llvm-project

f26c9add - [NVPTX] Add lowering for v2i32/v2f16/v2f32 scalar_to_vector operations (#175712)

[NVPTX] Add lowering for v2i32/v2f16/v2f32 scalar_to_vector operations (#175712) #153478 revealed additional cases where these scalar_to_vector operations can appear at the lowering stage. Generally the DAGCombiner transforms these to something else. Without it running however these can directly manifest (hence the unit test disabling DAGCombine for simplicity) but we can map them directly to real instructions so add the proper lowering.