Stream: git-wasmtime

Topic: wasmtime / PR #6907 x64: Fix incorrect lowering of vector...


view this post on Zulip Wasmtime GitHub notifications bot (Aug 24 2023 at 22:17):

alexcrichton opened PR #6907 from alexcrichton:fix-x64-bitselect to bytecodealliance:main:

Unconditionally use pblendvb which will provide the semantics desired for bitselect since the all_ones_or_all_zeros predicate guarantees that byte-size-chunks are all-zeros-or-all-ones and pblendvb works at the byte level. As suggested by Jamey in #6890.

Closes #6890

<!--
Please make sure you include the following information:

Our development process is documented in the Wasmtime book:
https://docs.wasmtime.dev/contributing-development-process.html

Please ensure all communication follows the code of conduct:
https://github.com/bytecodealliance/wasmtime/blob/main/CODE_OF_CONDUCT.md
-->

view this post on Zulip Wasmtime GitHub notifications bot (Aug 24 2023 at 22:17):

alexcrichton requested abrown for a review on PR #6907.

view this post on Zulip Wasmtime GitHub notifications bot (Aug 24 2023 at 22:17):

alexcrichton requested wasmtime-compiler-reviewers for a review on PR #6907.

view this post on Zulip Wasmtime GitHub notifications bot (Aug 28 2023 at 14:14):

abrown submitted PR review:

This looks right to me. all_ones_or_all_zeros would have to get more complex for the previous logic to work; this is simpler. I'm also interested in @alexcrichton's comment elsewhere that it might be worthwhile to look into using a shuffle.

view this post on Zulip Wasmtime GitHub notifications bot (Aug 28 2023 at 15:04):

abrown merged PR #6907.


Last updated: Dec 23 2024 at 12:05 UTC