[llvm] [AMDGPU] Generate s_absdiff_i32 (PR #164835)

[via llvm-commits]

================
@@ -0,0 +1,12 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc -mtriple=amdgcn-amd-amdpal -mcpu=gfx900 < %s | FileCheck %s
+
+define amdgpu_ps i32 @absdiff_v1(i32 inreg %arg, i32 inreg %arg2) {
+; CHECK-LABEL: absdiff_v1:
+; CHECK:       ; %bb.0:
+; CHECK-NEXT:    s_absdiff_i32 s0, s0, s1
+; CHECK-NEXT:    ; return to shader part epilog
+  %diff = sub i32 %arg, %arg2
+  %res = call i32 @llvm.abs.i32(i32 %diff, i1 false)
+  ret i32 %res
+}
----------------
LU-JOHN wrote:

Added negative test for multi-use case.

https://github.com/llvm/llvm-project/pull/164835