jlb6740 commented on issue #3209:
Assertion was unnecessary and incorrect. It assumed input types had to align with a very specific Wasm instruction which is not the right thing to do.
jlb6740 commented on issue #3209:
fix for #3161
jlb6740 commented on issue #3209:
Also seems to fix #3160
jlb6740 commented on issue #3209:
Thanks @alexcrichton. This really just needed refactoring to avoid entering prematurely the branch that implements F64x2ConvertLowI32x4U. Above is a fix .. hopefully nothing is amiss.
alexcrichton commented on issue #3209:
I'll defer to @cfallin's review of the code here, I'm not super familiar with the x64 backend myself
(although I do think it would be good to add some tests)
jlb6740 commented on issue #3209:
This looks good to me now -- thanks for bearing with us through all the discussion!
No, thank you. Simply needed to refactor after interpreting correctly what the bug was saying. Highlights the need for fuzz testing when lowering becomes branchy.
jlb6740 edited a comment on issue #3209:
This looks good to me now -- thanks for bearing with us through all the discussion!
@cfallin No, thank you. Simply needed to refactor after interpreting correctly what the bug was saying. Highlights the need for fuzz testing when lowering becomes branchy.
Last updated: Dec 23 2024 at 12:05 UTC