Add support for SIMD NaN directives #686
Conversation
|
cc: @alexcrichton: thanks for |
|
Nice! Is it possible to enable some more simd tests as a result of this change? |
Not immediately, but that's where I'm headed. It allows me to see other problems, e.g., the ones I'm trying to fix in bytecodealliance/cranelift#1272. Re: the build failures, I wonder if they are due to endianness; I assumed I could convert |
abrown
force-pushed
the
add-simd-nan-asserts
branch
from
31adcad to
0cefe47
Compare
Never mind, it was just buggy coding 😄. In any case, after thinking about endianness I remembered that Wasm values are little-endian so that should not be an issue here. |
|
👍 |
…nking pass hasn't been run; Also: - make sure to apply diversions when determining offsets for code relaxation. - select the smallest encoding when selecting a relaxed branch instruction.
Until WebAssembly/spec#1104 is merged and the changes are implemented downstream, I would like to be able to run the SIMD spec tests, some of which contain directives like those added in bytecodealliance/wat#33. This change implements those variants. I am not opposed to refactoring the directive-matching code to reduce duplication but I figured that the current changes will be sufficient until all of these directives are merged into
assert_return.