After adding general parsing of measurement statistics, I can now also create a statistical summary across all ~170 benchmarks as shown below. This lets me see for example the minimum IPC, maximum IPC, mean IPC and standard deviation. This will then provide some information whether a particular workload is “low” or “high” in a metric and how significantly.<\/p>\n\n\n\n
The statistics below come from the workload statistics with AMD metrics first followed by Intel metrics. For example, based on this table we can see mean values for topdown metrics:<\/p>\n\n\n\n
These numbers are recalculated as the reports are re-generated but with 170 workloads mostly included are a good first overview of how the workloads operate on my AMD 7840.<\/p>\n\n\n\n
Some next steps including flagging the outliers in the metrics and seeing how I can create histograms for different fields below.<\/p>\n\n\n\n
| metric<\/th> | count<\/th> | min<\/th> | max<\/th> | median<\/th> | mean<\/th> | stddev<\/th>\n<\/tr>\n<\/thead>\n |
|---|---|---|---|---|---|---|
| elapsed<\/td> | 174<\/td> | 2.5<\/td> | 8.8e+03<\/td> | 554<\/td> | 1.25e+03<\/td> | 1.62e+03<\/td>\n<\/tr>\n |
| on_cpu<\/td> | 174<\/td> | 0<\/td> | 16<\/td> | 7.26<\/td> | 7.31<\/td> | 5.53<\/td>\n<\/tr>\n |
| inblock<\/td> | 172<\/td> | 0<\/td> | 5.47e+06<\/td> | 0<\/td> | 3.19e+04<\/td> | 4.16e+05<\/td>\n<\/tr>\n |
| onblock<\/td> | 172<\/td> | 0.46<\/td> | 4e+05<\/td> | 133<\/td> | 9.97e+03<\/td> | 3.54e+04<\/td>\n<\/tr>\n |
| page-fault<\/td> | 174<\/td> | 4.33<\/td> | 1.24e+05<\/td> | 2.36e+03<\/td> | 1.14e+04<\/td> | 2.04e+04<\/td>\n<\/tr>\n |
| context-switch<\/td> | 174<\/td> | 1.51<\/td> | 5.12e+04<\/td> | 75<\/td> | 2.18e+03<\/td> | 7.73e+03<\/td>\n<\/tr>\n |
| IPC<\/td> | 174<\/td> | 0.03<\/td> | 4.63<\/td> | 1.44<\/td> | 1.64<\/td> | 0.88<\/td>\n<\/tr>\n |
| GHz<\/td> | 174<\/td> | 0<\/td> | 4.62<\/td> | 1.98<\/td> | 1.91<\/td> | 1.39<\/td>\n<\/tr>\n |
| retire-rate<\/td> | 174<\/td> | 0.8<\/td> | 76.2<\/td> | 29.2<\/td> | 32.3<\/td> | 15.6<\/td>\n<\/tr>\n |
| frontend-stall<\/td> | 174<\/td> | 0.1<\/td> | 73<\/td> | 19.1<\/td> | 22.5<\/td> | 17.5<\/td>\n<\/tr>\n |
| backend-stall<\/td> | 174<\/td> | 4.1<\/td> | 97.1<\/td> | 36.8<\/td> | 41.6<\/td> | 21.3<\/td>\n<\/tr>\n |
| spec-stall<\/td> | 174<\/td> | 0<\/td> | 21.2<\/td> | 2.7<\/td> | 3.56<\/td> | 3.95<\/td>\n<\/tr>\n |
| retire-ucode<\/td> | 174<\/td> | 0<\/td> | 1<\/td> | 0<\/td> | 0.0736<\/td> | 0.131<\/td>\n<\/tr>\n |
| retire-fastpath<\/td> | 174<\/td> | 0.7<\/td> | 76.2<\/td> | 24.7<\/td> | 27.8<\/td> | 14.5<\/td>\n<\/tr>\n |
| float-density<\/td> | 174<\/td> | 0.013<\/td> | 676<\/td> | 67.5<\/td> | 133<\/td> | 154<\/td>\n<\/tr>\n |
| frontend-latency<\/td> | 174<\/td> | 0.1<\/td> | 58.3<\/td> | 9.1<\/td> | 13.6<\/td> | 12.6<\/td>\n<\/tr>\n |
| frontend-bandwidth<\/td> | 174<\/td> | 0<\/td> | 28.7<\/td> | 5.5<\/td> | 6.06<\/td> | 4.86<\/td>\n<\/tr>\n |
| opcache-miss<\/td> | 87<\/td> | 52.4<\/td> | 54.7<\/td> | 53.8<\/td> | 53.8<\/td> | 0.401<\/td>\n<\/tr>\n |
| icache-miss<\/td> | 87<\/td> | 8.2<\/td> | 9.6<\/td> | 8.5<\/td> | 8.57<\/td> | 0.294<\/td>\n<\/tr>\n |
| backend-cpu<\/td> | 174<\/td> | 0.7<\/td> | 64<\/td> | 9.3<\/td> | 12.3<\/td> | 10.9<\/td>\n<\/tr>\n |
| backend-memory<\/td> | 174<\/td> | 0.4<\/td> | 95.5<\/td> | 19.9<\/td> | 23.5<\/td> | 17.4<\/td>\n<\/tr>\n |
| amd-l2-miss<\/td> | 174<\/td> | 0.08<\/td> | 59.8<\/td> | 16.3<\/td> | 17.3<\/td> | 11.7<\/td>\n<\/tr>\n |
| amd-l2-density<\/td> | 174<\/td> | 0.036<\/td> | 470<\/td> | 38.2<\/td> | 49.7<\/td> | 58.9<\/td>\n<\/tr>\n |
| spec-branch<\/td> | 174<\/td> | 0<\/td> | 21.1<\/td> | 2.1<\/td> | 3.07<\/td> | 3.8<\/td>\n<\/tr>\n |
| spec-pipeline<\/td> | 174<\/td> | 0<\/td> | 1.4<\/td> | 0<\/td> | 0.11<\/td> | 0.205<\/td>\n<\/tr>\n |
| branch-miss<\/td> | 174<\/td> | 0.01<\/td> | 14.8<\/td> | 1.96<\/td> | 2.79<\/td> | 2.99<\/td>\n<\/tr>\n |
| branch-density<\/td> | 174<\/td> | 8.68<\/td> | 276<\/td> | 125<\/td> | 130<\/td> | 61.9<\/td>\n<\/tr>\n |
| branch-cond<\/td> | 174<\/td> | 5.66<\/td> | 271<\/td> | 92.7<\/td> | 98.4<\/td> | 48.9<\/td>\n<\/tr>\n |
| branch-ind<\/td> | 174<\/td> | 0.003<\/td> | 29.8<\/td> | 3.07<\/td> | 4.53<\/td> | 5.4<\/td>\n<\/tr>\n |
| smt-contention<\/td> | 174<\/td> | 0<\/td> | 45.4<\/td> | 12.5<\/td> | 13.3<\/td> | 13.1<\/td>\n<\/tr>\n |
| elapsed<\/td> | 169<\/td> | 1.49<\/td> | 1.44e+04<\/td> | 750<\/td> | 1.71e+03<\/td> | 2.45e+03<\/td>\n<\/tr>\n |
| on_cpu<\/td> | 169<\/td> | 0<\/td> | 15.8<\/td> | 9.05<\/td> | 7.76<\/td> | 5.6<\/td>\n<\/tr>\n |
| inblock<\/td> | 166<\/td> | 0<\/td> | 8.53e+04<\/td> | 53.9<\/td> | 2.33e+03<\/td> | 9.24e+03<\/td>\n<\/tr>\n |
| onblock<\/td> | 166<\/td> | 0.37<\/td> | 4.25e+05<\/td> | 38.1<\/td> | 9.06e+03<\/td> | 3.59e+04<\/td>\n<\/tr>\n |
| page-fault<\/td> | 169<\/td> | 4.05<\/td> | 1.17e+05<\/td> | 1.73e+03<\/td> | 1.02e+04<\/td> | 1.91e+04<\/td>\n<\/tr>\n |
| context-switch<\/td> | 169<\/td> | 1.81<\/td> | 9.27e+04<\/td> | 72.6<\/td> | 2.44e+03<\/td> | 1e+04<\/td>\n<\/tr>\n |
| IPC<\/td> | 169<\/td> | 0.05<\/td> | 5.54<\/td> | 1.82<\/td> | 2<\/td> | 0.975<\/td>\n<\/tr>\n |
| GHz<\/td> | 169<\/td> | 0<\/td> | 3.06<\/td> | 1.46<\/td> | 1.3<\/td> | 0.885<\/td>\n<\/tr>\n |
| retire-rate<\/td> | 169<\/td> | 3.5<\/td> | 87.6<\/td> | 42.8<\/td> | 44.3<\/td> | 14.7<\/td>\n<\/tr>\n |
| frontend-stall<\/td> | 169<\/td> | 1<\/td> | 50.4<\/td> | 18.3<\/td> | 19.4<\/td> | 11.1<\/td>\n<\/tr>\n |
| backend-stall<\/td> | 169<\/td> | 1.3<\/td> | 93.5<\/td> | 23.8<\/td> | 27.6<\/td> | 18.1<\/td>\n<\/tr>\n |
| spec-stall<\/td> | 169<\/td> | 0<\/td> | 46.7<\/td> | 6.8<\/td> | 9.16<\/td> | 8.55<\/td>\n<\/tr>\n |
| retire-ucode<\/td> | 169<\/td> | 0<\/td> | 16.7<\/td> | 3<\/td> | 3.37<\/td> | 2.4<\/td>\n<\/tr>\n |
| retire-fastpath<\/td> | 169<\/td> | 2.3<\/td> | 83.4<\/td> | 39.5<\/td> | 40.9<\/td> | 14.1<\/td>\n<\/tr>\n |
| frontend-latency<\/td> | 169<\/td> | 0.3<\/td> | 35.1<\/td> | 9.5<\/td> | 10.3<\/td> | 6.63<\/td>\n<\/tr>\n |
| frontend-bandwidth<\/td> | 169<\/td> | 0.4<\/td> | 25.8<\/td> | 8.4<\/td> | 9.09<\/td> | 6<\/td>\n<\/tr>\n |
| backend-cpu<\/td> | 169<\/td> | 0.6<\/td> | 67.7<\/td> | 10.4<\/td> | 13.8<\/td> | 11<\/td>\n<\/tr>\n |
| backend-memory<\/td> | 169<\/td> | 0<\/td> | 90.4<\/td> | 10.1<\/td> | 13.8<\/td> | 14.1<\/td>\n<\/tr>\n |
| l1-stall<\/td> | 77<\/td> | 0<\/td> | 24.4<\/td> | 5.2<\/td> | 5.45<\/td> | 4.88<\/td>\n<\/tr>\n |
| l2-stall<\/td> | 77<\/td> | 0<\/td> | 57.1<\/td> | 8.6<\/td> | 9.01<\/td> | 9.09<\/td>\n<\/tr>\n |
| l3-stall<\/td> | 77<\/td> | 0<\/td> | 35<\/td> | 2.6<\/td> | 4.07<\/td> | 5.51<\/td>\n<\/tr>\n |
| dram-stall<\/td> | 77<\/td> | 0<\/td> | 49<\/td> | 6<\/td> | 8.86<\/td> | 10.6<\/td>\n<\/tr>\n |
| store-stall<\/td> | 77<\/td> | 0<\/td> | 28.3<\/td> | 0.9<\/td> | 1.56<\/td> | 3.53<\/td>\n<\/tr>\n |
| intel-l2-miss<\/td> | 169<\/td> | 0.59<\/td> | 92.9<\/td> | 26.8<\/td> | 28.2<\/td> | 17.1<\/td>\n<\/tr>\n |
| intel-l2-density<\/td> | 169<\/td> | 0.029<\/td> | 370<\/td> | 26.9<\/td> | 35.7<\/td> | 45.2<\/td>\n<\/tr>\n |
| spec-branch<\/td> | 169<\/td> | 0<\/td> | 46.7<\/td> | 6.2<\/td> | 8.71<\/td> | 8.58<\/td>\n<\/tr>\n |
| spec-pipeline<\/td> | 169<\/td> | 0<\/td> | 6.2<\/td> | 0.3<\/td> | 0.456<\/td> | 0.701<\/td>\n<\/tr>\n |
| branch-miss<\/td> | 169<\/td> | 0<\/td> | 20.4<\/td> | 1.03<\/td> | 1.81<\/td> | 2.58<\/td>\n<\/tr>\n |
| branch-density<\/td> | 169<\/td> | 6.24<\/td> | 275<\/td> | 126<\/td> | 127<\/td> | 60.8<\/td>\n<\/tr>\n |
| branch-cond<\/td> | 169<\/td> | 6.24<\/td> | 275<\/td> | 126<\/td> | 127<\/td> | 60.8<\/td>\n<\/tr>\n |
| branch-ind<\/td> | 169<\/td> | 0.035<\/td> | 83<\/td> | 21.1<\/td> | 22.6<\/td> | 17<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n","protected":false},"excerpt":{"rendered":" After adding general parsing of measurement statistics, I can now also create a statistical summary across all ~170 benchmarks as shown below. This lets me see for example the minimum IPC, maximum IPC, mean IPC and standard deviation. This will …<\/span> Continue reading →<\/span><\/a><\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[11,3],"tags":[23,26],"class_list":["post-1689","post","type-post","status-publish","format-standard","hentry","category-experiment","category-website","tag-benchmarks","tag-metrics"],"_links":{"self":[{"href":"https:\/\/mvermeulen.org\/perf\/wp-json\/wp\/v2\/posts\/1689","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/mvermeulen.org\/perf\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/mvermeulen.org\/perf\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/mvermeulen.org\/perf\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/mvermeulen.org\/perf\/wp-json\/wp\/v2\/comments?post=1689"}],"version-history":[{"count":2,"href":"https:\/\/mvermeulen.org\/perf\/wp-json\/wp\/v2\/posts\/1689\/revisions"}],"predecessor-version":[{"id":1692,"href":"https:\/\/mvermeulen.org\/perf\/wp-json\/wp\/v2\/posts\/1689\/revisions\/1692"}],"wp:attachment":[{"href":"https:\/\/mvermeulen.org\/perf\/wp-json\/wp\/v2\/media?parent=1689"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/mvermeulen.org\/perf\/wp-json\/wp\/v2\/categories?post=1689"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/mvermeulen.org\/perf\/wp-json\/wp\/v2\/tags?post=1689"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}} |