add intrinsics for FP arithmetic add/div/mul/sub include rounding versions.
also add "undefined" intrinsics
Details
Details
Diff Detail
Diff Detail
- Repository
- rL LLVM
Event Timeline
Comment Actions
I think that we are inconsistent with "setzero" and "undef" for PathThru values in this file. GCC uses "undef" and it is right.
lib/Headers/avx512fintrin.h | ||
---|---|---|
52 ↗ | (On Diff #30159) | This comment should be moved down, it is related to setzero. |
889 ↗ | (On Diff #30159) | In all these cases we also should use "undefined". Right? |
lib/Headers/avx512fintrin.h | ||
---|---|---|
55 ↗ | (On Diff #30159) | What we need here is an undef value. My gut feeling is that the way to properly get an undef value would be to use a __builtin that evaluates to undef. |
lib/Headers/avx512fintrin.h | ||
---|---|---|
889 ↗ | (On Diff #30159) | yes, it should use "undefined". |