github-actions[bot] commented on issue #4977:
Subscribe to Label Action
cc @cfallin, @fitzgen
<details>
This issue or pull request has been labeled: "cranelift", "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>
cfallin commented on issue #4977:
good confirmation that the changes are benign
I spoke too soon actually; it seems that the
simd_lane
test is failing, but hopefully it's an easy reordering/fix...
elliottt commented on issue #4977:
good confirmation that the changes are benign
I spoke too soon actually; it seems that the
simd_lane
test is failing, but hopefully it's an easy reordering/fix...Yep, I'll make sure to add a filetest once I diagnose the problem :+1:
elliottt commented on issue #4977:
Fixed in 3b7b4ce4fff71159176b112b86122c45a60a46bf. I was trying to avoid using a priority when I could add a negative assertion on a flag value, but given that the branch I was trying to assert the negation of used an
and
with two flags, this just wasn't sound: I would need anor
here to make it work.
elliottt commented on issue #4977:
Performance seems pretty unaffected:
Here's a measurement on my laptop (which is admittedly noisy) for compiling the bz2 benchmark:
% hyperfine --runs=100 --warmup=10 --show-output "taskset 0x1 ./target/release/wasmtime compile ../sightglass/benchmarks/bz2/benchmark.wasm" Benchmark 1: taskset 0x1 ./target/release/wasmtime compile ../sightglass/benchmarks/bz2/benchmark.wasm Time (mean ± σ): 102.2 ms ± 1.2 ms [User: 96.9 ms, System: 5.3 ms] Range (min … max): 99.7 ms … 105.1 ms 100 runs
And here's the same measurement on
main
:% hyperfine --runs=100 --warmup=10 --show-output "taskset 0x1 ./target/release/wasmtime compile ../sightglass/benchmarks/bz2/benchmark.wasm" Benchmark 1: taskset 0x1 ./target/release/wasmtime compile ../sightglass/benchmarks/bz2/benchmark.wasm Time (mean ± σ): 100.8 ms ± 1.2 ms [User: 95.1 ms, System: 5.6 ms] Range (min … max): 98.9 ms … 103.6 ms 100 runs
I got pretty large variance between individual runs, but both were pretty consistently close.
Last updated: Dec 23 2024 at 13:07 UTC