[llvm] [AMDGPU] Handle negated f16 and fp conversion DAG combines (PR #213202)

[via llvm-commits]

================
@@ -0,0 +1,117 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 6
+; RUN: llc -mtriple=amdgpu9.0a-amd-amdhsa < %s | FileCheck %s
+
+define amdgpu_kernel void @fneg_fpext_half(ptr addrspace(1) %out, ptr addrspace(1) %in) {
+; CHECK-LABEL: fneg_fpext_half:
+; CHECK:       ; %bb.0: ; %entry
+; CHECK-NEXT:    s_load_dwordx4 s[0:3], s[8:9], 0x0
+; CHECK-NEXT:    v_mov_b32_e32 v0, 0
+; CHECK-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-NEXT:    global_load_ushort v1, v0, s[2:3]
+; CHECK-NEXT:    s_waitcnt vmcnt(0)
+; CHECK-NEXT:    v_cvt_f32_f16_e64 v1, -v1
+; CHECK-NEXT:    global_store_dword v0, v1, s[0:1]
+; CHECK-NEXT:    s_endpgm
+entry:
+  %h = load half, ptr addrspace(1) %in, align 2
+  %f = fpext half %h to float
+  %neg = fneg float %f
----------------
HalfBloodPrince010 wrote:

Added a new test case for the FP16_TO_FP case.

https://github.com/llvm/llvm-project/pull/213202