[PATCH] D79073: [AMDGPU] For PAL, make sure Scratch Buffer Descriptor do not clobber GIT pointer
Matt Arsenault via Phabricator via llvm-commits
llvm-commits at lists.llvm.org
Fri May 1 17:16:06 PDT 2020
arsenm added inline comments.
================
Comment at: llvm/lib/Target/AMDGPU/SIFrameLowering.cpp:491
.addMemOperand(MMO);
- } else if (ST.isMesaGfxShader(Fn) ||
- (PreloadedScratchRsrcReg == AMDGPU::NoRegister)) {
+ } else if (ST.isMesaGfxShader(Fn) || (!PreloadedScratchRsrcReg)) {
assert(!ST.isAmdHsaOrMesa(Fn));
----------------
Extra parentheses
================
Comment at: llvm/test/CodeGen/AMDGPU/SRSRC-GIT-clobber-check.mir:1
+# NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py
+# RUN: llc -mtriple=amdgcn--amdpal -mcpu=gfx1010 -run-pass=prologepilog -o - %s | FileCheck %s
----------------
Comment seems to be a lie. The label and some instructions are missing
================
Comment at: llvm/test/CodeGen/AMDGPU/SRSRC-GIT-clobber-check.mir:9
+
+ define amdgpu_gs void @shader(i32 inreg %mergedGroupInfo) #0 {
+ ret void
----------------
Dead "#0"
================
Comment at: llvm/test/CodeGen/AMDGPU/SRSRC-GIT-clobber-check.mir:28
+ ; CHECK: $sgpr4 = S_MOV_B32 $sgpr8, implicit-def $sgpr4_sgpr5_sgpr6_sgpr7
+ ; CHECK: $sgpr4_sgpr5_sgpr6_sgpr7 = S_LOAD_DWORDX4_IMM $sgpr4_sgpr5, 0, 0, 0, implicit-def $sgpr4_sgpr5_sgpr6_sgpr7 :: (dereferenceable invariant load 16, align 4, addrspace 4)
+ bb.0:
----------------
Odd that we ended up with a def and implicit def of the same register
CHANGES SINCE LAST ACTION
https://reviews.llvm.org/D79073/new/
https://reviews.llvm.org/D79073
More information about the llvm-commits
mailing list