mikemccand commented on PR #12311: URL: https://github.com/apache/lucene/pull/12311#issuecomment-1574989612
I ran @rmuir's vectorbench on a new Raptor Lake build (i9-13900K). Note that this CPU does NOT seem to support AVX-512: ``` flags : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss ht tm pbe syscall nx pdpe1gb rdtscp lm constant_tsc art arch_perfmon pebs bts rep_good nopl xtopology nonstop_tsc cpuid aperfmperf tsc_known_freq pni pclmulqdq dtes64 monitor ds_cpl vmx smx est tm2 ssse3 sdbg fma cx16 xtpr pdcm sse4_1 sse4_2 x2apic movbe popcnt tsc_deadline_timer aes xsave avx f16c rdrand lahf_lm abm 3dnowprefetch cpuid_fault ssbd ibrs ibpb stibp ibrs_enhanced tpr_shadow vnmi flexpriority ept vpid ept_ad fsgsbase tsc_adjust bmi1 avx2 smep bmi2 erms invpcid rdseed adx smap clflushopt clwb intel_pt sha_ni xsaveopt xsavec xgetbv1 xsaves split_lock_detect avx_vnni dtherm ida arat pln pts hwp hwp_notify hwp_act_window hwp_epp hwp_pkg_req hfi umip pku ospke waitpkg gfni vaes vpclmulqdq tme rdpid movdiri movdir64b fsrm md_clear serialize pconfig arch_lbr ibt flush_l1d arch_capabilities vmx flags : vnmi preemption_timer posted_intr invvpid ept_x_only ept_ad ept_1gb flexpriority apicv tsc_offset vtpr mtf vapic ept vpid unrestricted_guest vapic_reg vid ple shadow_vmcs ept_mode_based_exec tsc_scaling usr_wait_pause ``` Benchy results: ``` Benchmark (size) Mode Cnt Score Error Units BinaryCosineBenchmark.cosineDistanceNew 1 thrpt 5 128.894 ± 0.021 ops/us BinaryCosineBenchmark.cosineDistanceNew 128 thrpt 5 64.367 ± 0.062 ops/us BinaryCosineBenchmark.cosineDistanceNew 207 thrpt 5 37.563 ± 0.030 ops/us BinaryCosineBenchmark.cosineDistanceNew 256 thrpt 5 36.636 ± 0.004 ops/us BinaryCosineBenchmark.cosineDistanceNew 300 thrpt 5 31.305 ± 0.012 ops/us BinaryCosineBenchmark.cosineDistanceNew 512 thrpt 5 20.769 ± 0.008 ops/us BinaryCosineBenchmark.cosineDistanceNew 702 thrpt 5 13.757 ± 0.017 ops/us BinaryCosineBenchmark.cosineDistanceNew 1024 thrpt 5 10.201 ± 0.008 ops/us BinaryCosineBenchmark.cosineDistanceOld 1 thrpt 5 128.889 ± 0.095 ops/us BinaryCosineBenchmark.cosineDistanceOld 128 thrpt 5 14.177 ± 0.064 ops/us BinaryCosineBenchmark.cosineDistanceOld 207 thrpt 5 8.967 ± 0.023 ops/us BinaryCosineBenchmark.cosineDistanceOld 256 thrpt 5 7.295 ± 0.014 ops/us BinaryCosineBenchmark.cosineDistanceOld 300 thrpt 5 6.215 ± 0.034 ops/us BinaryCosineBenchmark.cosineDistanceOld 512 thrpt 5 3.709 ± 0.003 ops/us BinaryCosineBenchmark.cosineDistanceOld 702 thrpt 5 2.714 ± 0.005 ops/us BinaryCosineBenchmark.cosineDistanceOld 1024 thrpt 5 1.872 ± 0.001 ops/us BinaryDotProductBenchmark.dotProductNew 1024 thrpt 5 23.562 ± 0.025 ops/us BinaryDotProductBenchmark.dotProductOld 1024 thrpt 5 3.885 ± 0.082 ops/us BinarySquareBenchmark.squareDistanceNew 1 thrpt 5 522.630 ± 4.081 ops/us BinarySquareBenchmark.squareDistanceNew 128 thrpt 5 101.951 ± 4.427 ops/us BinarySquareBenchmark.squareDistanceNew 207 thrpt 5 65.050 ± 0.254 ops/us BinarySquareBenchmark.squareDistanceNew 256 thrpt 5 60.495 ± 0.922 ops/us BinarySquareBenchmark.squareDistanceNew 300 thrpt 5 51.767 ± 0.042 ops/us BinarySquareBenchmark.squareDistanceNew 512 thrpt 5 32.832 ± 0.022 ops/us BinarySquareBenchmark.squareDistanceNew 702 thrpt 5 23.786 ± 0.018 ops/us BinarySquareBenchmark.squareDistanceNew 1024 thrpt 5 17.062 ± 0.145 ops/us BinarySquareBenchmark.squareDistanceOld 1 thrpt 5 529.882 ± 1.503 ops/us BinarySquareBenchmark.squareDistanceOld 128 thrpt 5 32.478 ± 0.037 ops/us BinarySquareBenchmark.squareDistanceOld 207 thrpt 5 20.901 ± 0.023 ops/us BinarySquareBenchmark.squareDistanceOld 256 thrpt 5 16.644 ± 0.070 ops/us BinarySquareBenchmark.squareDistanceOld 300 thrpt 5 14.502 ± 0.111 ops/us BinarySquareBenchmark.squareDistanceOld 512 thrpt 5 8.703 ± 0.050 ops/us BinarySquareBenchmark.squareDistanceOld 702 thrpt 5 6.473 ± 0.013 ops/us BinarySquareBenchmark.squareDistanceOld 1024 thrpt 5 4.454 ± 0.016 ops/us FloatCosineBenchmark.cosineNew 1 thrpt 5 395.222 ± 3.364 ops/us FloatCosineBenchmark.cosineNew 4 thrpt 5 275.572 ± 2.528 ops/us FloatCosineBenchmark.cosineNew 6 thrpt 5 217.377 ± 0.561 ops/us FloatCosineBenchmark.cosineNew 8 thrpt 5 192.311 ± 1.492 ops/us FloatCosineBenchmark.cosineNew 13 thrpt 5 143.959 ± 0.061 ops/us FloatCosineBenchmark.cosineNew 16 thrpt 5 127.340 ± 0.181 ops/us FloatCosineBenchmark.cosineNew 25 thrpt 5 116.471 ± 0.219 ops/us FloatCosineBenchmark.cosineNew 32 thrpt 5 117.458 ± 0.031 ops/us FloatCosineBenchmark.cosineNew 64 thrpt 5 100.845 ± 0.016 ops/us FloatCosineBenchmark.cosineNew 100 thrpt 5 73.392 ± 0.129 ops/us FloatCosineBenchmark.cosineNew 128 thrpt 5 79.363 ± 0.012 ops/us FloatCosineBenchmark.cosineNew 207 thrpt 5 48.170 ± 0.027 ops/us FloatCosineBenchmark.cosineNew 256 thrpt 5 43.298 ± 0.103 ops/us FloatCosineBenchmark.cosineNew 300 thrpt 5 36.302 ± 0.017 ops/us FloatCosineBenchmark.cosineNew 512 thrpt 5 26.113 ± 0.076 ops/us FloatCosineBenchmark.cosineNew 702 thrpt 5 19.034 ± 0.008 ops/us FloatCosineBenchmark.cosineNew 1024 thrpt 5 16.945 ± 0.052 ops/us FloatCosineBenchmark.cosineOld 1 thrpt 5 398.987 ± 0.675 ops/us FloatCosineBenchmark.cosineOld 4 thrpt 5 279.282 ± 4.223 ops/us FloatCosineBenchmark.cosineOld 6 thrpt 5 220.884 ± 5.144 ops/us FloatCosineBenchmark.cosineOld 8 thrpt 5 196.722 ± 0.389 ops/us FloatCosineBenchmark.cosineOld 13 thrpt 5 146.701 ± 0.832 ops/us FloatCosineBenchmark.cosineOld 16 thrpt 5 130.186 ± 0.280 ops/us FloatCosineBenchmark.cosineOld 25 thrpt 5 87.526 ± 0.083 ops/us FloatCosineBenchmark.cosineOld 32 thrpt 5 70.398 ± 0.124 ops/us FloatCosineBenchmark.cosineOld 64 thrpt 5 35.020 ± 0.007 ops/us FloatCosineBenchmark.cosineOld 100 thrpt 5 21.121 ± 0.009 ops/us FloatCosineBenchmark.cosineOld 128 thrpt 5 16.276 ± 0.008 ops/us FloatCosineBenchmark.cosineOld 207 thrpt 5 10.017 ± 0.002 ops/us FloatCosineBenchmark.cosineOld 256 thrpt 5 8.085 ± 0.002 ops/us FloatCosineBenchmark.cosineOld 300 thrpt 5 6.882 ± 0.001 ops/us FloatCosineBenchmark.cosineOld 512 thrpt 5 3.981 ± 0.009 ops/us FloatCosineBenchmark.cosineOld 702 thrpt 5 2.900 ± 0.008 ops/us FloatCosineBenchmark.cosineOld 1024 thrpt 5 1.990 ± 0.001 ops/us FloatDotProductBenchmark.dotProductNew 1 thrpt 5 482.634 ± 0.308 ops/us FloatDotProductBenchmark.dotProductNew 4 thrpt 5 358.350 ± 0.814 ops/us FloatDotProductBenchmark.dotProductNew 6 thrpt 5 299.456 ± 9.216 ops/us FloatDotProductBenchmark.dotProductNew 8 thrpt 5 282.228 ± 0.560 ops/us FloatDotProductBenchmark.dotProductNew 13 thrpt 5 237.520 ± 0.758 ops/us FloatDotProductBenchmark.dotProductNew 16 thrpt 5 226.653 ± 0.598 ops/us FloatDotProductBenchmark.dotProductNew 25 thrpt 5 203.128 ± 0.136 ops/us FloatDotProductBenchmark.dotProductNew 32 thrpt 5 234.430 ± 2.885 ops/us FloatDotProductBenchmark.dotProductNew 64 thrpt 5 199.576 ± 1.049 ops/us FloatDotProductBenchmark.dotProductNew 100 thrpt 5 143.859 ± 0.351 ops/us FloatDotProductBenchmark.dotProductNew 128 thrpt 5 163.681 ± 1.253 ops/us FloatDotProductBenchmark.dotProductNew 207 thrpt 5 98.173 ± 1.241 ops/us FloatDotProductBenchmark.dotProductNew 256 thrpt 5 95.487 ± 0.053 ops/us FloatDotProductBenchmark.dotProductNew 300 thrpt 5 64.907 ± 0.056 ops/us FloatDotProductBenchmark.dotProductNew 512 thrpt 5 62.076 ± 0.192 ops/us FloatDotProductBenchmark.dotProductNew 702 thrpt 5 33.659 ± 0.053 ops/us FloatDotProductBenchmark.dotProductNew 1024 thrpt 5 29.892 ± 0.188 ops/us FloatDotProductBenchmark.dotProductOld 1 thrpt 5 558.982 ± 14.484 ops/us FloatDotProductBenchmark.dotProductOld 4 thrpt 5 425.068 ± 3.916 ops/us FloatDotProductBenchmark.dotProductOld 6 thrpt 5 398.557 ± 1.348 ops/us FloatDotProductBenchmark.dotProductOld 8 thrpt 5 359.623 ± 0.203 ops/us FloatDotProductBenchmark.dotProductOld 13 thrpt 5 262.966 ± 0.098 ops/us FloatDotProductBenchmark.dotProductOld 16 thrpt 5 229.867 ± 0.083 ops/us FloatDotProductBenchmark.dotProductOld 25 thrpt 5 165.441 ± 0.115 ops/us FloatDotProductBenchmark.dotProductOld 32 thrpt 5 152.221 ± 0.138 ops/us FloatDotProductBenchmark.dotProductOld 64 thrpt 5 85.443 ± 0.270 ops/us FloatDotProductBenchmark.dotProductOld 100 thrpt 5 53.636 ± 0.020 ops/us FloatDotProductBenchmark.dotProductOld 128 thrpt 5 42.828 ± 0.023 ops/us FloatDotProductBenchmark.dotProductOld 207 thrpt 5 26.981 ± 0.055 ops/us FloatDotProductBenchmark.dotProductOld 256 thrpt 5 21.944 ± 0.157 ops/us FloatDotProductBenchmark.dotProductOld 300 thrpt 5 18.856 ± 0.011 ops/us FloatDotProductBenchmark.dotProductOld 512 thrpt 5 11.330 ± 0.011 ops/us FloatDotProductBenchmark.dotProductOld 702 thrpt 5 8.150 ± 0.006 ops/us FloatDotProductBenchmark.dotProductOld 1024 thrpt 5 5.814 ± 0.006 ops/us FloatSquareBenchmark.squareNew 1 thrpt 5 479.897 ± 0.414 ops/us FloatSquareBenchmark.squareNew 4 thrpt 5 347.548 ± 5.824 ops/us FloatSquareBenchmark.squareNew 6 thrpt 5 320.104 ± 3.070 ops/us FloatSquareBenchmark.squareNew 8 thrpt 5 272.376 ± 2.516 ops/us FloatSquareBenchmark.squareNew 13 thrpt 5 236.600 ± 1.357 ops/us FloatSquareBenchmark.squareNew 16 thrpt 5 225.289 ± 0.361 ops/us FloatSquareBenchmark.squareNew 25 thrpt 5 201.074 ± 0.363 ops/us FloatSquareBenchmark.squareNew 32 thrpt 5 222.044 ± 1.173 ops/us FloatSquareBenchmark.squareNew 64 thrpt 5 192.298 ± 2.776 ops/us FloatSquareBenchmark.squareNew 100 thrpt 5 131.676 ± 0.082 ops/us FloatSquareBenchmark.squareNew 128 thrpt 5 144.401 ± 1.032 ops/us FloatSquareBenchmark.squareNew 207 thrpt 5 85.532 ± 0.490 ops/us FloatSquareBenchmark.squareNew 256 thrpt 5 79.800 ± 0.023 ops/us FloatSquareBenchmark.squareNew 300 thrpt 5 67.323 ± 0.287 ops/us FloatSquareBenchmark.squareNew 512 thrpt 5 43.781 ± 0.029 ops/us FloatSquareBenchmark.squareNew 702 thrpt 5 27.687 ± 0.008 ops/us FloatSquareBenchmark.squareNew 1024 thrpt 5 20.829 ± 0.131 ops/us FloatSquareBenchmark.squareOld 1 thrpt 5 479.053 ± 0.635 ops/us FloatSquareBenchmark.squareOld 4 thrpt 5 345.476 ± 3.422 ops/us FloatSquareBenchmark.squareOld 6 thrpt 5 320.319 ± 0.476 ops/us FloatSquareBenchmark.squareOld 8 thrpt 5 347.901 ± 0.762 ops/us FloatSquareBenchmark.squareOld 13 thrpt 5 223.134 ± 0.950 ops/us FloatSquareBenchmark.squareOld 16 thrpt 5 213.299 ± 0.439 ops/us FloatSquareBenchmark.squareOld 25 thrpt 5 141.332 ± 1.297 ops/us FloatSquareBenchmark.squareOld 32 thrpt 5 117.027 ± 0.792 ops/us FloatSquareBenchmark.squareOld 64 thrpt 5 63.886 ± 0.034 ops/us FloatSquareBenchmark.squareOld 100 thrpt 5 42.088 ± 0.016 ops/us FloatSquareBenchmark.squareOld 128 thrpt 5 33.163 ± 0.007 ops/us FloatSquareBenchmark.squareOld 207 thrpt 5 19.855 ± 0.098 ops/us FloatSquareBenchmark.squareOld 256 thrpt 5 16.870 ± 0.043 ops/us FloatSquareBenchmark.squareOld 300 thrpt 5 14.326 ± 0.973 ops/us FloatSquareBenchmark.squareOld 512 thrpt 5 8.837 ± 0.047 ops/us FloatSquareBenchmark.squareOld 702 thrpt 5 6.299 ± 0.042 ops/us FloatSquareBenchmark.squareOld 1024 thrpt 5 4.351 ± 0.062 ops/us ``` -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: issues-unsubscr...@lucene.apache.org For queries about this service, please contact Infrastructure at: us...@infra.apache.org --------------------------------------------------------------------- To unsubscribe, e-mail: issues-unsubscr...@lucene.apache.org For additional commands, e-mail: issues-h...@lucene.apache.org
