JIT: tolerate edge profile inconsistencies better. #50213
Merged
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
We should not blow up jitting just because edge counts are inconsistent. Also, make sure we're not doing any guarded devirt if we disable pgo. And, add an option JitEnablePgoRange to selectively enable use of PGO for problem isolation.
dotnet-issue-labeler
bot
added
the
area-CodeGen-coreclr
|
cc @dotnet/jit-contrib Two diffs across SPMI, both places where we were enabling guarded PGO even though fetching PGO data had failed. |
BruceForstall
approved these changes
Mar 25, 2021
src/coreclr/jit/fgprofile.cpp
Outdated
| @@ -2521,6 +2521,11 @@ void Compiler::fgIncorporateEdgeCounts() | |||
| // slop - profile slush fund | |||
| // wbUsedSlop [out] - true if we tapped into the slush fund | |||
| // | |||
| // Returns: | |||
| // true if the edge weight was adjusted | |||
| // false if the edge weight update was inconsisten with the | |||
src/coreclr/jit/fgprofile.cpp
Outdated
| @@ -2620,6 +2620,11 @@ bool flowList::setEdgeWeightMinChecked(BasicBlock::weight_t newWeight, | |||
| // slop - profile slush fund | |||
| // wbUsedSlop [out] - true if we tapped into the slush fund | |||
| // | |||
| // Returns: | |||
| // true if the edge weight was adjusted | |||
| // false if the edge weight update was inconsisten with the | |||
| @@ -463,7 +463,10 @@ CONFIG_INTEGER(JitClassProfiling, W("JitClassProfiling"), 1) | |||
| CONFIG_INTEGER(JitEdgeProfiling, W("JitEdgeProfiling"), 1) | |||
|
|
|||
| // Profile consumption options | |||
| CONFIG_INTEGER(JitDisablePGO, W("JitDisablePGO"), 0) // Ignore pgo data | |||
| CONFIG_INTEGER(JitDisablePgo, W("JitDisablePgo"), 0) // Ignore pgo data for all methods | |||
There was a problem hiding this comment.
Would this be easier to understand in reverse? E.g., JitEnablePgo=0? Like we have COMPlus_TieredCompilation=0.
Doesn't matter to me, but sometimes negatives get confusing. We certainly aren't consistent here.
There was a problem hiding this comment.
Since PGO is on by default and this option is mainly to do top-level triage/workaround for PGO issues, I think "disable" is less confusing.
Though when I needed to binary search I found having an "enable" less confusing.... so am not self-consistent here either.
ghost
locked as resolved and limited conversation to collaborators
Sign up for free
to subscribe to this conversation on GitHub.
Already have an account?
Sign in.
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
We should not blow up jitting just because edge counts are inconsistent.
Also, make sure we're not doing any guarded devirt if we disable pgo.
And, add an option JitEnablePgoRange to selectively enable use of PGO for
problem isolation.