[PATCH] D133972: [AMDGPU] Fix size of SOPK instructions to 4 bytes

Matt Arsenault via Phabricator via llvm-commits llvm-commits at lists.llvm.org
Mon Sep 19 06:17:39 PDT 2022


arsenm added inline comments.


================
Comment at: llvm/test/CodeGen/AMDGPU/sopk-no-literal.ll:1
+; RUN: llc -O0 -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1030 -debug-only=branch-relaxation %s -o - 2>&1 | FileCheck %s
+
----------------
If you're going to rely on -debug-only, this needs REQUIRES: asserts


================
Comment at: llvm/test/CodeGen/AMDGPU/sopk-no-literal.ll:11
+; CHECK: foo:
+; Function Attrs: convergent noinline nounwind
+define void @foo(i32 %a, i32 %b, float %c, float %d) unnamed_addr #8 {
----------------
Don't need attrs comment


================
Comment at: llvm/test/CodeGen/AMDGPU/sopk-no-literal.ll:12
+; Function Attrs: convergent noinline nounwind
+define void @foo(i32 %a, i32 %b, float %c, float %d) unnamed_addr #8 {
+; CHECK-NEXT: bb.0:
----------------
I'm not seeing where there's a sopk instruction in this function


================
Comment at: llvm/test/CodeGen/AMDGPU/sopk-no-literal.ll:89-90
+; CHECK-NEXT: s_setpc_b64 s[30:31]
+  %18 = fadd reassoc nnan ninf nsz contract float %c, %d
+  %19 = fadd reassoc nnan ninf nsz contract float %18, %c
+  ret void
----------------
Don't need all the instruction flags. Should also use named values


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D133972/new/

https://reviews.llvm.org/D133972



More information about the llvm-commits mailing list