1685998 - Implement SIMD "i32x4 -> i64x2 widen" instructions

Status: RESOLVED FIXED

(Core :: JavaScript: WebAssembly, enhancement, P3)

Description

[Lars T Hansen [:lth]]

See https://github.com/WebAssembly/simd/pull/290. This has been accepted into the spec, and V8, LLVM, and emscripten all support it now; binaryen is unclear. Checklist:

[ x ] Test cases
[ x ] Wast
[ x ] Baseline (x86, x64, arm64)
[ x ] Ion (x86, x64)

Ion arm64 will land separately, with all the other Ion arm64 SIMD support (bug 1687629).

Comment 1

[Lars T Hansen [:lth]]

V8 opcodes: https://chromium-review.googlesource.com/c/v8/v8/+/2441369/4/src/wasm/wasm-opcodes.h

Comment 2

[Lars T Hansen [:lth]]

wat changes: https://github.com/bytecodealliance/wasm-tools/pull/196

Comment 3

[Lars T Hansen [:lth]]

Attachment: Bug 1685998 - Implement i64x2.widen_*_i32x4. r?jseward

Implement the 32->64 widening operations.

Drive-by fixes to existing test cases for 8->16 widening: the initial
shifting of the input values made them meaningless, making everything
zero. This masked the fact that the output arrays were the wrong
element width.

Depends on D102612

Comment 4

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/integration/autoland/rev/d5051e03652c
Implement i64x2.widen_*_i32x4.  r=jseward

Comment 5

[Sandor Molnar[:smolnar]]

https://hg.mozilla.org/mozilla-central/rev/d5051e03652c