# Invocation command line: # /autobench/sources/speccpu2006/bin/runspec --config cpu2006.p7.xlcver11.16cores.cfg --action run --rate 1 --copies 64 --parallel_setup 64 --parallel_setup_type submit --reportable --nobuild --tune all --iterations 3 --define numcores=16 --define rhel=1 --define mybasearchp7=1 --define mybaseopt=5 --extension xl11-100904-100903a-O5p7 int # output_root was not used for this run ############################################################################ ####################################################################### # # SPEC CPU2006 Benchmark Config file IBM eServer pSeries Linux # ####################################################################### # THIS CONFIG FILE IS FOR BUILDING on P6 # # --define (-S) options that are passed to the runspec command # that dynamically affect this config file # # numcores=n The number of cores in the system # to determine how many copies to run in rate # # mybaseopt=x Allows override where x=3,4,5 (-O3, -O4, -O5) for base runs # # nolibhuge Allows override to not use extra libraries on base runs # nosmartheap # # debug Adds debug ('-g') compiler option %define P7 -qarch=pwr7 -qtune=pwr7 %define LHBDT -B/usr/share/libhugetlbfs/ -tl -Wl,--hugetlbfs-align %define LHLIB -lhugetlbfs ###################################### %define FDPROPTS -O4 -omullX %define FDPRINT %define make_j 3 #====================================================================== # Global Options apply to all runs #====================================================================== # # For just a report generated without a raw file or any runs required # Do a "runspec --reportonly" (also --action=report or --fakereport) # # Just set up the build or run directories, use --buildsetup or --setup # on the runspec command # Sets nobuild as the default - you will need -D to force a compile # nobuild = 1 # -D forces a build (recompile) # -N prohibits a build (recompile) #action = validate # -a tune = base # -T iterations = 1 # -n output_format = text,html,csv # -o #basepeak = 0 reportable = 0 # -l turns off ignore_errors = 1 # -I mean_anyway = 1 allow_extension_override = 1 ext = XL-ppc64 # -e size = ref post_setup = sync # Include the run-time derived system information so that each # system run has the correct information #include: SUT.inc # ----- Begin inclusion of 'SUT.inc' ############################################################################ # company_name = IBM Corporation hw_vendor = IBM Corporation license_num = 11 sw_state = Run Level 3 (multi-user) hw_cpu_mhz = 3556 hw_cpu_name = POWER7 hw_memory = 256 GB (32x8 GB) DDR3 1066 MHz # # # hw_disk = 4x146.8 GB Software RAID-0 SAS SFF 15K RPM # # sw_os000 = Red Hat Enterprise Linux Server release 6.0 sw_os001 = (ppc64), Kernel 2.6.32-71.el6.ppc64 sw_file = ext2 # # hw_avail = Sep-2010 sw_avail = Nov-2010 # sw_compiler000 = IBM XL C/C++ for Linux, V11.1 sw_compiler001 = Updated with the Nov2010 PTF sw_other000= -Post-Link Optimization for Linux on sw_other001= POWER, Version 5.5.0-3 sw_other002= -MicroQuill SmartHeap 9 # hw_model000 = IBM Power 740 Express (3.55 GHz, 16 core, hw_model001 = RedHat) # hw_ncpuorder = 8,16 cores hw_nchips = 2 hw_ncores = 16 hw_nthreadspercore = 4 hw_ncoresperchip = 8 hw_pcache = 32 KB I + 32 KB D on chip per core hw_fpu = Integrated hw_other = None # # hw_scache = 256 KB I+D on chip per core hw_tcache = 4 MB I+D on chip per core hw_ocache = None # # # # # # # ppc64_cpu output # ---- End inclusion of '/autobench/sources/speccpu2006/config/SUT.inc' # This tells runspec to set enviornment variables of the form # ENV_xxx if found in here. Forced to 0 for reportable runs env_vars = 0 # Control if the embedded md5 string of compiler options is checked for # recompile if the options have changed check_md5 = 0 # Controls if the log file info is echo'd to the screen teeout = yes # If 1, make copy of the config file before modifing with MD5 # signitures of the compile options. backup_config = 0 # verbose -> -v #verbose = 1 to 99, 5 is default, 91 is good for debugging verbose = 1 %ifdef %{make_j} makeflags = -j%{make_j} %endif # We alway use submit to ensure there is a bindprocessor done. # use_submit_for_speed If set, use submit commands for speed as well as rate use_submit_for_speed = 1 submit = echo "$command" > run.sh ; numactl --membind=\$SPECCOPYNUM --physcpubind=\$SPECCOPYNUM sh run.sh %if %{numcores} == 16 # Juno machine: 16 cores, 2 NUMA nodes and 8 processors per socket # So processors 0-31 should use NUMA node 0 for memory allocation and processors 32-63 NUMA node 1 submit = echo "$command" > run.sh; let "MYCPU=\$SPECCOPYNUM"; MYMEM=`printf '%d' \$((\$MYCPU/32))` ; numactl --physcpubind=\$SPECCOPYNUM --membind=\$MYMEM sh run.sh %endif sw_auto_parallel = No #--prevent log file lines that are too long for vi log_line_width = 130 # Point to where the usr/compiler/OS Flags file is flagsurl000 = http://www.spec.org/cpu2006/flags/IBM-Linux-XL.20101123.01.xml CC_PATH = /opt/ibmcmp/vac/11.1/bin CXX_PATH = /opt/ibmcmp/vacpp/11.1/bin XLF_PATH = /opt/ibmcmp/xlf/13.1/bin #=============================================================================== # How the compilers are called # The settings used here are for IBM C/C++ and xl Fortran compilers #=============================================================================== default=default=default=default: CC = $(CC_PATH)/xlc -qlanglvl=extc99 CXX = $(CXX_PATH)/xlC FC = $(XLF_PATH)/xlf95 #====================================================================== # Base Level Optimizations # No limit to number of flags, must be same for all of each Language. #====================================================================== default=base=default=default: %if %{mymode64} sw_base_ptrsize = 64-bit %else sw_base_ptrsize = 32-bit %endif %ifndef %{debug} %define DEBUG %else %define DEBUG -g %endif %define BASEARCH %{P7} %define MODE %if %{mymode64} %define MODE -q64 %endif %define BASEOPT -O5 %{BASEARCH} %{MODE} %{DEBUG} -qipa=threads #Check for BASEOPT overrides %if %{mybaseopt} == 3 %define BASEOPT -O3 %{BASEARCH} %{MODE} %{DEBUG} %endif #====================================================================== # INT and FP Portability Flags # no optimizations allowed here #====================================================================== default=default=default=default: %if %{mymode64} CPORTABILITY = -DSPEC_CPU_LP64 CXXPORTABILITY = -DSPEC_CPU_LP64 FPPPORTABILITY = -DSPEC_CPU_LP64 400.perlbench=default=default=default: CPORTABILITY = -DSPEC_CPU_LINUX_PPC -DSPEC_CPU_LP64 403.gcc=peak=default=default: CPORTABILITY = -DSPEC_CPU_LP64 410.bwaves=default=default=default: FPORTABILITY = -qfixed 416.gamess=default=default=default: FPORTABILITY = -qfixed 434.zeusmp=default=default=default: FPORTABILITY = -qfixed 435.gromacs=default=default=default: FPORTABILITY = -qfixed -qextname 436.cactusADM=default=default=default: FPORTABILITY = -qfixed -qextname CPORTABILITY = -DSPEC_CPU_LP64 FPPPORTABILITY = -DSPEC_CPU_LP64 437.leslie3d=default=default=default: FPORTABILITY = -qfixed 447.dealII=default=default=default: srcalt=apache_stdcxx_4_2_1 453.povray=default=default=default: CXXPORTABILITY = -DSPEC_CPU_LP64 454.calculix=default=default=default: FPORTABILITY = -qfixed -qextname 462.libquantum=default=default=default: CPORTABILITY = -DSPEC_CPU_LINUX -DSPEC_CPU_LP64 464.h264ref=default=default=default: CPORTABILITY = -qchars=signed -DSPEC_CPU_LP64 481.wrf=default=default=default: CPORTABILITY = -DNOUNDERSCORE -DSPEC_CPU_LP64 482.sphinx3=default=default=default: CPORTABILITY = -qchars=signed -DSPEC_CPU_LP64 483.xalancbmk=default=default=default: CXXPORTABILITY = -DSPEC_CPU_LINUX -DSPEC_CPU_LP64 %else 400.perlbench=default=default=default: CPORTABILITY = -DSPEC_CPU_LINUX_PPC 462.libquantum=default=default=default: CPORTABILITY = -DSPEC_CPU_LINUX 464.h264ref=default=default=default: CPORTABILITY = -qchars=signed 483.xalancbmk=default=default=default: CXXPORTABILITY = -DSPEC_CPU_LINUX 410.bwaves=default=default=default: FPORTABILITY = -qfixed 416.gamess=default=default=default: FPORTABILITY = -qfixed 434.zeusmp=default=default=default: FPORTABILITY = -qfixed 435.gromacs=default=default=default: FPORTABILITY = -qfixed -qextname 436.cactusADM=default=default=default: FPORTABILITY = -qfixed -qextname 437.leslie3d=default=default=default: FPORTABILITY = -qfixed 447.dealII=default=default=default: srcalt=apache_stdcxx_4_2_1 454.calculix=default=default=default: FPORTABILITY = -qfixed -qextname 481.wrf=default=default=default: CPORTABILITY = -DNOUNDERSCORE 482.sphinx3=default=default=default: CPORTABILITY = -qchars=signed %endif # mymode64 #====================================================================== # INT and FP Base Flags #====================================================================== int=base=default=default: COPTIMIZE = %{BASEOPT} -qalias=noansi -qalloca CXXOPTIMIZE = %{BASEOPT} -qrtti EXTRA_CLIBS = %{LHLIB} EXTRA_CXXLIBS = -lsmartheap EXTRA_LDFLAGS = %{MODE} fp=base=default=default: COPTIMIZE = %{BASEOPT} %{LHBDT} CXXOPTIMIZE = %{BASEOPT} -qrtti -qstaticlink EXTRA_CXXLIBS = -Wl,--whole-archive /usr/lib/libhugetlbfs.a -Wl,--no-whole-archive FOPTIMIZE = %{BASEOPT} -qsmallstack=dynlenonheap -qalias=nostd %{LHBDT} EXTRA_LDFLAGS = %{MODE} #====================================================================== # # Default peak flags. # #====================================================================== default=peak=default=default: sw_peak_ptrsize = 32/64-bit %define O2 -O2 %{BASEARCH} %{DEBUG} %define O3 -O3 %{BASEARCH} %{DEBUG} %define O4 -O4 %{BASEARCH} %{DEBUG} -qipa=threads %define O5 -O5 %{BASEARCH} %{DEBUG} -qipa=threads #====================================================================== # INT peak Flags #====================================================================== int=peak=default=default: 400.perlbench=peak=default=default: # C only Benchmark # #fdpr 5503 failed with -O4 -vrox OPTIMIZE = %{O4} -qalias=noansi -qipa=level=2 EXTRA_LIBS = -lsmartheap LDCFLAGS = -Wl,-q PASS1_CFLAGS = -qpdf1 PASS1_LDFLAGS = -qpdf1 PASS2_CFLAGS = -qpdf2 PASS2_LDFLAGS = -qpdf2 fdo_pre2 = cp ${baseexe} ${baseexe}.orig ; /opt/bin/fdprpro -a instr -imullX ${baseexe}; cp ${baseexe}.instr ${baseexe} fdo_run2 = $command fdo_post2 = cp ${baseexe}.orig ${baseexe} ; /opt/bin/fdprpro -a opt -f ${baseexe}.nprof -O4 -omullX ${baseexe} ; cp ${baseexe}.fdpr ${baseexe} %ifdef %{64} CPORTABILITY = -DSPEC_CPU_LINUX_PPC -DSPEC_CPU_LP64 COPTIMIZE = -q64 %else CPORTABILITY = -DSPEC_CPU_LINUX_PPC %endif 401.bzip2=peak=default=default: # C only Benchmark OPTIMIZE = %{O3} EXTRA_LIBS = %{LHLIB} LDCFLAGS = -Wl,-q PASS1_CFLAGS = -qpdf1 PASS1_LDFLAGS = -qpdf1 PASS2_CFLAGS = -qpdf2 PASS2_LDFLAGS = -qpdf2 fdo_pre2 = cp ${baseexe} ${baseexe}.orig ; /opt/bin/fdprpro -a instr %{FDPRINT} ${baseexe}; cp ${baseexe}.instr ${baseexe} fdo_run2 = $command fdo_post2 = cp ${baseexe}.orig ${baseexe} ; /opt/bin/fdprpro -a opt -f ${baseexe}.nprof -O4 -vrox ${baseexe} ; cp ${baseexe}.fdpr ${baseexe} %ifdef %{64} COPTIMIZE = -q64 %endif 403.gcc=peak=default=default: # C only Benchmark OPTIMIZE = %{O4} -qalloca EXTRA_LIBS = %{LHLIB} LDCFLAGS = -Wl,-q PASS1_CFLAGS = -qpdf1 PASS1_LDFLAGS = -qpdf1 PASS2_CFLAGS = -qpdf2 PASS2_LDFLAGS = -qpdf2 fdo_pre2 = cp ${baseexe} ${baseexe}.orig ; /opt/bin/fdprpro -a instr %{FDPRINT} ${baseexe}; cp ${baseexe}.instr ${baseexe} fdo_run2 = $command fdo_post2 = cp ${baseexe}.orig ${baseexe} ; /opt/bin/fdprpro -a opt -f ${baseexe}.nprof -O4 -nodp -rtb ${baseexe} ; cp ${baseexe}.fdpr ${baseexe} %ifdef %{64} CPORTABILITY = -DSPEC_CPU_LP64 COPTIMIZE = -q64 %endif 429.mcf=peak=default=default: # C only Benchmark %if %{numcores} == 16 copies = 32 submit = echo "$command" > run.sh; let "CHECK=\$SPECCOPYNUM%2" ; if [[ \$CHECK -eq 0 ]]; then let "MYCPU=2*\$SPECCOPYNUM" ; fi; if [[ \$CHECK -eq 1 ]]; then let "MYCPU=2*\$SPECCOPYNUM-1" ; fi; MYMEM=`printf '%d' \$((\$MYCPU/32))` ; numactl --physcpubind=\$MYCPU --membind=\$MYMEM sh run.sh %endif OPTIMIZE = %{O5} EXTRA_LIBS = %{LHLIB} LDCFLAGS = -Wl,-q fdo_pre1 = cp ${baseexe} ${baseexe}.orig ; /opt/bin/fdprpro -a instr %{FDPRINT} ${baseexe}; cp ${baseexe}.instr ${baseexe} fdo_run1 = $command fdo_post1 = cp ${baseexe}.orig ${baseexe} ; /opt/bin/fdprpro -a opt -f ${baseexe}.nprof -O3 ${baseexe} ; cp ${baseexe}.fdpr ${baseexe} %ifdef %{64} COPTIMIZE = -q64 %endif 445.gobmk=peak=default=default: # C only Benchmark OPTIMIZE = %{O4} EXTRA_LIBS = %{LHLIB} LDCFLAGS = -Wl,-q PASS1_CFLAGS = -qpdf1 PASS1_LDFLAGS = -qpdf1 PASS2_CFLAGS = -qpdf2 PASS2_LDFLAGS = -qpdf2 fdo_pre2 = cp ${baseexe} ${baseexe}.orig ; /opt/bin/fdprpro -a instr %{FDPRINT} ${baseexe}; cp ${baseexe}.instr ${baseexe} fdo_run2 = $command fdo_post2 = cp ${baseexe}.orig ${baseexe} ; /opt/bin/fdprpro -a opt -f ${baseexe}.nprof -O3 ${baseexe} ; cp ${baseexe}.fdpr ${baseexe} %ifdef %{64} COPTIMIZE = -q64 %endif 456.hmmer=peak=default=default: # C only Benchmark OPTIMIZE = %{O5} -qsimd -qassert=refalign -qipa=inline=threshold=2888 -qipa=inline=limit=11880 EXTRA_LIBS = %{LHLIB} LDCFLAGS = -Wl,-q fdo_pre1 = cp ${baseexe} ${baseexe}.orig ; /opt/bin/fdprpro -a instr %{FDPRINT} ${baseexe}; cp ${baseexe}.instr ${baseexe} fdo_run1 = $command fdo_post1 = cp ${baseexe}.orig ${baseexe} ; /opt/bin/fdprpro -a opt -f ${baseexe}.nprof -O4 -nodp -m power7 ${baseexe} ; cp ${baseexe}.fdpr ${baseexe} %ifdef %{64} COPTIMIZE = -q64 %endif 458.sjeng=peak=default=default: # C only Benchmark OPTIMIZE = %{O5} EXTRA_LIBS = %{LHLIB} LDCFLAGS = -Wl,-q PASS1_CFLAGS = -qpdf1 PASS1_LDFLAGS = -qpdf1 PASS2_CFLAGS = -qpdf2 PASS2_LDFLAGS = -qpdf2 fdo_pre2 = cp ${baseexe} ${baseexe}.orig ; /opt/bin/fdprpro -a instr %{FDPRINT} ${baseexe}; cp ${baseexe}.instr ${baseexe} fdo_run2 = $command fdo_post2 = cp ${baseexe}.orig ${baseexe} ; /opt/bin/fdprpro -a opt -f ${baseexe}.nprof -O3 ${baseexe} ; cp ${baseexe}.fdpr ${baseexe} %ifdef %{64} COPTIMIZE = -q64 %endif 462.libquantum=peak=default=default: # C only Benchmark %if %{numcores} == 16 copies = 16 submit = echo "$command" > run.sh; let "MYCPU=4*\$SPECCOPYNUM"; MYMEM=`printf '%d' \$((\$MYCPU/32))` ; numactl --physcpubind=\$MYCPU --membind=\$MYMEM sh run.sh %endif %define 64 OPTIMIZE = %{O5} EXTRA_LIBS = %{LHLIB} LDCFLAGS = -Wl,-q PASS1_CFLAGS = -qpdf1 PASS1_LDFLAGS = -qpdf1 PASS2_CFLAGS = -qpdf2 PASS2_LDFLAGS = -qpdf2 fdo_pre2 = cp ${baseexe} ${baseexe}.orig ; /opt/bin/fdprpro -a instr %{FDPRINT} ${baseexe}; cp ${baseexe}.instr ${baseexe} fdo_run2 = $command fdo_post2 = cp ${baseexe}.orig ${baseexe} ; /opt/bin/fdprpro -a opt -f ${baseexe}.nprof -O4 -vrox -nodp ${baseexe} ; cp ${baseexe}.fdpr ${baseexe} %ifdef %{64} COPTIMIZE = -q64 %endif %undef 64 464.h264ref=peak=default=default: # C only Benchmark #%define 64 OPTIMIZE = %{O5} EXTRA_LIBS = %{LHLIB} LDCFLAGS = -Wl,-q PASS1_CFLAGS = -qpdf1 PASS1_LDFLAGS = -qpdf1 PASS2_CFLAGS = -qpdf2 PASS2_LDFLAGS = -qpdf2 fdo_pre2 = cp ${baseexe} ${baseexe}.orig ; /opt/bin/fdprpro -a instr ${baseexe}; cp ${baseexe}.instr ${baseexe} fdo_run2 = $command fdo_post2 = cp ${baseexe}.orig ${baseexe} ; /opt/bin/fdprpro -a opt -f ${baseexe}.nprof -O4 -vrox -nodp -rtb ${baseexe} ; cp ${baseexe}.fdpr ${baseexe} %ifdef %{64} COPTIMIZE = -q64 %endif #%undef 64 471.omnetpp=peak=default=default: # C++ Benchmark #OPTIMIZE = %{O5} -qrtti CXXOPTIMIZE = %{O4} -qrtti EXTRA_LIBS = -lsmartheap LDCXXFLAGS = -Wl,-q PASS1_CXXFLAGS = -qpdf1 PASS1_LDFLAGS = -qpdf1 PASS2_CXXFLAGS = -qpdf2 PASS2_LDFLAGS = -qpdf2 fdo_pre2 = cp ${baseexe} ${baseexe}.orig ; /opt/bin/fdprpro -a instr ${baseexe}; cp ${baseexe}.instr ${baseexe} fdo_run2 = $command fdo_post2 = cp ${baseexe}.orig ${baseexe} ; /opt/bin/fdprpro -a opt -f ${baseexe}.nprof -O3 -lu -1 -nodp -sdp 9 ${baseexe} ; cp ${baseexe}.fdpr ${baseexe} %ifdef %{64} CXXOPTIMIZE = -q64 %endif 473.astar=peak=default=default: # C++ Benchmark OPTIMIZE = %{O4} EXTRA_LIBS = %{LHLIB} -lsmartheap LDCXXFLAGS = -Wl,-q PASS1_CXXFLAGS = -qpdf1 PASS1_LDFLAGS = -qpdf1 PASS2_CXXFLAGS = -qpdf2 PASS2_LDFLAGS = -qpdf2 fdo_pre2 = cp ${baseexe} ${baseexe}.orig ; /opt/bin/fdprpro -a instr %{FDPRINT} ${baseexe}; cp ${baseexe}.instr ${baseexe} fdo_run2 = $command fdo_post2 = cp ${baseexe}.orig ${baseexe} ; /opt/bin/fdprpro -a opt -f ${baseexe}.nprof -O3 ${baseexe} ; cp ${baseexe}.fdpr ${baseexe} %ifdef %{64} CXXOPTIMIZE = -q64 %endif 483.xalancbmk=peak=default=default: # C++ Benchmark OPTIMIZE = %{O4} -qipa=partition=large EXTRA_LIBS = -lsmartheap LDCXXFLAGS = -Wl,-q fdo_pre1 = cp ${baseexe} ${baseexe}.orig ; /opt/bin/fdprpro -a instr %{FDPRINT} ${baseexe}; cp ${baseexe}.instr ${baseexe} fdo_run1 = $command fdo_post1 = cp ${baseexe}.orig ${baseexe} ; /opt/bin/fdprpro -a opt -f ${baseexe}.nprof -O3 -m power7 ${baseexe} ; cp ${baseexe}.fdpr ${baseexe} %ifdef %{64} CXXOPTIMIZE = -q64 %endif 999.specrand=peak=default=default: # C only Benchmark - Must run, but not reported unless fails OPTIMIZE = %{O5} EXTRA_LIBS = %{LHLIB} PASS1_CFLAGS = -qpdf1 PASS1_LDFLAGS = -qpdf1 PASS2_CFLAGS = -qpdf2 PASS2_LDFLAGS = -qpdf2 #====================================================================== # FP peak Flags #====================================================================== fp=peak=default=default: 410.bwaves=peak=default=default: # Fortran Benchmark #got 3% gain with -qsmallstack in single-threaded. %if %{numcores} == 16 copies = 16 submit = echo "$command" > run.sh; let "MYCPU=4*\$SPECCOPYNUM"; MYMEM=`printf '%d' \$((\$MYCPU/32))` ; numactl --physcpubind=\$MYCPU --membind=\$MYMEM sh run.sh %endif %if %{numcores} == 128 copies = 128 submit = echo "$command" > run.sh; let "MYCPU=4*\$SPECCOPYNUM"; MYMEM=`printf '%d' \$((\$MYCPU/32))` ; numactl --physcpubind=\$MYCPU --membind=\$MYMEM sh run.sh %endif %define 64 OPTIMIZE = %{O4} -qsmallstack=dynlenonheap EXTRA_LIBS = %{LHLIB} PASS1_FFLAGS = -qpdf1 PASS1_LDFLAGS = -qpdf1 PASS2_FFLAGS = -qpdf2 PASS2_LDFLAGS = -qpdf2 %ifdef %{64} FOPTIMIZE = -q64 %endif %undef 64 416.gamess=peak=default=default: # Fortran Benchmark OPTIMIZE = %{O5} -qalias=nostd EXTRA_LIBS = %{LHLIB} PASS1_FFLAGS = -qpdf1 PASS1_LDFLAGS = -qpdf1 PASS2_FFLAGS = -qpdf2 PASS2_LDFLAGS = -qpdf2 %ifdef %{64} FOPTIMIZE = -q64 %endif 433.milc=peak=default=default: # C only Benchmark %if %{numcores} == 16 copies = 16 submit = echo "$command" > run.sh; let "MYCPU=4*\$SPECCOPYNUM"; MYMEM=`printf '%d' \$((\$MYCPU/32))` ; numactl --physcpubind=\$MYCPU --membind=\$MYMEM sh run.sh %endif OPTIMIZE = %{O5} EXTRA_LIBS = %{LHLIB} LDCFLAGS = -Wl,-q fdo_pre1 = cp ${baseexe} ${baseexe}.orig ; /opt/bin/fdprpro -a instr %{FDPRINT} ${baseexe}; cp ${baseexe}.instr ${baseexe} fdo_run1 = $command fdo_post1 = cp ${baseexe}.orig ${baseexe} ; /opt/bin/fdprpro -a opt -f ${baseexe}.nprof -O4 -nodp ${baseexe} ; cp ${baseexe}.fdpr ${baseexe} %ifdef %{64} COPTIMIZE = -q64 %endif 434.zeusmp=peak=default=default: # Fortran Benchmark OPTIMIZE = %{O3} -qxlf90=nosignedzero EXTRA_LIBS = %{LHBDT} PASS1_FFLAGS = -qpdf1 PASS1_LDFLAGS = -qpdf1 PASS2_FFLAGS = -qpdf2 PASS2_LDFLAGS = -qpdf2 LDFFLAGS = -Wl,-q fdo_pre2 = cp ${baseexe} ${baseexe}.orig ; /opt/bin/fdprpro -a instr %{FDPRINT} ${baseexe}; cp ${baseexe}.instr ${baseexe} fdo_run2 = $command fdo_post2 = cp ${baseexe}.orig ${baseexe} ; /opt/bin/fdprpro -a opt -f ${baseexe}.nprof -O4 -vrox -nodp ${baseexe} ; cp ${baseexe}.fdpr ${baseexe} %ifdef %{64} FOPTIMIZE = -q64 %endif 435.gromacs=peak=default=default: # Fortran & C Benchmark OPTIMIZE = %{O4} -qsimd EXTRA_LIBS = %{LHLIB} LDCFLAGS = -Wl,-q LDFFLAGS = -Wl,-q PASS1_FFLAGS = -qpdf1 PASS1_CFLAGS = -qpdf1 PASS1_LDFLAGS = -qpdf1 PASS2_FFLAGS = -qpdf2 PASS2_CFLAGS = -qpdf2 PASS2_LDFLAGS = -qpdf2 fdo_pre2 = cp ${baseexe} ${baseexe}.orig ; /opt/bin/fdprpro -a instr %{FDPRINT} ${baseexe}; cp ${baseexe}.instr ${baseexe} fdo_run2 = $command fdo_post2 = cp ${baseexe}.orig ${baseexe} ; /opt/bin/fdprpro -a opt -f ${baseexe}.nprof -O4 -nodp ${baseexe} ; cp ${baseexe}.fdpr ${baseexe} %ifdef %{64} COPTIMIZE = -q64 FOPTIMIZE = -q64 %endif 436.cactusADM=peak=default=default: # Fortran & C Benchmark %if %{numcores} == 16 copies = 32 submit = echo "$command" > run.sh; let "CHECK=\$SPECCOPYNUM%2" ; if [[ \$CHECK -eq 0 ]]; then let "MYCPU=2*\$SPECCOPYNUM" ; fi; if [[ \$CHECK -eq 1 ]]; then let "MYCPU=2*\$SPECCOPYNUM-1" ; fi; MYMEM=`printf '%d' \$((\$MYCPU/32))` ; numactl --physcpubind=\$MYCPU --membind=\$MYMEM sh run.sh %endif %define 64 OPTIMIZE = %{O4} -qsimd -qnostrict FOPTIMIZE = -qsmallstack=dynlenonheap -qalias=nostd EXTRA_LIBS = %{LHLIB} %ifdef %{64} FPORTABILITY = -qfixed -qextname FPPPORTABILITY = -DSPEC_CPU_LP64 CPORTABILITY = -DSPEC_CPU_LP64 FOPTIMIZE = -q64 COPTIMIZE = -q64 %else FPORTABILITY = -qfixed -qextname %endif %undef 64 437.leslie3d=peak=default=default: # Fortran Benchmark %if %{numcores} == 16 copies = 16 submit = echo "$command" > run.sh; let "MYCPU=4*\$SPECCOPYNUM"; MYMEM=`printf '%d' \$((\$MYCPU/32))` ; numactl --physcpubind=\$MYCPU --membind=\$MYMEM sh run.sh %endif %if %{numcores} == 128 copies = 128 submit = echo "$command" > run.sh; let "MYCPU=4*\$SPECCOPYNUM"; MYMEM=`printf '%d' \$((\$MYCPU/32))` ; numactl --physcpubind=\$MYCPU --membind=\$MYMEM sh run.sh %endif %define 64 OPTIMIZE = %{O5} EXTRA_LIBS = %{LHBDT} LDFFLAGS = -Wl,-q fdo_pre1 = cp ${baseexe} ${baseexe}.orig ; /opt/bin/fdprpro -a instr %{FDPRINT} ${baseexe}; cp ${baseexe}.instr ${baseexe} fdo_run1 = $command fdo_post1 = cp ${baseexe}.orig ${baseexe} ; /opt/bin/fdprpro -a opt -f ${baseexe}.nprof -O3 -lu -1 -nodp -sdp 9 ${baseexe} ; cp ${baseexe}.fdpr ${baseexe} %ifdef %{64} FOPTIMIZE = -q64 %endif %undef 64 444.namd=peak=default=default: # C++ Benchmark OPTIMIZE = %{O5} EXTRA_LIBS = %{LHLIB} LDCXXFLAGS = -Wl,-q PASS1_CXXFLAGS = -qpdf1 PASS1_LDFLAGS = -qpdf1 PASS2_CXXFLAGS = -qpdf2 PASS2_LDFLAGS = -qpdf2 fdo_pre2 = cp ${baseexe} ${baseexe}.orig ; /opt/bin/fdprpro -a instr %{FDPRINT} ${baseexe}; cp ${baseexe}.instr ${baseexe} fdo_run2 = $command fdo_post2 = cp ${baseexe}.orig ${baseexe} ; /opt/bin/fdprpro -a opt -f ${baseexe}.nprof -O3 -lu -1 -nodp -sdp 9 ${baseexe} ; cp ${baseexe}.fdpr ${baseexe} %ifdef %{64} CXXOPTIMIZE = -q64 %endif 447.dealII=peak=default=default: # C++ Benchmark OPTIMIZE = %{O4} -qrtti EXTRA_CXXFLAGS = -qcpp_stdinc=/autobench/sources/stdcxx-4.2.1/dist/include/ansi:/autobench/sources/stdcxx-4.2.1/dist/include EXTRA_CXXLIBS = -L/autobench/sources/stdcxx-4.2.1/dist/lib -R/autobench/sources/stdcxx-4.2.1/dist/lib -lstd8d EXTRA_LIBS = -lsmartheap %ifdef %{64} CXXOPTIMIZE = -q64 %endif 450.soplex=peak=default=default: # C++ Benchmark %if %{numcores} == 16 copies = 32 submit = echo "$command" > run.sh; let "CHECK=\$SPECCOPYNUM%2" ; if [[ \$CHECK -eq 0 ]]; then let "MYCPU=2*\$SPECCOPYNUM" ; fi; if [[ \$CHECK -eq 1 ]]; then let "MYCPU=2*\$SPECCOPYNUM-1" ; fi; MYMEM=`printf '%d' \$((\$MYCPU/32))` ; numactl --physcpubind=\$MYCPU --membind=\$MYMEM sh run.sh %endif %define 64 OPTIMIZE = %{O3} EXTRA_LIBS = %{LHLIB} LDCXXFLAGS = -Wl,-q PASS1_CXXFLAGS = -qpdf1 PASS1_LDFLAGS = -qpdf1 PASS2_CXXFLAGS = -qpdf2 PASS2_LDFLAGS = -qpdf2 fdo_pre2 = cp ${baseexe} ${baseexe}.orig ; /opt/bin/fdprpro -a instr %{FDPRINT} ${baseexe}; cp ${baseexe}.instr ${baseexe} fdo_run2 = $command fdo_post2 = cp ${baseexe}.orig ${baseexe} ; /opt/bin/fdprpro -a opt -f ${baseexe}.nprof -O4 -nodp ${baseexe} ; cp ${baseexe}.fdpr ${baseexe} %ifdef %{64} CXXOPTIMIZE = -q64 %endif %undef 64 453.povray=peak=default=default: # C++ Benchmark %define 64 OPTIMIZE = %{O4} -qsimd EXTRA_LIBS = -lsmartheap64 LDCXXFLAGS = -Wl,-q PASS1_CXXFLAGS = -qpdf1 PASS1_LDFLAGS = -qpdf1 PASS2_CXXFLAGS = -qpdf2 PASS2_LDFLAGS = -qpdf2 #FDPR XL11 + 5.5.0.2 brokes it #FDPR XL11 + 5.5.0.3 looping with ref input #fdo_pre2 = cp ${baseexe} ${baseexe}.orig ; /opt/bin/fdprpro -a instr %{FDPRINT} ${baseexe}; cp ${baseexe}.instr ${baseexe} #fdo_run2 = $command #fdo_post2 = cp ${baseexe}.orig ${baseexe} ; /opt/bin/fdprpro -a opt -f ${baseexe}.nprof -O3 ${baseexe} ; cp ${baseexe}.fdpr ${baseexe} %ifdef %{64} CXXPORTABILITY = -DSPEC_CPU_LP64 CXXOPTIMIZE = -q64 %endif %undef 64 454.calculix=peak=default=default: # Fortran & C Benchmark OPTIMIZE = %{O5} EXTRA_LIBS = %{LHBDT} ## fdpr 5503 failed when using with LHBDT #LDCFLAGS = -Wl,-q #LDFFLAGS = -Wl,-q PASS1_FFLAGS = -qpdf1 PASS1_CFLAGS = -qpdf1 PASS1_LDFLAGS = -qpdf1 PASS2_FFLAGS = -qpdf2 PASS2_CFLAGS = -qpdf2 PASS2_LDFLAGS = -qpdf2 #fdo_pre2 = cp ${baseexe} ${baseexe}.orig ; /opt/bin/fdprpro -a instr ${baseexe}; cp ${baseexe}.instr ${baseexe} #fdo_run2 = $command #fdo_post2 = cp ${baseexe}.orig ${baseexe} ; /opt/bin/fdprpro -a opt -f ${baseexe}.nprof -O3 ${baseexe} ; cp ${baseexe}.fdpr ${baseexe} %ifdef %{64} FOPTIMIZE = -q64 COPTIMIZE = -q64 %endif 459.GemsFDTD=peak=default=default: # Fortran Benchmark %if %{numcores} == 16 copies = 16 submit = echo "$command" > run.sh; let "MYCPU=4*\$SPECCOPYNUM"; MYMEM=`printf '%d' \$((\$MYCPU/32))` ; numactl --physcpubind=\$MYCPU --membind=\$MYMEM sh run.sh %endif %if %{numcores} == 128 copies = 128 submit = echo "$command" > run.sh; let "MYCPU=4*\$SPECCOPYNUM"; MYMEM=`printf '%d' \$((\$MYCPU/32))` ; numactl --physcpubind=\$MYCPU --membind=\$MYMEM sh run.sh %endif OPTIMIZE = %{O4} -qsimd EXTRA_LIBS = %{LHBDT} %ifdef %{64} FOPTIMIZE = -q64 %endif 465.tonto=peak=default=default: # Fortran Benchmark OPTIMIZE = %{O5} -qsimd # SmartHeap version 9 is not supported with XL11 EXTRA_LIBS = %{LHLIB} PASS1_FFLAGS = -qpdf1 PASS1_LDFLAGS = -qpdf1 PASS2_FFLAGS = -qpdf2 PASS2_LDFLAGS = -qpdf2 LDFFLAGS = -Wl,-q fdo_pre2 = cp ${baseexe} ${baseexe}.orig ; /opt/bin/fdprpro -a instr ${baseexe}; cp ${baseexe}.instr ${baseexe} fdo_run2 = $command fdo_post2 = cp ${baseexe}.orig ${baseexe} ; /opt/bin/fdprpro -a opt -f ${baseexe}.nprof -O4 ${baseexe} ; cp ${baseexe}.fdpr ${baseexe} %ifdef %{64} FOPTIMIZE = -q64 %endif 470.lbm=peak=default=default: # C Benchmark basepeak=yes #%if %{numcores} == 16 #copies = 32 #submit = echo "$command" > run.sh; let "CHECK=\$SPECCOPYNUM%2" ; if [[ \$CHECK -eq 0 ]]; then let "MYCPU=2*\$SPECCOPYNUM" ; fi; if [[ \$CHECK -eq 1 ]]; then let "MYCPU=2*\$SPECCOPYNUM-1" ; fi; MYMEM=`printf '%d' \$((\$MYCPU/32))` ; numactl --physcpubind=\$MYCPU --membind=\$MYMEM sh run.sh #%endif #%if %{numcores} == 128 #copies = 128 #submit = echo "$command" > run.sh; let "MYCPU=4*\$SPECCOPYNUM"; MYMEM=`printf '%d' \$((\$MYCPU/32))` ; numactl --physcpubind=\$MYCPU --membind=\$MYMEM sh run.sh #%endif #OPTIMIZE = %{O5} #EXTRA_LIBS = %{LHBDT} #PASS1_CFLAGS = -qpdf1 #PASS1_LDFLAGS = -qpdf1 #PASS2_CFLAGS = -qpdf2 #PASS2_LDFLAGS = -qpdf2 #%ifdef %{64} #COPTIMIZE = -q64 #%endif 481.wrf=peak=default=default: # Fortran & C-for netCDF Benchmark #with or w/o smallstack make no diff # -O3 is better than O5 %define 64 OPTIMIZE = %{O3} EXTRA_LIBS = %{LHLIB} %ifdef %{64} COPTIMIZE = -q64 FOPTIMIZE = -q64 %endif %undef 64 482.sphinx3=peak=default=default: # C only Benchmark %if %{numcores} == 16 copies = 16 submit = echo "$command" > run.sh; let "MYCPU=4*\$SPECCOPYNUM"; MYMEM=`printf '%d' \$((\$MYCPU/32))` ; numactl --physcpubind=\$MYCPU --membind=\$MYMEM sh run.sh %endif OPTIMIZE = %{O4} EXTRA_LIBS = %{LHLIB} LDCFLAGS = -Wl,-q PASS1_CFLAGS = -qpdf1 PASS1_LDFLAGS = -qpdf1 PASS2_CFLAGS = -qpdf2 PASS2_LDFLAGS = -qpdf2 fdo_pre2 = cp ${baseexe} ${baseexe}.orig ; /opt/bin/fdprpro -a instr %{FDPRINT} ${baseexe}; cp ${baseexe}.instr ${baseexe} fdo_run2 = $command fdo_post2 = cp ${baseexe}.orig ${baseexe} ; /opt/bin/fdprpro -a opt -f ${baseexe}.nprof -O4 -nodp ${baseexe} ; cp ${baseexe}.fdpr ${baseexe} %ifdef %{64} COPTIMIZE = -q64 %endif 998.specrand=peak=default=default: # C only Benchmark - Must run, but not reported unless fails OPTIMIZE = %{O5} # The following section was added automatically, and contains settings that # did not appear in the original configuration file, but were added to the # raw file after the run. default: test_sponsor = IBM Corporation tester = IBM Corporation hw_cpu_char000 = Intelligent Energy Optimization hw_cpu_char001 = enabled, up to 3.86 GHz notes_peak_000 =IBM Post-Link Optimization tool with notes_peak_005 =options "-O4 -omullX" used for notes_peak_010 = 400.perlbench notes_peak_015 =options "-O4 -vrox" used for notes_peak_020 = 401.bzip2 notes_peak_025 =options "-O4 -nodp -rtb" notes_peak_030 = 403.gcc notes_peak_035 =options "-O3" used for notes_peak_040 = 429.mcf 445.gobmk 458.sjeng 473.astar notes_peak_045 =options "-O4 -nodp -m power7" used for notes_peak_050 = 456.hmmer notes_peak_055 =options "-O4 -vrox -nodp" used for notes_peak_060 = 462.libquantum notes_peak_065 =options "-O4 -vrox -nodp -rtb" used for notes_peak_070 = 464.h264ref notes_peak_075 =options "-O3 -lu -1 -nodp -sdp 9" used for notes_peak_080 = 471.omnetpp notes_peak_085 =options "-O3 -m power7" used for notes_peak_090 = 483.xalancbmk notes_peak_095 =Whenever option "-omullX" was used during the optimization phase, notes_peak_100 =option "-imullX" was also used during the instrumentation phase. notes_os_000 =ulimit -s (stack) set to 1048576. notes_os_005 =Large pages reserved as follows by root user: notes_os_010 = echo 4224 > /proc/sys/vm/nr_hugepages notes_os_015 =The following environment varibles were set before the runspec command: notes_os_020 = XLFRTEOPTS=intrinthds=1 notes_os_025 = HUGETLB_VERBOSE=0 notes_os_030 = HUGETLB_MORECORE=yes notes_os_035 = HUGETLB_ELFMAP=RW notes_submit_000 = Benchmarks bound to a processor using numactl on the submit command.