alexcrichton requested afonso360 for a review on PR #7252.
alexcrichton requested elliottt for a review on PR #7252.
alexcrichton opened PR #7252 from alexcrichton:rv64-zbs
to bytecodealliance:main
:
This commit extends the riscv64 backend with instructions from the
Zbs
extension. I did this awhile back for the bmi{1,2} instructions for x64 and this is mostly for completeness's sake rather than performance or anything like that, and because I had a long plane ride!
alexcrichton requested wasmtime-compiler-reviewers for a review on PR #7252.
afonso360 submitted PR review:
Nice! Some small suggestions if you'd like to expand on this.
With so many patterns here I think it might be worth eventually creating intermediate instructions in CLIF and moving some of these rules to the egraphs layer. But this looks good to me!
afonso360 submitted PR review:
Nice! Some small suggestions if you'd like to expand on this.
With so many patterns here I think it might be worth eventually creating intermediate instructions in CLIF and moving some of these rules to the egraphs layer. But this looks good to me!
afonso360 created PR review comment:
Also, perhaps we should add equivalent rules for these with swapped arguments. i.e.
(band (bnot ...) y)
we have them for the other instructions but not forbclr
.
afonso360 created PR review comment:
A small suggestion here would be to change this to
(u64_sub (ty_bits ty) 1)
so that we could also apply these patterns to i16 and i8.
afonso360 edited PR review comment.
afonso360 submitted PR review:
Nice! Some small suggestions if you'd like to expand on this.
With so many patterns here I think it might be worth eventually creating intermediate instructions in CLIF and moving some of these rules to the egraphs layer. But this looks good to me!
I've also been running the fuzzer on this branch and it hasn't found anything yet.
afonso360 submitted PR review:
Nice! Some small suggestions if you'd like to expand on this.
With so many patterns here I think it might be worth eventually creating intermediate instructions in CLIF and moving some of these rules to the egraphs layer. But this looks good to me! (cc: #1067)
I've also been running the fuzzer on this branch and it hasn't found anything yet.
afonso360 submitted PR review:
Nice! Some small suggestions if you'd like to expand on this.
With so many patterns here I think it might be worth eventually creating intermediate instructions in CLIF and moving some of these rules to the egraphs layer. But this looks good to me!
I've also been running the fuzzer on this branch and it hasn't found anything yet.
alexcrichton updated PR #7252.
alexcrichton has enabled auto merge for PR #7252.
alexcrichton merged PR #7252.
Last updated: Nov 22 2024 at 16:03 UTC