Implement llvm.ctpop.v* intrinsics
#3072
Conversation
Tested through x86 avx512vpopcntdq and avx512bitalg functions.
| let (op, op_len) = this.operand_to_simd(op)?; | ||
| let (dest, dest_len) = this.place_to_simd(dest)?; | ||
|
|
||
| assert_eq!(dest_len, op_len); |
This comment was marked as outdated.
This comment was marked as outdated.
Sorry, something went wrong.
There was a problem hiding this comment.
We generally have the policy that ICEing when intrinsics are used incorrectly is fine. I think we should apply that to LLVM intrinsics as well. If rust-lang/rust#116093 lands, people will get an appropriate warning when enabling the feature.
|
Thanks! |
|
☀️ Test successful - checks-actions |
There was a problem hiding this comment.
These test files are getting a bit too fine-grained IMO. What about just using intrinsics-x86-other for "everything except the SSE/SSE2 intrinsics"? llvm.x86.addcarry.64 should also be tested there, probably.
There was a problem hiding this comment.
Ah, we already have tests/pass/intrinsics-x86.rs. So this could be added there?
There was a problem hiding this comment.
The tests are grouped by what CPU features they use, and the avx512 CPU features are significantly more fine-grained than the others. I wouldn't mind lumping all the avx512 intrinsics together, but I don't think they belong in "other" with llvm.x86.rdtsc.
There was a problem hiding this comment.
Ah, that's fair. But the two test files added by this PR should IMO be merged.
@eduardosm can you wile a follow-up PR?
Tested through x86 avx512vpopcntdq and avx512bitalg functions.
I picked them from #2057 (comment), which looked easy.