# Invocation command line: # /home/cpu2017-1.1.5-amd-aocc300-milan-B1/bin/harness/runcpu --configfile amd_speed_aocc300_milan_B1.cfg --tune all --reportable --iterations 3 --nopower --runmode speed --tune base:peak --size test:train:refspeed intspeed # output_root was not used for this run ############################################################################ ################################################################################ # AMD AOCC 300 SPEC CPU2017 V1.1.5 Speed Configuration File for 64-bit Linux # # File name : amd_speed_aocc300_milan_B1.cfg # Creation Date : February 23, 2021 # CPU2017 Version : 1.1.5 # Supported benchmarks : All Speed benchmarks (intspeed, fpspeed) # Compiler name/version : AOCC 3.0.0 # Operating system version : OpenSUSE 15.2 # Supported OS's : Ubuntu 20.04, RHEL 8.3, SLES 15 SP2 # Hardware : AMD Milan, Rome, Naples (AMD64) # FP Base Pointer Size : 64-bit # FP Peak Pointer Size : 64-bit # INT Base Pointer Size : 64-bit # INT Peak Pointer Size : 64-bit # Auto Parallization : No # # Note: DO NOT EDIT THIS FILE, the only edits required to properly run these # binaries are made in the ini Python file. Please consult Readme.amd_speed_aocc300_milan_B1.txt # for a few uncommon exceptions which require edits to this file. # # Description: # # This binary package automates away many of the complexities necessary to set # up and run SPEC CPU2017 under optimized conditions on AMD Milan/Rome/Naples-based # server platforms within Linux (AMD64). # # The binary package was built specifically for AMD Milan/Rome/Naples microprocessors and # is not intended to run on other products. # # Please install the binary package by following the instructions in # "Readme.amd_speed_aocc300_milan_B1.txt" under the "How To Use the Binaries" section. # # The binary package is designed to work without alteration on two socket AMD # Milan/Rome/Naples-based servers with 64 cores per socket, SMT enabled and 1 TiB of DDR4 # memory distributed evenly among all 16 channels using 32 GiB DIMMs. # # To run the binary package on other Milan/Rome/Naples configurations, please review # "Readme.amd_speed_aocc300_milan_B1.txt". In general, Milan/Rome or Naples CPUs # should be autodetected with no action required by the user. # # In most cases, it should be unnecessary to edit "amd_speed_aocc300_milan_B1.cfg" or any # other file besides "ini_amd_speed_aocc300_milan_B1.py" where reporting fields # and run conditions are set. # # The run script automatically sets the optimal number of speed copies and binds # them appropriately. # # The run script and accompanying binary package are designed to work on Ubuntu # 20.04, RHEL 8.3 and SLES 15 SP2. # # Important! If you write your own run script, please set the stack size to # "unlimited" when executing this binary package. Failure to do so may cause # some benchmarks to overflow the stack. For example, to set stack size within # the bash shell, include the following line somewhere at the top of your run # script before the runcpu invocation: # # ulimit -s unlimited # # Modification of this config file should only be necessary if you intend to # rebuild the binaries. General instructions for rebuilding the binaries are # found in-line below. # ### add power options power_analyzer = WIN:9888 temp_meter = WIN:9889 ################################################################################ # Modifiable macros: ################################################################################ # Change the following line to true if you intend to REBUILD the binaries (AMD # does not support this). Valid values are "true" or "false" (no quotes). %define allow_build false # Only change these macros if you are rebuilding the binary package: %define compiler_name aocc300 %define binary_package_name amd_speed_%{compiler_name}_milan_B %define binary_package_revision 1 %define build_path /sppo/bin/cpu2017v115aocc3-b1/ %define flags_file_name %{compiler_name}-flags-B1.xml # To enable the platform file, be sure to uncomment the flagsurl02 header line # below. %define platform_file_name INVALID_platform_%{binary_package_name}.xml # You should never have to change binary_package_full_name: %define binary_package_full_name %{binary_package_name}%{binary_package_revision} ################################################################################ # Include file name ################################################################################ # The include file contains fields that are commonly changed. This file is auto- # generated based upon INI file settings and should not need user modification # for runs. %define inc_file_name %{binary_package_full_name}.inc ################################################################################ # Binary label extension and "allow_build"" switch ################################################################################ # Only modify the binary label extension if you plan to rebuild the binaries. %define ext %{binary_package_name} # If you plan to recompile these CPU2017 binaries, please choose a new extension # name (ext above) to avoid confusion with the current binary set on your system # under test, and to avoid confusion for SPEC submission reviewers. You will # also need to set "allow_build" to true below. Finally, you must modify the # Paths section below to point to your library locations if the paths are not # already set up in your build environment. ################################################################################ # Paths and Environment Variables # ** MODIFY AS NEEDED (modification should not be necessary for runs) ** ################################################################################ # Allow environment variables to be set before runs: preenv = 1 # Necessary to avoid gcc out-of-memory exceptions on certain SUTs: preENV_MALLOC_CONF = retain:true # OpenMP environment variables: preENV_OMP_SCHEDULE = static preENV_OMP_DYNAMIC = false preENV_OMP_STACKSIZE = 128M # Define the name of the directory that holds AMD library files: %define lib_dir %{binary_package_name}_lib # Set the shared object library path for runs and builds: preENV_LD_LIBRARY_PATH = $[top]/%{lib_dir}/64;$[top]/%{lib_dir}/32:%{ENV_LD_LIBRARY_PATH} # Define 32-bit library build paths: # Do not use $[top] with the 32-bit libraries because doing so will cause an # options checksum error triggering a xalanc recompile attempt on SUTs having # different file paths: JEMALLOC_LIB32_PATH = %{build_path}%{lib_dir}/32 %if '%{allow_build}' eq 'false' # The include file is only needed for runs, but not for builds. # include: %{inc_file_name} # ----- Begin inclusion of 'amd_speed_aocc300_milan_B1.inc' ############################################################################ ################################################################################ ################################################################################ # File name: amd_speed_aocc300_milan_B1.inc # File generation code date: January 25, 2021 # File generation date/time: June 22, 2018 / 19:23:10 # # This file is automatically generated during a SPEC CPU2017 run. # # To modify inc file generation, please consult the readme file or the run # script. ################################################################################ ################################################################################ ################################################################################ ################################################################################ # The following macros are generated for use in the cfg file. ################################################################################ ################################################################################ %define logical_core_count 256 %define physical_core_count 128 %define physical_core_max 127 %define logical_core_max 255 ################################################################################ ################################################################################ # The following inc blocks set the speed thread counts and affinity settings. # # intspeed benchmarks: 600.perlbench_s,602.gcc_s,605.mcf_s,620.omnetpp_s, # 623.xalancbmk_s,625.x264_s,631.deepsjeng_s,641.leela_s,648.exchange2_s, # 657.xz_s # fpspeed benchmarks: 603.bwaves_s,607.cactuBSSN_s,619.lbm_s,621.wrf_s, # 627.cam4_s,628.pop2_s,638.imagick_s,644.nab_s,649.fotonik3d_s, # 654.roms_s # # Selected thread counts from '7763' section of CPU info ################################################################################ # default preENV thread settings: default: preENV_OMP_THREAD_LIMIT = 256 preENV_GOMP_CPU_AFFINITY = 0-255 ################################################################################ ################################################################################ # intspeed base thread counts: intspeed=base: threads = 128 ENV_GOMP_CPU_AFFINITY = 0-127 bind0 = numactl --physcpubind=0-127 submit = echo "$command" > run.sh ; $BIND bash run.sh ################################################################################ ################################################################################ # fpspeed base thread counts: fpspeed=base: threads = 128 ENV_GOMP_CPU_AFFINITY = 0-127 bind0 = numactl --physcpubind=0-127 submit = echo "$command" > run.sh ; $BIND bash run.sh ################################################################################ ################################################################################ # peak thread counts: 1 600.perlbench_s,602.gcc_s,605.mcf_s,620.omnetpp_s,623.xalancbmk_s,625.x264_s,631.deepsjeng_s,641.leela_s,648.exchange2_s=peak: threads = 1 ENV_GOMP_CPU_AFFINITY = 0 bind0 = numactl --physcpubind=0 submit = echo "$command" > run.sh ; $BIND bash run.sh ################################################################################ ################################################################################ # peak thread counts: 128 603.bwaves_s,607.cactuBSSN_s,619.lbm_s,621.wrf_s,627.cam4_s,628.pop2_s,638.imagick_s,649.fotonik3d_s,654.roms_s,657.xz_s=peak: threads = 128 ENV_GOMP_CPU_AFFINITY = 0-127 bind0 = numactl --physcpubind=0-127 submit = echo "$command" > run.sh ; $BIND bash run.sh ################################################################################ ################################################################################ # peak thread counts: 256 644.nab_s=peak: threads = 256 ENV_GOMP_CPU_AFFINITY = 0 128 1 129 2 130 3 131 4 132 5 133 6 134 7 135 8 136 9 137 10 138 11 139 12 140 13 141 14 142 15 143 16 144 17 145 18 146 19 147 20 148 21 149 22 150 23 151 24 152 25 153 26 154 27 155 28 156 29 157 30 158 31 159 32 160 33 161 34 162 35 163 36 164 37 165 38 166 39 167 40 168 41 169 42 170 43 171 44 172 45 173 46 174 47 175 48 176 49 177 50 178 51 179 52 180 53 181 54 182 55 183 56 184 57 185 58 186 59 187 60 188 61 189 62 190 63 191 64 192 65 193 66 194 67 195 68 196 69 197 70 198 71 199 72 200 73 201 74 202 75 203 76 204 77 205 78 206 79 207 80 208 81 209 82 210 83 211 84 212 85 213 86 214 87 215 88 216 89 217 90 218 91 219 92 220 93 221 94 222 95 223 96 224 97 225 98 226 99 227 100 228 101 229 102 230 103 231 104 232 105 233 106 234 107 235 108 236 109 237 110 238 111 239 112 240 113 241 114 242 115 243 116 244 117 245 118 246 119 247 120 248 121 249 122 250 123 251 124 252 125 253 126 254 127 255 bind0 = numactl --physcpubind=0-255 submit = echo "$command" > run.sh ; $BIND bash run.sh ################################################################################ ################################################################################ ################################################################################ # Switch back to default: default: ################################################################################ ################################################################################ ################################################################################ # The remainder of this file defines CPU2017 report parameters. ################################################################################ ################################################################################ ################################################################################ # SPEC CPU 2017 report header ################################################################################ license_num =9017 tester =Lenovo Global Technology test_sponsor =Lenovo Global Technology hw_vendor =Lenovo Global Technology hw_model000 =ThinkSystem SR665 hw_model001 =2.45 GHz, AMD EPYC 7763 #--------- If you install new compilers, edit this section -------------------- sw_compiler =C/C++/Fortran: Version 3.0.0 of AOCC ################################################################################ ################################################################################ # Hardware, firmware and software information ################################################################################ hw_avail =Mar-2021 sw_avail =Mar-2021 hw_cpu_name =AMD EPYC 7763 hw_cpu_nominal_mhz =2450 hw_cpu_max_mhz =3500 hw_ncores =128 hw_nthreadspercore =2 hw_ncpuorder =1,2 chips hw_other =None # Other perf-relevant hw, or "None" fw_bios =Lenovo BIOS Version D8E115B 2.00 released Feb-2021 sw_base_ptrsize =64-bit sw_peak_ptrsize = 64-bit hw_pcache =32 KB I + 32 KB D on chip per core hw_scache =512 KB I+D on chip per core hw_tcache000 =256 MB I+D on chip per chip, hw_tcache001 = 32 MB shared / 8 cores hw_ocache =None ################################################################################ # Notes ################################################################################ # Enter notes_000 through notes_100 here. notes_000 =Binaries were compiled on a system with 2x AMD EPYC 7742 CPU + 1TiB Memory using openSUSE 15.2 notes_005 = notes_010 =NA: The test sponsor attests, as of date of publication, that CVE-2017-5754 (Meltdown) notes_015 =is mitigated in the system as tested and documented. notes_020 =Yes: The test sponsor attests, as of date of publication, that CVE-2017-5753 (Spectre variant 1) notes_025 =is mitigated in the system as tested and documented. notes_030 =Yes: The test sponsor attests, as of date of publication, that CVE-2017-5715 (Spectre variant 2) notes_035 =is mitigated in the system as tested and documented. notes_040 = notes_submit_000 ='numactl' was used to bind copies to the cores. notes_submit_005 =See the configuration file for details. notes_os_000 ='ulimit -s unlimited' was used to set environment stack size notes_os_005 ='ulimit -l 2097152' was used to set environment locked pages in memory limit notes_os_010 = notes_os_015 =runcpu command invoked through numactl i.e.: notes_os_020 =numactl --interleave=all runcpu notes_os_025 = notes_os_030 ='echo 8 > /proc/sys/vm/dirty_ratio' run as root to limit dirty cache to 8% of notes_os_035 =memory. notes_os_040 ='echo 1 > /proc/sys/vm/swappiness' run as root to limit swap usage to minimum notes_os_045 =necessary. notes_os_050 ='echo 1 > /proc/sys/vm/zone_reclaim_mode' run as root to free node-local memory notes_os_055 =and avoid remote memory usage. notes_os_060 ='sync; echo 3 > /proc/sys/vm/drop_caches' run as root to reset filesystem caches. notes_os_065 ='sysctl -w kernel.randomize_va_space=0' run as root to disable address space layout notes_os_070 =randomization (ASLR) to reduce run-to-run variability. notes_os_075 = notes_os_016 =For peak, transparent huge pages set to 'never' for 627, 644, 649, 654 notes_os_080 ='echo always > /sys/kernel/mm/transparent_hugepage/enabled' and notes_os_085 ='echo always > /sys/kernel/mm/transparent_hugepage/defrag' run as root to enable notes_os_090 =Transparent Hugepages (THP) for this run. notes_os_095 ='echo madvise > /sys/kernel/mm/transparent_hugepage/enabled' run as root for peak notes_os_100 =runs of 628.pop2_s and 638.imagick_s to enable THP only on request. notes_comp_000 =The AMD64 AOCC Compiler Suite is available at notes_comp_005 =http://developer.amd.com/amd-aocc/ notes_comp_010 = notes_jemalloc_000 =jemalloc: configured and built with GCC v4.8.2 in RHEL 7.4 (No options specified) notes_jemalloc_005 =jemalloc 5.1.0 is available here: notes_jemalloc_010 =https://github.com/jemalloc/jemalloc/releases/download/5.1.0/jemalloc-5.1.0.tar.bz2 notes_jemalloc_015 = sw_other =jemalloc: jemalloc memory allocator library v5.1.0 ################################################################################ # The following note fields describe platorm settings. ################################################################################ # example: (uncomment as necessary) # notes_plat_000 =BIOS settings: # notes_plat_002 = cTDP: 280 # notes_plat_004 = Determinism Slider set to Power # notes_plat_006 = Package Power: 280 # notes_plat_008 = EDC: 300 # notes_plat_010 = NPS: 1 # notes_plat_014 = 4-link xGMI max speed: 16Gbps # notes_plat_015 = Fan Speed: Maximum ################################################################################ # The following are custom fields: ################################################################################ # Use custom_fields to enter lines that are not listed here. For example: # notes_plat_100 = Energy Bias set to Max Performance # new_field = Ambient temperature set to 10C ################################################################################ # The following fields must be set here for only Int benchmarks. ################################################################################ intrate: sw_peak_ptrsize =64-bit ################################################################################ # The following fields must be set here for FP benchmarks. ################################################################################ fprate: sw_peak_ptrsize =64-bit ################################################################################ # The following fields must be set here or they will be overwritten by sysinfo. ################################################################################ intrate,fprate: hw_disk =unknown hw_memory000 =1 TB (16 x 64 GB 4Rx4 PC4-3200V-L) hw_memory001 =running at 2933 hw_memory002 = hw_nchips =2 prepared_by =prepared by unknown sw_file =unknown file sw_os000 =unknown os000 sw_os001 =kernel version # ex: Kernel 4.4.0-87-generic sw_state =Run level 3 (multi-user) ################################################################################ # End of inc file ################################################################################ # Switch back to the default block after the include file: default: # ---- End inclusion of '/home/cpu2017-1.1.5-amd-aocc300-milan-B1/config/amd_speed_aocc300_milan_B1.inc' # Switch back to default block after the include file: default: fail_build = 1 %elif '%{allow_build}' eq 'true' # If you intend to rebuild, be sure to set the library paths either in the # build script or here: preENV_LIBRARY_PATH = $[top]/%{lib_dir}/64;$[top]/%{lib_dir}/32:%{ENV_LIBRARY_PATH} % define build_ncpus 64 # controls number of simultaneous compiles fail_build = 0 makeflags = --jobs=%{build_ncpus} --load-average=%{build_ncpus} %else % error The value of "allow_build" is %{allow_build}, but it can only be "true" or "false". This error was generated %endif ################################################################################ # Enable automated data collection per benchmark ################################################################################ # Data collection is not enabled for reportable runs. # teeout is necessary to get data collection stdout into the logs. Best # practices for the individual data collection items would be to have # them store important output in separate files. Filenames could be # constructed from $SPEC (environment), $lognum (result number from runcpu), # and benchmark name/number. teeout = yes # Run runcpu with '-v 35' (or greater) to log lists of variables which can # be used in substitutions as below. # For CPU2006, change $label to $ext %define data-collection-parameters benchname='$name' benchnum='$num' benchmark='$benchmark' iteration=$iter size='$size' tune='$tune' label='$label' log='$log' lognum='$lognum' from_runcpu='$from_runcpu' %define data-collection-start $[top]/data-collection/data-collection start %{data-collection-parameters} %define data-collection-stop $[top]/data-collection/data-collection stop %{data-collection-parameters} monitor_specrun_wrapper = %{data-collection-start} ; $command ; %{data-collection-stop} ################################################################################ # Header settings ################################################################################ backup_config = 0 # set to 0 if you do not want backup files bench_post_setup = sync # command_add_redirect: If set, the generated ${command} will include # redirection operators (stdout, stderr), which are passed along to the shell # that executes the command. If this variable is not set, specinvoke does the # redirection. command_add_redirect = yes env_vars = yes flagsurl000 = http://www.spec.org/cpu2017/flags/Lenovo-Platform-SPECcpu2017-Flags-V1.2-Milan-C.xml flagsurl001 = http://www.spec.org/cpu2017/flags/aocc300-flags-A1.xml #flagsurl02 = $[top]/%{platform_file_name} # label: User defined extension string that tags your binaries & directories: label = %{ext} line_width = 1020 log_line_width = 1020 mean_anyway = yes output_format = all reportable = yes size = test,train,ref teeout = yes teerunout = yes tune = base,peak use_submit_for_speed = yes ################################################################################ # Compilers ################################################################################ default: CROSSPLAT_PORT_OPTS = -mno-adx -mno-sse4a CC = clang -m64 $[CROSSPLAT_PORT_OPTS] CXX = clang++ -m64 -std=c++98 $[CROSSPLAT_PORT_OPTS] FC = flang -m64 $[CROSSPLAT_PORT_OPTS] CLD = clang -m64 CXXLD = clang++ -m64 FLD = flang -m64 CC_VERSION_OPTION = --version CXX_VERSION_OPTION = --version FC_VERSION_OPTION = --version ################################################################################ # Portability Flags ################################################################################ default:# data model applies to all benchmarks ################################################################################ # Default Flags ################################################################################ EXTRA_LIBS = -fopenmp=libomp -lomp -ljemalloc -lamdlibm -lm MATHLIBOPT = #clearing this variable or else SPEC will set it to -lm VECMATHLIB = -fveclib=AMDLIBM OPT_ROOT = -march=znver3 $(VECMATHLIB) -ffast-math OPT_ROOT_BASE = -O3 $(OPT_ROOT) OPT_ROOT_PEAK = -Ofast $(OPT_ROOT) -flto ################################################################################ # Portability Flags ################################################################################ default: EXTRA_PORTABILITY = -DSPEC_LP64 # *** Benchmark-specific portability *** # Anything other than the data model is only allowed where a need is proven. # (ordered by last 2 digits of benchmark number) 600.perlbench_s: #lang='C' PORTABILITY = -DSPEC_LINUX_X64 621.wrf_s: #lang='F,C' CPORTABILITY = -DSPEC_CASE_FLAG FPORTABILITY = -Mbyteswapio 623.xalancbmk_s: #lang='CXX' PORTABILITY = -DSPEC_LINUX 627.cam4_s: #lang='F,C' PORTABILITY = -DSPEC_CASE_FLAG 628.pop2_s: #lang='F,C' CPORTABILITY = -DSPEC_CASE_FLAG FPORTABILITY = -Mbyteswapio ################################################################################ # Tuning Flags ################################################################################ ##################### # Base tuning flags # ##################### default=base: COPTIMIZE = $(OPT_ROOT_BASE) -flto -fstruct-layout=5 \ -mllvm -unroll-threshold=50 \ -mllvm -inline-threshold=1000 -fremap-arrays \ -mllvm -function-specialize -flv-function-specialization \ -mllvm -enable-gvn-hoist \ -mllvm -global-vectorize-slp=true \ -mllvm -enable-licm-vrp \ -mllvm -reduce-array-computations=3 \ -Wno-unused-command-line-argument CXXOPTIMIZE = $(OPT_ROOT_BASE) -flto \ -mllvm -enable-partial-unswitch \ -mllvm -unroll-threshold=100 \ -finline-aggressive -flv-function-specialization \ -mllvm -loop-unswitch-threshold=200000 \ -mllvm -reroll-loops \ -mllvm -aggressive-loop-unswitch \ -mllvm -extra-vectorizer-passes \ -mllvm -reduce-array-computations=3 \ -mllvm -global-vectorize-slp=true \ -Wno-unused-command-line-argument \ -mllvm -convert-pow-exp-to-int=false FOPTIMIZE = -Hz,1,0x1 $(OPT_ROOT_BASE) -Mrecursive \ -mllvm -fuse-tile-inner-loop -funroll-loops \ -mllvm -extra-vectorizer-passes \ -mllvm -lsr-in-nested-loop \ -mllvm -enable-licm-vrp \ -mllvm -reduce-array-computations=3 \ -mllvm -global-vectorize-slp=true \ -Wno-unused-command-line-argument LDCXXFLAGS = -Wl,-mllvm -Wl,-x86-use-vzeroupper=false EXTRA_LDFLAGS = -Wl,-mllvm -Wl,-region-vectorize \ -Wl,-mllvm -Wl,-function-specialize \ -Wl,-mllvm -Wl,-align-all-nofallthru-blocks=6 \ -Wl,-mllvm -Wl,-reduce-array-computations=3 LDFFLAGS = -Wl,-mllvm -Wl,-enable-X86-prefetching \ -Wl,-mllvm -Wl,-enable-licm-vrp #other libraries # Put OpenMP and math libraries here: # -lm needed at the end for some transcendental functions: EXTRA_LIBS = -fopenmp=libomp -lomp -lamdlibm -ljemalloc -lflang -lflangrti -lm EXTRA_FLIBS = # Don't put the AMD and mvec math libraries in MATHLIBOPT because it will trigger a reporting issue # because GCC won't use them. Forcefeed all benchmarks the math libraries in EXTRA_LIBS and clear # out MATHLIBOPT. MATHLIBOPT = # The following is necessary for 502/602 gcc: LDOPTIMIZE = -z muldefs # The following is necessary for 502/602 gcc: EXTRA_OPTIMIZE = -DSPEC_OPENMP -fopenmp -Wno-return-type ######################## # intspeed tuning flags # ######################## intspeed: FOPTIMIZE = $(OPT_ROOT_BASE) -flto EXTRA_FFLAGS = -mllvm -unroll-aggressive \ -mllvm -unroll-threshold=150 EXTRA_CXXFLAGS = -mllvm -do-block-reorder=aggressive \ -fvirtual-function-elimination -fvisibility=hidden LDCFLAGS = -Wl,-allow-multiple-definition -Wl,-mllvm \ -Wl,-enable-licm-vrp LDCXXFLAGS = -Wl,-mllvm -Wl,-do-block-reorder=aggressive LDFFLAGS = -Wl,-mllvm -Wl,-inline-recursion=4 \ -Wl,-mllvm -Wl,-lsr-in-nested-loop \ -Wl,-mllvm -Wl,-enable-iv-split intspeed=base: submit = echo always > /sys/kernel/mm/transparent_hugepage/enabled; $BIND $command intspeed=peak: submit = echo always > /sys/kernel/mm/transparent_hugepage/enabled; $BIND $command ######################## # fpspeed tuning flags # ######################## fpspeed: CXX = clang++ -m64 -std=c++98 $[CROSSPLAT_PORT_OPTS] fpspeed=base: submit = echo always > /sys/kernel/mm/transparent_hugepage/enabled; $BIND $command fpspeed=peak: submit = echo always > /sys/kernel/mm/transparent_hugepage/enabled; $BIND $command ##################### # Peak tuning flags # ##################### default=peak: COPTIMIZE = $(OPT_ROOT_PEAK) -fstruct-layout=5 \ -mllvm -unroll-threshold=50 -fremap-arrays \ -flv-function-specialization -mllvm \ -inline-threshold=1000 -mllvm -enable-gvn-hoist \ -mllvm -global-vectorize-slp=true -mllvm \ -function-specialize -mllvm -enable-licm-vrp \ -mllvm -reduce-array-computations=3 \ -Wno-unused-command-line-argument CXXOPTIMIZE = $(OPT_ROOT_PEAK) -finline-aggressive \ -mllvm -unroll-threshold=100 \ -flv-function-specialization -mllvm -enable-licm-vrp \ -mllvm -reroll-loops -mllvm \ -aggressive-loop-unswitch -mllvm \ -reduce-array-computations=3 -mllvm \ -global-vectorize-slp=true \ -Wno-unused-command-line-argument FOPTIMIZE = $(OPT_ROOT_PEAK) -Mrecursive \ -mllvm -reduce-array-computations=3 \ -mllvm -global-vectorize-slp=true \ -mllvm -enable-licm-vrp \ -Wno-unused-command-line-argument EXTRA_LDFLAGS = -Wl,-mllvm -Wl,-function-specialize \ -Wl,-mllvm -Wl,-align-all-nofallthru-blocks=6 \ -Wl,-mllvm -Wl,-reduce-array-computations=3 LDFFLAGS = -Wl,-mllvm -Wl,-enable-X86-prefetching \ -Wl,-mllvm -Wl,-enable-licm-vrp LDCXXFLAGS = -Wl,-mllvm -Wl,-x86-use-vzeroupper=false \ -Wl,-mllvm -Wl,-enable-licm-vrp EXTRA_LIBS = -fopenmp=libomp -lomp -lamdlibm -ljemalloc -lflang -lm EXTRA_OPTIMIZE = -DSPEC_OPENMP -fopenmp -Wno-return-type feedback = 0 PASS1_CFLAGS = -fprofile-instr-generate PASS2_CFLAGS = -fprofile-instr-use PASS1_FFLAGS = -fprofile-generate PASS2_FFLAGS = -fprofile-use PASS1_CXXFLAGS = -fprofile-instr-generate PASS2_CXXFLAGS = -fprofile-instr-use PASS1_LDFLAGS = -fprofile-instr-generate PASS2_LDFLAGS = -fprofile-instr-use fdo_run1 = $command ; llvm-profdata merge --output=default.profdata *.profraw # Int benchmark specific peak tuning flags: # FP benchmark specific peak tuning flags: 603.bwaves_s=peak: FOPTIMIZE = -Ofast $(OPT_ROOT) -Mrecursive \ -mllvm -reduce-array-computations=3 \ -mllvm -global-vectorize-slp=true \ -mllvm -enable-licm-vrp \ -Wno-unused-command-line-argument submit = echo always > /sys/kernel/mm/transparent_hugepage/enabled; $BIND $command 607.cactuBSSN_s=peak: submit = echo always > /sys/kernel/mm/transparent_hugepage/enabled; $BIND $command 621.wrf_s=peak: FOPTIMIZE = -Hz,1,0x1 $(OPT_ROOT_BASE) -Mrecursive \ -mllvm -fuse-tile-inner-loop -funroll-loops \ -mllvm -extra-vectorizer-passes \ -mllvm -lsr-in-nested-loop \ -mllvm -enable-licm-vrp \ -mllvm -reduce-array-computations=3 \ -mllvm -global-vectorize-slp=true \ -Wno-unused-command-line-argument submit = echo always > /sys/kernel/mm/transparent_hugepage/enabled; numactl --interleave=all --physcpubind=0-%{physical_core_max} $command 627.cam4_s=peak: submit = echo never > /sys/kernel/mm/transparent_hugepage/enabled; $BIND $command 628.pop2_s=peak: FOPTIMIZE = $(OPT_ROOT) -Ofast -Mrecursive \ -mllvm -reduce-array-computations=3 \ -mllvm -global-vectorize-slp=true \ -mllvm -enable-licm-vrp \ -Wno-unused-command-line-argument submit = echo madvise > /sys/kernel/mm/transparent_hugepage/enabled; $BIND $command 638.imagick_s=peak: submit = echo madvise > /sys/kernel/mm/transparent_hugepage/enabled; $BIND $command 644.nab_s=peak: EXTRA_LDFLAGS = -Wl,-mllvm -Wl,-region-vectorize \ -Wl,-mllvm -Wl,-function-specialize submit = echo never > /sys/kernel/mm/transparent_hugepage/enabled; $BIND $command 649.fotonik3d_s=peak: ENV_PGHPF_ZMEM =yes submit = echo never > /sys/kernel/mm/transparent_hugepage/enabled; $BIND $command 654.roms_s=peak: FOPTIMIZE = -Ofast $(OPT_ROOT) -Mrecursive \ -mllvm -reduce-array-computations=3 \ -mllvm -global-vectorize-slp=true \ -mllvm -enable-licm-vrp \ -Wno-unused-command-line-argument submit = echo never > /sys/kernel/mm/transparent_hugepage/enabled; $BIND $command # The following settings were obtained by running the sysinfo_program # 'specperl $[top]/bin/sysinfo' (sysinfo:SHA:60a26e139a7df7ba5521c983304469c762a79f3394ac112dddae4bac7d1a4f55) default: notes_plat_sysinfo_000 = notes_plat_sysinfo_005 = Sysinfo program /home/cpu2017-1.1.5-amd-aocc300-milan-B1/bin/sysinfo notes_plat_sysinfo_010 = Rev: r6538 of 2020-09-24 e8664e66d2d7080afeaa89d4b38e2f1c notes_plat_sysinfo_015 = running on localhost.localdomain Fri Jun 22 19:23:19 2018 notes_plat_sysinfo_020 = notes_plat_sysinfo_025 = SUT (System Under Test) info as seen by some common utilities. notes_plat_sysinfo_030 = For more information on this section, see notes_plat_sysinfo_035 = https://www.spec.org/cpu2017/Docs/config.html#sysinfo notes_plat_sysinfo_040 = notes_plat_sysinfo_045 = From /proc/cpuinfo notes_plat_sysinfo_050 = model name : AMD EPYC 7763 64-Core Processor notes_plat_sysinfo_055 = 2 "physical id"s (chips) notes_plat_sysinfo_060 = 256 "processors" notes_plat_sysinfo_065 = cores, siblings (Caution: counting these is hw and system dependent. The following notes_plat_sysinfo_070 = excerpts from /proc/cpuinfo might not be reliable. Use with caution.) notes_plat_sysinfo_075 = cpu cores : 64 notes_plat_sysinfo_080 = siblings : 128 notes_plat_sysinfo_085 = physical 0: cores 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 notes_plat_sysinfo_090 = 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 notes_plat_sysinfo_095 = 53 54 55 56 57 58 59 60 61 62 63 notes_plat_sysinfo_100 = physical 1: cores 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 notes_plat_sysinfo_105 = 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 notes_plat_sysinfo_110 = 53 54 55 56 57 58 59 60 61 62 63 notes_plat_sysinfo_115 = notes_plat_sysinfo_120 = From lscpu: notes_plat_sysinfo_125 = Architecture: x86_64 notes_plat_sysinfo_130 = CPU op-mode(s): 32-bit, 64-bit notes_plat_sysinfo_135 = Byte Order: Little Endian notes_plat_sysinfo_140 = CPU(s): 256 notes_plat_sysinfo_145 = On-line CPU(s) list: 0-255 notes_plat_sysinfo_150 = Thread(s) per core: 2 notes_plat_sysinfo_155 = Core(s) per socket: 64 notes_plat_sysinfo_160 = Socket(s): 2 notes_plat_sysinfo_165 = NUMA node(s): 4 notes_plat_sysinfo_170 = Vendor ID: AuthenticAMD notes_plat_sysinfo_175 = CPU family: 25 notes_plat_sysinfo_180 = Model: 1 notes_plat_sysinfo_185 = Model name: AMD EPYC 7763 64-Core Processor notes_plat_sysinfo_190 = Stepping: 1 notes_plat_sysinfo_195 = CPU MHz: 2950.832 notes_plat_sysinfo_200 = CPU max MHz: 2450.0000 notes_plat_sysinfo_205 = CPU min MHz: 1500.0000 notes_plat_sysinfo_210 = BogoMIPS: 4890.47 notes_plat_sysinfo_215 = Virtualization: AMD-V notes_plat_sysinfo_220 = L1d cache: 32K notes_plat_sysinfo_225 = L1i cache: 32K notes_plat_sysinfo_230 = L2 cache: 512K notes_plat_sysinfo_235 = L3 cache: 32768K notes_plat_sysinfo_240 = NUMA node0 CPU(s): 0-31,128-159 notes_plat_sysinfo_245 = NUMA node1 CPU(s): 32-63,160-191 notes_plat_sysinfo_250 = NUMA node2 CPU(s): 64-95,192-223 notes_plat_sysinfo_255 = NUMA node3 CPU(s): 96-127,224-255 notes_plat_sysinfo_260 = Flags: fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov notes_plat_sysinfo_265 = pat pse36 clflush mmx fxsr sse sse2 ht syscall nx mmxext fxsr_opt pdpe1gb rdtscp lm notes_plat_sysinfo_270 = constant_tsc rep_good nopl nonstop_tsc cpuid extd_apicid aperfmperf pni pclmulqdq notes_plat_sysinfo_275 = monitor ssse3 fma cx16 pcid sse4_1 sse4_2 x2apic movbe popcnt aes xsave avx f16c notes_plat_sysinfo_280 = rdrand lahf_lm cmp_legacy svm extapic cr8_legacy abm sse4a misalignsse 3dnowprefetch notes_plat_sysinfo_285 = osvw ibs skinit wdt tce topoext perfctr_core perfctr_nb bpext perfctr_llc mwaitx cpb notes_plat_sysinfo_290 = cat_l3 cdp_l3 invpcid_single hw_pstate sme ssbd mba sev ibrs ibpb stibp vmmcall notes_plat_sysinfo_295 = fsgsbase bmi1 avx2 smep bmi2 erms invpcid cqm rdt_a rdseed adx smap clflushopt clwb notes_plat_sysinfo_300 = sha_ni xsaveopt xsavec xgetbv1 xsaves cqm_llc cqm_occup_llc cqm_mbm_total notes_plat_sysinfo_305 = cqm_mbm_local clzero irperf xsaveerptr wbnoinvd amd_ppin arat npt lbrv svm_lock notes_plat_sysinfo_310 = nrip_save tsc_scale vmcb_clean flushbyasid decodeassists pausefilter pfthreshold notes_plat_sysinfo_315 = v_vmsave_vmload vgif umip pku ospke vaes vpclmulqdq rdpid overflow_recov succor smca notes_plat_sysinfo_320 = notes_plat_sysinfo_325 = /proc/cpuinfo cache data notes_plat_sysinfo_330 = cache size : 512 KB notes_plat_sysinfo_335 = notes_plat_sysinfo_340 = From numactl --hardware WARNING: a numactl 'node' might or might not correspond to a notes_plat_sysinfo_345 = physical chip. notes_plat_sysinfo_350 = available: 4 nodes (0-3) notes_plat_sysinfo_355 = node 0 cpus: 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 notes_plat_sysinfo_360 = 28 29 30 31 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 notes_plat_sysinfo_365 = 147 148 149 150 151 152 153 154 155 156 157 158 159 notes_plat_sysinfo_370 = node 0 size: 128760 MB notes_plat_sysinfo_375 = node 0 free: 127950 MB notes_plat_sysinfo_380 = node 1 cpus: 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 notes_plat_sysinfo_385 = 57 58 59 60 61 62 63 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 notes_plat_sysinfo_390 = 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 notes_plat_sysinfo_395 = node 1 size: 128884 MB notes_plat_sysinfo_400 = node 1 free: 128243 MB notes_plat_sysinfo_405 = node 2 cpus: 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 notes_plat_sysinfo_410 = 89 90 91 92 93 94 95 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 notes_plat_sysinfo_415 = 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 notes_plat_sysinfo_420 = node 2 size: 128910 MB notes_plat_sysinfo_425 = node 2 free: 128519 MB notes_plat_sysinfo_430 = node 3 cpus: 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 notes_plat_sysinfo_435 = 115 116 117 118 119 120 121 122 123 124 125 126 127 224 225 226 227 228 229 230 231 232 notes_plat_sysinfo_440 = 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 notes_plat_sysinfo_445 = 255 notes_plat_sysinfo_450 = node 3 size: 128873 MB notes_plat_sysinfo_455 = node 3 free: 128486 MB notes_plat_sysinfo_460 = node distances: notes_plat_sysinfo_465 = node 0 1 2 3 notes_plat_sysinfo_470 = 0: 10 12 32 32 notes_plat_sysinfo_475 = 1: 12 10 32 32 notes_plat_sysinfo_480 = 2: 32 32 10 12 notes_plat_sysinfo_485 = 3: 32 32 12 10 notes_plat_sysinfo_490 = notes_plat_sysinfo_495 = From /proc/meminfo notes_plat_sysinfo_500 = MemTotal: 528171092 kB notes_plat_sysinfo_505 = HugePages_Total: 0 notes_plat_sysinfo_510 = Hugepagesize: 2048 kB notes_plat_sysinfo_515 = notes_plat_sysinfo_520 = /sbin/tuned-adm active notes_plat_sysinfo_525 = Current active profile: throughput-performance notes_plat_sysinfo_530 = notes_plat_sysinfo_535 = /sys/devices/system/cpu/cpu*/cpufreq/scaling_governor has notes_plat_sysinfo_540 = performance notes_plat_sysinfo_545 = notes_plat_sysinfo_550 = From /etc/*release* /etc/*version* notes_plat_sysinfo_555 = os-release: notes_plat_sysinfo_560 = NAME="Red Hat Enterprise Linux" notes_plat_sysinfo_565 = VERSION="8.3 (Ootpa)" notes_plat_sysinfo_570 = ID="rhel" notes_plat_sysinfo_575 = ID_LIKE="fedora" notes_plat_sysinfo_580 = VERSION_ID="8.3" notes_plat_sysinfo_585 = PLATFORM_ID="platform:el8" notes_plat_sysinfo_590 = PRETTY_NAME="Red Hat Enterprise Linux 8.3 (Ootpa)" notes_plat_sysinfo_595 = ANSI_COLOR="0;31" notes_plat_sysinfo_600 = redhat-release: Red Hat Enterprise Linux release 8.3 (Ootpa) notes_plat_sysinfo_605 = system-release: Red Hat Enterprise Linux release 8.3 (Ootpa) notes_plat_sysinfo_610 = system-release-cpe: cpe:/o:redhat:enterprise_linux:8.3:ga notes_plat_sysinfo_615 = notes_plat_sysinfo_620 = uname -a: notes_plat_sysinfo_625 = Linux localhost.localdomain 4.18.0-240.el8.x86_64 #1 SMP Wed Sep 23 05:13:10 EDT 2020 notes_plat_sysinfo_630 = x86_64 x86_64 x86_64 GNU/Linux notes_plat_sysinfo_635 = notes_plat_sysinfo_640 = Kernel self-reported vulnerability status: notes_plat_sysinfo_645 = notes_plat_sysinfo_650 = CVE-2018-12207 (iTLB Multihit): Not affected notes_plat_sysinfo_655 = CVE-2018-3620 (L1 Terminal Fault): Not affected notes_plat_sysinfo_660 = Microarchitectural Data Sampling: Not affected notes_plat_sysinfo_665 = CVE-2017-5754 (Meltdown): Not affected notes_plat_sysinfo_670 = CVE-2018-3639 (Speculative Store Bypass): Mitigation: Speculative Store notes_plat_sysinfo_675 = Bypass disabled via prctl and notes_plat_sysinfo_680 = seccomp notes_plat_sysinfo_685 = CVE-2017-5753 (Spectre variant 1): Mitigation: usercopy/swapgs notes_plat_sysinfo_690 = barriers and __user pointer notes_plat_sysinfo_695 = sanitization notes_plat_sysinfo_700 = CVE-2017-5715 (Spectre variant 2): Mitigation: Full AMD retpoline, notes_plat_sysinfo_705 = IBPB: conditional, IBRS_FW, STIBP: notes_plat_sysinfo_710 = always-on, RSB filling notes_plat_sysinfo_715 = CVE-2020-0543 (Special Register Buffer Data Sampling): Not affected notes_plat_sysinfo_720 = CVE-2019-11135 (TSX Asynchronous Abort): Not affected notes_plat_sysinfo_725 = notes_plat_sysinfo_730 = run-level 3 Jun 22 19:12 notes_plat_sysinfo_735 = notes_plat_sysinfo_740 = SPEC is set to: /home/cpu2017-1.1.5-amd-aocc300-milan-B1 notes_plat_sysinfo_745 = Filesystem Type Size Used Avail Use% Mounted on notes_plat_sysinfo_750 = /dev/sda4 xfs 819G 102G 717G 13% /home notes_plat_sysinfo_755 = notes_plat_sysinfo_760 = From /sys/devices/virtual/dmi/id notes_plat_sysinfo_765 = Vendor: Lenovo notes_plat_sysinfo_770 = Product: ThinkSystem SR665 MB notes_plat_sysinfo_775 = Product Family: ThinkSystem notes_plat_sysinfo_780 = Serial: 1234567890 notes_plat_sysinfo_785 = notes_plat_sysinfo_790 = Additional information from dmidecode follows. WARNING: Use caution when you interpret notes_plat_sysinfo_795 = this section. The 'dmidecode' program reads system data which is "intended to allow notes_plat_sysinfo_800 = hardware to be accurately determined", but the intent may not be met, as there are notes_plat_sysinfo_805 = frequent changes to hardware, firmware, and the "DMTF SMBIOS" standard. notes_plat_sysinfo_810 = Memory: notes_plat_sysinfo_815 = 16x Samsung M393A4K40DB3-CWE 32 GB 2 rank 3200 notes_plat_sysinfo_820 = 16x Unknown Unknown notes_plat_sysinfo_825 = notes_plat_sysinfo_830 = BIOS: notes_plat_sysinfo_835 = BIOS Vendor: Lenovo notes_plat_sysinfo_840 = BIOS Version: D8E115B-2.00 notes_plat_sysinfo_845 = BIOS Date: 02/02/2021 notes_plat_sysinfo_850 = BIOS Revision: 2.0 notes_plat_sysinfo_855 = Firmware Revision: 3.0 notes_plat_sysinfo_860 = notes_plat_sysinfo_865 = (End of data from sysinfo program) hw_cpu_name = AMD EPYC 7763 hw_disk = 1 x 960 GB SATA SSD hw_memory = 512 GB (16 x 32 GB 2Rx4 PC4-3200AA-R) hw_nchips = 2 prepared_by = Lenovo Global Technology sw_file = xfs sw_os000 = Red Hat Enterprise Linux release 8.3 sw_os001 = (Ootpa) sw_os002 = Kernel 4.18.0-240.el8.x86_64 sw_state = Run level 3 (multi-user) # End of settings added by sysinfo_program 631.deepsjeng_s: # The following setting was inserted automatically as a result of # post-run basepeak application. basepeak = 1 # 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: power_management000 = BIOS set to prefer performance at the cost of power_management001 = additional power usage notes_plat_form_000 =BIOS configuration: notes_plat_form_005 =Operating Mode set to Maximum Performance and then set it to Custom Mode notes_plat_form_010 =4-Link xGMI Max Speed set to 16Gbps notes_plat_form_015 =SOC P-States set to P0 notes_plat_form_020 =NUMA nodes per socket set to NPS2