[llvm] [AMDGPU] Add scalar_to_vector pattern for v2bf16 (PR #225303)

[Changpeng Fang via llvm-commits]

================
@@ -0,0 +1,213 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 6
+; RUN: llc -O0 -mtriple=amdgpu12.50 -mattr=-real-true16 < %s | FileCheck -check-prefix=GFX1250-FAKE16 %s
+; RUN: llc -O0 -mtriple=amdgpu12.50 -mattr=+real-true16 < %s | FileCheck -check-prefix=GFX1250-TRUE16 %s
----------------
changpeng wrote:

> Maybe worth to add GISel run line, and then fix it or do whatever we need in a follow-up.

There are unrelated failures on gisel runs for the test. Will look whether it is actual issue that deserves to fix.

https://github.com/llvm/llvm-project/pull/225303