IBM AIX Flag Disclosure
SPEC CPU2000 & OMP2001
For use with AIX submissions with the IBM XL compilers.
Last Revised 22 August, 2005
Source Level Portability Options
================================
-DHOST_WORDS_BIG_ENDIAN (176.gcc)
Host system is big-endian.
-DAIX (186.crafty)
Sets some basic parameters like endian-ess, OS type, and ANSI
language extensions to be compatible with an AIX system.
-DNDEBUG (252.eon)
SPEC default for C++ compiler but also needed explicitly by
some linkers. Defining this disables any assert macros used for
debugging.
-DNEED_EXPLICIT_SPECIALIZATION (252.eon)
Supply function definitions with explicit types in two cases
where templatized versions fail to compile.
-DSPEC_CPU2000_AIX (253.perlbmk)
Compile the SPEC CPU2000 modified perl for an AIX system.
-DSYS_IS_BSD (254.gap)
Compile gap for a BSDish system.
-DSYS_STRING_H (254.gap)
Do not explicitly include string.h
-DSYS_HAS_TIME_PROTO (254.gap)
Do not supply prototypes for the time(), times() and getrusage()
functions.
-DSYS_HAS_MALLOC_PROTO (254.gap)
Do not supply prototypes for malloc() and free().
-DSYS_HAS_CALLOC_PROTO (254.gap)
Do not supply a prototype for calloc().
-DHAVE_SIGNED_CHAR (300.twolf)
System allows signed char type.
Compiler Invocation
===================
xlc Invokes the compiler for C source files with a default language level
of ansi and specifies that it allow type-based aliasing.
cc Invokes the compiler for C source files with a default language of
extended and specifies that it provide compatibility with older IBM
compilers and allow placement of string literals or constant values in
read/write storage. cc does not conform to the ISO/ANSI C standard.
xlC Invokes the compiler for C++ source files with a default language level
of ansi and specifies that it allow type-based aliasing.
xlc_r The same as "xlc" except that it generates a threadsafe executable,
compliant with the POSIX pthreads API.
xlf Invokes the compiler for Fortran source files with a default language
of Fortran 77.
xlf90 Invokes the compiler for Fortran source files with a default language
of Fortran 90.
xlf90_r The same as "xlf90" except that it generates a threadsafe executable,
compliant with the POSIX pthreads API.
cleanpdf Erase the information in the PDF directory if any exists to ensure
no feedback information is reused between compilations.
Compiler Options
================
-ma Use built-in alloca() function.
-O Performs optimizations that the compiler developers considered
the best combination for compilation speed and runtime
performance.
-O3 Perform some memory and compile time intensive optimizations in
addition to those executed with -O. The -O3 specific
optimizations have the potential to slightly alter the semantics
of a user's program. Optimizations may include, but are not
limited to: Aggressive code motion, and scheduling on
computations that have the potential to raise an exception, but
no valid exceptions will be suppressed; Relaxed conformance to
IEEE rules in cases where the difference in the results is not
important to an application; Rewriting of floating point expressions.
-O4 Equivalent to -O3 -qipa -qhot with automatic generation of
architecture ( -qarch= )and tuning ( -qtune= )options ideal for
that platform. The qipa level defaults to level=1.
-O5 Equivalent to -O3 -qipa=level=2 -qhot with automatic generation
of architecture ( -qarch= ) and tuning ( -qtune= ) options ideal
for that platform.
-D_ILS_MACROS Defined in /usr/include/ctype.h to use the macro version of the
string classification functions (e.g. isupper()).
-Q, -qinline The -Q option without any list inlines all appropriate
procedures, subject to limits on the number of inlined calls and
the amount of code size increase as a result. -qinline is an
alias for -Q.
-Q=xxx Inline all functions that contain less than xxx lines of
abstract code units.
-q64 Selects 64-bit compiler mode.
-qalign=struct=natural The compiler maps structure members to their
-qalign=natural natural boundaries. The first form is used by
the Fortran compiler; the second form is used
by the C compiler and is a deprecated form for
the Fortran compiler.
-qansialias Use type-based aliasing during optimization
-qarch=ppc Produces object code containing instructions that will run on
any of the 32-bit PowerPC hardware platforms.
-qarch=pwr3 Produces object code containing instructions that will run on
power3 processors.
-qarch=pwr4 Produces object code containing instructions that will run on
power4/power4+ processors.
-qarch=pwr5 Produces object code containing instructions that will run on
power5 processors.
-qarch=rs64b Produces object code containing instructions that will run on
RS64-II processors.
-qdatalocal Changes the default to assume that all variables ar local.
-qessl Specifies that, if either -lessl or -lesslsmp are also
specified, then Engineering and Scientific Subroutine Library
(ESSL) routines should be used in place of some Fortran 90
intrinsic procedures when there is a safe opportunity to do so.
-qlibessl Specifies that all functions whose names match ESSL library-
functions are, in fact, the library functions.
-qfdpr Collect information about programs for use with the AIX fdpr
(Feedback Directed Program Restructuring) performance-tuning utility.
-qfixed Indicates that the input source program is in fixed form. Allows
fixed format Fortran 77 programs to be compiled using the xlf90
compiler invocation.
-qfixed=<num> States that Fortran code is in fixed source form, with optional
argument specifying the maximum line length.
-qfloat=rsqrt Changes a division by the result of a square root operation into a
multiply by the reciprocal of the square root.
-qhot Perform high-order transformations on loops during optimization.
-qhot=arraypad Pad the sizes of arrays to align better in cache.
-qipa=level=1 Turns on interprocedural analysis with inlining, limited alias
analysis, and limited call-site tailoring. This is the default
level of -qipa.
-qipa=level=2 Turns on interprocedural analysis with inlining, cloning, full
alias analysis, constant propagation, call-site tailoring, and
dead code removal.
-qipa=noobject Do not generate object files during the first stage of inter-
procedural analysis.
-qinline Alias for -Q. See -Q.
-qipa=partition=large Specifies the size of the regions within the program to
analyze. Larger partitions contain more procedures,
which result in better interprocedural analysis but
require more storage to optimize.
-qlanglvl=ansi Compilation conforms to the ANSI standard.
-qlargepage Indicates that a program, designed to execute in a large page
memory environment, can take advantage of large 16 MB pages
provided on POWER4 or better CPUs.
-qmaxmem=-1 Allows the compiler to use as much memory as it needs to execute.
-qpdf1/pdf2 Profile directed feedback optimization
-qsave Sets the default storage class for local variables to STATIC.
-qsmp=omp Enable OpenMP parallelization directives.
-qsuffix=f=f90 Sets the suffix for source files to be .f90. The .f90 suffix is
required by xlf90 to compile Fortran 90 programs.
-qtune=604 Instruction selection, scheduling, and other implementation
dependent performance enhancements for the PowerPC 604/604e
processor.
-qtune=pwr3 Instruction selection, scheduling, and other implementation
dependent performance enhancements for the Power3 processor.
-qtune=pwr4 Instruction selection, scheduling, and other implementation
dependent performance enhancements for the Power4/Power4+
processors.
-qtune=rs64b Instruction selection, scheduling, and other implementation
dependent performance enhancements for the RS64-II processor.
-qunroll=n Unrolls inner loops in th program by a factor of n.
-w Suppress warning messages from the C, C++, and Fortran compilers.
Linker Options
==============
-Ldir Link looks in the directory that is specified by the
option "dir".
-lblacssmp Link the Parallel ESSL SMP BLACS Library.
-lessl Link the Engineering and Scientifc Subroutine Library (ESSL).
-lesslsmp Link the threadsafe version of the ESSL library.
-lpesslsmp Link the threadsafe, parallelized version of the ESSL library.
-lmass Link the mathematical acceleration subsystem libraries (MASS),
which contain libraries of tuned mathematical intrinsic
functions. See
http://techsupport.services.ibm.com/server/mass?fetch=home.html
-lhmu Link fast malloc libraries. These libraries are part of the
memdbg package that is included with IBM C compilers.
-lpdf Routines used in the first pass of the profile directed
feedback process. Routines from this library are not used in
building the final executable. In newer compilers, -qpdf1 does
this automatically, so using this in conjunction with -qpdf1
is redundant.
-blpdata Sets the bit in the file's XCOFF header indicating that this
executable will request the use of large pages when they are
available on the system and when the user has an appropriate
privilege
-bmaxdata:0x........ Sets the maximum combined size of the program's stack-
and data- segments to this number of byes, specified
in hexadecimal, when the default is too small.
-bnso Brings referenced library procedures into the object file
-bI:/lib/syscalls.exp Create statically linked object files (syscalls.exp
supplies the names of the routines that can be
imported).
FDPR:
=====
The fdpr (feedback directed program restructuring) program optimizes the
executable image of a program by collecting information on the behavior of
the program while the program is used for some typical workload, and then
creating a new version. It is available on AIX Version 4 and 5 systems as part
of the Performance Toolbox for AIX.
Options:
-o OutFile Specifies the name of the output file from the optimizer.
-p ProgramName The name of the executable program to optimize.
-v Selects verbose output during processing/compilation
-x Command Specifies the command used for invoking the instrumented
program. All the arguments after the -x flag are used
for the invocation.
-O2 Employ a program-reordering technique in which the
original structure of the program, including traceback
entries, is preserved.
-O3 Employ global reordering techniques that do not preserve
debug information.
The compilers include an optional "-qfdpr" flag that assists FDPR analysis but
is not required for it.
Large Page Settings:
====================
vmo command options (AIX 5.2 & above):
-r Apply changes at the system boot.
-o lgpg_regions=# Specifies the number of large pages to reserve.
Example: #=200 allows 200 large pages to be reserved.
-o lgpg_size=# Specifies the size in bytes of the hardware-supported
large pages. Example: #=16777216 is a 16M page size.
vmtune/vmtune64 command options (AIX 5.1 Only):
-g Sets the page size for the large page.
Example: -g 16777216 for 16M page.
-L Sets the number of large pages.
Example: -L 200 allows 200 large pages.
-y1 Enables the memory affinity.
chuser capabilities=CAP_BYPASS_RAC_VMM,CAP_PROPAGATE $USER
Allows $USER (non-root ID) to access the large pages that are
available.
bosboot -a Creates a boot image used on the next system reboot.
shutdown -r Halt the operating system and reboot.
Memory Affinity:
================
vmo command options (AIX 5.2):
-r Apply changes at the system boot
-o memory_affinity=1 Enable the VMM to restrict the memory frames attached to
the executing MCM
Note that the system needs to be rebooted to activate the Memory Affinity feature,
i.e. "bosboot -a; shutdown -r" as described above, and the Large Page and Memory
Affinity options can be used together. Memory Affinity is active by default in
AIX Version 5.2 5765-H62 (05/2003) and above. In all cases the "MEMORY_AFFINITY"
environment variable, defined below, needs to be set for the job that is running.
AIX Environment Variables:
==========================
LDR_CNTRL=LARGE_PAGE_DATA=M Asserts that there are sufficient large-pages available
for program data, allowing them to be allocated on first
reference, instead of allocating all of them at load time.
MEMORY_AFFINITY=MCM Turn on Memory Affinity which has been enabled with the
vmo command.
MALLOCMULTIHEAP=1 Maintains multiple heaps in the process, for servicing simultaneous
"malloc" requests.
OMP_DYNAMIC=FALSE Disables dynamic adjustment of the number of available threads.
OMP_NUM_THREADS=... The exact number of threads available to be used, or if OMP_DYNAMIC
is TRUE, the upper limit on the number of available threads.
XLFRTEOPTS=NAMELIST=OLD Allows a newly compiled program to read the namelist from a
binary compiled with the older namelist format.
XLSMPOPTS A list of runtime settings affecting SMP execution. Here are
some of the possibilities:
SCHEDULE=STATIC Work is scheduled to threads round-robin.
SPINS=0 Allows work-requests to spin indefinitely without
the thread having to yield the time-slice.
STACK=.... Specifies the largest allowable size of a thread's
stack.
YIELDS=0 Allows the thread to yield an indefinite number
of times without being driven into a sleep state.
STARTPROC=0 When assigning threads to CPU's, begin with thread
0 on CPU 0.
STRIDE=X When assigning the next thread to a CPU, add X to
the current CPU index instead of using (CPU+1).
Firmware Configuration Information:
===================================
PRIVATE mode L3-Caches: Enter L3 Mode Menu; if the "Private/Shared L3 Mode" item reads
"Currently Shared (/Private)", type "1" to toggle the mode to
PRIVATE, otherwise do nothing.
Process Management:
===================
The following commands are used to bind processes to processors in SPEC/CPU runs.
The SPEC/CPU harness uses the $SPECUSERNUM variable to enumerate the different
processes in a rate-run; in the text of the SPEC/CPU config-file, this is expressed
as "\$SPECUSERNUM" in order for the variable-name to be evaluated at runtime.
bindprocessor X Y AIX command, binding process X to CPU Y.
schedule.8 \$SPECUSERNUM $command Shell script that binds processes 0,1,2,3 to even-numbered
CPU's 0,2,4,6 and processes 4,5,6,7 to odd-numbered CPU's
1,3,5,7. This is written using the "bindprocessor" primitive.
schedule.16 ... Analogous scripts, binding the first 8, 16, 32, or 64
schedule.32 ... processes to the even-numbered CPU's and the second
schedule.64 ... 8, 16, 32, or 64 processes to the odd-numbered CPU's.
schedule.128 ...