JIT: Improve BBJ_COND to BBJ_ALWAYS conversion in fgUpdateFlowGraph #97681
Conversation
This file contains hidden or 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
|
Tagging subscribers to this area: @JulieLeeMSFT, @jakobbotsch Issue DetailsFollow-up to #97488. In
|
Diff results for #97681Throughput diffsThroughput diffs for linux/arm64 ran on windows/x64Overall (+0.02% to +0.06%)
FullOpts (+0.03% to +0.07%)
Throughput diffs for linux/x64 ran on windows/x64Overall (+0.02% to +0.07%)
FullOpts (+0.04% to +0.08%)
Throughput diffs for osx/arm64 ran on windows/x64Overall (+0.02% to +0.06%)
FullOpts (+0.03% to +0.06%)
Throughput diffs for windows/arm64 ran on windows/x64Overall (+0.02% to +0.06%)
MinOpts (-0.00% to +0.01%)
FullOpts (+0.03% to +0.06%)
Throughput diffs for windows/x64 ran on windows/x64Overall (+0.02% to +0.06%)
FullOpts (+0.04% to +0.06%)
Details here |
Diff results for #97681Assembly diffsAssembly diffs for linux/arm64 ran on windows/x64Diffs are based on 2,259,470 contexts (1,008,044 MinOpts, 1,251,426 FullOpts). MISSED contexts: 159 (0.01%) Overall (-42,568 bytes)
FullOpts (-42,568 bytes)
Assembly diffs for linux/x64 ran on windows/x64Diffs are based on 2,249,703 contexts (981,298 MinOpts, 1,268,405 FullOpts). MISSED contexts: 134 (0.01%) Overall (-47,589 bytes)
FullOpts (-47,589 bytes)
Assembly diffs for windows/x64 ran on windows/x64Diffs are based on 2,098,526 contexts (926,221 MinOpts, 1,172,305 FullOpts). MISSED contexts: 138 (0.01%) Overall (-41,586 bytes)
FullOpts (-41,586 bytes)
Details here Assembly diffs for windows/x86 ran on windows/x86Diffs are based on 2,290,721 contexts (838,165 MinOpts, 1,452,556 FullOpts). MISSED contexts: base: 808 (0.04%), diff: 842 (0.04%) Overall (-20,881 bytes)
FullOpts (-20,881 bytes)
Details here Throughput diffsThroughput diffs for linux/arm ran on windows/x86Overall (+0.01% to +0.03%)
FullOpts (+0.01% to +0.04%)
Throughput diffs for windows/x86 ran on windows/x86Overall (+0.02% to +0.06%)
FullOpts (+0.03% to +0.06%)
Details here Throughput diffs for linux/arm64 ran on linux/x64Overall (+0.02% to +0.07%)
FullOpts (+0.03% to +0.08%)
Throughput diffs for linux/x64 ran on linux/x64Overall (+0.02% to +0.08%)
FullOpts (+0.04% to +0.09%)
Details here |
Diff results for #97681Assembly diffsAssembly diffs for windows/x64 ran on linux/x64Diffs are based on 2,512,204 contexts (997,391 MinOpts, 1,514,813 FullOpts). MISSED contexts: 8 (0.00%) Overall (-52,115 bytes)
FullOpts (-52,115 bytes)
Details here Throughput diffsThroughput diffs for linux/arm64 ran on windows/x64Overall (+0.02% to +0.06%)
FullOpts (+0.03% to +0.07%)
Throughput diffs for linux/x64 ran on windows/x64Overall (+0.02% to +0.07%)
FullOpts (+0.04% to +0.08%)
Throughput diffs for osx/arm64 ran on windows/x64Overall (+0.02% to +0.06%)
FullOpts (+0.03% to +0.06%)
Throughput diffs for windows/arm64 ran on windows/x64Overall (+0.02% to +0.06%)
MinOpts (-0.01% to +0.00%)
FullOpts (+0.03% to +0.06%)
Throughput diffs for windows/x64 ran on windows/x64Overall (+0.02% to +0.07%)
FullOpts (+0.03% to +0.07%)
Details here Throughput diffs for windows/x86 ran on linux/x86Overall (+0.02% to +0.06%)
FullOpts (+0.03% to +0.06%)
Details here |
Diff results for #97681Assembly diffsAssembly diffs for linux/x64 ran on windows/x64Diffs are based on 2,517,901 contexts (991,070 MinOpts, 1,526,831 FullOpts). MISSED contexts: 8 (0.00%) Overall (-50,401 bytes)
FullOpts (-50,401 bytes)
Assembly diffs for windows/x64 ran on windows/x64Diffs are based on 2,512,204 contexts (997,391 MinOpts, 1,514,813 FullOpts). MISSED contexts: 8 (0.00%) Overall (-52,115 bytes)
FullOpts (-52,115 bytes)
Details here Assembly diffs for windows/x86 ran on windows/x86Diffs are based on 2,293,423 contexts (839,658 MinOpts, 1,453,765 FullOpts). MISSED contexts: base: 38 (0.00%), diff: 73 (0.00%) Overall (-22,900 bytes)
FullOpts (-22,900 bytes)
Details here Throughput diffsThroughput diffs for linux/arm64 ran on windows/x64Overall (+0.02% to +0.06%)
FullOpts (+0.03% to +0.07%)
Throughput diffs for linux/x64 ran on windows/x64Overall (+0.02% to +0.07%)
FullOpts (+0.04% to +0.08%)
Throughput diffs for osx/arm64 ran on windows/x64Overall (+0.02% to +0.06%)
FullOpts (+0.03% to +0.06%)
Throughput diffs for windows/arm64 ran on windows/x64Overall (+0.02% to +0.06%)
MinOpts (-0.01% to +0.00%)
FullOpts (+0.03% to +0.06%)
Throughput diffs for windows/x64 ran on windows/x64Overall (+0.02% to +0.07%)
FullOpts (+0.03% to +0.07%)
Details here Throughput diffs for linux/arm ran on windows/x86Overall (+0.01% to +0.04%)
FullOpts (+0.01% to +0.04%)
Throughput diffs for windows/x86 ran on windows/x86Overall (+0.02% to +0.06%)
FullOpts (+0.03% to +0.06%)
Details here Throughput diffs for linux/arm64 ran on linux/x64Overall (-0.14% to +0.00%)
FullOpts (-0.16% to +0.00%)
Throughput diffs for linux/x64 ran on linux/x64Overall (-0.12% to +0.00%)
FullOpts (-0.13% to +0.00%)
Details here |
Contributor
Author
|
cc @dotnet/jit-contrib, @AndyAyersMS PTAL. |
Diff results for #97681Assembly diffsAssembly diffs for linux/x64 ran on windows/x64Diffs are based on 2,517,901 contexts (991,070 MinOpts, 1,526,831 FullOpts). MISSED contexts: 8 (0.00%) Overall (-50,401 bytes)
FullOpts (-50,401 bytes)
Assembly diffs for windows/x64 ran on windows/x64Diffs are based on 2,512,204 contexts (997,391 MinOpts, 1,514,813 FullOpts). MISSED contexts: 8 (0.00%) Overall (-52,115 bytes)
FullOpts (-52,115 bytes)
Details here Assembly diffs for windows/x86 ran on windows/x86Diffs are based on 2,293,423 contexts (839,658 MinOpts, 1,453,765 FullOpts). MISSED contexts: base: 38 (0.00%), diff: 73 (0.00%) Overall (-22,900 bytes)
FullOpts (-22,900 bytes)
Details here Throughput diffsThroughput diffs for linux/arm ran on windows/x86Overall (+0.01% to +0.04%)
FullOpts (+0.01% to +0.04%)
Throughput diffs for windows/x86 ran on windows/x86Overall (+0.02% to +0.06%)
FullOpts (+0.03% to +0.06%)
Details here Throughput diffs for linux/arm64 ran on linux/x64Overall (-0.14% to +0.00%)
FullOpts (-0.16% to +0.00%)
Throughput diffs for linux/x64 ran on linux/x64Overall (-0.12% to +0.00%)
FullOpts (-0.13% to +0.00%)
Details here |
Contributor
Author
|
I've put this work on hold in favor of the successor edge items in #93020. I'll come back to this. |
Sign up for free
to subscribe to this conversation on GitHub.
Already have an account?
Sign in.
2 participants
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.
Follow-up to #97488. In
fgUpdateFlowGraph, we already have a check for convertingBBJ_CONDblocks toBBJ_ALWAYSblocks when the true and false targets are the same. This PR marginally improves the number of cases we can optimize by checking if the true/false target is aBBJ_ALWAYSto the false/true target. This pattern can occur inadvertently if a block's false target is an emptyBBJ_ALWAYSto the "real" false target (in other words, theBBJ_ALWAYSwas inserted to maintain the implicit fallthrough invariant during block reordering), and other flowgraph optimizations (like compacting) cause theBBJ_ALWAYSblock to point to the true target. In such cases, we can convert theBBJ_CONDblock to aBBJ_ALWAYSblock, and potentially delete the emptyBBJ_ALWAYS.