[llvm] 215f92b - [AMDGPU] Fix crash in the SILoadStoreOptimizer (#93862)

[via llvm-commits]

Author: Stanislav Mekhanoshin
Date: 2024-05-30T14:27:33-07:00
New Revision: 215f92b979277a7ea4d331ae25a3ab9b77c9da0e

URL: https://github.com/llvm/llvm-project/commit/215f92b979277a7ea4d331ae25a3ab9b77c9da0e
DIFF: https://github.com/llvm/llvm-project/commit/215f92b979277a7ea4d331ae25a3ab9b77c9da0e.diff

LOG: [AMDGPU] Fix crash in the SILoadStoreOptimizer (#93862)

It does not properly handle situation when address calculation uses
V_ADDC_U32 0, 0, carry-in (i.e. with both src0 and src1 immediates).

Added: 
    llvm/test/CodeGen/AMDGPU/load-store-opt-addc0.mir

Modified: 
    llvm/lib/Target/AMDGPU/SILoadStoreOptimizer.cpp

Removed: 
    


################################################################################
diff  --git a/llvm/lib/Target/AMDGPU/SILoadStoreOptimizer.cpp b/llvm/lib/Target/AMDGPU/SILoadStoreOptimizer.cpp
index 83f922fb09ae5..774920aac2f08 100644
--- a/llvm/lib/Target/AMDGPU/SILoadStoreOptimizer.cpp
+++ b/llvm/lib/Target/AMDGPU/SILoadStoreOptimizer.cpp
@@ -2034,7 +2034,7 @@ void SILoadStoreOptimizer::processBaseWithConstOffset(const MachineOperand &Base
   if (Src0->isImm())
     std::swap(Src0, Src1);
 
-  if (!Src1->isImm())
+  if (!Src1->isImm() || Src0->isImm())
     return;
 
   uint64_t Offset1 = Src1->getImm();

diff  --git a/llvm/test/CodeGen/AMDGPU/load-store-opt-addc0.mir b/llvm/test/CodeGen/AMDGPU/load-store-opt-addc0.mir
new file mode 100644
index 0000000000000..0f22be7968458
--- /dev/null
+++ b/llvm/test/CodeGen/AMDGPU/load-store-opt-addc0.mir
@@ -0,0 +1,26 @@
+# NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py UTC_ARGS: --version 5
+# RUN: llc -march=amdgcn -mcpu=gfx900 -run-pass=si-load-store-opt -o - %s | FileCheck --check-prefix=GCN %s
+
+# This used to crash
+
+---
+name:            analyze_addc_0_0
+body:             |
+  bb.1.entry:
+    liveins: $vgpr0
+
+    ; GCN-LABEL: name: analyze_addc_0_0
+    ; GCN: liveins: $vgpr0
+    ; GCN-NEXT: {{  $}}
+    ; GCN-NEXT: [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr0
+    ; GCN-NEXT: [[V_ADD_CO_U32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_CO_U32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADD_CO_U32_e64 [[COPY]], 16, 0, implicit $exec
+    ; GCN-NEXT: [[V_ADDC_U32_e64_:%[0-9]+]]:vgpr_32, dead [[V_ADDC_U32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADDC_U32_e64 0, 0, killed [[V_ADD_CO_U32_e64_1]], 0, implicit $exec
+    ; GCN-NEXT: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_CO_U32_e64_]], %subreg.sub0, [[V_ADDC_U32_e64_]], %subreg.sub1
+    ; GCN-NEXT: [[GLOBAL_LOAD_DWORDX4_:%[0-9]+]]:vreg_128 = GLOBAL_LOAD_DWORDX4 [[REG_SEQUENCE]], 0, 0, implicit $exec
+    %0:vgpr_32 = COPY $vgpr0
+    %1:vgpr_32, %2:sreg_64_xexec = V_ADD_CO_U32_e64 %0, 16, 0, implicit $exec
+    %3:vgpr_32, dead %26:sreg_64_xexec = V_ADDC_U32_e64 0, 0, killed %2, 0, implicit $exec
+    %4:vreg_64 = REG_SEQUENCE %1, %subreg.sub0, %3, %subreg.sub1
+    %5:vreg_128 = GLOBAL_LOAD_DWORDX4 %4, 0, 0, implicit $exec
+
+...