Renumber SIMD instructions based on final spec
Categories
(Core :: JavaScript: WebAssembly, enhancement, P2)
Tracking
()
Tracking | Status | |
---|---|---|
firefox88 | --- | fixed |
People
(Reporter: yury, Assigned: yury)
References
(Blocks 1 open bug)
Details
Attachments
(1 file)
The https://github.com/WebAssembly/simd/pull/452 was merged. The encoding of SIMD ops in Firefox has to be adjusted based on https://github.com/WebAssembly/simd/blob/master/proposals/simd/BinarySIMD.md .
The wasm-tools library was modified already https://github.com/bytecodealliance/wasm-tools/pull/222
Assignee | ||
Comment 1•4 years ago
|
||
Renumbers instructions based on final SIMD spec.
Removes I16x8AnyTrue / I32x4AnyTrue
Disables tests with (struct.narrow)
Comment 2•4 years ago
|
||
We should probably not land this until we're sure that upstream tools (emscripten, binaryen etc) have been updated to produce new opcodes?
Assignee | ||
Comment 3•4 years ago
|
||
The only risk the final codes will be changed during course of upstream tools implementation. We can also monitor status of op renumbering in the other JS engines, and land if any of the above will commit to new encoding.
Probably we also need to sync with particular FF version release for ease of troubleshooting or user agent version/feature identification.
Comment 4•4 years ago
|
||
Yeah, it wasn't risk I was thinking about - I think we're all pretty happy with the new opcodes - as much as upsetting people who are using older tools while we accept only newer opcodes, eg, Bergamot.
Assignee | ||
Comment 5•4 years ago
•
|
||
Found another TODO before landing this one: update or disable ARM64 cranelift for changed ops. Failing try tests https://treeherder.mozilla.org/jobs?repo=try&revision=13c6862e79f5944fd0a65e6d1aa18fb01ab78f3f&selectedTaskRun=aBcoGFgLR1KHGHLFVWY6zQ.0
Assignee | ||
Comment 6•4 years ago
|
||
Tracking proposal status across vendors at https://github.com/WebAssembly/simd/issues/480
Comment 8•4 years ago
|
||
bugherder |
Description
•