[llvm] [AMDGPU] Add custom lowering of llvm.convert.from.arbitrary.fp for FP8 (PR #194144)

[Shilei Tian via llvm-commits]

================
@@ -0,0 +1,342 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 6
+; RUN: llc -global-isel=0 < %s -mtriple=amdgcn-amd-amdhsa -mcpu=gfx942 | FileCheck -check-prefix=GFX942 %s
+; RUN: llc -global-isel=0 < %s -mtriple=amdgcn-amd-amdhsa -mcpu=gfx950 | FileCheck -check-prefix=GFX950 %s
+; RUN: llc -global-isel=0 < %s -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1170 | FileCheck -check-prefix=GFX1170 %s
+; RUN: llc -global-isel=0 < %s -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1250 | FileCheck -check-prefix=GFX1250 %s
+
+; llvm.convert.from.arbitrary.fp on FP8ConversionInsts targets.
+
+declare float @llvm.convert.from.arbitrary.fp.f32.i8(i8, metadata)
----------------
shiltian wrote:

nit: no need for intrinsic declaration

https://github.com/llvm/llvm-project/pull/194144