JIT: Seed stress modes by string hash #83771
Conversation
Instead of using the numeric value of each stress mode inside the enum, calculate a hash based on the string name. This has the benefit that it won't change based on new members being added/reordered and reduces correlation of which stress modes get enabled together. Fix dotnet#83733
ghost
added
the
area-CodeGen-coreclr
|
Tagging subscribers to this area: @JulieLeeMSFT, @jakobbotsch, @kunalspathak Issue DetailsInstead of using the numeric value of each stress mode inside the enum calculate a hash based on the string name. This has the benefit that it won't change based on new members being added/reordered and reduces correlation of which stress modes get enabled together. Fix #83733
|
|
/azp run runtime-coreclr jitstress, runtime-coreclr libraries-jitstress |
|
Azure Pipelines successfully started running 2 pipeline(s). |
|
Looks like this exposes some new stress related issue: Assertion failed '(emitThisGCrefRegs & regMask) == 0' in 'System.String:Concat[ushort](System.Collections.Generic.IEnumerable`1[ushort]):System.String' during 'Emit code' |
|
The above issue also hits in libraries-jitstress-random, I opened #83963 for it. Will convert this to a draft until we get a fix for that. |
|
/azp run runtime-coreclr jitstress, runtime-coreclr libraries-jitstress |
|
Azure Pipelines successfully started running 2 pipeline(s). |
With side effects that may go on top, it may be block morph's responsibility to set this flag. Also set the type of the reused GT_COMMA nodes; haven't actually seen any issues here, but seems prudent to do this. Fixes some stress issues seen over in dotnet#83771.
| if (verbose) | ||
| { | ||
| printf("\n\n*** JitStress: %ws ***\n\n", s_compStressModeNames[stressArea]); | ||
| printf("\n\n*** JitStress: %s ***\n\n", s_compStressModeNames[stressArea]); |
There was a problem hiding this comment.
This also fixes this printf for PAL (it was just printing as "JitStress: %ws" on non-Windows builds)
With side effects that may go on top, it may be block morph's responsibility to set this flag. Also set the type of the reused GT_COMMA nodes; haven't actually seen any issues here, but seems prudent to do this. Fixes some stress issues seen over in #83771.
|
Stress failures should be fixed by #84352, will rerun... |
|
/azp run runtime-coreclr jitstress, runtime-coreclr libraries-jitstress |
|
Azure Pipelines successfully started running 2 pipeline(s). |
|
/azp run runtime-coreclr jitstress, runtime-coreclr libraries-jitstress |
|
Azure Pipelines successfully started running 2 pipeline(s). |
|
/azp run runtime-coreclr jitstress, runtime-coreclr libraries-jitstress |
|
Azure Pipelines successfully started running 2 pipeline(s). |
|
/azp run runtime-coreclr jitstress, runtime-coreclr libraries-jitstress |
|
Azure Pipelines successfully started running 2 pipeline(s). |
|
/azp run runtime-coreclr jitstress, runtime-coreclr libraries-jitstress |
|
Azure Pipelines successfully started running 2 pipeline(s). |
|
Yay, stress is finally green. Going to remove the stress mode I pointed at in #85975 (comment) and then merge. |
ghost
locked as resolved and limited conversation to collaborators
2 participants
Instead of using the numeric value of each stress mode inside the enum calculate a hash based on the string name. This has the benefit that it won't change based on new members being added/reordered and reduces correlation of which stress modes get enabled together.
Fix #83733