Stream: git-wasmtime

Topic: wasmtime / issue #4061 Cranelift: fix #3953: rework singl...

Wasmtime GitHub notifications bot (Apr 21 2022 at 05:59):

github-actions[bot] commented on issue #4061:

Subscribe to Label Action

cc @cfallin, @fitzgen

<details>
This issue or pull request has been labeled: "cranelift", "cranelift:area:aarch64", "cranelift:area:machinst", "cranelift:area:x64", "isle"

Thus the following users have been cc'd because of the following labels:

• cfallin: isle
• fitzgen: isle

To subscribe or unsubscribe from this label, edit the <code>.github/subscribe-to-label.json</code> configuration file.

Learn more.
</details>

Wasmtime GitHub notifications bot (Apr 21 2022 at 06:12):

cfallin commented on issue #4061:

Some Sightglass results:

• Execution time (wallclock speedups, higher is faster):
  • pulldown-cmark: no diff
  • `spidermonkey: 1.01 - 1.02x (99% conf)
  • bz2: no diff
  • meshoptimizer: 1.16x - 1.19x (99% conf) (!)
  • blake3-scalar: no diff
  • blake3-simd: no diff
• Compilation time (wallclock speedups, higher is faster):
  • pulldown-cmark: no diff
  • spidermonkey: 0.89x - 0.95x (99% conf) (I will profile this tomorrow)
  • bz2: no diff
  • meshoptimizer: no diff
  • blake3-scalar: no diff
  • blake3-simd: no diff

So a speedup on two benchmarks, SpiderMonkey (1-2%) and meshoptimizer (16-19%); the latter I suspect is due to some opportunities for compare-load or compare-immediate merging in hot blocks (can double-check with hotblocks tool tomorrow).

The compile time regression on SpiderMonkey I'll take a look at and resolve -- we don't want to take that hit. I suspect the multiple-use analysis can be made faster if needed.

Wasmtime GitHub notifications bot (Apr 21 2022 at 06:12):

cfallin edited a comment on issue #4061:

Some Sightglass results:

• Execution time (wallclock speedups, higher is faster):
  • pulldown-cmark: no diff
  • spidermonkey: 1.01 - 1.02x (99% conf)
  • bz2: no diff
  • meshoptimizer: 1.16x - 1.19x (99% conf) (!)
  • blake3-scalar: no diff
  • blake3-simd: no diff
• Compilation time (wallclock speedups, higher is faster):
  • pulldown-cmark: no diff
  • spidermonkey: 0.89x - 0.95x (99% conf) (I will profile this tomorrow)
  • bz2: no diff
  • meshoptimizer: no diff
  • blake3-scalar: no diff
  • blake3-simd: no diff

So a speedup on two benchmarks, SpiderMonkey (1-2%) and meshoptimizer (16-19%); the latter I suspect is due to some opportunities for compare-load or compare-immediate merging in hot blocks (can double-check with hotblocks tool tomorrow).

The compile time regression on SpiderMonkey I'll take a look at and resolve -- we don't want to take that hit. I suspect the multiple-use analysis can be made faster if needed.

Wasmtime GitHub notifications bot (Apr 22 2022 at 01:24):

cfallin commented on issue #4061:

@abrown I went ahead and reworked the ValueUseState analysis for better compile time; Sightglass now shows no difference in compile time with this PR vs baseline for SpiderMonkey (or the other benchmarks). Could you take another look at the new algorithm?

Last updated: Oct 23 2024 at 20:03 UTC