Stream: git-wasmtime

Topic: wasmtime / issue #5097 Remove redundant branch and select...

view this post on Zulip Wasmtime GitHub notifications bot (Oct 24 2022 at 17:03):

github-actions[bot] commented on issue #5097:

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", "cranelift:meta", "isle"

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

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

Learn more.
</details>

view this post on Zulip Wasmtime GitHub notifications bot (Oct 24 2022 at 18:35):

abrown commented on issue #5097:

I'm taking a look at this as well...

view this post on Zulip Wasmtime GitHub notifications bot (Oct 24 2022 at 19:03):

elliottt commented on issue #5097:

I think overall this is a good change! I looked primarily at the x64 optimizations and I see that the lowering optimizations for br* and *cmp are already in place here. I guess my only concern is whether the "forced duplicate comparisons" will become a performance issue — can you do a sightglass run of a few benchmarks and post it here? (I wonder, and feel free to say "no," whether it might be helpful to add an x64 ISA test to show (and lock in) that br using an icmp will in fact result in only two instructions and not materialize the comparison result?).

I'll run sightglass benchmarks for this branch, and also add the test to verify that we're not materializing the comparison result. Thanks!

view this post on Zulip Wasmtime GitHub notifications bot (Oct 24 2022 at 20:37):

elliottt commented on issue #5097:

@abrown I ran sightglass benchmarks on a dedicated machine, and detected no change between this branch and main. At the time I ran this benchmark, 85f864d95e2836dbf0fbfc339a0bd1d2f96e883e was the head of this branch.

$ ./target/release/sightglass-cli benchmark benchmarks/bz2/benchmark.wasm \
  --engine ../main.so \
  --engine ../85f864d95e2836dbf0fbfc339a0bd1d2f96e883e.so \
  --iterations-per-process=100 \
  --processes=1

instantiation :: cycles :: benchmarks/bz2/benchmark.wasm

  No difference in performance.

  [120656 189764.02 320940] 85f864d95e2836dbf0fbfc339a0bd1d2f96e883e.so
  [123442 198471.42 317082] main.so

compilation :: cycles :: benchmarks/bz2/benchmark.wasm

  No difference in performance.

  [223741976 367702912.90 592279790] 85f864d95e2836dbf0fbfc339a0bd1d2f96e883e.so
  [220267650 361527444.68 508238136] main.so

execution :: cycles :: benchmarks/bz2/benchmark.wasm

  No difference in performance.

  [130806336 193406633.08 232802798] 85f864d95e2836dbf0fbfc339a0bd1d2f96e883e.so
  [130420804 192605053.70 231372888] main.so


$ ./target/release/sightglass-cli benchmark benchmarks/spidermonkey/benchmark.wasm \
  --engine ../main.so \
  --engine ../85f864d95e2836dbf0fbfc339a0bd1d2f96e883e.so \
  --iterations-per-process=10 \
  --processes=1

execution :: cycles :: benchmarks/spidermonkey/benchmark.wasm

  No difference in performance.

  [1410446964 2020809040.60 2306033394] 85f864d95e2836dbf0fbfc339a0bd1d2f96e883e.so
  [1359735690 2191588531.20 2308222376] main.so

instantiation :: cycles :: benchmarks/spidermonkey/benchmark.wasm

  No difference in performance.

  [611792 859291.40 951768] 85f864d95e2836dbf0fbfc339a0bd1d2f96e883e.so
  [597436 820546.20 932754] main.so

compilation :: cycles :: benchmarks/spidermonkey/benchmark.wasm

  No difference in performance.

  [9393256300 9918239033.80 10205566518] 85f864d95e2836dbf0fbfc339a0bd1d2f96e883e.so
  [8561503454 9716444659.00 10357871886] main.so

Last updated: Nov 22 2024 at 17:03 UTC