Open DrewScoggins opened 10 months ago
Tagging subscribers to this area: @JulieLeeMSFT, @jakobbotsch See info in area-owners.md if you want to be subscribed.
Author: | DrewScoggins |
---|---|
Assignees: | - |
Labels: | `tenet-performance`, `tenet-performance-benchmarks`, `area-CodeGen-coreclr`, `untriaged`, `needs-area-label` |
Milestone: | - |
@EgorBo Can you take a look?
cc @tannergooding @khushal1996
cc @anthonycanino @DeepakRajendrakumaran
I don't see any codegen diffs for the given commit range locally
Method | Job | Toolchain | Mean | Error | StdDev | Ratio | Code Size |
---|---|---|---|---|---|---|---|
Interpol_AVX | Job-HJGTCR | \Core_Root\corerun.exe | 3.959 us | 0.6729 us | 0.0369 us | 1.01 | 906 B |
Interpol_AVX | Job-DVKPPT | \Core_Root_base\corerun.exe | 3.913 us | 1.0288 us | 0.0564 us | 1.00 | 906 B |
no perf or asm diffs between these commits on AMD 7950X + x86. Tried to enable/disable avx512, still no diffs. Could be some JCC erratum/intel specific? Moving to 9.0 since x86 has lower priority
This was found during the manual 7.0 -> 8.0 RC1 comparison. Linked below is a link to the test history.
https://pvscmdupload.blob.core.windows.net/reports/allTestHistory%2frefs%2fheads%2fmain_x86_Windows%2010.0.18362%2fBilinearTest.Interpol_AVX.html
The suspected commit affecting this is, https://github.com/dotnet/runtime/pull/84384.