Stream: git-wasmtime

Topic: wasmtime / PR #1564 aarch64: correctly pass f32/f64 stack...


view this post on Zulip Wasmtime GitHub notifications bot (Apr 21 2020 at 13:46):

bnjbvr opened PR #1564 from aarch64-fp-stack-args to master:

Thanks to a particularly argument-heavy Spidermonkey test case, I noticed that f32/f64 stack arguments were incorrectly passed as GPR. Here's a fix.

@cfallin, PTAL.

view this post on Zulip Wasmtime GitHub notifications bot (Apr 21 2020 at 13:46):

bnjbvr requested cfallin for a review on PR #1564.

view this post on Zulip Wasmtime GitHub notifications bot (Apr 21 2020 at 14:06):

bnjbvr submitted PR Review.

view this post on Zulip Wasmtime GitHub notifications bot (Apr 21 2020 at 14:06):

bnjbvr created PR Review Comment:

nit: store_stack

view this post on Zulip Wasmtime GitHub notifications bot (Apr 21 2020 at 14:08):

bnjbvr updated PR #1564 from aarch64-fp-stack-args to master:

Thanks to a particularly argument-heavy Spidermonkey test case, I noticed that f32/f64 stack arguments were incorrectly passed as GPR. Here's a fix.

@cfallin, PTAL.

view this post on Zulip Wasmtime GitHub notifications bot (Apr 21 2020 at 14:56):

cfallin submitted PR Review.

view this post on Zulip Wasmtime GitHub notifications bot (Apr 21 2020 at 14:56):

cfallin submitted PR Review.

view this post on Zulip Wasmtime GitHub notifications bot (Apr 21 2020 at 14:56):

cfallin created PR Review Comment:

I just realized: should this 16 be dependent on the calling convention? IIRC the Baldrdash/Wasm stack frame is actually 24 bytes, not 16 (and maybe would be padded to 32 to maintain 16-alignment, I'm not sure)?

view this post on Zulip Wasmtime GitHub notifications bot (Apr 21 2020 at 15:04):

bnjbvr submitted PR Review.

view this post on Zulip Wasmtime GitHub notifications bot (Apr 21 2020 at 15:04):

bnjbvr created PR Review Comment:

Yep, this is handled in #1530 :)

view this post on Zulip Wasmtime GitHub notifications bot (Apr 21 2020 at 15:15):

bnjbvr updated PR #1564 from aarch64-fp-stack-args to master:

Thanks to a particularly argument-heavy Spidermonkey test case, I noticed that f32/f64 stack arguments were incorrectly passed as GPR. Here's a fix.

@cfallin, PTAL.


Last updated: Dec 23 2024 at 13:07 UTC