Stream: git-wasmtime

Topic: wasmtime / PR #9897 Make Pulley pass `simd_f32x4_arith.wast`

Wasmtime GitHub notifications bot (Dec 23 2024 at 20:23):

tyoeer opened PR #9897 from tyoeer:main to bytecodealliance:main:

Helping towards issue #9783.

This PR makes Pulley pass spec_testsuite/simd_f32x4_arith.wast, by adding f32x4 instructions for subtraction, negation, multiplication, and division.

Random notes:

• Added instructions placement and doc comments in pulley/src/lib.rs are based on what is done with vtrunc32x4.
• Added instructions implementation is based on vmuli32x4.
• How instructions are generally grouped together/ordered appears to vary, e.g. ftrunc32, vtrunc32x4, and vtrunc64x2 are together, but ftrunc64 isn't near there.
• It also appears inconsistent if a SIMD float instruction has an f at the end, e.g. vtrunc32x4 vs. vabsf32x4.

Wasmtime GitHub notifications bot (Dec 23 2024 at 20:23):

tyoeer requested wasmtime-compiler-reviewers for a review on PR #9897.

Wasmtime GitHub notifications bot (Dec 23 2024 at 20:23):

tyoeer requested fitzgen for a review on PR #9897.

Wasmtime GitHub notifications bot (Dec 23 2024 at 20:23):

tyoeer requested wasmtime-core-reviewers for a review on PR #9897.

Wasmtime GitHub notifications bot (Dec 23 2024 at 20:23):

tyoeer requested wasmtime-default-reviewers for a review on PR #9897.

Wasmtime GitHub notifications bot (Dec 23 2024 at 21:45):

github-actions[bot] commented on PR #9897:

Subscribe to Label Action

cc @fitzgen

<details>
This issue or pull request has been labeled: "cranelift", "pulley"

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

• fitzgen: pulley

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 (Dec 27 2024 at 05:45):

alexcrichton submitted PR review:

Thanks!

Wasmtime GitHub notifications bot (Dec 27 2024 at 05:45):

alexcrichton created PR review comment:

Mind using f32x4 in the opcodes (e.g. vsubf32x4) to clearly indicate that it's for floats and not integers?

Wasmtime GitHub notifications bot (Jan 02 2025 at 13:08):

tyoeer updated PR #9897.

Wasmtime GitHub notifications bot (Jan 02 2025 at 13:09):

tyoeer submitted PR review.

Wasmtime GitHub notifications bot (Jan 02 2025 at 13:09):

tyoeer created PR review comment:

Done

Wasmtime GitHub notifications bot (Jan 02 2025 at 13:09):

tyoeer edited PR review comment.

Wasmtime GitHub notifications bot (Jan 02 2025 at 14:10):

tyoeer requested alexcrichton for a review on PR #9897.

Wasmtime GitHub notifications bot (Jan 06 2025 at 15:40):

alexcrichton commented on PR #9897:

@tyoeer I think this might need a rebase?

Wasmtime GitHub notifications bot (Jan 06 2025 at 16:07):

tyoeer updated PR #9897.

Wasmtime GitHub notifications bot (Jan 06 2025 at 16:13):

alexcrichton has enabled auto merge for PR #9897.

Wasmtime GitHub notifications bot (Jan 06 2025 at 16:13):

alexcrichton submitted PR review.

Wasmtime GitHub notifications bot (Jan 06 2025 at 16:18):

tyoeer commented on PR #9897:

@tyoeer I think this might need a rebase?

It has been rebased. Another PR has also already added vdivf32x4, so that one was removed from this PR during merging.
https://docs.wasmtime.dev/contributing-development-process.html#focused-commits-or-squashing mentions that it's preferred to not rebase when making changes, which is why I waited until I was specifically asked to.

Wasmtime GitHub notifications bot (Jan 06 2025 at 16:18):

tyoeer edited PR #9897:

Helping towards issue #9783.

This PR makes Pulley pass spec_testsuite/simd_f32x4_arith.wast, by adding f32x4 instructions for subtraction, negation, multiplication, and division (division already got added by another PR).

Random notes:

• Added instructions placement and doc comments in pulley/src/lib.rs are based on what is done with vtrunc32x4.
• Added instructions implementation is based on vmuli32x4.
• How instructions are generally grouped together/ordered appears to vary, e.g. ftrunc32, vtrunc32x4, and vtrunc64x2 are together, but ftrunc64 isn't near there.
• It also appears inconsistent if a SIMD float instruction has an f at the end, e.g. vtrunc32x4 vs. vabsf32x4.

Wasmtime GitHub notifications bot (Jan 06 2025 at 16:22):

alexcrichton commented on PR #9897:

Oh no worries! This falls into the "various technical reasons" category where this couldn't be merged due to a merge conflict. I'll go update those docs though to mention that, thanks for dropping the link!

Wasmtime GitHub notifications bot (Jan 06 2025 at 16:35):

alexcrichton merged PR #9897.

Last updated: Jan 24 2025 at 00:11 UTC