6 vk imm6 vk imm6 vk aj ak aj aj ak sk aj sk imm6 ak imm6 ak imm6 sk imm6 sk
00
01
00
00
01
00
01
01
00
01
00
01
00
01 t
01
00 f Syntax
011100 ai{,d:,l:} vk,aj
011100 ai,{s:w} vk,aj
011000 si{,d:,l:} vk,aj
011000 si,{s:w} vk,aj
011110 ai{,d:,l:} vk,imm6
011110 ai,{s:w} vk,imm6
011010 si{,d:,l:} vk,imm6
011010 si,{s:w} vk,imm6
011101 vi,aj ak{,d:,l:}
011101 vi,aj sk,{s:w}
011001 vi,aj sk{,d:,l:}
011001 vi,aj sk,{s:w}
011111 vi,imm6 ak{,d:,l:}
011111 vi,imm6 ak,{s:w}
011011 vi,imm6 sk{,d:,l:}
011011 vi,imm6 sk,{s:w}
8.2.2 Jump Instructions g
101111
101111
101111 i ai ai ai aj aj j aj k
000000
000000
000000 t
00
00
00 f
000000
000000
000000
Syntax j, ai aj j, ai aj,sr j, ai aj,rt
Description
V to A move 64-bit
V to A move 32-bit
V to S move 64-bit
V to S move 32-bit
V to A move 64-bit imm6
V to A move 32-bit imm6
V to S move 64-bit imm6
V to S move 32-bit imm6
A to V move 64-bit
A to V move 32-bit
S to V move 64-bit
S to V move 32-bit
A to V move 64-bit imm6
A to V move 32-bit imm6
S to V move 64-bit imm6
S to V move 32-bit imm6
Description
Jump ai<-PC+4;PC<-aj
Jump subroutine ai<-PC+4;PC<-aj
Jump return ai<-PC+4;PC<-aj
S–2314–50 191
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual g
100010
100011
100011
100010
100011
100011
100010
100010
100010
100010
100010
8.3 A Register Instructions
The following A register instructions are described in this section:
• A Register Integer Instructions (see Section 8.3.1, page 192).
• A Register Logical Instructions (see Section 8.3.2, page 193).
• A Register Shift Instructions (see Section 8.3.3, page 193).
• A Register Immediate Instructions (see Section 8.3.4, page 194).
• A Register Integer Compare Instructions (see Section 8.3.5, page 194).
• A Register Byte and Halfword Instructions (see Section 8.3.6, page 194).
• Other A Register Instructions (see Section 8.3.7, page 195).
• A Register Branch Instructions (see Section 8.3.8, page 195).
• A Register Memory Access Instructions (see Section 8.3.9, page 195).
8.3.1 A Register Integer Instructions ai ai ai ai i ai ai ai ai ai ai ai aj aj aj aj j aj aj aj aj aj aj aj k ak imm8 imm8 ak imm8 imm8 ak ak ak
000000
000000 t wl wl wl
11 wl wl wl f
000000
000000
101000
000101
000101
101101
011000
011001
001000
001010
001011
Syntax ai{,w:,l:} aj+ak ai{,l:} aj+imm8 ai,w aj+imm8 ai{w:,l:} aj-sk ai{,l:} aj-imm8 ai,w aj-imm8 ai{,w:,l:} aj*ak ai{,l:} hi(aj*ak) ai{,w:,l:} ak/aj ai{,w:,l:} lz(aj) ai{,w:,l:} pop(aj)
Description
A int add
A int add long imm8
A int add word imm8
A int subtract
A int subtract long imm8
A int subtract word imm8
A int multiply
A int multiply high
A int divide
A leading zero count
A population count
192 S–2314–50
Cray X1 Instruction Set and Encoding [8]
8.3.2 A Register Logical Instructions g
100010
100011
100010
100011
100010
100010
100011
100010 ai ai ai ai ai ai i ai ai aj aj aj aj aj aj j aj aj k ak imm8 ak imm8 ak ak imm8 ak t
11
11
11
11
11 f
010000
010000
010001
010001
010010
010011
010011
010100
Syntax ai{,l:} aj&ak
Description
A logical AND ai{,l:} aj&imm8 A logical AND imm8 ai{,l:} aj|ak A logical OR ai{,l:} aj|imm8 A logical OR imm8 ai{,l:} ~aj^ak A logical equivalence ai{,l:} aj^ak A logical exclusive OR ai{,l:} aj^imm8 A logical exclusive OR imm8 ai{,l:} ~aj&ak A logical ANDNOT
8.3.3 A Register Shift Instructions g
100010
100011
100011
100010
100011
100011
100010
100011 i ai ai ai ai ai ai ai ai k ak imm8 imm8 ak imm8 imm8 ak imm8 aj aj aj aj j aj aj aj aj f wl wl
11 t
011100
011100
101110
011110
011110
101110
011111
011111
Syntax ai{,w:,l:} aj<<ak ai{,l:} aj<<imm8 ai,w aj<<imm8 ai{,w:,l:} aj>>ak ai{,l:} aj>>imm8 ai,w aj>>imm8 ai{,l:} +aj>>ak ai{,l:} +aj>>imm8
Description
A shift left logical
A shift left logical long imm8
A shift left logical word imm8
A shift right logical
A shift right logical long imm8
A shift right logical word imm8
A shift right arithmetic
A shift right arithmetic imm8
S–2314–50 193
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual
8.3.4 A Register Immediate Instructions g
100000
100000
100000
100000 i ai ai ai ai j
00hw
01hw
10hw
11hw k imm16 imm16 imm16 imm16 t f Syntax ai:{a:b:c:d} imm16 ai imm16:{a:b:c:d} ai:{ab:abc} imm16 ai imm16:{ab:abc:abcd}
Description
A insert imm16
A clear and enter imm16
A insert sign_ext(imm16)
A clear and enter sign_ext(imm16)
8.3.5 A Register Integer Compare Instructions g
100010
100011
100010
100011 i ai ai ai ai j aj aj aj aj k ak imm8 ak imm8 t
11
11 f
110010
110010
111010
111010
Syntax ai{,l:} aj< ak ai{,l:} aj<imm8
Description
A int compare less than
A int compare less than imm8 ai,u{l:} aj<ak A int compare less than unsigned ai,u{l:} aj<imm8 A int compare less than unsigned imm8
8.3.6 A Register Byte and Halfword Instructions g
000101
000101
000101
000101
000101
000101
194 i ai ai ai ai ai ai j aj aj aj aj aj aj ak ak ak k ak ak ak t
11
11
11
11
11
11 f
100000
100001
100100
100101
100010
100011
Syntax ai{,l:} extb(aj, ak) ai{,l:} exth(aj, ak) ai,{l:} insb(aj, ak) ai,{l:} insh(aj, ak) ai,{l:} mskb(aj, ak) ai,{l:} mskh(aj, ak)
Description
A extract byte
A extract halfword
A insert byte
A insert halfword
A mask byte
A mask halfword
S–2314–50
Cray X1 Instruction Set and Encoding [8]
8.3.7 Other A Register Instructions g
100010
100010
100010
100011 i ai ai ai ai j aj k ak aj ak
000000 ak
000000 imm8
11 t
11
11 f
010111
010110
000111
Syntax ai{,l:} aj?ai:ak ai{,l:} aj?ak:ai ai{,l:} cvl(ak)
000111 ai{,l:} cvl(imm8)
Description
A move if zero
A move if non-zero
A compute vector length
A compute vector length imm8
8.3.8 A Register Branch Instructions g
101000
101001
101010
101011
101100
101101 i ai ai ai ai ai ai k imm20 imm20 imm20 imm20 imm20 imm20
Syntax bz ai,imm20 bn ai,imm20 ble ai,imm20 blt ai,imm20 bge ai,imm20 bgt ai,imm20
Description
A branch (ai==0)
PC<-PC+4+(imm20<<2)
A branch (ai!=0)
PC<-PC+4+(imm20<<2)
A branch (ai<=0)
PC<-PC+4+(imm20<<2)
A branch (ai<0)
PC<-PC+4+(imm20<<2)
A branch (ai>=0)
PC<-PC+4+(imm20<<2)
A branch (ai>0)
PC<-PC+4+(imm20<<2)
8.3.9 A Register Memory Access Instructions
The following A register memory access instructions are described in this section:
• Load and Store Instructions (see Section 8.3.9.1, page 196).
• Prefetch Instructions (see Section 8.3.9.2, page 196).
• Atomic Memory Instructions (see Section 8.3.9.3, page 196).
S–2314–50 195
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual
8.3.9.1 Load and Store Instructions g
100101
100100
100001
100001
100111
100110
100001
100001
100001
100001 i ai ai ai ai ai ai ai ai ai ai j aj aj aj aj aj aj aj aj aj aj k imm14 imm14 ak ak imm14 imm14 ak ak
000000
000000 t
01
00
01
00
01
01 f
000000
000000
000001
000001
000010
000011
Syntax ai{,d:,l:} [aj+imm14] ai,{s:w} [aj+imm14] ai{,d:,l:} [aj+ak] ai,{s:w} [aj+ak]
[aj+imm14] ai{,d:,l:}
[aj+imm14] ai,{s:w}
[aj+ak] ai{,d:,l:}
[aj+ak] ai,{s:w} ai{,d:,l:} [aj],ua
[aj] ai{,d:,l:},ua
Description
A load 64-bit imm14
A load 32-bit, sign-extended imm14
A load 64-bit indexed
A load 32-bit sign-extended indexed
A store 64-bit imm14
A store 32-bit imm14
A store 64-bit indexed
A store 32-bit indexed
A load 64-bit unaligned
A store 64-bit unaligned
8.3.9.2 Prefetch Instructions g
100101
100001 i j
000000 aj
000000 aj k imm14 ak t
01 f
000000
Syntax pref [aj+imm14] pref [aj+ak]
Description
Prefetch imm14
Prefetch indexed
8.3.9.3 Atomic Memory Instructions
In the following instructions, function codes beginning with hnt (hints) advise the hardware on the anticipated access patterns for the referenced memory location.
They have the following meanings: ex (000) Exclusive. Read misses should allocate the line exclusively, anticipating a subsequent write by the same MSP. Write misses
196 S–2314–50
Cray X1 Instruction Set and Encoding [8] g
000100
000100
000100
000100
000100 na (010) aj aj j aj aj aj ai ai i ai ai ai k ak ak ak
000000 ak t
01
01
01
01
01 should allocate the line (and must do so exclusively). If no hint is specified, the hint defaults to ex.
Non-allocate. Read and write references should not allocate space for the line in the local cache. This hint should be used when no temporal locality is expected (to avoid cache pollution), or the reference is to memory that a different MSP will access subsequently (explicit communication). If the referenced item is present in the cache, the cache should be accessed normally (this is not a cache bypass mode).
f
hnt000
hnt001
hnt010
hnt100
hnt101
Syntax ai{,d:,l:}
[aj],afadd,ak{,hnt}
Description
Atomic fetch and add ai{,d:,l:}
[aj],afax,ak{,hnt} ai{,d:,l:}
[aj],acswap,ak{,hnt}
Atomic fetch and exclusive OR
Atomic compare and swap
[aj] ai{,d:,l:},aadd{,hnt} Atomic add
[aj] ai{,d:,l:},aax,ak{,hnt}
Atomic and exclusive
OR
8.4 S Register Instructions
The following S register instructions are described in this section:
• S Register Integer Instructions (see Section 8.4.1, page 198).
• S Register Logical Instructions (see Section 8.4.2, page 198).
• S Register Shift Instructions (see Section 8.4.3, page 199).
• S Register Immediate Instructions (see Section 8.4.4, page 199).
• S Register Integer Compare Instructions (see Section 8.4.5, page 200).
• Other S Register Instructions (see Section 8.4.6, page 200).
• S Register Branch Instructions (see Section 8.4.7, page 200).
• S Register Floating Point Instructions (see Section 8.4.8, page 201).
S–2314–50 197
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual
• S Register Floating Point Compare Instructions (see Section 8.4.9, page 201).
• S Register Conversion Instructions (see Section 8.4.10, page 202).
• S Register Memory Instructions (see Section 8.4.11, page 203).
8.4.1 S Register Integer Instructions g
110010
110011
110011
110010
110011
110011
110010
110010
110010
110010
110010 si si si si i si si si si si si si sj sj sj sj j sj sj sj sj sj sj sj k sk imm8 imm8 sk imm8 imm8 sk sk sk
000000
000000 t wl wl wl
11 wl wl wl f
000000
000000
101000
000101
000101
101101
011000
011001
001000
001010
001011
Syntax Description si{,w:,l:} sj+sk S int add si{,l:} sj+imm8 S int add long imm8 si,w sj+imm8 si{,w:,l:} sj-sk
S int add word imm8
S int subtract si{,l:} sj-imm8 si,w sj-imm8 si{,w:,l:} sj*sk si{,l:} hi(sj*sk) si{,w:,l:} sk/sj
S int subtract long imm8
S int subtract word imm8
S int multiply
S int multiply high
S int divide si{,w:,l:} lz(sj) S leading zero count si{,w:,l:} pop(sj) S population count
8.4.2 S Register Logical Instructions g
110010
110011
110010
110011
110010 i si si si si si j sj sj sj sj sj k sk imm8 sk imm8 sk t
11
11
11 f
010000
010000
010001
010001
010010
Syntax Description si{,l:} sj&sk S logical AND si{,l:} sj&imm8 S logical AND imm8 si{,l:} sj|sk S logical OR si{,l:} sj|imm8 S logical OR imm8 si{,l:} ~sj^sk S logical equivalence
198 S–2314–50
Cray X1 Instruction Set and Encoding [8] g
110010
110011
110010 i si si si j sj sj sj k sk imm8 sk t
11
11 f
010011
010011
010100
8.4.3 S Register Shift Instructions g
110010
110011
110011
110010
110011
110011
110010
110011 i si si si si si si si si sj sj sj j sj sj sj sj sj k sk imm8 t wl imm8 sk imm8 wl imm8 sk imm8
11 f
011100
011100
101100
011110
011110
101110
011111
011111
Syntax si{,l:} sj^sk si{,l:} ~sj&sk
Description
S logical exclusive OR si{,l:} sj^imm8 S logical exclusive OR imm8
S logical ANDNOT
Syntax si{,w:,l:} sj<<sk si{,l:} sj<<imm8 si,w sj<<imm8 si{,w:,l:} sj>>sk si{,l:} sj>>imm8 si,w sj>>imm8 si{,l:} +sj>>sk si{,l:} +sj>>imm8
Description
S shift left logical
S shift left logical long imm8
S shift left logical word imm8
S shift right logical
S shift right logical long imm8
S shift right logical word imm8
S shift right arithmetic
S shift right arithmetic imm8
8.4.4 S Register Immediate Instructions g
110000
110000
110000
110000 i si si si si j
00hw
01hw
10hw
11hw k imm16 imm16 imm16 imm16 t
S–2314–50 f Syntax si:{a:b:c:d} imm16 si imm16:{a:b:c:d} si:{ab:abc} imm16 si imm16:{ab:abc:abcd}
Description
S insert imm16
S clear and enter imm16
S insert sign_ext(imm16)
S clear and enter sign_ext(imm16)
199
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual
8.4.5 S Register Integer Compare Instructions g
110010
110011
110010
110011
200 i si si si si j sj sj sj sj k sk imm8 sk imm8
8.4.7 S Register Branch Instructions j g
111000
111001
111010
111011 si si i si si k imm20 imm20 imm20 imm20 t t
11
11 f
110010
110010
111010
111010
Syntax si{,l:} sj< sk
Description
S int compare less than si{,l:} sj<imm8 S int compare less than imm8 si,u{l:} sj<sk S int compare less than unsigned si,u{l:} sj<imm8 S int compare less than unsigned imm8
8.4.6 Other S Register Instructions g
110010
110010
110010
110010
110010 si si si i si si j sj sj sj sj
000000 k sk sk sk sk sk t
11
11
01
00
11 f
010111
010110
001100
001100
101111
Syntax si{,l:} sj?si:sk si{,l:} sj?sk:si si{,d:,l:} cpys(sj,sk) si{,s:w} cpys(sj,sk) si bmm(sk) f Syntax bz si,imm20 bn si,imm20 ble si,imm20 blt si,imm20
Description
S move if zero
S move if non-zero
S copy sign 64-bit
S copy sign 32-bit
S bit matrix multiply
Description
S branch (si=0)
PC<-PC+4+(imm20<<2)
S branch (si!=0)
PC<-PC+4+(imm20<<2)
S branch (si<=0)
PC<-PC+4+(imm20<<2)
S branch (si<0)
PC<-PC+4+(imm20<<2)
S–2314–50
Cray X1 Instruction Set and Encoding [8] g
111100
111101 i si si j k imm20 imm20 t f Syntax bge si,imm20 bgt si,imm20
8.4.8 S Register Floating Point Instructions g
110010
110010
110010
110010
110010
110010 i si si si si si si j sj sj sj sj sj sj k sk sk sk sk
000000
000000 sd sd sd sd t sd sd f
000000
000101
011000
001000
001001
001111
Syntax si,{s:d} sj+sk si,{s:d} sj-sk si,{s:d} sj*sk
Description
S floating point add
S floating point subtract
S floating point multiply
S floating point divide si,{s:d} sj/sk si,{s:d} sqrt(sj) S floating point square root si,{s:d} abs(sj) S floating point absolute value
8.4.9 S Register Floating Point Compare Instructions g
110010
110010
110010 i si si si j sj sj sj k sk sk sk t sd sd sd f
110001
111110
110010
Syntax si,{s:d} sj=sk si,{s:d} sj!=sk si,{s:d} sj<sk
Description
S branch (si>=0)
PC<-PC+4+(imm20<<2)
S branch (si>0)
PC<-PC+4+(imm20<<2)
Description
S floating point compare equal
S floating point compare not equal
S floating point compare less than
S–2314–50 201
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual g
110010
110010 i si si j sj sj k sk sk t sd sd f
110011
110000
Syntax si,{s:d} sj<=sk si,{s:d} sj?sk
Description
S floating point compare less than or equal
S floating point compare unordered si si si si si si si si si si si si si si si i si g
110010
110010
110010
110010
110010
110010
110010
110010
110010
110010
110010
110010
110010
110010
110010
110010
8.4.10 S Register Conversion Instructions sj sj sj sj sj sj sj sj sj sj sj sj sj sj sj j sj
00
00
00
11
11
10
10
11
11
10
10
11
11 t
01
01
01 k
000000
000000
000000
000000
000000
000000
000000
000000
000000
000000
000000
000000
000000
000000
000000
000000 f
101010
101011
101000
101010
101011
101001
101000
101001
101000
101001
100000
100001
100000
100001
100010
100011
Syntax si,d sj,w si,d sj,l si,d sj,s si,s sj,w si,s sj,l si,s sj,d si,l sj,s si,l sj,d si,w sj,s si,w sj,d si,l round(sj),s si,l round(sj),d si,w round(sj),s si,w round(sj),d si,l trunc(sj),s si,l trunc(sj),d
Description
S convert word to double
S convert long to double
S convert single to double
S convert word to single
S convert long to single
S convert double to single
S convert single to long
S convert double to long
S convert single to word
S convert double to word
S round single to long
S round double to long
S round single to word
S round double to word
S truncate single to long
S truncate double to long
202 S–2314–50
Cray X1 Instruction Set and Encoding [8] g
110010
110010
110010
110010
110010
110010
110010
110010
110010
110010 i si si si si si si si si si si j sj sj sj sj sj sj sj sj sj sj k
000000
000000
000000
000000
000000
000000
000000
000000
000000
000000
11
11
10
10
11
11
10
10 t
10
10
100100
100101
100100
100101
100110
100111
100110
100111 f Syntax
1000100 si,w trunc(sj),s
100011 si,w trunc(sj),d si,l ceil(sj),s si,l ceil(sj),d si,w ceil(sj),s si,w ceil(sj),d si,l floor(sj),s si,l floor(sj),d si,w floor(sj),s si,w floor(sj),d
Description
S truncate single to word
S truncate double to word
S ceiling single to long
S ceiling double to long
S ceiling single to word
S ceiling double to word
S floor single to long
S floor double to long
S floor single to word
S floor double to word
8.4.11 S Register Memory Instructions g
110101
110100
110001
110001
110111
110110
110001
110001 i si si si si si si si si j aj aj aj aj aj aj aj aj k imm14 imm14 ak ak imm14 imm14 ak ak t
01
00
01
00 f Syntax si{,d:,l:} [aj+imm14] si,{s:w} [aj+imm14]
000000
000000 si{,d:,l:} [aj+ak] si,{s:w} [aj+ak]
000001
000001
[aj+imm14] si{,d:,l:}
[aj+imm14] si,{s:w}
[aj+ak] si{,d:,l:}
[aj+ak] si,{s:w}
Description
S load 64-bit imm14
S load 32-bit sign_extend imm14
S load 64-bit indexed
S load 32-bit sign_extend indexed
S store 64-bit imm14
S store 32-bit imm14
S store 64-bit indexed
S store 32-bit indexed
8.5 Vector Register Instructions
The instructions in the following sections use the vector registers:
S–2314–50 203
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual g i
0100m m vi
0101m m vi
0100m m vi
0101m m vi
0100m m vi
0101m m vi
0100m m vi
0101m m vi
0100m m vi
0101m m vi
204
• Vector Register Integer Instructions (see Section 8.5.1, page 204).
• Vector Register Logical Instructions (see Section 8.5.2, page 205).
• Vector Register Shift Instructions (see Section 8.5.3, page 205).
• Vector Register Integer Compare Instructions (see Section 8.5.4, page 206).
• Vector Register Floating Point Instructions (see Section 8.5.5, page 207).
• Vector Register Floating Point Compare Instructions (see Section 8.5.6, page
• Vector Register Conversion Instructions (see Section 8.5.7, page 209).
• Other Vector Register Instructions (see Section 8.5.8, page 210).
• Vector Mask Instructions (see Section 8.5.9, page 211).
• Vector Memory Instructions (see Section 8.5.10, page 211).
• Privileged Instructions (see Section 8.5.11, page 213).
The notation mm in the vector syntax refers to one of the vector mask registers.
8.5.1 Vector Register Integer Instructions j vj vj vj vj vj vj vj vj vj vj k vk sk vk sk vk sk vk sk vk sk t wl wl
11
11 wl wl
11
11 wl wl f Syntax
000000 vi{,w:,l:} vj+vk,mm
000000 vi,vc vj+sk,mm
000010 vi{,w:,l:} vj+vk,mm
000010 vi,vc vj+sk,mm
000001 vi{,w:,l:} vk-vj,mm
000001 vi{,w:,l:} sk-vj,mm
000011 vi, vc vk-vj,mm
000011 vi,cv sk-vj,mm
011000 vi{,w:,l:} vj*vk,mm
011000 vi{,w:,l:} vj*sk,mm
Description
V int add
V/S int add
V int add with carry
V/S int add with carry
V int subtract
V/S int subtract
V int subtract with borrow
V/S int subtract with borrow
V int multiply
V/S int multiply
S–2314–50
Cray X1 Instruction Set and Encoding [8] g
0101m
0100m
0100m m m m i
0100m m vi vi vi vi j vj vj vj vj k vk sk t
11
11
000000 wl
000000 wl f Syntax
011001 vi{,l:} hi(vj*vk),mm
011001 vi{,l:} hi(vj*sk),mm
001010 vi{,w:,l:} lz(vj),mm
001011 vi{,w:,l:} pop(vj),mm
Description
V int multiply high
V/S int multiply high
V leading zero count
V population count
8.5.2 Vector Register Logical Instructions g
0100mm
0101mm
0100mm
0101mm
0100mm
0101mm
0100mm
0101mm
0100mm
0101mm i vi vi vi vi vi vi vi vi vi vi j vj vj vj vj vj vj vj vj vj vj vk sk vk sk k vk sk vk sk vk sk wl wl wl wl wl wl wl t wl wl wl f
010000
010000
010001
010001
010010
010010
010011
010011
010100
010110
Syntax vi{,w:,l:} vj&vk,mm vi{,w:,l:} vj&sk,mm vi{,w:,l:} vj|vk,mm vi{,w:,l:} vj|sk,mm vi{,w:,l:} ~vj^vk,mm vi{,w:,l:} ~vj^sk,mm vi{,w:,l:} vj^vk,mm vi{,w:,l:} vj^sk,mm vi{,w:,l:} ~vj&vk,mm vi{,w:,l:} ~vj&sk,mm
Description
V logical AND
V/S logical AND
V logical OR
V/S logical OR
V logical equivalence
V/S logical equivalence
V logical exclusive OR
V/S logical exclusive
OR
V logical ANDNOT
V/S logical ANDNOT
8.5.3 Vector Register Shift Instructions g
0100mm
0101mm i vi
0101mm vi
0100mm vi vi j vj vj vj vj k vk sk vk sk t wl wl wl wl f
011100
011100
011110
011110
Syntax vi{,w:,l:} vj<<vk,mm vi{,w:,l:} vj<<sk,mm vi{,w:,l:} vj>>vk,mm vi{,w;,l:} vj>>sk,mm
S–2314–50
Description
V shift left logical
V/S shift left logical
V shift right logical
V/S shift right logical
205
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual g
0100mm
0101mm i vi vi j vj vj k vk sk t wl wl f
011111
011111
Syntax vi{,w:,l:} +vj>>vk,mm vi{,w:,l:} +vj>>sk,mm
Description
V shift right arithmetic
V/S shift right arithmet
8.5.4 Vector Register Integer Compare Instructions g i
0100mm 000mi
0101mm 000mi
0100mm 000mi
0101mm 000mi
0100mm 000mi
0101mm 000mi
0100mm 000mi
0101mm 000mi
0100mm 000mi
0101mm 000mi
0100mm 000mi vj vj vj vj vj vj vj vj j vj vj vj k vk sk vk sk vk sk vk sk vk sk vk
0101mm 000mi vj sk wl wl wl wl wl t wl wl wl wl wl wl wl f
110001
110001
111110
111110
110010
110010
110011
110011
111011
111011
111011
111011
Syntax mi{,w:,l:} vj== vk,mm mi{,w:,l:} vj== sk,mm mi{,w:,l:} vj!=vk,mm mi{,w:,l:} vj!= sk,mm mi{,w:,l:} vj<vk,mm mi{,w:,l:} vj<sk,mm mi,u{,w:,l:} vj<vk,mm mi,u{,w:,l:} vj<sk,mm
Description
V int compare equal
V/S int compare equal
V int compare not equal
V/S int compare not equal
V int compare less than
V/S int compare less than
V int compare less than unsigned
V/S int compare less than unsigned mi{,w:,l:} vj<= vk,mm V int compare less than or equal mi{,w:,l:} vj<= sk,mm V/S int compare less than or equal mi,u{,w:,l:} vj<= vk,mm V int compare less than or equal unsigned mi,u{,w:,l:} vj<= sk,mm V/S int compare less than or equal unsigned
206 S–2314–50
Cray X1 Instruction Set and Encoding [8] g
0101mm
0101mm
0101mm
0101mm i
000mi
000mi
000mi
000mi j vj vj vj vj k sk sk sk sk t wl wl wl wl f
110100
111100
110101
111101
Syntax mi{,w:,l:} vj>sk,mm
Description
V/S int compare greater than mi,u{,w:,l:} vj> sk,mm V/S int compare greater than unsigned mi{,w:,l:} vj>= sk,mm V/S int compare greater than or equal mi,u{,w:,l:} vj>= sk,mm V/S int compare greater than or equal unsigned
8.5.5 Vector Register Floating Point Instructions g i
0100mm vi
0101mm vi
0100mm vi
0101mm vi
0100m vi
0101mm vi
0100mm vi
0101mm vi
0100mm vi
0100mm vi j vj vj vj vj vj vj vj vj vj vj sk vk k vk sk vk vk vk sk
000000 sd
000000 sd sd sd sd sd sd t sd sd sd f
000000
000000
000001
000001
011000
011000
001000
001000
001001
001111
Syntax vi,{s:d} vj+sk,mm vi,{s:d} vj+vk,mm vi,{s:d} vj-vk,mm vi,{s:d} sk-vj,mm vi,{s:d} vj*vk,mm vi,{s:d} vj*vk,mm vi,{s:d} vk/vj,mm vi,{s:d} sk/vj,mm vi,{s:d} sqrt(vj),mm vi,{s:d} abs(vj),mm
Description
V floating point add
V/S floating point add
V floating point subtract
V/S floating point subtract
V floating point multiply
V/S floating point multiply
V floating point divide
V/S floating point divide
V floating point square root
V floating point absolute value
S–2314–50 207
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual
8.5.6 Vector Register Floating Point Compare Instructions g
0100mm
0101mm
0100mm
0101mm
0100mm
0101mm
0100mm i
00mi
00mi
00mi
00mi
00mi
00mi
00mi j vj vj vj vj vj vj vj k vk sk vk sk vk sk vk t sd sd sd sd sd sd sd f
110001
110001
111110
111110
110010
110010
110011
Syntax mi,{s:d} vj==vk,mm mi,{s:d} vj==sk,mm mi,{s:d} vj!=vk,mm mi,{s:d} vj!=sk,mm mi,{s:d} vj<vk,mm mi,{s:d} vj<sk,mm mi,{s:d} vj<=vk,mm
0101mm
0101mm
0101mm
0100mm
0101mm
00mi
00mi
00mi
00mi
00mi vj vj vj vj vj sk sk sk vk sk sd sd sd sd sd
110010
110100
110101
111000
111000 mi,{s:d} vj<=sk,mm mi,{s:d} vj>sk,mm mi,{s:d} vj>=sk,mm mi,{s:d} vj?vk,mm mi,{s:d} vj?sk,mm
Description
V floating point compare equal
V/S floating point compare equal
V floating point compare not equal
V/S floating point compare not equal
V floating point compare less than
V/S floating point compare less than
V floating point compare less than or equal
V/S floating point compare less than or equal
V/S floating point compare greater than
V/S floating point compare greater than or equal
V floating point compare unordered
V/S floating point compare unordered
208 S–2314–50
Cray X1 Instruction Set and Encoding [8]
8.5.7 Vector Register Conversion Instructions g i
0100mm vi
0100mm vi
0100mm vi
0100mm vi
0100mm vi
0100mm vi
0100mm vi
0100mm vi
0100mm vi
0100mm vi
0100mm vi
0100mm vi
0100mm vi
0100mm vi
0100mm vi
0100mm vi
0100mm vi
0100mm vi vj vj vj vj vj vj j vj vj vj vj vj vj vj vj vj vj vj vj k
000000
000000
000000
000000
000000
000000
000000
000000
000000
000000
000000
000000
000000
000000
000000
000000
000000
000000 t f Syntax
01 101010 vi,d vj,w,mm
01 101011 vi,d vj,l,mm
01 101000 vi,d vj,s,mm
00 101010 vi,s vj,w,mm
00 101011 vi,s vj,l,mm
00 101001 vi,s vj,d,mm
11 101010 vi,l vj,w,mm
11 101000 vi,l vj,s,mm
11 101001 vi,l vj,d,mm
10 101011 vi,w vj,l,mm
10 100000 vi,w vj,s,mm
10 100001 vi,w vj,d,mm
11 100000 vi,l round(vj),s,mm
11 100001 vi,l round(vj),d,mm
10 100000 vi,w round(vj),s,mm
10 100001 vi,w round(vj),d,mm
11 100010 vi,l trunc(vj),s,mm
11 100011 vi,l trunc(vj),d,mm
Description
V convert word to double
V convert long to double
V convert single to double
V convert word to single
V convert long to single
V convert double to single
V convert word to long
V convert single to long
V convert double to long
V convert long to word
V convert single to word
V convert double to word
V round single to long
V round double to long
V round single to word
V round double to word
V truncate single to long
V truncate double to long
S–2314–50 209
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual g i
0100mm vi
0100mm vi
0100mm vi
0100mm vi
0100mm vi
0100mm vi
0100mm vi
0100mm vi
0100mm vi
0100mm vi j vj vj vj vj vj vj vj vj vj vj k
000000
000000
000000
000000
000000
000000
000000
000000
000000
000000 t f Syntax
10 100010 vi,w trunc(vj),s,mm
10 100011 vi,w trunc(vj),d,mm
11 100100 vi,l ceil(vj),s,mm
11 100101 vi,l ceil(vj),d,mm
10 100100 vi,w ceil(vj),s,mm
10 100100 vi,w ceil(vj),d,mm
11 100110 vi,l floor(vj),s,mm
11 100111 vi,l floor(vj),d,mm
10 100110 vi,w floor(vj),s,mm
10 100111 vi,w floor(vj),d,mm
Description
V truncate single to word
V truncate double to word
V ceiling single to long
V ceiling double to long
V ceiling single to word
V ceiling double to word
V floor single to long
V floor double to long
V floor single to word
V floor double to word
8.5.8 Other Vector Register Instructions g
011100 i
000000
0100mm vi
0100mm vi
0100mm vi
011100 000000 j
000000
000000 vj vj
000000 k vk vk vk vk
000000 t f Syntax
11 000010 bmm vk
11 101111 vi bmm(vk),mm
01 001100 vi{,d:,l:} cpys(vj,vk),mm
00 001100 vi,{s:w} cpys(vj,vk),mm
00 000100 vrip
Description
V load bit matrix multiply unit
V bit matrix multiply
V copy sign 64-bit
V copy sign 32-bit
Vector rest in peace
210 S–2314–50
Cray X1 Instruction Set and Encoding [8]
8.5.9 Vector Mask Instructions g i
0100mm vi
0101mm vi
0100mm vi j vj vj vj k vk
0101mm vi
010100 vi vj
000mj
010100 vi
010000 vi
000mj
000mj ak vk
011100 000mi 000000 ak sk vk sk ak
011100 ai
011100 ai
011100 ai
000000 000mk 00
000000 000mk 00
000000 000mk 00
011100 000mi 000mj 000mk 00
011100 00mi
011100 00mi
011100 00mi
011100 00mi
00mj
000mj
00mj
00mj
00mk
000mk 00
00mk
01
00
00
00
00mk 00 t
01
01
00
00
11
11 f Syntax
101110 vi{,d:,l:} mm?vk:vj
Description
V merge 64-bit
101110 vi{,d:,l:} mm?sk:vj
101110
101100 vi{,d:,l:} mm?vk:vj
101110 vi,{s:w} mm?sk:vj vi scan(ak,mj)
101101 vi{,l:} cidx(ak,mj)
V/S merge 64-bit
V merge 32-bit
V/S merge 32-bit
V iota continuous
V iota compressed
010101 vi{,d:,l:} cmprss(vk,mj) V compress
001000
001001 mi fill(ak) ai last(mk)
001010 ai first(mk)
001011 ai pop(mk)
VM fill
VM last element set
VM first element set
VM pop count
VM logical AND 010000 mi mj&mk
010001 mi mj|mk
010010 mi ~mj^mk
010011 mi mj^mk
010100 mi mj&mj
VM logical OR
VM logical equivalence
VM logical exclusive OR
VM logical ANDNOT
8.5.10 Vector Memory Instructions
The function codes beginning with hnt (hint) in this section advises the hardware on whether and how a data value should be cached. The values have the following meanings: ex (000) sh (001)
When you specify ex or do not specify a value for hnt you get the default, which is exclusive. Read misses allocate the cache line exclusively, anticipating a subsequent write by the same MSP. If the cache line is already shared, read misses allocate the line in shared mode. Write misses must allocate the line exclusively.
Shared. Read misses allocate the cache line in the shared state, anticipating a subsequent read by a different MSP. This hint is ignored by writes.
S–2314–50 211
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual g i
0110mm vi
0110mm vi
0110mm vi
0110mm vi
0110mm vi
0110mm vi
0110mm vi
0110mm vi
0110mm vi
0110mm vi
0110mm vi
0110mm vi na (010) aj aj aj aj aj aj aj aj aj aj aj j aj k ak t
01 ak 00 imm6 01 imm6 00 vk 01 vk ak
00 ak imm6 01
01
00 imm6 00 vk 01 vk 00
Non-allocate. Read and write references do not allocate space for the line in local cache. This hint is used when no temporal locality is expected (to avoid cache pollution), or when the reference is to memory that will be subsequently accessed by a different MSP
(explicit communication). If the referenced item is present in cache, the cache is accessed normally.
f
hnt000
hnt000
Syntax vi{,d:,l:} [aj,ak],mm{,hnt} vi{,s:,w:} [aj,ak],mm{,hnt}
Description
V load 64-bit ak scaled by 8
V load 32-bit ak scaled by 4
hnt010 vi{,d:,l:} [aj,imm6],mm{,hnt} V load 64-bit imm6 scaled by 8
hnt010 vi{,s:,w:}
[aj,imm6],mm{,hnt}
hnt100 vi{,d:,l:} [aj,vk],mm{,hnt}
V load 32-bit imm6 scaled by 4
V gather 64-bit vk scaled by 8
hnt100
hnt001 vi{,s:,w:} [aj,vk],mm{,hnt}
[aj,ak] vi{,d:,l:},mm{,hnt}
V gather 32-bit vk scaled by 4
V store 64-bit ak scaled by 8
hnt001 [aj,ak] vi{,s:,w:},mm{,hnt} V store 32-bit ak scaled by 4
hnt011 [aj,imm6] vi{,d:,l:},mm{,hnt} V store 64-bit imm6 scaled by 8
hnt011 [aj,imm6] vi{,s:,w:},mm{,hnt}
V store 32-bit imm6 scaled by 4
hnt101 [aj,vk] vi{,d:,l:},mm{,hnt}
hnt101 [aj,vk] vi{,s:,w:},mm{,hnt}
V scatter 64-bit vk scaled by 8
V scatter 32-bit vk scaled by 4
212 S–2314–50
Cray X1 Instruction Set and Encoding [8] g
0110mm
0110mm i vi vi j aj aj k vk vk t
01
00 f Syntax Description
hnt111 [aj,vk] vi{,d:,l:},mm,ord{,hnt} V ordered scatter 64-bit vk scaled by 8
hnt111 [aj,vk] vi{,s:,w:},mm,ord{,hnt}
V ordered scatter 32-bit vk scaled by 4
8.5.11 Privileged Instructions g
000110
000011
00011
011100 i
000000 ai
000000
000000 j
000000 aj aj
000000 k
000000
000000 ak
000000 t
00
01
01
00 f
000000
000100
000101
000101
Syntax eret ai hw[ai] hw[ai] ak vclean
Description
Return from exception
Read hardware state
Write hardware state
Clean hidden vector state
S–2314–50 213
S–2314–50
CAL Defined Sequences [9]
Defined sequences are named sequences of instructions that can be saved for assembly later in the source program. Defined sequences have several functional similarities.
The four types of defined sequences are specified by the MACRO, OPDEF, DUP, and
ECHO pseudo instructions. The ENDM, ENDDUP, and STOPDUP pseudo instructions terminate defined sequences. The LOCAL and OPSYN pseudo instructions are associated with definitions and are included in this chapter.
The defined sequence pseudo instructions are as follows:
Pseudo
MACRO
OPDEF
DUP
ECHO
ENDM
ENDDUP
STOPDUP
LOCAL
OPSYN
EXITM
Description
A sequence of source program instructions saved by the assembler for inclusion in a program when called for by the macro name. The macro call resembles a pseudo instruction.
A sequence of source program instructions saved by the assembler for inclusion in a program called for by the OPDEF pseudo instruction. The opdef resembles a symbolic machine instruction.
Introduces a sequence of code that is assembled repetitively a specified number of times; the duplicated code immediately follows the DUP pseudo instruction.
Introduces a sequence of code that is assembled repetitively until an argument list is exhausted.
Ends a macro or opdef definition.
Terminates a DUP or ECHO sequence of code.
Stops the duplication of a code sequence by overriding the repetition condition.
Specifies unique strings that are usually used as symbols within a
MACRO , OPDEF, DUP, or ECHO pseudo instruction.
Defines a label field name that is the same as a specified operation in the operand field name.
Terminates the innermost nested MACRO or OPDEF expansion.
215
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual
9.1 Similarities Among Defined Sequences
Defined sequences have the following functional similarities:
• Editing
• Definition format
• Formal parameters
• Instruction calls
• Interact with the INCLUDE pseudo instruction
9.1.1 Editing
216
Assembler editing is disabled at definition time. The body of the definition
(see Section 9.1.2, page 217) is saved before micros and concatenation marks
are edited.
If editing is enabled, editing of the definition occurs during assembly each time it is called. The ENDDUP, ENDM, END, INCLUDE, and LOCAL pseudo instructions and prototype statements should not contain micros or concatenation characters because they may not be recognized at definition time.
When a sequence is defined, editing is disabled and cannot be explicitly enabled.
When a sequence is called, the assembler performs the following operations:
• Checks all parameter substitutions marked at definition time
• Edits the statement if editing is enabled
• Processes the statement
By disabling editing at definition time (default) and specifying the INCLUDE pseudo instruction with embedded underscores, a saving in program overhead is achieved. Because editing is disabled at definition time, concatenation does not occur until the macro is called. If editing is enabled when the macro is called, the file is included at that time. This technique is demonstrated in the following example:
MACRO
INC
.
.
.
IN_CLUDE MYFILE ; INCLUDE pseudo instruction with an embedded
S–2314–50
CAL Defined Sequences [9]
.
.
.
ENDM
; underscore
Embedding underscores in an INCLUDE pseudo instruction becomes desirable when the INCLUDE pseudo instruction identifies large files. Because files are included when the macro is called and not at definition time, embedding underscores in the INCLUDE pseudo instruction can reduce the overhead required for a program.
9.1.2 Definition Format
S–2314–50
MACRO , OPDEF, DUP, and ECHO pseudo instructions use the same definition format. The format consists of a header, body, and end.
The header consists of a MACRO, OPDEF, DUP, or ECHO pseudo instruction, a prototype statement for a MACRO or OPDEF definition, and, optionally, LOCAL pseudo instructions. For a macro, the prototype statement provides a macro functional definition and a list of formal parameters. For an opdef, the prototype statement supplies the syntax and the formal parameters.
LOCAL pseudo instructions identify parameter names that the assembler must make unique to the assembly each time the definition sequence is placed in a program segment. Asterisk comments can be placed in the header and do not affect the way the assembler scans the header. Asterisk comments are dropped
from the definition. To force asterisk comments into a definition, see Section
The body of the definition begins with the first statement following the header.
The body can consist of a series of CAL instructions other than an END pseudo instruction. The body of a definition can be empty, or it can include other definitions and calls. A definition used within another definition is not recognized, however, until the definition in which it is contained is called; therefore, an inner definition cannot be called before the outer definition is called for the first time.
A comment statement identified by an asterisk in column l is ignored in the definition header and the definition body. Such comments are not saved as a part of the definition sequence. Comment fields on other statements in the body of a definition are saved.
An ENDM pseudo instruction with the proper name in the label field ends a macro or opdef definition. A statement count or an ENDDUP pseudo instruction with
217
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual the proper name in the label field ends a dup definition. An ENDDUP pseudo instruction with the proper name in the label field ends an echo definition.
9.1.3 Formal Parameters
Formal parameters are defined in the definition header and recognized in the definition body. Four types of formal parameters are recognized as follows:
• Positional
• Keyword
• Echo
• Local
The characters that identify positional, keyword, echo, and local parameters must all have unique names within a given definition. Positional, keyword, and echo parameters are also case-sensitive. To be recognized, you must specify these parameters in the body of the definition exactly as specified in the definition header. Parameter names must meet the requirements for identifiers as described
You can embed a formal parameter name within the definition body; however, embedded parameters must satisfy the following requirements:
• The first character of an embedded parameter must begin with a legal initial-identifier-character.
• An embedded parameter cannot be preceded by an initial-identifier-character
(for example, PARAM is a legally embedded parameter within the
ABC_PARAM_DEF string because it is preceded by an underscore character).
PARAM is not a legally embedded character within the string ABCPARAMDEF because it is preceded by an initial-identifier-character (C).
• An embedded parameter must not be followed by an identifier-character.
In the following example, the embedded parameter is legal because it is followed by an element separator (blank character):
PARAM678
In the following example, the embedded parameter is illegal because it is followed by the identifier-character 9:
PARAM6789
218 S–2314–50
CAL Defined Sequences [9]
• Embedded parameters must contain 8 or less characters. PARAM6789 is illegal because it contains 9 characters. The character that follows an embedded parameter (9) cannot be an identifier-character.
• If and only if the new format is specified, an embedded parameter must occur before the first comment character (;) of each statement within the body.
• An embedded parameter must have a matching formal parameter name in the definition header.
• Formal parameter names should not be END, ENDM, ENDDUP, LOCAL, or
INCLUDE pseudo instructions. If any of these are used as parameter names, substitution of actual arguments occurs when these names are contained in any inner definition reference.
Note: If the file is included at expansion time, arguments are not substituted for formal parameters into statements within included files.
9.1.4 Instruction Calls
Each time a definition sequence of code is called, an entry is added to a list of currently active defined sequences within the assembler. The most recent entry indicates the current source of statements to be assembled. When a definition is called within a definition sequence that is being assembled, another entry is made to the list of defined sequences, and assembly continues with the new definition sequence belonging to the inner, or nested, call.
At the end of a definition sequence, the most recent list entry is removed and assembly continues with the previous list entry. When the list of defined sequences is exhausted, assembly continues with statements from the source file.
An inner nested call can be recursive; that is, it can reference the same definition that is referenced by an outer call. The depth of nested calls permitted by CAL is limited only by the amount of memory available.
The sequence field in the right margin of the listing shows the definition name and nesting depth for defined sequences being assembled. Nesting depth numbers begin in column 89 and can be one of the following: :1, :2, :3, :4, :5,
:6 , :7, :8, :9, :*.
If the nesting depth is greater than 9, the assembler keeps track of the current nesting level; an asterisk represents nesting depths of 10 or more. Nesting depth numbers are restricted to two characters so that only the two rightmost character positions are overwritten.
S–2314–50 219
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual
220
If the sequence field (columns 73 through 90) of the source file is not empty, the assembler copies the existing field for a call into every statement expanded by the call reserving columns 89 and 90 for the nesting level. For example, if the sequence field for MCALL was LQ5992A.112, the sequence field for a statement expanded from MCALL would read as follows:
LQ5992A.1l2
:l
Additional nested calls within MCALL would change the nesting level, but the sequence field would be unchanged during MCALL. For example:
LQ5992A.112
LQ5992A.l12
LQ5992A.112
LQ5992A.112
LQ5992A.112
LQ5992A.112
:2
:2
:2
:3
:*
:l
If the sequence field (columns 73 through 90) of the source file is empty, the assembler inserts the name of the definition, as follows:
Name
Macro
Opdef
Dup
Echo
Description
The inserted name in the sequence field is the name found in the result field of the macro prototype statement.
The inserted name in the sequence field is the name used in the label field of the OPDEF pseudo instruction itself.
The inserted name in the sequence field is the name used in the label field of the DUP pseudo instruction, or if the count is specified and name is not, the name is *Dup.
The inserted name in the sequence field is the name used in the label field of the ECHO pseudo instruction.
In all cases, the first two columns of the sequence field contain asterisks (**) to indicate that the assembler has generated the sequence field. Columns 89 and
90 of the sequence field are reserved for the nesting level. If, for example, the sequence field is missing for MCALL, it would read as follows:
** MCALL :1
Additional nested calls within MCALL would change the nesting level, but the sequence field number would be unchanged for the duration of MCALL.
The following example illustrates how the assembler tracks the nesting sequence:
S–2314–50
CAL Defined Sequences [9]
*MCALL
*MCALL
*MCALL
*MCALL
*MCALL
** MCALL
** MCALL
:1
:2
:2
:2
:3
:*
:1
9.1.5 Interaction with the INCLUDE Pseudo Instruction
The INCLUDE pseudo instruction operates with defined sequences, as follows:
Sequence
MACRO
OPDEF
DUP
ECHO
Description
INCLUDE pseudo instructions are expanded at definition time.
INCLUDE pseudo instructions are expanded at definition time.
INCLUDE pseudo instructions are expanded at definition time. If count is specified, the INCLUDE pseudo instruction statement itself is not included in the statements being counted.
INCLUDE pseudo instructions are expanded at definition time.
9.2 Macros (
MACRO
)
S–2314–50
A macro definition identifies a sequence of statements. This sequence of statements is saved by the assembler for inclusion elsewhere in a program. A macro is referenced later in the source program by the macro call. Each time the macro call occurs, the definition sequence is placed into the source program.
You can specify the MACRO pseudo instruction anywhere within a program segment. If the MACRO pseudo instruction is found within a definition, it is defined. If the MACRO pseudo instruction is found within a skipping sequence, it is skipped and is not recognized as a pseudo instruction.
If a macro definition occurs within the global definitions part of a program segment, it is defined as global. If macro definitions occur within a program module (an IDENT, END sequence), they are local. A global definition can be redefined locally, however, at the end of the program module, it is re-enabled and the local definition is discarded. A global definition can be referenced from anywhere within the assembler program following the definition.
The following example illustrates a macro definition:
MACRO
221
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual
GLOBAL
* GLOBAL DEFINITION IS USED.
GLOBAL ENDM
LIST MAC
; Globally defined.
GLOBAL
* GLOBAL DEFINITION IS USED.
IDENT TEST
GLOBAL
; Call to global definition.
; Call to global definition.
* GLOBAL DEFINITION IS USED.
MACRO
GLOBAL
; Locally defined.
; Attempted global definition.
* Redefinition warning message is issued
* LOCAL DEFINITION IS USED.
GLOBAL ENDM
GLOBAL
* LOCAL DEFINITION IS USED.
END
IDENT TEST2
GLOBAL
* GLOBAL DEFINITION IS USED.
END
; Call to local definition.
; Local definitions discarded
; Call to global definition.
9.2.1 Macro Definition
222
The macro definition header consists of the MACRO pseudo instruction, a prototype statement, and optional LOCAL pseudo instructions. The prototype statement provides a name for the macro and a list of formal parameters and default arguments.
A comment statement, identified by an asterisk in column 1, is ignored in the definition header or definition body. Such comments are not saved as a part of the definition sequence. Comment fields on other statements in the body of a definition are saved.
The end of a macro definition is signaled by an ENDM pseudo instruction with a name that matches the name in the result field of the macro prototype statement.
For a description of the ENDM pseudo instruction, see Section 9.6, page 254.
The following macro definition transfers an integer from an A register to an S register and converts it to a normalized floating-point number: macro intconv p2 p1,p2 ; pl=A reg, p2=S reg.
+f_pl ; Transfer with special exp and sign
S–2314–50
S–2314–50
CAL Defined Sequences [9] intconv p2 endm
; extension.
+f_p1 ; Normalize the S register.
; End of macro definition.
As with every macro, INTCONV begins with the MACRO pseudo instruction. The second statement is the prototype statement, which names the macro and defines the parameters. The next three statements are definition statements that identify what the macro does. The ENDM pseudo instruction ends the macro definition.
The format of the macro definition is as follows: ignored
[location] functional
MACRO functional
LOCAL
.
.
.
ENDM ignored parameters
[[name],[name]]
The variables in the above macro definition are described as follows:
• location
The location variable specifies an optional label field parameter. It must be terminated by a space and it must meet the requirements for names as
described in Section 6.3, page 76.
• functional
The functional variable specifies the name of the macro. It must be a valid identifier or the equal sign. If functional is the same as a currently defined pseudo instruction or macro, this definition redefines the operation associated with functional, and a message is issued.
• parameters
The parameters variable specifies positional and/or keyword parameters.
Positional parameters must be entered before keyword parameters. Keyword
223
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual
224 parameters do not have to follow positional parameters. The syntax of the
parameter variable is as follows:
positional-parameters[,[keyword-parameters]]
The syntax for positional-parameters is described as follows:
[[|] [*]name] [,positional-parameters]
The variables that comprise the positional parameter are described as follows:
- |/*
The vertical bar (|) is optional. If it is not included, the positional-parameter’s argument can be an embedded argument, character string, or null string. If the vertical bar (|) is included, the parameter can be a syntactically valid expression or a null string.
A left parenthesis signals the beginning of an embedded argument and must be terminated by a matching right parenthesis. An embedded argument can contain an argument or pairs of matching left and right parentheses. If an asterisk precedes the positional parameter name, the embedded argument is used in its entirety. If an asterisk does not precede the positional parameter name, the outermost parentheses are stripped from the embedded argument and the remaining string is used as the argument.
A character string can be any character up to but not including a legal terminator
(space, tab, or semicolon for new format) or an element separator (comma).
If the assembler finds an open parenthesis (character string) with no closing parenthesis (which would make it an embedded-argument), the following warning-level message is issued:
Embedded argument was not found.
A syntactically valid expression can include a legal terminator (space, tab, or semicolon for new format) or an element separator (comma). The syntactically valid expression satisfies the requirements for an expression, but it is used only as an argument and is not evaluated in the macro call itself. If the syntactically valid expression is an embedded argument, then, as long as an asterisk precedes the positional-parameter name, the embedded argument is used in its entirety.
If an asterisk does not precede the positional-parameter name, the outermost parentheses are stripped from the embedded argument and the remaining string
S–2314–50
S–2314–50
CAL Defined Sequences [9] is used as the argument. Use of the syntactically valid expression permits you to enter a string (=‘, ’R) of characters that may contain one or more spaces or a comma.
The null string is an empty string.
• positional-parameters
positional-parameters must be specified with valid and unique names and they
must meet the requirements for names as described in Section 6.3, page 76. There
can be none, one, or more positional parameters. The default argument for a
positional-parameter is an empty string.
The positional parameters defined in the macro definition are case-sensitive.
Positional parameters that are specified in the definition body must identically match positional parameters defined by the macro prototype statement.
The syntax for keyword-parameters can be any of the following:
[*]name=[expression-argument-value]
[,[keyword-parameters]]
[*] | name=[expression-argument-value]
[,[keyword-parameters]]
[*]name=[string-argument-value]
[,[keyword-parameters]]
The elements of keyword-parameters are described as follows:
• keyword-parameters
keyword-parameters must be specified with valid and unique names. Names within keyword-parameter must meet the requirements for names as described in
There can be zero, one, or more keyword-parameters . You can enter names within keyword-parameters in any order. Default arguments can be provided for each keyword-parameter at definition time, and they are used if the keyword is not specified at call time.
225
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual
226
The keyword-parameters defined in a macro definition are case-sensitive.
The keyword-parameters specified in the macro body must match the
positional-parameters specified in the macro prototype statement.
The | is optional. If the | is not included, the -s option argument can be an embedded argument, a character string, or a null string. If the | is included, the parameter be either a syntactically valid expression or a null string.
Embedded argument. A left parenthesis signals the beginning of an embedded argument and it must be matched by a right parenthesis. An embedded argument can also contain pairs of matching left and right parentheses. If an asterisk precedes the positional parameter name, the embedded argument is used in its entirety; otherwise, the outermost parentheses are stripped from the embedded argument and the remaining string is used as the argument.
Character string. Any character up to but not including a legal terminator (space, tab, or semicolon for new format) or an element separator. If the assembler finds an open parenthesis (character string) with no closing parenthesis (which would make it an embedded argument), the following warning-level listing message is issued:
Embedded argument was not found.
The null argument is an empty string.
Syntactically valid expression. An expression can include a legal terminator
(space, tab, or semicolon for new format) or an element separator (comma). The syntactically valid expression is a legal expression, but it is used only as an argument and is not evaluated in the macro call itself.
If the syntactically valid expression is an embedded argument and if an asterisk precedes the positional parameter name, the embedded argument is used in its entirety. If an asterisk does not precede the positional-parameter name, the outermost parentheses are stripped from the embedded argument and the remaining string is used as the argument.
If a default is provided for a keyword-parameter, it must meet the preceding requirements.
The following example illustrates the use of positional parameters:
.
.
.
MACRO
JUSTIFY |PARAM
; Macro prototype.
S–2314–50
CAL Defined Sequences [9]
9.2.2 Macro Calls
S–2314–50
JUSTIFY ENDM
JUSTIFY
JUSTIFY
‘,’R ; Macro call
‘ ’R; Macro call
When the following macro is called, the positional parameter p1 receives a value of v1 because an asterisk does not precede the parameter on the prototype statement. The positional parameter p2, however, receives a value of (v2) because an asterisk precedes the parameter on the prototype statement.
.
.
macro paren p1,p2 ; Macro prototype.
.
paren endm paren (v1),(v2) ; Macro call.
An instruction of the following format can call a macro definition:
[locarg] functional positional-arguments ["," [keyword-arguments]]
[locarg] functional keyword-arguments
The elements of the macro call are described as follows:
• locarg
The locarg element specifies an optional label field argument. locarg must be terminated by a space or a tab (new format only). locarg can be any character up to but not including a space. If a label field parameter is specified on the macro definition, you can specify a matching label field parameter on the macro call. locarg is substituted wherever the label field parameter occurs in the definition. If no label field parameter is specified in the definition, this field must be empty.
• functional
The functional element specifies the macro name. It must be an identifier or an equal sign. functional must match the name specified in the macro definition.
• positional-arguments
227
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual
228
Positional-arguments specify an actual argument string that corresponds to a positional-parameter that is specified in the definition prototype statement. The requirements for positional-arguments are specified by the corresponding positional-parameter in the macro definition prototype statement. Positional-arguments are not case-sensitive to positional-parameters on the macro call.
The first positional-argument is substituted for the first positional-parameter in the prototype operand field, the second positional-argument string is substituted for the second positional-parameter in the prototype operand field, and so on. If the number of positional-arguments is less than the number of
positional-parameters in the prototype operand field, null argument strings are used for the missing positional-arguments.
Two consecutive commas indicate a null (empty) positional-argument string.
• keyword-arguments
keyword-arguments are an actual argument string that corresponds to a
keyword-parameter specified in the macro definition prototype statement.
The requirements for keyword-arguments are specified by the corresponding
keyword-parameter in the macro definition prototype statement.
keyword-arguments are not recognized until after n subfields (n commas); n is the number of positional parameters in the operand field of the macro definition.
You can list keyword-arguments in any order; matching the order in which
keyword-parameters are listed on the macro prototype statement is unnecessary.
However, because the keyword-parameter is case-sensitive, it must be specified in the macro call exactly as specified in the macro prototype statement to be recognized.
The default keyword-parameters specified in the macro prototype statement are used as the actual keyword-arguments for missing keyword-arguments.
All arguments must meet the requirements of the corresponding parameters as specified in the macro definition prototype statement.
Note: The | and * are not permitted on the macro call statement. These characters specified in the prototype statement for positional-parameters or
keyword-parameters are remembered by the assembler when the macro is called.
To call a macro, use its name in a code sequence. The INTCONV macro is called as follows:
MACRO
S–2314–50
S–2314–50
CAL Defined Sequences [9]
INTCONV
P2
P2
INTCO ENDM
LIST
P1,P2 ; P1=A reg, P2=Sreg
+F_P1 ; Transfer with special expression
; and sign extension.
+F_P2 ; Normalize the S register.
; End of macro definition.
MAC
Call and expansion of the INTCONV macro:
INTCONV
S2
S2
A1,S3 ; Macro call.
+FA1 ; Transfer with special expression
; and sign extension.
+FS2 ; Normalize the S register.
Note: Comments preceded by an underscore and an asterisk are included in the definition bodies of the following macro examples. These comments are included to illustrate the way in which parameters are passed from the macro call to the macro definition. Because comments are not assembled,
_* comments allow arguments to be shown without regard to hardware differences or available machine instructions.
The following examples show the use of positional-parameters and
keyword-parameters.
The macro table contains positional and keyword parameters.
tabn macro table tables section con con con con section table endm list tabn,val1=#0,val2=,val3=0 data
‘tabn’1 val1 val2 val3
* : Resume use of previous section.
mac
The following shows the call and expansion of the table macro:
; Macro call.
table tables section taba con con con con a
4 taba,val3=4,val2=a data
‘taba’1
~
229
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual
230 section * : Resume use of previous section.
Macro noorder demonstrates that keyword-parameters are not order dependent.
macro noorder s1 s2 s3 s4 noorder endm list param1,param2,param3=,param4=b param1 param2 param3 param4 mac
The call and expansion of the noorder macro is as follows: noorder s1 s2 s3 s4
2 d
(1),2,param4=dog,param3=d
1 dog
Macros ONE, two, and THREE demonstrate that the number of parameters specified in the macro call may form the number of parameters specified in the macro definition.
MACRO
ONE
_*PARAMETER1:
_*PARAMETER2:
_*PARAMETER3:
PARAM1,PARAM2,PARAM3
PARAM1
; SYM1 corresponds to PARAM1.
PARAM2
; Null string.
PARAM3
; Null string.
ONE ENDM
LIST MAC
The call and expansion of the ONE macro using one parameter is as follows:
ONE
* PARAMETER 1:
* PARAMETER 2:
* PARAMETER 3: macro two
_* Parameter 1:
SYM1 ; Call using one parameter.
SYM1 ; SYM1 corresponds to PARAM1.
; Null string.
; Null string.
param1,param2,param3 param1
S–2314–50
S–2314–50
CAL Defined Sequences [9]
_* Parameter 2:
_* Parameter 3:
; SYM1 corresponds to param1.
param2
; SYM2 corresponds to param2.
param3
; Null string.
two endm list mac
The call and expansion of the two macro using two parameters is as follows: two
* Parameter 1:
* Parameter 2:
* Parameter 3:
MACRO
THREE
_*PARAMETER 1: sym1,sym2 ; Call using two parameters.
sym1 ; sym1 corresponds to param1.
sym2 ; sym2 corresponds to param2.
; Null string.
_*PARAMETER 2:
_*PARAMETER 3:
PARAM1,PARAM2,PARAM3
PARAM1
;SYM1 corresponds to PARAM1.
PARAM2
;SYM2 corresponds to PARAM2.
PARAM3
;SYM3 corresponds to PARAM3.
THREE ENDM
LIST MAC
The call and expansion of the THREE macro using prototype parameters is as follows:
THREE
* PARAMETER 1:
* PARAMETER 2:
* PARAMETER 3:
SYM1,SYM2,SYM3 ; Call matching prototype.
SYM1 ; SYM1 corresponds to PARAM1.
SYM2 ; SYM2 corresponds to PARAM2.
SYM3 ; SYM3 corresponds to PARAM3.
The following examples demonstrate the use of the optional |.
Macro BANG demonstrates the use of the embedded argument (1,2), syntactically valid expressions for positional-parameters (‘abc,def’), keyword-parameters
(PARAM3=1+2), and the null string.
MACRO
BANG
_* PARAMETER 1:
_* PARAMETER 2:
PARAMl,|PARAM2,|PARAM3=,PARAM4=
PARAM1
; Embedded argument.
PARAM2
; Syntactically valid expression
231
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual
232
_* PARAMETER 3:
_* PARAMETER 4:
PARAM3
; Syntactically valid expression
PARAM4
; Null string.
BANG ENDM
LIST MAC
The call and expansion of the BANG macro is as follows:
BANG
* PARAMETER 1:
* PARAMETER 2:
* PARAMETER 3:
* PARAMETER 4:
(1,2),‘abc,def’,PARAM3=1+2
; Macro call.
1,2
; Embedded argument.
‘abc,def’
; Syntactically valid expression.
1+2
; Syntactically valid expression.
; Null string.
In the previous example:
• If the argument for PARAM1 had been (((1,2))), S1 would have received
((1,2)) at expansion.
• The | specified on PARAM2 and PARAM3 permits commas and spaces to be embedded within strings ’abc,def’ and allows expressions to be expanded without evaluation 1+2.
• PARAM4 passes a null string. A space or comma following the equal sign specifies a null or empty character string as the default argument.
In the following macro, called remem, the | is remembered from the macro definition when it is called: macro remem |param1=’ ’r ; Prototype statement includes | s1 param1 remem endm list mac
The call and expansion of the remem macro is as follows: remem param1=‘,’r s1 ‘,’r
; Macro call does not include |
S–2314–50
S–2314–50
CAL Defined Sequences [9]
The NULL and nullparm macros that follow demonstrate the effect of null strings when parameters are passed.
NULL demonstrates the effect of a null string on macro expansions. P2 is passed a null string. When NULL is expanded, the resulting line is left-shifted two spaces, which is the difference between the length of the parameter (P2) and the null string.
NULL
MACRO
NULL P1,P2,P3
S1 P1
S2 P2
S3 P3
ENDM
LIST MAC
; Left shifted two places.
The call and expansion of the NULL macro is as follows:
; Macro call.
*
NULL 1,,3
S1 1
S2
S3 3
; Left shifted two places.
Macro nullparm demonstrates how a macro is expanded when the macro call does not include the label field name specified on the macro definition.
macro nullparm longparm
; Prototype statement.
1 longparm = nullparm endm list mac
The call and expansion of the nulparm macro is as follows:
= 1 nullparm
233
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual
234
Note: The label field parameter was omitted on the macro call in the previous example. The result and operand fields of the first line of the expansion were shifted left 8 character positions because a null argument was substituted for the 8-character parameter, LONGPARM.
If the old format is used, only one space appears between the label field parameter and result field in the macro definition. If a null argument is substituted for the location parameter, the result field is shifted into the label field in column 2. Therefore, at least two spaces should always appear between a parameter in the label field and the first character in the result field in a definition. If the new format is used, the result field is never shifted into the label field.
The following macro, DEFAULT, illustrates how defaults are assigned for keywords when the macro is expanded:
MACRO
DEFAULT
_* PARAM 1
_* PARAM 2
_* PARAM 3
DEFAULT ENDM
LIST
PARAM1=(ABC DEF,GHI),PARAM2=ABC,PARAM3=
MAC
The following illustrates calls and expansions of the DEFAULT macro:
DEFAULT PARAM1=ARG1,PARAM2=ARG2,PARAM3=ARG3
; Macro call.
*ARG1
*ARG2
*ARG3
DEFAULT PARAM1=,PARAM2=(ARG2),PARAM3=ARG3
*ARG2
*ARG3
DEFAULT PARAM1=((ARG1)),PARAM2=,PARAM3=ARG3
; Macro call.
* (*ARG1)
* ARG3
The following examples illustrate the correct and incorrect way to specify a literal string in a macro definition.
Macro WRONG shows the incorrect way to specify a literal string in a macro definition. The comments in the expansion are writer comments and are not part of the expansion.
S–2314–50
S–2314–50
CAL Defined Sequences [9]
MACRO
WRONG PARAM1=’ ’R ; Prototype statement.
_* PARAM1
WRONG ENDM
LIST MAC
; End of macro definition.
; List expansion.
The call and expansion of WRONG is as follows (the assembler erroneously expands WRONG; ’ ’R was intended):
WRONG ; Macro call
* ’
Macro right shows the correct way to specify a literal string in a macro definition.
macro right |param1=’ ’r ; Prototype statement.
* param1 right endm list mac
; End of macro definition.
; List expansion.
The expansion of right is as follows (the assembler expands right as intended because of the |):
* ’ ’r right ; Macro call.
The following macros demonstrate the wrong and right methods for replacing parameters on the prototype statement with parameters on the macro call statement.
Macro BAD demonstrates the wrong method of replacing parameters.
MACRO
BAD PARAM1,PARAM2,PARAM3=JJJ
* PARAMETER 1:
* PARAMETER 2:
PARAM1
PARAM2
* PARAMETER 3:
BAD ENDM
LIST MAC
PARAM3
; End of macro definition.
; Listing expansion.
The call and expansion of the BAD macro is as follows:
BAD
* PARAMETER 1:
PARAM3=XKK ; Macro call.
PARAM3=KKK
* PARAMETER 2:
235
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual
236
* PARAMETER 3: JJJ
Macro good demonstrates the correct method for replacing parameters.
macro good param1,param2,param3=jjj
_* parameter 1: _param1
; Null string.
_* parameter 2:
_* parameter 3: good endm list mac param2
; Null string.
param3
; End of macro definition.
; Listing expansion.
The call and expansion of the good macro is as follows: good
* parameter 1:
,param3=kkk ; Macro call.
; Null string.
* parameter 2:
* parameter 3: kkk
; Null string.
Macro ALPHA demonstrates the specification of an embedded parameter.
MACRO
ALPHA |PARAM
_* FORMAL PARM:
_* EMBEDDED PARM:
ALPHA ENDM
LIST MAC
; EDIT=ON
; Appending a string.
PARAM
ABC_PARAM_DEFG
; Concatenation off at call time.
; End of macro definition.
; List expansion.
The call and expansion of the ALPHA macro is as follows:
ALPHA 1
* FORMAL PARM:
* EMBEDDED PARM:
; Macro call.
1
ABC1DEFG
The assembler processes the embedded parameter in macro ALPHA, as follows:
1. The assembler scans the string to identify the parameter. ABC _ cannot be a parameter because the underscore character is not defined as an identifier character for a parameter.
2. The assembler identifies PARAM as the parameter when the second underscore character is encountered.
S–2314–50
CAL Defined Sequences [9]
3. 1 is substituted for PARAM, producing string ABC_1_DEFG.
4. If editing is enabled, the underscore characters are removed and the resulting string is ABClDEFG.
If editing is disabled, the string is ABC_1_DEFG.
5. The assembler processes the statement.
9.3 Operation Definitions (
OPDEF
)
An operation definition (OPDEF) identifies a sequence of statements to be called later in the source program by an opdef call. Each time the opdef call occurs, the definition sequence is placed into the source program.
Opdefs resemble machine instructions and can be used to define new machine instructions or to redefine current machine instructions. Machine instructions map into opcodes that represent some hardware operation. When an operation is required that is not available through the hardware, an opdef can be written to perform that operation. When the opdef is called, the opdef maps into the opdef definition body and the operation is performed by the defined sequence specified in the definition body.
You can replace any existing machine instruction with an opdef. Although opdef definitions should conform to meaningful operations that are supported by the hardware, they are not restricted to such operations.
The opdef definition sets up the parameters into which the arguments specified in the opdef call are substituted. Opdef parameters are always expressed in terms of registers or expressions. The opdef call passes arguments to the parameters in the opdef definition. The syntax for the opdef definition and the opdef call are identical with two exceptions:
• The complex register has been redefined for the opdef definition prototype statement as follows:
register_mnemonic.register_parameter
• Expressions have been redefined for the opdef definition prototype statement, as follows:
@ [expression_parameter]
S–2314–50 237
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual
238
These two exceptions allow you to specify parameters in the place of registers and expressions for an opdef definition.
The syntax defining a register_parameter and an expression_parameter is case-sensitive. Every character that identifies the parameter in the opdef prototype statement must be identical to every character in the body of the opdef definition. This includes the case (uppercase, lowercase, or mixed case) of each character.
Because the opdef can accept arguments in many forms, it can be more flexible than a macro. Opdefs place a greater responsibility for parsing arguments on the assembler. When a macro is specified, the responsibility for parsing arguments is placed on the user in many cases. Parsing a macro argument can involve numerous micro substitutions, which greatly increase the number of statements required to perform a similar operation with an opdef.
Defined sequences (macros, opdefs, dups, and echos) are costly in terms of assembler efficiency. As the number of statements in a defined sequence increases, the speed of the assembler decreases. This decrease in speed is directly related to the number of statements expanded and the number of times a defined sequence is called.
Limiting the number of statements in a defined sequence improves the performance of the assembler. In some cases, an opdef can perform the same operation as a macro and use fewer statements in the process.
The following example illustrates that an opdef can accept many different kinds of arguments from the opdef call:
MANYCALL OPDEF
A.REG1
A.REG2|A.REG3
; Opdef prototype statement.
S1 A.REG2
S2 A.REG3
A.REG1
S3
MANYCALL ENDM
; OR of registers S1 and S2.
; End of opdef definition.
The following example illustrates the calls and expansions of the previous example:
; First call to opdef MANYCALL.
A1
S1
S2
A2|A3
A.2
A.3
S3 S1|S2
A.1
S3
A.1
A.2|A.3
; OR of registers S1 and S2.
; Second call to opdef MANYCALL.
S–2314–50
S–2314–50
CAL Defined Sequences [9]
S1
S2
A.2
A.3
S3 S1|S2
A.1
S3
ONE =
TWO =
1
2
; OR of registers S1 and S2.
; Define symbols.
THREE = 3
A.ONE A.TWO|A.THREE
; Third call to opdef MANYCALL.
S1
S2
A.2
A.3
S3 S1|S2
A.ONE S3
A1 A.2|A.THREE
S1
S2
A.2
A.3
S3 S1|S2
A.1
s3
; OR of registers S1 and S2.
; Fourth call to opdef MANYCALL.
; OR of registers S1 and S2.
In the first and second calls to opdef MANYCALL, the arguments passed to
REG1 , REG2, and REG3 are 1, 2, and 3, respectively. In the third call to opdef
MANYCALL , the arguments passed to REG1, REG2, and REG3 are ONE, TWO, and
THREE , respectively. The fourth call to opdef MANYCALL demonstrates that the form of the arguments can vary within one call to an opdef if they take a valid form. The arguments passed to REG1, REG2, and REG3 in the fourth call are 1,
2 , and THREE, respectively.
The following example illustrates how to use an opdef to limit the number of statements required in a defined sequence:
.
.
MACRO
$IF
.
$IF ENDM
.
.
.
.
.
.
$IF
$ELSE
.
REG1,COND,REG2 ; Macro prototype statement.
S6,EQ,S.3
; End of macro definition.
; Macro call.
239
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual
240
.
.
$ENDIF
Parsing the parameters (S6,EQ,S3) passed to the definition requires many micro substitutions within the definition body. These micros increase the number of statements within the definition body.
The same function is performed in the following example, but an opdef is specified instead of a macro. In this instance, specifying an opdef rather than a macro reduces the number of statements required for the function.
Because an opdef is called by its form, it is more flexible than a macro in accepting arguments. The opdef expects to be passed two S registers and the EQ mnemonic. You can specify the arguments for the registers in a number of ways and still be recognized as S register arguments by the opdef.
opdef example $if s.reg1,eq,s.reg2 ; Opdef definition statement.
_* Register1: reg1
_* Register2: reg2 example endm list mac
; End of opdef definition.
; Listing expansion.
The following are the calls and expansions of the preceding example: s6,eq,s.3
$if
* Register1: 6
* Register2: 3
If an opdef occurs within the global definitions part of a program segment, it is defined as global. Opdef definitions are local if they occur within a program module (an IDENT, END sequence). A global definition can be redefined locally, but the global definition is re-enabled and the local definition is discarded at the end of the program module. You can reference a global definition anywhere within an assembler program after it has been defined.
You can specify the OPDEF pseudo instruction anywhere within a program segment. If the OPDEF pseudo instruction is found within a definition, it is defined. If the OPDEF pseudo instruction is found within a skipping sequence, it is skipped and is not recognized as a pseudo instruction.
In the following example, the operand and comment fields of the expanded line are shifted two positions to the left (difference between reg and 1): example opdef
S–2314–50
CAL Defined Sequences [9] s.reg
a.reg
example endm list
@exp ; Prototype statement.
@exp ; New machine instruction.
; End of opdef definition.
mac ; Listing expansion.
The following are the calls and expansions of the preceding example: s1 a.1
2
2
; Opdef call.
; New machine instruction.
9.3.1 Opdef Definition
S–2314–50
The OPDEF pseudo instruction is the first statement of an opdef definition.
Although an opdef is constructed much like a macro, an opdef is defined by an opdef statement, not by a name.
Opdef syntax is uniquely defined on the result field alone, in which case, the operand field is not specified or on the result and operand fields. The OPDEF prototype permits up to five subfields within the result and operand fields.
At least one subfield must be present within the result field. No subfields are required in the operand field.
The syntax for each of the subfields within the result and operand fields of the opdef prototype statement is identical. No special syntax forms exist for any of the subfields. The rules that apply for the first subfield in the result field apply to the remainder of the subfields within the result field and to all subfields within the operand field.
The format of the opdef definition is as follows: name
[loc]
OPDEF defsynres
LOCAL defsynop
[name][,[name]]
.
.
.
name ENDM
241
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual
242
The variables in the opdef definition are described as follows:
• name
name identifies the opdef definition and has no association with functionals that appear in the result field of instructions. name must match the name in the label field of the ENDM pseudo instruction, which ends the definition.
• loc
loc specifies an optional label field parameter. loc must meet the requirements
for names as described in Section 6.3, page 76.
• defsynres
defsynres specifies the definition syntax for the result field. It can be one to five subfields specifying a valid result field syntax. The result field must be a symbolic.
Valid result subfields for opdefs can be one of the following:
– Initial register
– Mnemonic
– Initial expression
To specify an initial register on the opdef prototype statement, use one of the following syntax forms for initial-registers:
[prefix][register-prefix] register[register-expression-separatorregister-ending][suffix]
[prefix][register-prefix] register[register-expression-separatorexpression-ending][suffix]
Note: The formal grammar for the Cray X1 assembly language has changed fairly significantly from the Cray PVP grammar. This section has been partially modified to reflect that, but more changes will be made. The intent is that the grammar reflected in the Cray X1 instruction set be generalized for opdefs. A formal grammar will be presented in the near future.
The elements of an initial register definition are as follows:
• prefix
prefix is optional and can be either a left parenthesis or a left bracket ([).
• register-prefix
S–2314–50
S–2314–50
CAL Defined Sequences [9]
register-prefix is optional. It can be specified as any of the following characters:
+ – ~ !
• register
register is required. It can be any simple or complex register. The only simple register is VL.
The complex registers are designated in the opdef definition in the form:
register_designator.register_parameter. The register_designator for complex registers can be any of the following: A C M S T.
The register-parameter is a 1- to 8-character identifier composed of identifier characters.
When you specify a simple register or a complex register mnemonic on an opdef call, it is recognized by the opdef definition without regard to the case
(uppercase, lowercase, or mixed case) in which it was entered.
The optional register-expression-separator can be designated by any of the following:
& | ^ ~ << >> + * / ?
: = <= >=
• register-ending
The optional register-ending is specified using the following syntax form:
register [register-expression-separator register-or-expression]
The register and register-expression-separator elements are described previously under initial-register.
The optional register-or-expression can be a register or an expression. If register is not specified, expression is required. If expression is not specified, register is required.
expression has been redefined for the opdef prototype statement, as
expression-parameter. expression-parameter is an identifier that must begin with the at symbol (@). The @ can be followed by 0 to 7 identifier characters.
• expression-ending
expression-ending is specified as follows:
243
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual
244
expression [register_
expression_separator [register-or-expression]
expression is required and has been redefined for the opdef prototype statement as follows: expression-parameter
expression-parameter is an identifier that must begin with the at symbol (@). The @ can be followed by 0 to 7 identifier characters.
register_expression-separator is defined above.
The optional register-or-expression can be a register or an expression. If register is not specified, expression is required. If expression is not specified, register is required.
A mnemonic is a 1- to 8-character identifier that must begin with a letter (A through Z or a through z), a decimal digit (0 through 9), or one of the following characters: $, %, &, ’,*, +, -, ., /, :, =, ?,, ‘, |, or ~. Optional characters 2 through 8 can be the at symbol (@) or any of the previously mentioned characters.
Mnemonics are case-insensitive.
Initial-expression specifies an initial-expression on the opdef prototype statement, use one of the following syntax forms for initial-expressions:
prefixl [expression-prefix] expression [expression-separator [register-ending]]
[prefix] [expression-prefix] expression [expression-separator [expression-ending]]
expression [expression-separator [register-ending]]
expression [expression-separator [expression-ending]]
The elements of the initial expression are described as follows:
• prefix
prefix is optional and can be either a right parenthesis or a right bracket ([).
• expression-prefix
expression-prefix is optional and can be any of the following:
<, >, #<, or #>
S–2314–50
CAL Defined Sequences [9]
9.3.2 Opdef Calls
S–2314–50
• expression
expression is required and has been redefined for the opdef prototype statement, as follows: expression-parameter
expression-parameter is an identifier that must begin with the at symbol (@). The
@ can be followed from 0 to 7 identifier characters.
• expression-separator
expression-separator is optional and can be one of the following:
), ], &, |, \, <, >, #<, or #>
• register-ending and expression-ending
register-ending and expression-ending are the same for initial expressions as for initial registers.
• defsynop
Definition syntax for the operand field; can be zero, one, or two subfields specifying a valid operand field syntax. If a subfield exists in the result field, the first subfield in the operand field must be a symbolic.
The definition syntax for the operand field of an opdef is the same as the definition syntax for the result field of an opdef. See the definition of defsynres, earlier in this subsection.
An opdef definition is called by an instruction that matches the syntax of the result and operand fields as specified in the opdef prototype statement.
The arguments on the opdef call are passed to the parameters on the opdef prototype statement. The special syntax for registers and expressions that was required on the opdef definition does not extend to the opdef call.
The format of the opdef call is as follows: locarg callsynres callsynop
The variables associated with the opdef call are described as follows:
245
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual
246
• locarg
locarg is an optional label field argument. It can consist of any characters and is terminated by a space (embedded spaces are illegal).
If a label field parameter is specified on the opdef definition, a matching label field parameter can be specified on the opdef call. locarg is substituted wherever the label field parameter occurs in the definition. If no label field parameter is specified in the definition, this field must be empty.
• callsynres
callsynres specifies the result field syntax for the opdef call. It can consist of one, two, or three subfields and must have the same syntax as specified in the result field of the opdef definition prototype statement.
The syntax of the result field call is the same as the syntax of the result field definition with two exceptions. The special syntax rules that are in effect for registers and expressions on the opdef definition do not apply to the opdef call. The syntax for registers and expressions used on the opdef call is the same as the syntax for registers and expressions.
The subfields in the result field on the opdef call can be specified with one of the following:
– Initial-register
– Mnemonic
– Initial-expression
For a description of the syntax for the result field of the opdef call, see the syntax for the result field of the opdef definition.
• callsynop
callsynop specifies the operand field syntax for the opdef call. It can consist of zero, one, two, or three subfields, and it must have the same syntax as specified in the operand field of the opdef definition prototype statement.
The syntax of the operand field call is the same as the syntax of the operand field definition with two exceptions. The special syntax rules that are in effect for registers and expressions on the opdef definition do not apply to the opdef call. The syntax for registers and expressions used on the opdef call is the same as the syntax for registers and expressions.
The subfields in the operand field on the opdef call can be specified with one of the following:
S–2314–50
S–2314–50
CAL Defined Sequences [9]
– Initial-register
– Mnemonic
– Initial-expression
For a description of the syntax for the operand field of the opdef call, see the syntax for the result field of the opdef definition.
The following rules apply for opdef calls:
• The character strings callsynres and callsynop must be exactly as specified in the opdef definition.
• An expression must appear whenever an expression in the form @ exp is indicated in the prototype statement. The actual argument string is substituted in the definition sequence wherever the corresponding formal parameter @ exp occurs.
• The actual argument string consisting of a complex-register mnemonic followed by a period (.) followed by a register-parameter. A register-designator followed by a register-parameter must appear wherever the register-designator
A.
register-parameter, B. register-parameter, SB. register-parameter, S.
register-parameter, T. register-parameter, ST. register-parameter, SM.
register-parameter, or V. register-parameter, respectively, appeared in the prototype statement.
– If the register-parameter is of the form octal-integer, the actual argument is the octal-integer part. The octal-integer is restricted to 4 octal digits.
– If the register-parameter is of the form . integer-constant or . symbol, the actual argument is an integer-constant or a symbol.
The following opdef definition shows a scalar floating-point divide sequence: fdv opdef ; Scalar floating-point divide
; prototype statement.
L
L s.r1
errif errif s.r1
s.r2
s.r3
s.r1
fdv endm s.r2/fs.r3
r1,eq,r2 r1,eq,r3
/hs.r3
s.r2*fs.r1
s.r3*is.r1
s.r2*fs.r3
247
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual
The following example illustrates the opdef call and expansion of the preceding example: a ; Divide s3 by s2, result to s4.
a s4 s3/fs2 errif 4,eq,3 errif 4,eq,2 s.4
/hs.2
s.3
s.3*fs.4
s.2
s.2*is.4
s.4
s.3*fs.2
The following opdef definition, call, and expansion define a conditional jump where a jump occurs if the A register values are equal:
JEQ OPDEF
L
L
_*
JEQ
A0
JAZ
JEQ ENDM
A.A1,A.A2,@TAG ; Opdef prototype statement.
A_A1-A_A2
@TAG ; Expression is expected.
; End of opdef definition.
LIST MAC ; Listing expansion.
The following example illustrates the opdef call and expansion of the preceding example (the expansion starts on line 2.):
; Opdef call.
*
JEQ A3,A6,GO
A0 A3-A5
JAZ GO ; Expression is expected.
The opdef in the following example illustrates how an opdef can redefine an existing machine instruction:
EXAMPLE OPDEF
S.REG
A.REG
EXAMPLE ENDM
LIST MAC
@EXP ; Opdef protype instruction.
@EXP ; New instruction.
; End of opdef definition.
; Listing expansion.
The following example illustrates the opdef call and expansion of the preceding example:
S1 2
A.1
2
; Opdef call.
; New instruction.
248 S–2314–50
S–2314–50
CAL Defined Sequences [9]
The following example demonstrates how the expansion of an opdef is affected when the opdef call does not include a label that was specified in the opdef definition: regchg opdef lbl s.reg1 s.reg2
lbl = * s.reg2 s.reg1
regchg endm list mac
; Opdef prototype statement.
; Left-shift if lbl is left off.
; Register s2 gets register s1.
; End of opdef definition.
; Listing expansion.
The following example illustrates the opdef call and expansion of the preceding example: s1 s2
= * s.2
s.1
; Opdef call.
; Left-shift if lbl is left off.
; Register s2 gets register s1.
The label field parameter was omitted on the opdef call in the previous example.
The result and operand fields of the first line of the expansion were shifted left three character positions because a null argument was substituted for the
3-character parameter, lbl.
If the old format is used, only one space appears between the label field parameter and result field in the macro definition. If a null argument is substituted for the location parameter, the result field is shifted into the label field in column 2. Therefore, at least two spaces should always appear between a parameter in the label field and the first character in the result field in a definition.
If the new format is used, the result field is never shifted into the label field.
The following example illustrates the case insensitivity of the register and register-prefix:
CASE OPDEF
S1
.
.
.
CASE ENDM
#Pa2 ; Prototype statement.
The following example illustrates the opdef calls of the preceding example:
S1 #pa2
S1 #Pa2
; Recognized by CASE.
; Recognized by CASE.
249
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual
S1 #pA2
S1 #PA2 s1 #pa2 s1 #Pa2 s1 #pA2 s1 #PA2
; Recognized by CASE.
; Recognized by CASE.
; Recognized by CASE.
; Recognized by CASE.
; Recognized by CASE.
; Recognized by CASE.
9.4 Duplication (
DUP
)
The DUP pseudo instruction defines a sequence of code that is assembled repetitively immediately following the definition. The sequence of code is assembled the number of times specified on the DUP pseudo instruction. The sequence of code to be repeated consists of the statements following the DUP pseudo instruction and any optional LOCAL pseudo instructions. Comment statements are ignored. The sequence to be duplicated ends when the statement count is exhausted or when an ENDDUP pseudo instruction with a matching label field name is encountered.
The DUP pseudo instruction only accepts one type of formal parameter. That parameter must be specified with the LOCAL pseudo instruction.
You can specify the DUP pseudo instruction anywhere within a program segment.
If the DUP pseudo instruction is found within a definition, it is defined and is not recognized as a pseudo instruction. If the DUP pseudo instruction is found within a skipping sequence, it is skipped and is not recognized as a pseudo instruction.
The format of the DUP pseudo instruction is as follows:
[dupname] DUP expression[,[count]]
The variables associated with the DUP pseudo instruction are described as follows:
• dupname
dupname specifies an optional name for the dup sequence. It is required if the count field is null or missing. If no count field is present, dupname must match an ENDDUP name. The sequence field in the DUP pseudo instruction itself represents the nested dup level and appears in columns 89 and 90 on the listing. For a description of sequence field nest level numbering,
250 S–2314–50
S–2314–50
CAL Defined Sequences [9]
The dupname variable must meet the requirements for names as described in
• expression
expression is an absolute expression with a positive value that specifies the number of times to repeat the code sequence. All symbols, if any, must be defined previously. If the current base is mixed, octal is used for the expression. If the value is 0, the code is skipped. You can use a STOPDUP to override the given expression.
The expression operand must meet the requirements for expressions as
described in Section 6.9, page 94.
• count
count is an optional absolute expression with positive value that specifies the number of statements to be duplicated. All symbols (if any) must be defined previously. If the current base is mixed, octal is used for the expression.
LOCAL pseudo instructions and comment statements (* in column 1) are ignored for the purpose of this count. Statements are counted before expansion of nested macro or opdef calls, and dup or echo sequences.
The count operand must meet the requirements for expressions as described in
In the following example, the code sequence following the DUP pseudo instruction will be repeated 3 times. There are 5 statements in the sequence.
DUP 3,5
LOCAL SYM1,SYM2
*Asterisk comment; not counted
S1 1
*Asterisk comment; not counted
INCLUDE ALPHA
; LOCAL pseudo instruction not counted.
; First statement is definition.
; INCLUDE pseudo instruction not
; counted.
The following is the file, ALPHA:
S2 3
S4 4
*Asterisk comment
S5
S6
5
6
; Second statement in definition.
; Third statement in definition.
; not counted
; Fourth statement in definition.
; Fifth statement in definition.
251
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual
In the following example, the two con pseudo instructions are duplicated three times immediately following the definition: list example dup con con example enddup
1
2 dup
3 ; Definition.
The following example illustrates the expansion of the preceding example: con 1 con 2 con 1 con 2 con 1 con 2
9.5 Duplicate with Varying Argument (
ECHO
)
The ECHO pseudo instruction defines a sequence of code that is assembled zero or more times immediately following the definition. On each repetition, the actual arguments are substituted for the formal parameters until the longest argument list is exhausted. Null strings are substituted for the formal parameters after shorter argument lists are exhausted. The echo sequence to be repeated consists of statements following the ECHO pseudo instruction and any optional LOCAL pseudo instructions. Comment statements are ignored. The echo sequence ends with an ENDDUP that has a matching label field name.
You can use the STOPDUP pseudo instruction to override the repetition count determined by the number of arguments in the longest argument list.
You can specify the ECHO pseudo instruction anywhere within a program segment. If the ECHO pseudo instruction is found within a definition, it is defined and is not recognized as a pseudo instruction. If the ECHO pseudo instruction is found within a skipping sequence, it is skipped and is not recognized as a pseudo instruction.
The format of the ECHO pseudo instruction is as follows: dupname ECHO [name=argument],[[name=]argument]
252 S–2314–50
S–2314–50
CAL Defined Sequences [9]
The variables associated with the ECHO pseudo instruction are described as follows:
• dupname
dupname specifies the required name of the echo sequence. It must match the label field name in the ENDDUP instruction that terminates the echo sequence.
dupname must meet the requirements for names as described in Section 6.3, page 76.
• name
name specifies the formal parameter name. It must be unique. There can be none, one, or more formal parameters. name must meet the requirements for
names as described in Section 6.3, page 76.
• argument
argument specifies a list of actual arguments. The list can be one argument or a parenthesized list of arguments.
A single argument is any ASCII character up to but not including the element separator, a space, a tab (new format only), or a semicolon (new format only).
The first character cannot be a left parenthesis.
A parenthesized list can be a list of one or more actual arguments. Each actual argument can be one of the following:
– An ASCII character string can contain embedded arguments. If, however, an ASCII string is intended, the first character in the string cannot be a left parenthesis. A legal ASCII string is 4(5). An illegal ASCII string is (5)4(5).
– A null argument; an empty ASCII character string.
– An embedded argument that contains a list of arguments enclosed in matching parentheses. An embedded argument can contain blanks or commas and matched pairs of parentheses. The outermost parentheses are always stripped from an embedded argument when an echo definition is expanded.
An embedded argument must meet the requirements for embedded arguments
In the following example, the ECHO pseudo instruction is expanded twice immediately following the definition:
LIST
EXAMPLE ECHO
DUP
PARAM1=(1,3),PARAM2=(2,4)
253
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual
CON
CON
; Definition.
PARAM1
; Gets 1 and 3.
PARAM2
; Gets 2 and 4.
EXAMPLE ENDDUP
The following example illustrates the expansion of the preceding example:
CON 1
CON 2
CON 3
CON 3
CON 4
; Gets 1 and 3.
; Gets 2 and 4.
; Gets 1 and 3.
; Gets 1 and 3.
; Gets 2 and 4.
In the following example, the echo pseudo instruction is expanded once immediately following the definition with two null arguments.
list example echo
_*Parameter 1 is:
_*Parameter 2 is: example enddup dup param1=,param2=()
; ECHO with two null parameters.
’param1’
’param2’
The following illustrates the expansion of the preceding example:
*Parameter 1 is:
*Parameter 2 is:
’’
’’
9.6 Ending a Macro or Operation Definition (
ENDM
)
An ENDM pseudo instruction terminates the body of a macro or opdef definition.
If ENDM is used within a MACRO or OPDEF definition with a different name, it has no effect.
You can specify the ENDM pseudo instruction only within a macro or opdef definition. If the ENDM pseudo instruction is found within a skipping sequence, it is skipped and is not recognized as a pseudo instruction.
The format of the ENDM pseudo instruction is as follows: func ENDM ignored
254 S–2314–50
CAL Defined Sequences [9]
The func variable associated with the ENDM pseudo instruction identifies the name of the macro or opdef definition sequence. It must be a valid identifier or the equal sign. func must match the name that appears in the result field of the macro prototype or the label field name in an OPDEF instruction.
If the ENDM pseudo instruction is encountered within a definition but func does not match the name of an opdef or the name of a macro, the ENDM instruction is defined and does not terminate the opdef or macro definition in which it is found. func must meet the requirements for functionals.
9.7 Premature Exit from a Macro Expansion (
EXITM
)
The EXITM pseudo instruction immediately terminates the innermost nested macro or opdef expansion, if any, caused by either a macro or an opdef call. If files were included within this expansion and/or one or more dup or echo expansions are in progress within the innermost macro or opdef expansion they are also terminated immediately. If such an expansion does not exist, the EXITM pseudo instruction issues a caution level listing message and does nothing.
You can specify the EXITM pseudo instruction anywhere within a program segment. If the EXITM pseudo instruction is found within a definition, it is defined and is not recognized as a pseudo instruction. If the EXITM pseudo instruction is found within a skipping sequence, it is skipped and is not recognized as a pseudo instruction.
The format of the EXITM pseudo instruction is as follows: ignored EXITM ignored
S–2314–50
In the following example of a macro call, the macro expansion is terminated immediately by the EXITM pseudo instruction and the second comment is not included as part of the expansion: macro alpha
_*First comment exitm
_*Second comment alpha endm list mac
The following example illustrates the expansion of the preceding example:
255
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual alpha
*First comment exitm
; Macro call
9.8 Ending Duplicated Code (
ENDDUP
)
The ENDDUP pseudo instruction ends the definition of the code sequence to be repeated. An ENDDUP pseudo instruction terminates a dup or echo definition with the same name. If ENDDUP is used within a DUP or ECHO definition with a different label field name, it has no effect. ENDDUP has no effect on a dup definition terminated by a statement count; in this case, ENDDUP is counted.
The ENDDUP pseudo instruction is restricted to definitions (DUP or ECHO). If the ENDDUP pseudo instruction is found on a MACRO or OPDEF definition, it is defined and is not recognized as a pseudo instruction. If the ENDDUP pseudo instruction is found within a skipping sequence, it is skipped and is not recognized as a pseudo instruction.
The format of the ENDDUP pseudo instruction is as follows:
dupname ENDDUP ignored
The dupname variable associated with the ENDDUP pseudo instruction specifies the required name of a dup sequence. dupname must meet the requirements for
names as described in Section 6.3, page 76.
9.9 Premature Exit of the Current Iteration of Duplication Expansion (
NEXTDUP
)
The NEXTDUP pseudo instruction stops the current iteration of a duplication sequence indicated by a DUP or an ECHO pseudo instruction. Assembly of the current repetition of the dup sequence is terminated immediately and the next repetition, if any, is begun.
Assembly of the current iteration of the innermost duplication expansion with a matching label field name is terminated immediately. If the label field name is not present, assembly of the current iteration of the innermost duplication expansion is terminated immediately.
If other dup, echo, macro, or opdef expansions were included within the duplication expansion to be terminated, these expansions are also terminated immediately. If a file also is being included at expansion time within the duplication expansion it is terminated immediately.
256 S–2314–50
CAL Defined Sequences [9]
You can specify the NEXTDUP pseudo instruction anywhere within a program segment. If the NEXTDUP pseudo instruction is found within a definition, it is defined and is not recognized as a pseudo instruction. If the NEXTDUP pseudo instruction is found within a skipping sequence, it is skipped and is not recognized as a pseudo.
The format of the NEXTDUP pseudo instruction is as follows:
[dupname] NEXTDUP ignored
The optional dupname variable specifies the name of a dup sequence. If the name is present but does not match any existing duplication expansion, a caution-level listing message is issued and the pseudo instruction does nothing. If the name is not present and a duplication expansion does not currently exist, a caution-level listing message is issued and the pseudo instruction does nothing.
9.10 Stopping Duplication (
STOPDUP
)
The STOPDUP pseudo instruction stops duplication of a code sequence indicated by a DUP or ECHO pseudo instruction. STOPDUP overrides the repetition count.
Assembly of the current dup sequence is terminated immediately. STOPDUP terminates the innermost dup or echo sequence with the same name as found in the label field. If no label field name exists, STOPDUP will terminate the innermost dup or echo sequence. STOPDUP does not affect the definition of the code sequence that will be duplicated.
Assembly of the innermost duplication expansion with a matching label field name is terminated immediately; however, if the label field name is not present, assembly of the innermost duplication expansion is terminated immediately. If other dup, echo, macro, or opdef expansions were included within the duplication expansion that will be terminated, these expansions also are terminated immediately. If a file also is being included at expansion time within the duplication expansion that will be terminated, the inclusion of that file is terminated immediately.
You can specify the STOPDUP pseudo instruction anywhere within a program segment. If the STOPDUP pseudo instruction is found within a definition, it is defined and is not recognized as a pseudo instruction. If the STOPDUP pseudo instruction is found within a skipping sequence, it is skipped and is not recognized as a pseudo instruction.
S–2314–50 257
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual
258
The format of the STOPDUP pseudo instruction is as follows:
[dupname] STOPDUP ignored
The dupname variable associated with the STOPDUP pseudo instruction specifies the name of a dup sequence. If the name is present but does not match any existing duplication expansion, or, if the name is not present and a duplication expansion does not currently exist, a caution-level listing message is issued and the pseudo instruction does nothing. dupname must meet the requirements for
names as described in Section 6.3, page 76.
The following example uses a DUP pseudo instruction to define an array with values 0, 1, and 2:
S = W*
DUP 3,1
CON W.*-S
The following illustrates the expansion of the preceding example:
CON W.*-S
CON W.*-S
CON W.*-S
In the following example the ECHO and DUP pseudo instructions define a nested duplication:
ECHO ECHO
I SET
DUPI DUP
JK SET
DUPJK DUP
RI.I
JK SET
DUPJK ENDDUP
I SET
DUPI ENDDUP
ECHO ENDDUP
8
0
RI=(A,S),RJK=(B,T)
0
64
RJK.JK
JK+1
I+1
Note: The following expansion is not generated by the assembler, but it is included to show the expansion of the previously nested duplication expansion.
S–2314–50
S–2314–50
CAL Defined Sequences [9]
The following example illustrates the expansion of the preceding example:
.
.
A.0
.
.
.
.
B.0
A.0
B.64
.
.
.
.
.
.
A.1
B.0
.
.
.
.
.
.
A.7
B.64
.
.
S.0
.
.
.
.
T.O
.
.
S.0
T.64
.
.
S.8
.
.
T.64
; In the first call of the echo, the A
; and B parameters are used.
; DUPJK generates the A.0 gets register
; B.0 through register A.0 gets register
; B.64 instructions.
; DUPI increments the A register from
; A.1 to A.7 for succeeding passes
; through DUPJK.
; DUPJK generates register A.
i gets
; register B.0 through register A.
i gets
; register B.64 instructions.
; i is l to 7.
; In the second expansion of the echo
; pseudo instruction the S and T
; parameters are used.
; DUPJK and DUPI generate the same
; series of register instructions for
; the S and T registers that were
; generated for the A and B registers.
In the following example the STOPDUP pseudo instruction terminates duplication:
LIST
T SET
A DUP
T SET
IFE
A STOPDUP
CON
A ENDDUP
DUP
0
1000
T+1
T,EQ,3,1
T
; Terminate duplication when T=3.
The following example illustrates the expansion of the preceding example:
T
T
SET
CON
SET
T+1
T
T+1
259
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual
T
A
CON
SET
STOPDUP
T
T+1
In the following example a STOPDUP pseudo instruction is used to terminate a DUP immediately:
3 DNAME DUP
_* First comment
STOPDUP
_* Second comment
DNAME ENDDUP
The following example illustrates the expansion of the preceding example:
* First comment
STOPDUP
The following example is similar to the previous example except NEXTDUP replaces STOPDUP. The current iteration is terminated immediately when the
NEXTDUP pseudo instruction is encountered.
3 DNAME DUP
_* First comment
NEXTDUP
_* Second comment
DNAME ENDDUP
The following example illustrates the expansion of the preceding example:
* First comment
NEXTDUP
* First comment
NEXTDUP
* First comment
NEXTDUP
9.11 Specifying Local Unique Character String Replacements (
LOCAL
)
The LOCAL pseudo instruction specifies unique character string replacements within a program segment that are defined only within the macro, opdef, dup, or echo definition. These character string replacements are known only in the macro, opdef, dup, or echo at expansion time. The most common usage of the LOCAL pseudo instruction is for defining symbols, but the LOCAL
260 S–2314–50
S–2314–50
CAL Defined Sequences [9] pseudo instruction is not restricted to the definition of symbols. Local pseudo instructions within a macro, opdef, dup, or echo header are not part of the macro definition.
On each macro or opdef call and each repetition of a dup or echo definition sequence, the assembler creates a unique 8-character string (commonly used for the definition of symbols by the user) for each local parameter and substitutes the created string for the local parameter on each occurrence within the definition.
The unique character string created for local parameters has the form %% nnnnnn; where n is a decimal digit.
Zero or more LOCAL pseudo instructions can appear in the header of a macro, opdef, dup, or echo definition. The LOCAL pseudo instructions must immediately follow the macro or opdef prototype statement or DUP and ECHO pseudo instructions, except for intervening comment statements.
You can specify the LOCAL pseudo instruction only within a definition. If the
LOCAL pseudo instruction is found within a skipping sequence, it is skipped and is not recognized as a pseudo instruction.
The format of the LOCAL pseudo instruction is as follows: ignored LOCAL [name][,[name]]
The name variable associated with the LOCAL pseudo instruction specifies formal parameters that must be unique and will be rendered local to the definition. name
must meet the requirements for names as described in Section 6.3, page 76.
The following example demonstrates that all formal parameters must be unique:
.
.
MACRO
UNIQUE PARM2 ; PARM2 is defined within UNIQUE.
LOCAL PARM1,PARM2 ; ERROR: PARM2 previously defined as a
.
.
; parameter in the macro prototype
; statement.
.
.
UNIQUE ENDM
The following example demonstrates how a unique character string is generated for each parameter defined by the LOCAL pseudo instruction: macro string local param1,param2 ; Not part of the definition body.
261
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual param1 = 1 s1 param1 ; Register s1 gets the value defined by
; param1.
param2 = 2 s2 param2 ; Register s2 gets the value defined by
; param2.
string endm list mac
; End of macro definition.
; Listing expansion.
The following example illustrates the call and expansion from the preceding example: string
%%262144 = s1
%%131072 = s2
; Macro call.
1
%%262144
; Register s1 gets the value defined by
; param1.
2
%%131072
; Register s2 gets the value defined by
; param2.
The call to the macro string generates unique strings for param1 (%%262144) and for param2 (%%131072).
9.12 Synonymous Operations (
OPSYN
)
The OPSYN pseudo instruction defines an operation that is synonymous with another macro or pseudo instruction operation. The name in the label field is defined as being the same as the name in the operand field. You can redefine any pseudo instruction or macro in this manner.
The name in the label field can be a currently defined macro or pseudo instruction in which case, the current definition is replaced and a message is issued informing you that a redefinition has occurred.
An operation defined by OPSYN is global if the OPSYN pseudo instruction occurs within the global part of an assembler segment, and it is local if the OPSYN pseudo instruction appears within an assembler module of a segment. You can reference global operations in any program segment following the definition.
Every local operation is removed at the end of a program module, making any previous global definition with the same name available again.
262 S–2314–50
S–2314–50
CAL Defined Sequences [9]
If the OPSYN pseudo instruction occurs within a definition, it is defined and is not recognized as a pseudo instruction. If the OPSYN pseudo instruction is found within a skipping sequence, it is skipped and is not recognized as a pseudo instruction.
The format of the OPSYN pseudo instruction is as follows:
func1 OPSYN [func2]
The func1 variable associated with the OPSYN pseudo instruction specifies a required name. It must be a valid name. The name of a defined operation such as a pseudo instruction or macro, or the equal sign. func 1 must not be blank and must meet the requirements for names.
The func2 variable specifies an optional name. It must be the name of a defined operation or the equal sign. If func 2 is blank, func 1 becomes a do-nothing pseudo instruction.
In the following example, the macro definition includes the OPSYN pseudo instruction that redefines the IDENT pseudo instruction:
IDENTT OPSYN
MLEVEL
IDENT
ERROR ; Eliminates the warning error that is
; issued because the IDENT pseudo
; instruction is redefined.
MACRO
IDENT
LIST
NAME LIST
IDENTT
IDENT ENDM
NAME
LIS,OFF,NXRF
LIS,ON,XRF
; Processed if LIST=NAME on CAL control
; statement.
NAME
The following example illustrates the OPSYN call and expansion (The expansion starts on line 2.):
IDENT A
LIST LIS,OFF,NXRF
A LIST LIS,ON,XRF ; Processed if LIST=NAME on CAL control
; statement.
IDENTT A
263
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual
In the following example, the first macro illustrates that a functional can be redefined many times: macro first s1 s2 s3 first endm second opsyn
1
2 s1+2 third opsyn first
; second is the same as first.
second
; third is the same as second.
The following example includes the opdef calls and expansions from the preceding example:
; Macro call.
first s1 s2 s3 second s1 s2 s3 third s1 s2 s3
1
2 s1+s2 s1
1
2 s1+s2 s1
1
2 s1+s2
In the following example, the functional EQU is defined to perform the same operation as =:
EQU OPSYN = ; EQU is defined to
; perform the
; operation that the
; = pseudo
; instruction
; performs.
264 S–2314–50
S–2314–50
Calling Conventions [A]
Cray specifies a calling convention that the CAL programmer must follow to properly interface with Cray-provided libraries and compiler-generated code.
Note: (Deferred implementation) This appendix provides a brief overview of the calling conventions. Publication of the complete calling conventions is deferred.
To support the calling conventions, the assembler recognizes the following symbolic register designators: a.SP == a63 a.FP == a62 a.EA == a61 a.RA == a60
Register a.SP contains the stack pointer. It is always aligned to 8 bytes. The stack grows from higher addresses to lower addresses, and a.SP contains the address of the highest unused aligned Longword.
Register a.FP, contains the frame pointer. It is set from a.SP on entry to a subprogram and replaced in a.SP on exit. The old value of a.FP is saved to the new stack frame at offset -0x0 from the new a.FP.
Register a.EA, must contain the subprogram’s entry address on entry to a subprogram. It is saved to the new stack frame at offset -0x8 from the new a.FP
.
Register a.RA must contain the subprogram call’s return address on entry to a subprogram. It is saved to the new stack frame at offset -0x10 from the new a.FP
.
Arguments are passed in a1:a16, s1:s16, and in memory at the end of the caller’s stack frame, where an argument passing area of at least 128 bytes is always allocated. The first sixteen argument words are passed in registers, with floating-point value arguments in scalar registers and all others, including arguments by reference, in address registers. Register arguments are not compacted, so a subprogram expecting to receive an int, a float, and an int would take them from a1, s2, and a3. Additional argument words after the sixteenth must be passed in memory. Each argument occupies at least a Longword. Arguments less than 64 bits wide are right-justified and sign-extended.
265
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual
Return values <= 128 bytes long are returned in registers. Scalar registers beginning with s1 are used to return floating-point results. Other types <= 128 bytes long are returned in address registers beginning with a1. Return values >
128 bytes long are returned in a properly aligned memory area which is allocated by the caller and whose address is passed as a "hidden" initial argument.
Mask register m0 must have all bits set on entry to and exit from a subprogram.
Subroutine callers can assume that a32-a58, a62 (a.FP), s32-s63, and m0 are preserved across the call.
The compilers can place useful information about the subprogram prior to its instruction text, such as its name and source code location. The format of this information is compiler-dependent.
MSP codes have additional conventions not described here. Vector functions, also known as VFUNCTIONs, have alternative calling conventions.
266 S–2314–50
ASDEF Macros and Opdefs [B]
(Deferred implementation)
S–2314–50 267
CR
SO
SI
DLE
HT
LF
VT
FF
DC1
DC2
DC3
DC4
NAK
SYN
ETX
EOT
ENQ
ACK
BEL
BS
Character
NUL
SOH
STX
S–2314–50
Character Set [C]
Table 19 lists the character sets supported by CAL.
Table 19. Character Set
011/09
012/0A
013/0B
014/0C
015/0D
016/0E
017/0F
020/10
ASCII code (octal/hex)
000/00
001/01
002/02
003/03
004/04
005/05
006/06
007/07
010/08
021/11
022/12
023/13
024/14
025/15
026/16
269
/
0
-
.
1
2
+
,
)
*
’
(
%
&
"
#
$
RS
US
Space
!
SUB
ESC
FS
GS
Character
ETB
CAN
EM
270
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual
045/25
046/26
047/27
050/28
051/29
052/2A
053/2B
054/2C
055/2D
056/2E
057/2F
060/30
061/31
062/32
ASCII code (octal/hex)
027/17
030/18
031/19
032/1A
033/1B
034/1C
035/1D
036/1E
037/1F
040/20
041/21
042/22
043/23
044/24
S–2314–50
K
L
I
J
M
N
G
H
E
F
C
D
A
B
<
=
:
;
>
?
@
8
9
6
7
4
5
Character
3
S–2314–50
101/41
102/42
103/43
104/44
105/45
106/46
107/47
110/48
111/49
112/4A
113/4B
114/4C
115/4D
116/4E
ASCII code (octal/hex)
063/33
064/34
065/35
066/36
067/37
070/38
071/39
072/3A
073/3D
074/3C
075/3D
076/3E
077/3F
100/40
Character Set [C]
271
g h e f i j c d a b
_
’
]
^
X
Y
V
W
Z
[
T
U
R
S
P
Q
Character
O
272
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual
135/5D
136/5E
137/5F
140/69
141/61
142/62
143/63
144/64
145/65
146/66
147/67
150/68
151/69
152/6A
ASCII code (octal/hex)
117/4F
120/50
121/51
122/52
123/53
124/54
125/55
126/56
127/57
130/58
131/59
132/5A
133/5B
134/5C
S–2314–50
{
} y z
|
~
DEL t u r s v w x p q n o l m
Character k
S–2314–50
ASCII code (octal/hex)
153/6B
154/6C
155/6D
156/6E
157/6F
160/70
161/71
162/72
163/73
164/74
165/75
166/76
167/77
170/78
171/79
172/7A
173/7B
174/7C
175/7D
176/7E
177/7F
Character Set [C]
273
S–2314–50
Glossary
absolute address
1. A unique, explicit identification of a memory location, a peripheral device, or a location within a peripheral device. 2. A precise memory location that is an actual address number rather than an expression from which the address can be calculated. See also absolute memory address.
application node
A node that is used to run user applications. Application nodes are best suited for executing parallel applications and are managed by the strong application placement scheduling and gang scheduling mechanism psched. See also OS
node; support node.
argument
1. A value that determines the result or behavior of a software entity, such as a program or function. 2. A character string in a command line that supplies information for the preceding option. Sometimes called option-argument.
array
A data structure that contains a series of related data items arranged in rows and columns for convenient access. The C shell and the awk(1) command can store and process arrays. A Fortran array is a set of scalar data, all of the same type and type parameters. The rank of an array is at least 1 and at most 7. Arrays may be used as expression operands, procedure arguments, and function results, and they may appear in input/output (I/O) lists.
assemble
To prepare a machine language program from a symbolic language program, such as a Cray Assembly Language (CAL) program, by substituting machine code for symbolic code, and absolute or relocatable addresses for symbolic addresses. See also compiler and interpreter.
assembler
A computer program that creates a machine language program from a symbolic language program. The assembler substitutes machine operation codes for symbolic operation codes and substitutes absolute or relocatable addresses for symbolic instructions. See also compiler and interpreter.
275
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual assembly language
A programming language that uses symbolic notation to represent machine language instructions. It is a low-level language that is closely related to the internal architecture of the computer on which it runs. See also high-level programming language.
association
An association permits an entity to be referenced by different names in a scoping unit or by the same or different names in different scoping units. Several kinds of associations exist. The principal kinds of association are pointer association, argument association, host association, use association, and storage association.
See also scoping unit.
atomic
Not interruptible. An atomic operation occurs as a logical unit.
barrier
An obstacle within a program that provides a mechanism for synchronizing tasks. When a task encounters a barrier, it must wait until all specified tasks reach the barrier.
big-endian
A format for storage or transmission of binary data in which the most significant bit (or byte) has the lowest address (the word is stored big-end-first). The reverse convention is little-endian.
cache
In a processing unit, a high-speed buffer that is continually updated to contain recently accessed contents of memory. Its purpose is to reduce access time.
cache line
A division of cache. Each cache line can hold multiple data items. On the Cray
X1 system, a cache line is 32 bytes, which is the maximum size of a hardware message.
276 S–2314–50
S–2314–50
Glossary cache pollution
A delay that occurs when data that will not be used before it is evicted is loaded into cache, potentially displacing more useful data.
CPU
Processor. See MSP and SSP.
deferred implementation
The label used to introduce information about a feature that will not be implemented until a later release.
designator
Sometimes it is convenient to reference only part of an object, such as an element or section of an array, a substring of a character string, or a component of a structure. This requires the use of the name of the object followed by a selector that selects a part of the object. A name followed by a selector is called a designator.
distributed memory
1. Memory in which each processor has a separate share of the total memory. 2.
Memory that is physically distributed among several modules.
entry point
A location in a program or routine at which execution begins. A routine may have several entry points, each serving a different purpose. Linkage between program modules is performed when the linkage editor binds the external references of one group of modules to the entry points of another module.
environment
1. The set of hardware and software products on which an application is being run. 2. A set of values supported by the shell used to pass information between processes (specifically, from parent to child).
environment variable
A variable that stores a string of characters for use by your shell and the processes that execute under the shell. Some environment variables are predefined by the shell, and others are defined by an application or user.
277
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual
278
Shell-level environment variables let you specify the search path that the shell uses to locate executable files, the shell prompt, and many other characteristics of the operation of your shell. Most environment variables are described in the
ENVIRONMENT VARIABLES section of the man page for the affected command.
explicit communication
A programming style in which communication is language independent and in which communication occurs through library calls. The message-passing and explicit shared memory programming methods use products such as MPI or
SHMEM. See also implicit communication.
gather/scatter
An operation that copies data between remote and local memory or within local memory. A gather is any software operation that copies a set of data that is nonsequential in a remote (or local) processor, usually storing into a sequential (contiguous) area of local processor memory. A scatter copies data from a sequential, contiguous area of local processor memory) into nonsequential locations in a remote (or local) memory.
implicit communication
A programming style in which most communication is recognized implicitly by the compiler and run-time libraries from the Fortran source (including directives). The data-sharing and work-sharing programming methods use implicit communication. Also called global addressing or implicit shared memory. See explicit communication.
load
To create a binary executable file (an executable) from a binary relocatable object file (the object). This process adds library subprograms to the object and resolves the external references among subprograms. Executable files and the libraries and data they access are loaded into memory during the load step. Links are created among modules that must access each other. The command that performs a load is called a link-edit loader, or simply a loader.
loader
A generic term for the system software product that loads a compiled or assembled program into memory and prepares it for execution.
S–2314–50
S–2314–50
Glossary loader table
The form in which code is presented to the loader. Loader tables, which are generated by compilers and assemblers according to loader requirements, contain information required for loading. This information includes the type of code; names, types, and lengths of storage blocks; data to be stored; and so on.
mask
A machine word that specifies the parts of another machine word on which to operate. Bit masks are commonly used with the AND, OR, and exclusive OR instructions.
mode
1. A method of operation (for example, binary mode). 2. The permissions for a file (read, write, and execute for user, group,and other). The permissions mode is referred to by either an octal number (absolute mode) or a sequence of characters
(relative mode). The mode is used in conjunction with chmod to change permissions of files. 3. Processing mode (either background or foreground).
module file
A metafile that defines information specific to an application or collection of applications. (This term is not related to the module statement of the Fortran language; it is related to setting up the Cray X1 system environment.) For example, to define the paths, command names, and other environment variables to use the Cray Programming Environment, you use the module file PrgEnv, which contains the base information needed for application compilations. The module file mpt sets a number of environment variables needed for message passing and data passing application development.
Modules
Note: This term is not related to the module statement of the Fortran language, but is related to setting up the environment of the X1 system so a Cray product can be used. A package on the Cray X1 system that allows you to dynamically modify your user environment by using module files. The user interface to this package is the module command. The module command provides a number of capabilities to the user, including loading a module file, unloading a module file, listing which module files are loaded, determining which module files are available, and others. See also module file.
279
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual node
The configurable scalable building block for a Cray X1 mainframe. The actual hardware contents of a node are housed in four multichip modules (MCMs). This is the conceptual or software configuration view of a hardware unit called a node module. Physically, all nodes are the same; software controls how a node is used, such as for an OS node, application node, or support node. See also application
node; MCM, MSP, node module; OS node; SSP; support node.
node module
The physical node in a Cray X1 system.
operator
1. A symbolic expression that indicates the action to be performed in an expression; operator types include arithmetic, relational, and logical.
OS node
The node that provides kernel-level services, such as system calls, to all support nodes and application nodes. See also application node; node; support node.
program counter
A hardware element that contains the address of the instruction currently executing.
satisfy
A term used when describing adder operations. In an additive adder, 0 plus 0 equals 0; a 0 result in a bit position is a satisfy condition. It is so named because it prevents a carry up from propagating further, thus satisfying it.
scalar instruction
An instruction that requires the use of a scalar, rather than a vector, register.
scalar processing
A form of fine-grain serial processing whereby iterative operations are performed sequentially on the elements of an array, with each iteration producing one result. Compare to vector processing .
280 S–2314–50
S–2314–50
Glossary scalar register
The register that serves as source and destination for operands that use scalar arithmetic and logical instructions.
scope
The region of a program in which a variable is defined and can be referenced.
scoping unit
Part of a program in which a name has a fixed meaning. A program unit or subprogram generally defines a scoping unit. Type definitions and procedure interface bodies also constitute scoping units. Scoping units do not overlap, although one scoping unit may contain another in the sense that it surrounds it.
If a scoping unit contains another scoping unit, the outer scoping unit is referred to as the host scoping unit of the inner scoping unit.
shift
A basic mathematical operation that moves data right or left within a register or between two registers.
SHMEM
A library of optimized functions and subroutines that take advantage of shared memory to move data between the memories of processors. The routines can either be used by themselves or in conjunction with another programming style such as Message Passing Interface.
stack
1. A data structure that provides a dynamic, sequential data list that can be accessed from either end; a last-in, first-out (push-down, pop-up) stack is accessed from just one end. 2. The call stack consists of stack frames that hold return locations for called routines, routine arguments, local variables, and saved registers.
stack frame
An element of a stack that contains local variables, arguments, contents of the registers used by an individual routine, a frame pointer that points to the previous stack frame, and the value of the program counter at the time the routine was called. A stack frame is allocated when a reentrant subroutine is entered; it is deallocated on exit.
281
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual
282 stride
The relationship between the layout of an array’s elements in memory and the order in which those elements are accessed. A stride of 1 means that memory-adjacent array elements are accessed on successive iterations of an array-processing loop.
support node
The node that is used to run serial commands, such as shells, editors, and other user commands (ls, for example). See also application node; OS node; node; support
node.
symbol table
A table of symbolic names (for example, variables) used in a program to store their memory locations. The symbol table is part of the executable object generated by the compiler. Debuggers use it to help analyze the program.
type
A means for categorizing data. Each intrinsic and user-defined data type has four characteristics: a name, a set of values, a set of operators, and a means to represent constant values of the type in a program.
UNICOS
The operating system for Cray SV1 series systems.
UNICOS/mp
The operating system for Cray X1 systems.
vector
An array, or a subset of an array, on which a computer operates. When arithmetic, logical, or memory operations are applied to vectors, it is referred to as vector processing.
vector instruction
An instruction that operates on a series of elements by repeating the same function and producing a series of results; eliminates instruction startup time for all of the operands but the first.
S–2314–50
S–2314–50
Glossary vector length
The number of elements in a vector.
vector length register
A register that holds the number of elements of a vector register that are required for an operation.
vector processing
A technique whereby operations are performed simultaneously on the elements of an array instead of iteratively. Compare to scalar processing.
virtual address
The apparent location of a data element, such as the name of a variable or an array element, as referenced by a user’s program during a read or write operation. The hardware translates a virtual address into a physical address, which is the actual location in the machine memory.
283
Index
A
cross-reference listing format, 60
B
Binary definition files macros in, 64
micros in, 65 opdefs in, 65 opsyns in, 65
using
multiple references to a definition, 55–57
Binary Definition files using multiple references to a definition, 55
C
CAL program program segment, 42
S–2314–50
label field, 68 operand field, 68 result field, 68
Counters
location, 93, 176 origin, 93, 176 word-bit-position, 93, 176
Cray Assembly Language (CAL), 1
D
Data types
Defined sequences definition format, 217
Duplicate with varying argument, 252
285
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual
Ending macros and operation definitions, 254
premature exit from a macro expansion, 255
premature exit from code duplication, 256
E
statements, 97 concatenation, 97 micro substitution, 97
L
Listings lines per page control, 53
286
Location elements location counter, 92 origin counter, 92 word pointer, 92
M
MLEVEL pseudo instructions, 154
N
S–2314–50
O
Operation definitions (OPDEF), 237
P
case sensitivity, 33 classes, 33
S–2314–50
Index
287
Cray Assembly Language (CAL) for Cray X1™ Systems Reference Manual
Q
R
Relative attributes, 74 absolute, 74
external, 75 immobile, 75 relocatable, 75
S
Statements actual and edited, 100
U
W
Word boundary
Word-bit-position counter, 93, 176
288 S–2314–50
Download
Public link updated
The public link to your chat has been updated. You may view and manage your chat links after registration here.
Document public link
Advertisement
Key features
Assembly Language Reference
System Architecture
Instruction Set
Data Types
Registers
Frequently asked questions
CAL refers to Cray Assembly Language, a programming language for the Cray X1 Systems. This document provides a reference manual for the Cray Assembly Language (CAL).
The Cray X1 supports integer and floating point data types. Integers are stored in either 64-bit or 32-bit formats, while floating-point numbers are stored in 64-bit or 128-bit formats.
The Cray X1 supports several types of registers, including scalar, vector, mask, vector length, vector carry, and bit matrix multiply registers. Scalar registers are used for general-purpose operations. Vector registers are designed for high-performance vector processing.
