@@ -832,8 +832,10 @@ Performance
832832-----------
833833
834834Full pyperformance results comparing the frame-pointer build against an
835- identical build without frame pointers (geometric mean across 108
836- benchmarks):
835+ identical build without frame pointers (geometric mean across 80
836+ benchmarks [#missing_benchmarks ]_). For reproducibility,
837+ pyperformance JSON files can be found in
838+ [#benchmarks ]_. Benchmark visualization can be found in the Appendix:
837839
838840===================================== =======================
839841Machine Geometric mean overhead
@@ -844,7 +846,8 @@ AMD EPYC 9654 (x86-64) 1.01x slower
844846AWS Graviton c7g.16xlarge (aarch64) 1.02x slower
845847Ampere Altra Max (aarch64) 1.01x slower
846848Raspberry Pi (aarch64). 1.00x slower
847- macOS M3 (arm64) 1.00x slower
849+ macOS M3 Pro (arm64) 1.00x slower
850+ Intel i7 12700H (x86-64) 1.02x slower
848851===================================== =======================
849852
850853This overhead applies to both the interpreter and to C extensions that inherit
@@ -1031,11 +1034,22 @@ Footnotes
10311034 and Linux", by Markus Stange.
10321035 https://github.com/mstange/samply
10331036
1037+ .. [#benchmarks ] Python frame-pointer benchmark pyperformance JSON files
1038+ https://github.com/Fidget-Spinner/python-framepointer-bench
1039+
1040+ .. [#missing_benchmarks ] Some benchmarks are missing due to incompatibilities with
1041+ Python 3.15alpha.
1042+
10341043 .. _#96174 : https://github.com/python/cpython/issues/96174
10351044.. _python/cpython issue #96174 : https://github.com/python/cpython/issues/96174
10361045.. _#126910 : https://github.com/python/cpython/issues/126910
10371046
10381047
1048+ Appendix
1049+ ========
1050+
1051+ # TODO: KJ, once we have Diego's results.
1052+
10391053Copyright
10401054=========
10411055
0 commit comments