AMDGPU: Use getMergedLocation in SILoadStoreOptimizer (#156396)

This is merging loads and stores so use the combined DebugLoc.

Not sure if computeBase should be using the merged location from
all the involved instructions. I'm also not sure how to test this
sort of thing.

2 files changed, 130 insertions, 25 deletions

diff --git a/llvm/lib/Target/AMDGPU/SILoadStoreOptimizer.cpp b/llvm/lib/Target/AMDGPU/SILoadStoreOptimizer.cpp
index f0d1117..00aae2c9 100644
--- a/llvm/lib/Target/AMDGPU/SILoadStoreOptimizer.cpp
+++ b/llvm/lib/Target/AMDGPU/SILoadStoreOptimizer.cpp
@@ -233,10 +233,11 @@ private:
 
   void copyToDestRegs(CombineInfo &CI, CombineInfo &Paired,
                       MachineBasicBlock::iterator InsertBefore,
-                      AMDGPU::OpName OpName, Register DestReg) const;
+                      const DebugLoc &DL, AMDGPU::OpName OpName,
+                      Register DestReg) const;
   Register copyFromSrcRegs(CombineInfo &CI, CombineInfo &Paired,
                            MachineBasicBlock::iterator InsertBefore,
-                           AMDGPU::OpName OpName) const;
+                           const DebugLoc &DL, AMDGPU::OpName OpName) const;
 
   unsigned read2Opcode(unsigned EltSize) const;
   unsigned read2ST64Opcode(unsigned EltSize) const;
@@ -1367,10 +1368,9 @@ SILoadStoreOptimizer::checkAndPrepareMerge(CombineInfo &CI,
 // Paired.
 void SILoadStoreOptimizer::copyToDestRegs(
     CombineInfo &CI, CombineInfo &Paired,
-    MachineBasicBlock::iterator InsertBefore, AMDGPU::OpName OpName,
-    Register DestReg) const {
+    MachineBasicBlock::iterator InsertBefore, const DebugLoc &DL,
+    AMDGPU::OpName OpName, Register DestReg) const {
   MachineBasicBlock *MBB = CI.I->getParent();
-  DebugLoc DL = CI.I->getDebugLoc();
 
   auto [SubRegIdx0, SubRegIdx1] = getSubRegIdxs(CI, Paired);
 
@@ -1398,9 +1398,9 @@ void SILoadStoreOptimizer::copyToDestRegs(
 Register
 SILoadStoreOptimizer::copyFromSrcRegs(CombineInfo &CI, CombineInfo &Paired,
                                       MachineBasicBlock::iterator InsertBefore,
+                                      const DebugLoc &DL,
                                       AMDGPU::OpName OpName) const {
   MachineBasicBlock *MBB = CI.I->getParent();
-  DebugLoc DL = CI.I->getDebugLoc();
 
   auto [SubRegIdx0, SubRegIdx1] = getSubRegIdxs(CI, Paired);
 
@@ -1456,7 +1456,8 @@ SILoadStoreOptimizer::mergeRead2Pair(CombineInfo &CI, CombineInfo &Paired,
   const TargetRegisterClass *SuperRC = getTargetRegisterClass(CI, Paired);
   Register DestReg = MRI->createVirtualRegister(SuperRC);
 
-  DebugLoc DL = CI.I->getDebugLoc();
+  DebugLoc DL =
+      DebugLoc::getMergedLocation(CI.I->getDebugLoc(), Paired.I->getDebugLoc());
 
   Register BaseReg = AddrReg->getReg();
   unsigned BaseSubReg = AddrReg->getSubReg();
@@ -1484,7 +1485,7 @@ SILoadStoreOptimizer::mergeRead2Pair(CombineInfo &CI, CombineInfo &Paired,
           .addImm(0)                                 // gds
           .cloneMergedMemRefs({&*CI.I, &*Paired.I});
 
-  copyToDestRegs(CI, Paired, InsertBefore, AMDGPU::OpName::vdst, DestReg);
+  copyToDestRegs(CI, Paired, InsertBefore, DL, AMDGPU::OpName::vdst, DestReg);
 
   CI.I->eraseFromParent();
   Paired.I->eraseFromParent();
@@ -1541,7 +1542,8 @@ MachineBasicBlock::iterator SILoadStoreOptimizer::mergeWrite2Pair(
          (NewOffset0 != NewOffset1) && "Computed offset doesn't fit");
 
   const MCInstrDesc &Write2Desc = TII->get(Opc);
-  DebugLoc DL = CI.I->getDebugLoc();
+  DebugLoc DL =
+      DebugLoc::getMergedLocation(CI.I->getDebugLoc(), Paired.I->getDebugLoc());
 
   Register BaseReg = AddrReg->getReg();
   unsigned BaseSubReg = AddrReg->getSubReg();
@@ -1582,7 +1584,9 @@ MachineBasicBlock::iterator
 SILoadStoreOptimizer::mergeImagePair(CombineInfo &CI, CombineInfo &Paired,
                                      MachineBasicBlock::iterator InsertBefore) {
   MachineBasicBlock *MBB = CI.I->getParent();
-  DebugLoc DL = CI.I->getDebugLoc();
+  DebugLoc DL =
+      DebugLoc::getMergedLocation(CI.I->getDebugLoc(), Paired.I->getDebugLoc());
+
   const unsigned Opcode = getNewOpcode(CI, Paired);
 
   const TargetRegisterClass *SuperRC = getTargetRegisterClass(CI, Paired);
@@ -1607,7 +1611,7 @@ SILoadStoreOptimizer::mergeImagePair(CombineInfo &CI, CombineInfo &Paired,
 
   MachineInstr *New = MIB.addMemOperand(combineKnownAdjacentMMOs(CI, Paired));
 
-  copyToDestRegs(CI, Paired, InsertBefore, AMDGPU::OpName::vdata, DestReg);
+  copyToDestRegs(CI, Paired, InsertBefore, DL, AMDGPU::OpName::vdata, DestReg);
 
   CI.I->eraseFromParent();
   Paired.I->eraseFromParent();
@@ -1618,7 +1622,9 @@ MachineBasicBlock::iterator SILoadStoreOptimizer::mergeSMemLoadImmPair(
     CombineInfo &CI, CombineInfo &Paired,
     MachineBasicBlock::iterator InsertBefore) {
   MachineBasicBlock *MBB = CI.I->getParent();
-  DebugLoc DL = CI.I->getDebugLoc();
+  DebugLoc DL =
+      DebugLoc::getMergedLocation(CI.I->getDebugLoc(), Paired.I->getDebugLoc());
+
   const unsigned Opcode = getNewOpcode(CI, Paired);
 
   const TargetRegisterClass *SuperRC = getTargetRegisterClass(CI, Paired);
@@ -1639,7 +1645,7 @@ MachineBasicBlock::iterator SILoadStoreOptimizer::mergeSMemLoadImmPair(
   New.addImm(MergedOffset);
   New.addImm(CI.CPol).addMemOperand(combineKnownAdjacentMMOs(CI, Paired));
 
-  copyToDestRegs(CI, Paired, InsertBefore, AMDGPU::OpName::sdst, DestReg);
+  copyToDestRegs(CI, Paired, InsertBefore, DL, AMDGPU::OpName::sdst, DestReg);
 
   CI.I->eraseFromParent();
   Paired.I->eraseFromParent();
@@ -1650,7 +1656,9 @@ MachineBasicBlock::iterator SILoadStoreOptimizer::mergeBufferLoadPair(
     CombineInfo &CI, CombineInfo &Paired,
     MachineBasicBlock::iterator InsertBefore) {
   MachineBasicBlock *MBB = CI.I->getParent();
-  DebugLoc DL = CI.I->getDebugLoc();
+
+  DebugLoc DL =
+      DebugLoc::getMergedLocation(CI.I->getDebugLoc(), Paired.I->getDebugLoc());
 
   const unsigned Opcode = getNewOpcode(CI, Paired);
 
@@ -1680,7 +1688,7 @@ MachineBasicBlock::iterator SILoadStoreOptimizer::mergeBufferLoadPair(
         .addImm(0)            // swz
         .addMemOperand(combineKnownAdjacentMMOs(CI, Paired));
 
-  copyToDestRegs(CI, Paired, InsertBefore, AMDGPU::OpName::vdata, DestReg);
+  copyToDestRegs(CI, Paired, InsertBefore, DL, AMDGPU::OpName::vdata, DestReg);
 
   CI.I->eraseFromParent();
   Paired.I->eraseFromParent();
@@ -1691,7 +1699,9 @@ MachineBasicBlock::iterator SILoadStoreOptimizer::mergeTBufferLoadPair(
     CombineInfo &CI, CombineInfo &Paired,
     MachineBasicBlock::iterator InsertBefore) {
   MachineBasicBlock *MBB = CI.I->getParent();
-  DebugLoc DL = CI.I->getDebugLoc();
+
+  DebugLoc DL =
+      DebugLoc::getMergedLocation(CI.I->getDebugLoc(), Paired.I->getDebugLoc());
 
   const unsigned Opcode = getNewOpcode(CI, Paired);
 
@@ -1731,7 +1741,7 @@ MachineBasicBlock::iterator SILoadStoreOptimizer::mergeTBufferLoadPair(
           .addImm(0)            // swz
           .addMemOperand(combineKnownAdjacentMMOs(CI, Paired));
 
-  copyToDestRegs(CI, Paired, InsertBefore, AMDGPU::OpName::vdata, DestReg);
+  copyToDestRegs(CI, Paired, InsertBefore, DL, AMDGPU::OpName::vdata, DestReg);
 
   CI.I->eraseFromParent();
   Paired.I->eraseFromParent();
@@ -1742,12 +1752,13 @@ MachineBasicBlock::iterator SILoadStoreOptimizer::mergeTBufferStorePair(
     CombineInfo &CI, CombineInfo &Paired,
     MachineBasicBlock::iterator InsertBefore) {
   MachineBasicBlock *MBB = CI.I->getParent();
-  DebugLoc DL = CI.I->getDebugLoc();
+  DebugLoc DL =
+      DebugLoc::getMergedLocation(CI.I->getDebugLoc(), Paired.I->getDebugLoc());
 
   const unsigned Opcode = getNewOpcode(CI, Paired);
 
   Register SrcReg =
-      copyFromSrcRegs(CI, Paired, InsertBefore, AMDGPU::OpName::vdata);
+      copyFromSrcRegs(CI, Paired, InsertBefore, DL, AMDGPU::OpName::vdata);
 
   auto MIB = BuildMI(*MBB, InsertBefore, DL, TII->get(Opcode))
                  .addReg(SrcReg, RegState::Kill);
@@ -1789,7 +1800,9 @@ MachineBasicBlock::iterator SILoadStoreOptimizer::mergeFlatLoadPair(
     CombineInfo &CI, CombineInfo &Paired,
     MachineBasicBlock::iterator InsertBefore) {
   MachineBasicBlock *MBB = CI.I->getParent();
-  DebugLoc DL = CI.I->getDebugLoc();
+
+  DebugLoc DL =
+      DebugLoc::getMergedLocation(CI.I->getDebugLoc(), Paired.I->getDebugLoc());
 
   const unsigned Opcode = getNewOpcode(CI, Paired);
 
@@ -1807,7 +1820,7 @@ MachineBasicBlock::iterator SILoadStoreOptimizer::mergeFlatLoadPair(
        .addImm(CI.CPol)
        .addMemOperand(combineKnownAdjacentMMOs(CI, Paired));
 
-  copyToDestRegs(CI, Paired, InsertBefore, AMDGPU::OpName::vdst, DestReg);
+  copyToDestRegs(CI, Paired, InsertBefore, DL, AMDGPU::OpName::vdst, DestReg);
 
   CI.I->eraseFromParent();
   Paired.I->eraseFromParent();
@@ -1818,12 +1831,14 @@ MachineBasicBlock::iterator SILoadStoreOptimizer::mergeFlatStorePair(
     CombineInfo &CI, CombineInfo &Paired,
     MachineBasicBlock::iterator InsertBefore) {
   MachineBasicBlock *MBB = CI.I->getParent();
-  DebugLoc DL = CI.I->getDebugLoc();
+
+  DebugLoc DL =
+      DebugLoc::getMergedLocation(CI.I->getDebugLoc(), Paired.I->getDebugLoc());
 
   const unsigned Opcode = getNewOpcode(CI, Paired);
 
   Register SrcReg =
-      copyFromSrcRegs(CI, Paired, InsertBefore, AMDGPU::OpName::vdata);
+      copyFromSrcRegs(CI, Paired, InsertBefore, DL, AMDGPU::OpName::vdata);
 
   auto MIB = BuildMI(*MBB, InsertBefore, DL, TII->get(Opcode))
                  .add(*TII->getNamedOperand(*CI.I, AMDGPU::OpName::vaddr))
@@ -2094,12 +2109,13 @@ MachineBasicBlock::iterator SILoadStoreOptimizer::mergeBufferStorePair(
     CombineInfo &CI, CombineInfo &Paired,
     MachineBasicBlock::iterator InsertBefore) {
   MachineBasicBlock *MBB = CI.I->getParent();
-  DebugLoc DL = CI.I->getDebugLoc();
+  DebugLoc DL =
+      DebugLoc::getMergedLocation(CI.I->getDebugLoc(), Paired.I->getDebugLoc());
 
   const unsigned Opcode = getNewOpcode(CI, Paired);
 
   Register SrcReg =
-      copyFromSrcRegs(CI, Paired, InsertBefore, AMDGPU::OpName::vdata);
+      copyFromSrcRegs(CI, Paired, InsertBefore, DL, AMDGPU::OpName::vdata);
 
   auto MIB = BuildMI(*MBB, InsertBefore, DL, TII->get(Opcode))
                  .addReg(SrcReg, RegState::Kill);
diff --git a/llvm/test/CodeGen/AMDGPU/ds-read2-write2-debug-info.ll b/llvm/test/CodeGen/AMDGPU/ds-read2-write2-debug-info.ll
new file mode 100644
index 0000000..08730038
--- /dev/null
+++ b/llvm/test/CodeGen/AMDGPU/ds-read2-write2-debug-info.ll
@@ -0,0 +1,89 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 6
+; RUN: opt -passes=debugify < %s | llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx900 | FileCheck %s
+
+@lds = addrspace(3) global [512 x float] poison, align 4
+
+define amdgpu_kernel void @simple_write2_one_val_f32(ptr addrspace(1) %C, ptr addrspace(1) %in) #0 {
+; CHECK-LABEL: simple_write2_one_val_f32:
+; CHECK:       .Lfunc_begin0:
+; CHECK-NEXT:    .cfi_sections .debug_frame
+; CHECK-NEXT:    .cfi_startproc
+; CHECK-NEXT:  ; %bb.0:
+; CHECK-NEXT:    .file 1 "/" "<stdin>"
+; CHECK-NEXT:    .loc 1 1 1 prologue_end ; <stdin>:1:1
+; CHECK-NEXT:    s_load_dwordx2 s[0:1], s[8:9], 0x8
+; CHECK-NEXT:  .Ltmp0:
+; CHECK-NEXT:    ;DEBUG_VALUE: simple_write2_one_val_f32:1 <- $vgpr0
+; CHECK-NEXT:    ;DEBUG_VALUE: simple_write2_one_val_f32:5 <- [DW_OP_plus_uconst 8, DW_OP_stack_value] $vgpr0
+; CHECK-NEXT:    ;DEBUG_VALUE: simple_write2_one_val_f32:3 <- undef
+; CHECK-NEXT:    .loc 1 2 1 ; <stdin>:2:1
+; CHECK-NEXT:    v_lshlrev_b32_e32 v0, 2, v0
+; CHECK-NEXT:  .Ltmp1:
+; CHECK-NEXT:    ;DEBUG_VALUE: simple_write2_one_val_f32:4 <- $vgpr0
+; CHECK-NEXT:    ;DEBUG_VALUE: simple_write2_one_val_f32:2 <- undef
+; CHECK-NEXT:    .loc 1 3 1 ; <stdin>:3:1
+; CHECK-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-NEXT:    global_load_dword v1, v0, s[0:1]
+; CHECK-NEXT:  .Ltmp2:
+; CHECK-NEXT:    ;DEBUG_VALUE: simple_write2_one_val_f32:6 <- [DW_OP_plus_uconst 32, DW_OP_stack_value] $vgpr0
+; CHECK-NEXT:    .loc 1 0 0 is_stmt 0 ; <stdin>:0
+; CHECK-NEXT:    s_waitcnt vmcnt(0)
+; CHECK-NEXT:    ds_write2_b32 v0, v1, v1 offset1:8
+; CHECK-NEXT:    .loc 1 9 1 is_stmt 1 ; <stdin>:9:1
+; CHECK-NEXT:    s_endpgm
+; CHECK-NEXT:  .Ltmp3:
+  %x.i = tail call i32 @llvm.amdgcn.workitem.id.x() #1
+  %in.gep = getelementptr float, ptr addrspace(1) %in, i32 %x.i
+  %val = load float, ptr addrspace(1) %in.gep, align 4
+  %arrayidx0 = getelementptr inbounds [512 x float], ptr addrspace(3) @lds, i32 0, i32 %x.i
+  store float %val, ptr addrspace(3) %arrayidx0, align 4
+  %add.x = add nsw i32 %x.i, 8
+  %arrayidx1 = getelementptr inbounds [512 x float], ptr addrspace(3) @lds, i32 0, i32 %add.x
+  store float %val, ptr addrspace(3) %arrayidx1, align 4
+  ret void
+}
+
+define amdgpu_kernel void @simple_read2_f32(ptr addrspace(1) %out) #0 {
+; CHECK-LABEL: simple_read2_f32:
+; CHECK:       .Lfunc_begin1:
+; CHECK-NEXT:    .cfi_startproc
+; CHECK-NEXT:  ; %bb.0:
+; CHECK-NEXT:    .loc 1 11 1 prologue_end ; <stdin>:11:1
+; CHECK-NEXT:    v_lshlrev_b32_e32 v2, 2, v0
+; CHECK-NEXT:  .Ltmp4:
+; CHECK-NEXT:    ;DEBUG_VALUE: simple_read2_f32:8 <- $vgpr2
+; CHECK-NEXT:    .loc 1 0 0 is_stmt 0 ; <stdin>:0
+; CHECK-NEXT:    ds_read2_b32 v[0:1], v2 offset1:8
+; CHECK-NEXT:  .Ltmp5:
+; CHECK-NEXT:    ;DEBUG_VALUE: simple_read2_f32:9 <- undef
+; CHECK-NEXT:    ;DEBUG_VALUE: simple_read2_f32:11 <- [DW_OP_plus_uconst 32, DW_OP_stack_value] $vgpr2
+; CHECK-NEXT:    ;DEBUG_VALUE: simple_read2_f32:12 <- undef
+; CHECK-NEXT:    .loc 1 10 1 is_stmt 1 ; <stdin>:10:1
+; CHECK-NEXT:    s_load_dwordx2 s[0:1], s[8:9], 0x0
+; CHECK-NEXT:  .Ltmp6:
+; CHECK-NEXT:    ;DEBUG_VALUE: simple_read2_f32:7 <- undef
+; CHECK-NEXT:    ;DEBUG_VALUE: simple_read2_f32:10 <- [DW_OP_plus_uconst 8, DW_OP_stack_value] undef
+; CHECK-NEXT:    .loc 1 16 1 ; <stdin>:16:1
+; CHECK-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-NEXT:    v_add_f32_e32 v0, v0, v1
+; CHECK-NEXT:  .Ltmp7:
+; CHECK-NEXT:    ;DEBUG_VALUE: simple_read2_f32:13 <- $vgpr0
+; CHECK-NEXT:    ;DEBUG_VALUE: simple_read2_f32:14 <- undef
+; CHECK-NEXT:    .loc 1 18 1 ; <stdin>:18:1
+; CHECK-NEXT:    global_store_dword v2, v0, s[0:1]
+; CHECK-NEXT:    .loc 1 19 1 ; <stdin>:19:1
+; CHECK-NEXT:    s_endpgm
+; CHECK-NEXT:  .Ltmp8:
+  %x.i = tail call i32 @llvm.amdgcn.workitem.id.x() #1
+  %arrayidx0 = getelementptr inbounds [512 x float], ptr addrspace(3) @lds, i32 0, i32 %x.i
+  %val0 = load float, ptr addrspace(3) %arrayidx0, align 4
+  %add.x = add nsw i32 %x.i, 8
+  %arrayidx1 = getelementptr inbounds [512 x float], ptr addrspace(3) @lds, i32 0, i32 %add.x
+  %val1 = load float, ptr addrspace(3) %arrayidx1, align 4
+  %sum = fadd float %val0, %val1
+  %out.gep = getelementptr inbounds float, ptr addrspace(1) %out, i32 %x.i
+  store float %sum, ptr addrspace(1) %out.gep, align 4
+  ret void
+}
+
+attributes #0 = { nounwind }