# Invocation command line: # /spec/omp2001/bin/runspec -v 10 -c H64+.publish.cfg -n 3 -T all -i ref medium ############################################################################ ############################################################################ ############################################################################ ################################################################################################ ################################################################################################ # # Global options for runspec # runlist = all action = validate iterations = 3 output_format = asc,config ignore_errors = 0 reportable = 1 mean_anyway = 1 size = ref tune = base,peak ext = 570_64 env_vars = 1 deletework = 0 # # The following commands cause the process thread to bind to a correseponding # CPU. This improves performance by removing the overhead of threads "drifting" # between CPU's. # # # # General environment settings. # OS_LIBS=-L/usr/lib # ENV_XLSMPOPTS=SPINS=0:YIELDS=0:STACK=8000000 # ENV_OMP_NUM_THREADS=128 # ENV_OMP_DYNAMIC=TRUE # Helps "base" out a bit. ENV_XLSMPOPTS=SPINS=0:YIELDS=0:STACK=8000000:SCHEDULE=STATIC:STARTPROC=0:STRIDE=2 ENV_OMP_NUM_THREADS=64 ENV_OMP_DYNAMIC=FALSE ENV_MALLOCMULTIHEAP=1 ENV_XLFRTEOPTS=intrinthds=1 ENV_MEMORY_AFFINITY=MCM # #=============================================================================== # How the compilers are called #=============================================================================== # To remove all the warning messages that are put out, you can append the following # compile time message suppression flag to the compiler invocation lines: # # -w -qinfo=noinl -qsuppress=1500-036 # # To suppress the Fortran function compilation messages, add to the FC/F77/FXLF # lines: # -qsuppress=cmpmsg # # # To speed builds with ipa, use "-qipa=noobject" to suppress code generation # until the final IPA link step. # To speed builds with ipa, use "-qipa=threads:n" to direct ipa to use n threads. #=============================================================================== # C_COMPILER = F_COMPILER = CC = ${C_COMPILER}/usr/vac/bin/xlc_r -qsmp=omp FC = ${F_COMPILER}/usr/bin/xlf90_r -qsmp=omp # # The next line set up parallel make, the value of j is the number of threads. # makeflags = -j8 # # galgel & facerec do not have dependency rules to allow parallel make's. The # easiest way to handle them is to add "makeflags = -j1" to their portability # flags section. # # # # SPECompM2001 portability flags. These are required to make the programs # work with the given compilers and are not optimizations. # # If there aren't any flags for the benchmark, we leave an empty entry as # a placeholder. 310.wupwise_m=default=default=default: FPORTABILITY=-qfixed 312.swim_m=default=default=default: FPORTABILITY=-qfixed 314.mgrid_m=default=default=default: FPORTABILITY=-qfixed 316.applu_m=default=default=default: FPORTABILITY=-qfixed 318.galgel_m=default=default=default: makeflags = -j1 FPORTABILITY=-qfixed=80 -qsuffix=f=f90 320.equake_m=default=default=default: 324.apsi_m=default=default=default: FPORTABILITY=-qfixed 326.gafort_m=default=default=default: FPORTABILITY=-qsuffix=f=f90 328.fma3d_m=default=default=default: FPORTABILITY=-qsuffix=f=f90 330.art_m=default=default=default: 332.ammp_m=default=default=default: # # SPECompM2001 base-level optimizations. # # Note that the flags vary between languages (C & Fortran in this case). # We also use feedback for C and not Fortran, so we have to add some # overrides between them. # medium=base=default=default: # FOPTIMIZE = -q64 -O5 -qipa=noobject -qipa=partition=large -qmaxmem=-1 # FOPTIMIZE = -O5 -blpdata -qhot=arraypad -qipa=noobject -qipa=partition=large -qmaxmem=-1 # EXTRA_LDFLAGS = -bmaxdata:0x80000000 # FOPTIMIZE = -O5 -blpdata -qipa=noobject -qipa=partition=large -qmaxmem=-1 # EXTRA_LDFLAGS = -bmaxdata:0xD0000000/dsa FOPTIMIZE = -q64 -O5 -qipa=noobject -btextpsize:64K -bdatapsize:64K -bstackpsize:64K feedback = 0 # COPTIMIZE=-q64 -O5 -blpdata -qalign=natural -qhot=arraypad -Q fdo_pre0 = ${C_COMPILER}/usr/vac/bin/cleanpdf PASS1_CFLAGS = -qpdf1 -O5 -q64 -btextpsize:64K -bdatapsize:64K -bstackpsize:64K PASS1_LDCFLAGS = -qpdf1 -O5 -q64 -btextpsize:64K -bdatapsize:64K -bstackpsize:64K PASS2_CFLAGS = -qpdf2 -O5 -q64 -btextpsize:64K -bdatapsize:64K -bstackpsize:64K PASS2_LDCFLAGS = -qpdf2 -O5 -q64 -btextpsize:64K -bdatapsize:64K -bstackpsize:64K 320.equake_m=base=default=default: feedback = 1 330.art_m=base=default=default: feedback = 1 332.ammp_m=base=default=default: feedback = 1 # # # SPECompM2001 peak-level optimizations. # 310.wupwise_m=peak=default=default: ENV_OMP_DYNAMIC=FALSE ENV_XLSMPOPTS=SPINS=0:YIELDS=0:STACK=8000000:SCHEDULE=STATIC:STARTPROC=0:STRIDE=1 ENV_OMP_NUM_THREADS=128 OPTIMIZE = -O5 -qhot=arraypad -qipa=noobject -qipa=partition=large -qmaxmem=-1 -q64 -btextpsize:64K -bdatapsize:64K 312.swim_m=peak=default=default: srcalt=ompl.32 ENV_OMP_DYNAMIC=FALSE ENV_XLSMPOPTS=SPINS=0:YIELDS=0:STACK=8000000:SCHEDULE=STATIC:STARTPROC=0:STRIDE=2 ENV_OMP_NUM_THREADS=64 OPTIMIZE = -O5 -qarch=pwr3 -qtune=pwr3 -q64 -btextpsize:64K -bdatapsize:64K -bstackpsize:64K 314.mgrid_m=peak=default=default: ENV_OMP_DYNAMIC=FALSE ENV_XLSMPOPTS=SPINS=0:YIELDS=0:STACK=8000000:SCHEDULE=STATIC:STARTPROC=0:STRIDE=2 ENV_OMP_NUM_THREADS=64 OPTIMIZE = -O5 -qipa=partition=large -qalign=struct=natural -q64 -btextpsize:64K -bdatapsize:64K 316.applu_m=peak=default=default: srcalt=ompl.32 ENV_OMP_DYNAMIC=FALSE ENV_XLSMPOPTS=SPINS=0:YIELDS=0:STACK=8000000:SCHEDULE=STATIC:STARTPROC=0:STRIDE=1 ENV_OMP_NUM_THREADS=128 OPTIMIZE = -O3 -qarch=pwr4 -qtune=pwr4 -q64 -btextpsize:64K -bdatapsize:64K 318.galgel_m=peak=default=default: ENV_OMP_DYNAMIC=FALSE ENV_XLSMPOPTS=SPINS=0:YIELDS=0:STACK=8000000:SCHEDULE=STATIC:STARTPROC=0:STRIDE=2 ENV_OMP_NUM_THREADS=64 FOPTIMIZE = -O5 -blpdata -qipa=noobject -qipa=partition=large -qmaxmem=-1 EXTRA_LDFLAGS = -bmaxdata:0xD0000000/dsa 320.equake_m=peak=default=default: srcalt=ompl.32 ENV_OMP_DYNAMIC=FALSE ENV_XLSMPOPTS=SPINS=0:YIELDS=0:STACK=8000000:SCHEDULE=STATIC:STARTPROC=0:STRIDE=2 ENV_OMP_NUM_THREADS=64 fdo_pre0 = ${C_COMPILER}/usr/vac/bin/cleanpdf PASS1_CFLAGS = -qpdf1 -O5 -qfdpr -q64 -bdatapsize:64K PASS1_LDCFLAGS = -qpdf1 -O5 -qfdpr -q64 -bdatapsize:64K PASS2_CFLAGS = -qpdf2 -O5 -qfdpr -q64 -bdatapsize:64K PASS2_LDCFLAGS = -qpdf2 -O5 -qfdpr -q64 -bdatapsize:64K fdo_run2 = fdpr -q -O3 -p $commandexe -x $command fdo_post2 = cp ${baseexe}.fdpr $baseexe 324.apsi_m=peak=default=default: # # OPTIMIZE = -O5 -qalign=struct=natural -bdatapsize:64K -bstackpsize:64K # # EXTRA_LDFLAGS = -bmaxdata:0xD0000000/dsa ENV_OMP_DYNAMIC=FALSE ENV_XLSMPOPTS=SPINS=0:YIELDS=0:STACK=8000000:SCHEDULE=STATIC:STARTPROC=0:STRIDE=1 ENV_OMP_NUM_THREADS=128 OPTIMIZE = -q64 -O5 -qipa=noobject -qipa=partition=large -qmaxmem=-1 326.gafort_m=peak=default=default: # # OPTIMIZE = -O5 -qhot=arraypad -qipa=noobject -qipa=partition=large -qmaxmem=-1 -qalign=struct=natural -btextpsize:64K -bdatapsize:64K -bstackpsize:64K -lhmu -lmass -qessl -lesslsmp # # EXTRA_LDFLAGS = -bmaxdata:0xD0000000/dsa srcalt=ompl.32 ENV_OMP_DYNAMIC=FALSE ENV_XLSMPOPTS=SPINS=0:YIELDS=0:STACK=8000000:SCHEDULE=STATIC:STARTPROC=0:STRIDE=1 ENV_OMP_NUM_THREADS=128 OPTIMIZE = -q64 -O5 -qipa=noobject -qipa=partition=large -qmaxmem=-1 328.fma3d_m=peak=default=default: basepeak=1 # ENV_OMP_DYNAMIC=FALSE # ENV_XLSMPOPTS=SPINS=0:YIELDS=0:STACK=8000000:SCHEDULE=STATIC:STARTPROC=0:STRIDE=2 # ENV_OMP_NUM_THREADS=64 # OPTIMIZE = -q64 -O5 -qipa=noobject -qipa=partition=large -qmaxmem=-1 # # OPTIMIZE = -O5 -qhot=arraypad -qipa=noobject -qipa=partition=large -qmaxmem=-1 -btextpsize:64K -bdatapsize:64K -bstackpsize:64K -lmass -qessl -lesslsmp # # EXTRA_LDFLAGS = -bmaxdata:0xD0000000/dsa # # basepeak = 1 330.art_m=peak=default=default: ENV_OMP_DYNAMIC=FALSE ENV_XLSMPOPTS=SPINS=0:YIELDS=0:STACK=8000000:SCHEDULE=STATIC:STARTPROC=0:STRIDE=2 ENV_OMP_NUM_THREADS=64 EXTRA_CFLAGS = -DINTS_PER_CACHELINE=32 -DDBLS_PER_CACHELINE=16 fdo_pre0 = ${C_COMPILER}/usr/vac/bin/cleanpdf PASS1_CFLAGS = -qpdf1 -O5 -qfdpr -q64 -btextpsize:64K -bdatapsize:64K -lesslsmp PASS1_LDCFLAGS = -qpdf1 -O5 -qfdpr -q64 -btextpsize:64K -bdatapsize:64K -lesslsmp PASS2_CFLAGS = -qpdf2 -O5 -qfdpr -q64 -btextpsize:64K -bdatapsize:64K -lesslsmp PASS2_LDCFLAGS = -qpdf2 -O5 -qfdpr -q64 -btextpsize:64K -bdatapsize:64K -lesslsmp fdo_run2 = fdpr -q -O3 -p $commandexe -x $command fdo_post2 = cp ${baseexe}.fdpr $baseexe 332.ammp_m=peak=default=default: ENV_OMP_DYNAMIC=FALSE ENV_XLSMPOPTS=SPINS=0:YIELDS=0:STACK=8000000:SCHEDULE=STATIC:STARTPROC=0:STRIDE=1 ENV_OMP_NUM_THREADS=128 fdo_pre0 = ${C_COMPILER}/usr/vac/bin/cleanpdf PASS1_CFLAGS = -qpdf1 -O5 -qipa=partition=large -qmaxmem=-1 -q64 -btextpsize:64K -bdatapsize:64K PASS1_LDCFLAGS = -qpdf1 -O5 -qipa=partition=large -qmaxmem=-1 -q64 -btextpsize:64K -bdatapsize:64K PASS2_CFLAGS = -qpdf2 -O5 -qipa=partition=large -qmaxmem=-1 -q64 -btextpsize:64K -bdatapsize:64K PASS2_LDCFLAGS = -qpdf2 -O5 -qipa=partition=large -qmaxmem=-1 -q64 -btextpsize:64K -bdatapsize:64K # # # Notes for the Floating-Point benchmarks. # medium=default=default=default: notes0000= Portability Flags & Environment Variables notes0020= -qfixed used in: 310.wupwise_m, 312.swim_m, 314.mgrid_m, 316.applu_m, 324.apsi_m notes0030= -qfixed=80 used in: 318.galgel_m notes0040= -qsuffix=f=f90 used in: 318.galgel_m, 326.gafort_m, 328.fma3d_m notes0050= notes0060= Base Flags notes0070= C: -qpdf1/pdf2 notes0075= -q64 -O5 -blpdata -qalign=natural -qhot=arraypad -Q -qsmp=omp notes0080= FORTRAN: -O5 -qhot=arraypad -qipa=noobject -qipa=partition=large -qmaxmem=-1 -qsmp=omp notes0090= notes0100= Base & Peak User Environment: notes0110= OMP_NUM_THREADS=32 notes0120= OMP_DYNAMIC=FALSE notes0130= XLSMPOPTS=SPINS=0:YIELDS=0:STACK=8000000:SCHEDULE=STATIC:STARTPROC=0:STRIDE=1 notes0140= MALLOCMULTIHEAP=1 notes0150= notes0200= Peak Flags: notes0210= -qsmp=omp used in all cases notes0220= 310.wupwise_m: -O5 -qarch=pwr3 -qtune=pwr3 -qalign=struct=natural -qfdpr -q64 -btextpsize:64K -bdatapsize:64K notes0230= fdpr -q -O3 #notes0232= XLSMPOPTS=SPINS=0:YIELDS=0:STACK=8000000:SCHEDULE=STATIC:STARTPROC=0:STRIDE=2 #notes0235= ENV_OMP_NUM_THREADS=16 notes0240= 312.swim_m: -O5 -qarch=pwr3 -qtune=pwr3 -q64 -btextpsize:64K -bdatapsize:64K -bstackpsize:64K notes0260= 314.mgrid_m: -O5 -qipa=partition=large -qalign=struct=natural -q64 -btextpsize:64K -bdatapsize:64K notes0265= XLSMPOPTS=SPINS=0:YIELDS=0:STACK=8000000:SCHEDULE=STATIC:STARTPROC=0:STRIDE=2 notes0270= OMP_NUM_THREADS=16 notes0280= 316.applu_m: -O3 -qarch=pwr4 -qtune=pwr4 -q64 -btextpsize:64K -bdatapsize:64K #notes0282= XLSMPOPTS=SPINS=0:YIELDS=0:STACK=8000000:SCHEDULE=STATIC:STARTPROC=0:STRIDE=2 #notes0285= OMP_NUM_THREADS=16 notes0290= 318.galgel_m: -O5 -qarch=pwr3 -qtune=pwr3 -btextpsize:64K -bdatapsize:64K -bstackpsize:64K -lmass -qessl -lesslsmp notes0300= EXTRA_LDFLAGS=-bmaxdata:0x80000000 notes0302= XLSMPOPTS=SPINS=0:YIELDS=0:STACK=8000000:SCHEDULE=STATIC:STARTPROC=0:STRIDE=2 notes0305= OMP_NUM_THREADS=16 notes0310= 320.equake_m: -qpdf1/pdf2 notes0320= -O5 -qfdpr -q64 -bdatapsize:64K notes0330= fdpr -q -O3 notes0332= XLSMPOPTS=SPINS=0:YIELDS=0:STACK=8000000:SCHEDULE=STATIC:STARTPROC=0:STRIDE=2 notes0335= OMP_NUM_THREADS=16 notes0340= 325.apsi_m: -O5 -qalign=struct=natural -bdatapsize:64K notes0350= -bmaxdata:0x80000000 #notes0352= XLSMPOPTS=SPINS=0:YIELDS=0:STACK=8000000:SCHEDULE=STATIC:STARTPROC=0:STRIDE=2 #notes0355= OMP_NUM_THREADS=16 notes0360= 326.gafort_m: -O5 -qhot=arraypad -qipa=noobject -qipa=partition=large -qmaxmem=-1 -qalign=struct=natural notes0365= -btextpsize:64K -bdatapsize:64K -bstackpsize:64K -lhmu -lmass -qessl -lesslsmp notes0370= EXTRA_LDFLAGS=-bmaxdata:0x80000000 notes0380= 328.fma3d_m: -O5 -qhot=arraypad -qipa=noobject -qipa=partition=large -qmaxmem=-1 -btextpsize:64K -bdatapsize:64K notes0385= -bstackpsize:64K -lmass -qessl -lesslsmp notes0390= EXTRA_LDFLAGS=-bmaxdata:0x80000000 notes0400= 330.art_m: -qpdf1/pdf2 notes0410= -O5 -qfdpr -q64 -btextpsize:64K -bdatapsize:64K -lesslsmp notes0420= fdpr -q -O3 notes0430= EXTRA_CFLAGS= -DINTS_PER_CACHELINE=32 -DDBLS_PER_CACHELINE=16 notes0440= 332.ammp_m: -qpdf1/pdf2 notes0450= -O5 -qipa=partition=large -qmaxmem=-1 -q64 -btextpsize:64K -bdatapsize:64K notes0460= notes2220= Alternate sources: notes2230= Add critical region around update of linked list in parallel loop. notes2240= Approved src.alt available as ompm-purdue1-20040324.tar.gz notes2250= Used for 330.art_m, base and peak. notes2260= notes2270= Peak sources: notes2280= SPEC OMPL2001 source for 32bit systems modified for SPEC OMPM2001 used notes2290= with 312.swim_m, 316.applu_m, 320.equake_m, 326.gafort_m. notes2300= # # # Global options for reports # default=default=default=default: company_name =IBM Corporation license_num =HPG0005 tester_name =IBM test_site =Austin, TX test_date =Dec-2005 # # Information you need to fill in for each platform. # hw_avail =Feb-2006 hw_vendor =IBM Corporation hw_model =IBM OpenPower 570 (2200 MHz, 16 CPU) hw_cpu =Power5+ hw_cpu_mhz =2200 hw_fpu =Integrated hw_ncpu =16 cores, 8 chips, 2 cores/chip (SMT on) hw_ncpuorder=4,8,12,16 hw_pcache =64KBI+32KBD (on chip)/core hw_scache =1920KB unified (on chip)/core hw_tcache =36MB unified (off-chip)/DCM, 2 DCM/SUT hw_ocache =None hw_memory =64x2GB hw_disk =2x73GB SCSI, 15K RPM hw_other =None machine_name=perf135 hw_parallel =no sw_avail =Feb-2006 sw_parallel =OpenMP sw_os =AIX 5L V5.3 sw_compiler0 =XL C/C++ Enterprise Edition Version 8.0 for AIX sw_compiler1 =XL Fortran Enterprise Edition Version 10.1 for AIX sw_compiler2 =Other Software: ESSL 4.2.0.3 sw_file =AIX/JFS2 sw_state =Multi-user # # Closing notes on the sytem configuration. # notes2320= notes2330= The installed OS level is AIX 5L for POWER version 5.3 with the 5300-04 Recommended Maintenence package. notes2335= notes2340= SMT: Acronym for "Simultaneous Multi-Threading". A processor technology that allows notes2345= the simultaneous execution of multiple thread contexts within a single processor notes2350= core. (Enabled by default) notes2355= DCM: Acronym for "Dual-Chip Module" (one dual-core processor chip + one L3-cache chip) notes2360= SUT: Acronym for "System Under Test" notes2370= ESSL: Engineering and Scientific Subroutine Library notes2375= notes2380= Extended C: IBM XL C for AIX invoked as xlc_r notes2381= ANSI C89: IBM XL C for AIX invoked as xlc_r notes2395= Fortran 90: IBM XL Fortran for AIX invoked as xlf_r notes2400= notes2405= ulimits set to unlimited. notes2410= Large page mode and memory affinity were set as follows: notes2415= vmo -r -o lgpg_regions=2048 -o lgpg_size=16777216 notes2420= chuser capabilities=CAP_BYPASS_RAC_VMM,CAP_PROPAGATE $USER notes2425= reboot -q notes2430= export MEMORY_AFFINITY=MCM notes2435= notes2440= The following config-file entry was used to assign each benchmark process to a core: notes2445= submit = bindprocessor \$\$ \$SPECUSERNUM; $command notes2450= The "bindprocessor" AIX command binds a process to a CPU core. notes2460=