llvm-project

8b6145d2 - [NVPTX] Add intrinsics for ff/f16/bf16 to ue5m3 conversions (#218677)

[NVPTX] Add intrinsics for ff/f16/bf16 to ue5m3 conversions (#218677) This patch adds the following intrinsics for `ff/f16/bf16` to `ue5m3` conversions introduced in PTX 9.4: - `cvt{.rn,.rz,.rp}{.satfinite}.ue5m3x2.f32` - `cvt{.rn,.rz}{.satfinite}.scaled::n1::ue8m0.ue5m3x2.f32` - `cvt{.rn,.rz,.rp}{.satfinite}.ue5m3x2{.f16x2,.bf16x2}` - `cvt{.rn,.rz}{.satfinite}.scaled::n1::ue8m0.ue5m3x2{.f16x2,.bf16x2}` Tests have been verified through `ptxas-13.4`. PTX ISA Reference: https://docs.nvidia.com/cuda/developer-preview/13.4/parallel-thread-execution/index.html#data-movement-and-conversion-instructions-cvt --------- Signed-off-by: DharuniRAcharya <dharunira@nvidia.com>