[llvm] 06cfbe3 - [AMDPU] Add support for idxen and bothen buffer load/store merging in SILoadStoreOptimizer (#86285)
via llvm-commits
llvm-commits at lists.llvm.org
Mon Mar 25 07:44:26 PDT 2024
Author: David Stuttard
Date: 2024-03-25T14:44:22Z
New Revision: 06cfbe3cfd44cd2ca9eb970b8c0e5f4911468440
URL: https://github.com/llvm/llvm-project/commit/06cfbe3cfd44cd2ca9eb970b8c0e5f4911468440
DIFF: https://github.com/llvm/llvm-project/commit/06cfbe3cfd44cd2ca9eb970b8c0e5f4911468440.diff
LOG: [AMDPU] Add support for idxen and bothen buffer load/store merging in SILoadStoreOptimizer (#86285)
Added more buffer instruction merging support
Added:
llvm/test/CodeGen/AMDGPU/merge-buffer-gfx12.mir
llvm/test/CodeGen/AMDGPU/merge-buffer.mir
Modified:
llvm/lib/Target/AMDGPU/SILoadStoreOptimizer.cpp
llvm/test/CodeGen/AMDGPU/merge-tbuffer.mir
Removed:
################################################################################
diff --git a/llvm/lib/Target/AMDGPU/SILoadStoreOptimizer.cpp b/llvm/lib/Target/AMDGPU/SILoadStoreOptimizer.cpp
index 4ddee2f6d5befa..e2e70ba9733b5b 100644
--- a/llvm/lib/Target/AMDGPU/SILoadStoreOptimizer.cpp
+++ b/llvm/lib/Target/AMDGPU/SILoadStoreOptimizer.cpp
@@ -399,19 +399,35 @@ static InstClassEnum getInstClass(unsigned Opc, const SIInstrInfo &TII) {
switch (AMDGPU::getMUBUFBaseOpcode(Opc)) {
default:
return UNKNOWN;
+ case AMDGPU::BUFFER_LOAD_DWORD_BOTHEN:
+ case AMDGPU::BUFFER_LOAD_DWORD_BOTHEN_exact:
+ case AMDGPU::BUFFER_LOAD_DWORD_IDXEN:
+ case AMDGPU::BUFFER_LOAD_DWORD_IDXEN_exact:
case AMDGPU::BUFFER_LOAD_DWORD_OFFEN:
case AMDGPU::BUFFER_LOAD_DWORD_OFFEN_exact:
case AMDGPU::BUFFER_LOAD_DWORD_OFFSET:
case AMDGPU::BUFFER_LOAD_DWORD_OFFSET_exact:
+ case AMDGPU::BUFFER_LOAD_DWORD_VBUFFER_BOTHEN:
+ case AMDGPU::BUFFER_LOAD_DWORD_VBUFFER_BOTHEN_exact:
+ case AMDGPU::BUFFER_LOAD_DWORD_VBUFFER_IDXEN:
+ case AMDGPU::BUFFER_LOAD_DWORD_VBUFFER_IDXEN_exact:
case AMDGPU::BUFFER_LOAD_DWORD_VBUFFER_OFFEN:
case AMDGPU::BUFFER_LOAD_DWORD_VBUFFER_OFFEN_exact:
case AMDGPU::BUFFER_LOAD_DWORD_VBUFFER_OFFSET:
case AMDGPU::BUFFER_LOAD_DWORD_VBUFFER_OFFSET_exact:
return BUFFER_LOAD;
+ case AMDGPU::BUFFER_STORE_DWORD_BOTHEN:
+ case AMDGPU::BUFFER_STORE_DWORD_BOTHEN_exact:
+ case AMDGPU::BUFFER_STORE_DWORD_IDXEN:
+ case AMDGPU::BUFFER_STORE_DWORD_IDXEN_exact:
case AMDGPU::BUFFER_STORE_DWORD_OFFEN:
case AMDGPU::BUFFER_STORE_DWORD_OFFEN_exact:
case AMDGPU::BUFFER_STORE_DWORD_OFFSET:
case AMDGPU::BUFFER_STORE_DWORD_OFFSET_exact:
+ case AMDGPU::BUFFER_STORE_DWORD_VBUFFER_BOTHEN:
+ case AMDGPU::BUFFER_STORE_DWORD_VBUFFER_BOTHEN_exact:
+ case AMDGPU::BUFFER_STORE_DWORD_VBUFFER_IDXEN:
+ case AMDGPU::BUFFER_STORE_DWORD_VBUFFER_IDXEN_exact:
case AMDGPU::BUFFER_STORE_DWORD_VBUFFER_OFFEN:
case AMDGPU::BUFFER_STORE_DWORD_VBUFFER_OFFEN_exact:
case AMDGPU::BUFFER_STORE_DWORD_VBUFFER_OFFSET:
diff --git a/llvm/test/CodeGen/AMDGPU/merge-buffer-gfx12.mir b/llvm/test/CodeGen/AMDGPU/merge-buffer-gfx12.mir
new file mode 100644
index 00000000000000..d7f5d1a2378955
--- /dev/null
+++ b/llvm/test/CodeGen/AMDGPU/merge-buffer-gfx12.mir
@@ -0,0 +1,1154 @@
+# NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py UTC_ARGS: --version 2
+# RUN: llc -mtriple=amdgcn -mcpu=gfx1200 -verify-machineinstrs -run-pass si-load-store-opt -o - %s | FileCheck -check-prefixes=GFX12 %s
+
+---
+name: buffer_load_dword_dwordx3
+body: |
+ bb.0.entry:
+ ; GFX12-LABEL: name: buffer_load_dword_dwordx3
+ ; GFX12: [[COPY:%[0-9]+]]:sgpr_32 = COPY $sgpr0
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:sgpr_32 = COPY $sgpr1
+ ; GFX12-NEXT: [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr2
+ ; GFX12-NEXT: [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr3
+ ; GFX12-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3
+ ; GFX12-NEXT: [[BUFFER_LOAD_DWORDX4_VBUFFER_OFFSET:%[0-9]+]]:vreg_128 = BUFFER_LOAD_DWORDX4_VBUFFER_OFFSET [[REG_SEQUENCE]], $sgpr_null, 4, 0, 0, implicit $exec :: (dereferenceable load (s128), align 1, addrspace 4)
+ ; GFX12-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[BUFFER_LOAD_DWORDX4_VBUFFER_OFFSET]].sub0
+ ; GFX12-NEXT: [[COPY5:%[0-9]+]]:vreg_96 = COPY killed [[BUFFER_LOAD_DWORDX4_VBUFFER_OFFSET]].sub1_sub2_sub3
+ %0:sgpr_32 = COPY $sgpr0
+ %1:sgpr_32 = COPY $sgpr1
+ %2:sgpr_32 = COPY $sgpr2
+ %3:sgpr_32 = COPY $sgpr3
+ %5:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3
+ %7:vgpr_32 = BUFFER_LOAD_DWORD_VBUFFER_OFFSET %5:sgpr_128, $sgpr_null, 4, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+ %8:vreg_96 = BUFFER_LOAD_DWORDX3_VBUFFER_OFFSET %5:sgpr_128, $sgpr_null, 8, 0, 0, implicit $exec :: (dereferenceable load (s96), align 1, addrspace 4)
+...
+---
+
+name: buffer_load_dwordx3_dword
+body: |
+ bb.0.entry:
+ ; GFX12-LABEL: name: buffer_load_dwordx3_dword
+ ; GFX12: [[COPY:%[0-9]+]]:sgpr_32 = COPY $sgpr0
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:sgpr_32 = COPY $sgpr1
+ ; GFX12-NEXT: [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr2
+ ; GFX12-NEXT: [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr3
+ ; GFX12-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3
+ ; GFX12-NEXT: [[BUFFER_LOAD_DWORDX4_VBUFFER_OFFSET:%[0-9]+]]:vreg_128 = BUFFER_LOAD_DWORDX4_VBUFFER_OFFSET [[REG_SEQUENCE]], $sgpr_null, 4, 0, 0, implicit $exec :: (dereferenceable load (s128), align 1, addrspace 4)
+ ; GFX12-NEXT: [[COPY4:%[0-9]+]]:vreg_96 = COPY [[BUFFER_LOAD_DWORDX4_VBUFFER_OFFSET]].sub0_sub1_sub2
+ ; GFX12-NEXT: [[COPY5:%[0-9]+]]:vgpr_32 = COPY killed [[BUFFER_LOAD_DWORDX4_VBUFFER_OFFSET]].sub3
+ %0:sgpr_32 = COPY $sgpr0
+ %1:sgpr_32 = COPY $sgpr1
+ %2:sgpr_32 = COPY $sgpr2
+ %3:sgpr_32 = COPY $sgpr3
+ %5:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3
+ %7:vreg_96 = BUFFER_LOAD_DWORDX3_VBUFFER_OFFSET %5:sgpr_128, $sgpr_null, 4, 0, 0, implicit $exec :: (dereferenceable load (s96), align 1, addrspace 4)
+ %8:vgpr_32 = BUFFER_LOAD_DWORD_VBUFFER_OFFSET %5:sgpr_128, $sgpr_null, 16, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+...
+---
+
+name: buffer_load_dwordx2_dwordx2
+body: |
+ bb.0.entry:
+ ; GFX12-LABEL: name: buffer_load_dwordx2_dwordx2
+ ; GFX12: [[COPY:%[0-9]+]]:sgpr_32 = COPY $sgpr0
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:sgpr_32 = COPY $sgpr1
+ ; GFX12-NEXT: [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr2
+ ; GFX12-NEXT: [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr3
+ ; GFX12-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3
+ ; GFX12-NEXT: [[BUFFER_LOAD_DWORDX4_VBUFFER_OFFSET:%[0-9]+]]:vreg_128 = BUFFER_LOAD_DWORDX4_VBUFFER_OFFSET [[REG_SEQUENCE]], $sgpr_null, 4, 0, 0, implicit $exec :: (dereferenceable load (s128), align 1, addrspace 4)
+ ; GFX12-NEXT: [[COPY4:%[0-9]+]]:vreg_64 = COPY [[BUFFER_LOAD_DWORDX4_VBUFFER_OFFSET]].sub0_sub1
+ ; GFX12-NEXT: [[COPY5:%[0-9]+]]:vreg_64 = COPY killed [[BUFFER_LOAD_DWORDX4_VBUFFER_OFFSET]].sub2_sub3
+ %0:sgpr_32 = COPY $sgpr0
+ %1:sgpr_32 = COPY $sgpr1
+ %2:sgpr_32 = COPY $sgpr2
+ %3:sgpr_32 = COPY $sgpr3
+ %5:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3
+ %7:vreg_64 = BUFFER_LOAD_DWORDX2_VBUFFER_OFFSET %5:sgpr_128, $sgpr_null, 4, 0, 0, implicit $exec :: (dereferenceable load (s64), align 1, addrspace 4)
+ %8:vreg_64 = BUFFER_LOAD_DWORDX2_VBUFFER_OFFSET %5:sgpr_128, $sgpr_null, 12, 0, 0, implicit $exec :: (dereferenceable load (s64), align 1, addrspace 4)
+...
+---
+
+name: buffer_load_dword_dwordx2
+body: |
+ bb.0.entry:
+ ; GFX12-LABEL: name: buffer_load_dword_dwordx2
+ ; GFX12: [[COPY:%[0-9]+]]:sgpr_32 = COPY $sgpr0
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:sgpr_32 = COPY $sgpr1
+ ; GFX12-NEXT: [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr2
+ ; GFX12-NEXT: [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr3
+ ; GFX12-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3
+ ; GFX12-NEXT: [[BUFFER_LOAD_DWORDX3_VBUFFER_OFFSET:%[0-9]+]]:vreg_96 = BUFFER_LOAD_DWORDX3_VBUFFER_OFFSET [[REG_SEQUENCE]], $sgpr_null, 4, 0, 0, implicit $exec :: (dereferenceable load (s96), align 1, addrspace 4)
+ ; GFX12-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[BUFFER_LOAD_DWORDX3_VBUFFER_OFFSET]].sub0
+ ; GFX12-NEXT: [[COPY5:%[0-9]+]]:vreg_64 = COPY killed [[BUFFER_LOAD_DWORDX3_VBUFFER_OFFSET]].sub1_sub2
+ %0:sgpr_32 = COPY $sgpr0
+ %1:sgpr_32 = COPY $sgpr1
+ %2:sgpr_32 = COPY $sgpr2
+ %3:sgpr_32 = COPY $sgpr3
+ %5:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3
+ %7:vgpr_32 = BUFFER_LOAD_DWORD_VBUFFER_OFFSET %5:sgpr_128, $sgpr_null, 4, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+ %8:vreg_64 = BUFFER_LOAD_DWORDX2_VBUFFER_OFFSET %5:sgpr_128, $sgpr_null, 8, 0, 0, implicit $exec :: (dereferenceable load (s64), align 1, addrspace 4)
+...
+---
+
+name: buffer_load_dwordx2_dword
+body: |
+ bb.0.entry:
+ ; GFX12-LABEL: name: buffer_load_dwordx2_dword
+ ; GFX12: [[COPY:%[0-9]+]]:sgpr_32 = COPY $sgpr0
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:sgpr_32 = COPY $sgpr1
+ ; GFX12-NEXT: [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr2
+ ; GFX12-NEXT: [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr3
+ ; GFX12-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3
+ ; GFX12-NEXT: [[BUFFER_LOAD_DWORDX3_VBUFFER_OFFSET:%[0-9]+]]:vreg_96 = BUFFER_LOAD_DWORDX3_VBUFFER_OFFSET [[REG_SEQUENCE]], $sgpr_null, 4, 0, 0, implicit $exec :: (dereferenceable load (s96), align 1, addrspace 4)
+ ; GFX12-NEXT: [[COPY4:%[0-9]+]]:vreg_64 = COPY [[BUFFER_LOAD_DWORDX3_VBUFFER_OFFSET]].sub0_sub1
+ ; GFX12-NEXT: [[COPY5:%[0-9]+]]:vgpr_32 = COPY killed [[BUFFER_LOAD_DWORDX3_VBUFFER_OFFSET]].sub2
+ %0:sgpr_32 = COPY $sgpr0
+ %1:sgpr_32 = COPY $sgpr1
+ %2:sgpr_32 = COPY $sgpr2
+ %3:sgpr_32 = COPY $sgpr3
+ %5:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3
+ %7:vreg_64 = BUFFER_LOAD_DWORDX2_VBUFFER_OFFSET %5:sgpr_128, $sgpr_null, 4, 0, 0, implicit $exec :: (dereferenceable load (s64), align 1, addrspace 4)
+ %8:vgpr_32 = BUFFER_LOAD_DWORD_VBUFFER_OFFSET %5:sgpr_128, $sgpr_null, 12, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+...
+---
+
+
+name: buffer_load_dword_dword
+body: |
+ bb.0.entry:
+ ; GFX12-LABEL: name: buffer_load_dword_dword
+ ; GFX12: [[COPY:%[0-9]+]]:sgpr_32 = COPY $sgpr0
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:sgpr_32 = COPY $sgpr1
+ ; GFX12-NEXT: [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr2
+ ; GFX12-NEXT: [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr3
+ ; GFX12-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3
+ ; GFX12-NEXT: [[BUFFER_LOAD_DWORDX2_VBUFFER_OFFSET:%[0-9]+]]:vreg_64 = BUFFER_LOAD_DWORDX2_VBUFFER_OFFSET [[REG_SEQUENCE]], $sgpr_null, 4, 0, 0, implicit $exec :: (dereferenceable load (s64), align 1, addrspace 4)
+ ; GFX12-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[BUFFER_LOAD_DWORDX2_VBUFFER_OFFSET]].sub0
+ ; GFX12-NEXT: [[COPY5:%[0-9]+]]:vgpr_32 = COPY killed [[BUFFER_LOAD_DWORDX2_VBUFFER_OFFSET]].sub1
+ %0:sgpr_32 = COPY $sgpr0
+ %1:sgpr_32 = COPY $sgpr1
+ %2:sgpr_32 = COPY $sgpr2
+ %3:sgpr_32 = COPY $sgpr3
+ %5:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3
+ %7:vgpr_32 = BUFFER_LOAD_DWORD_VBUFFER_OFFSET %5:sgpr_128, $sgpr_null, 4, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+ %8:vgpr_32 = BUFFER_LOAD_DWORD_VBUFFER_OFFSET %5:sgpr_128, $sgpr_null, 8, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+...
+---
+
+name: buffer_load_dword_32
+body: |
+ bb.0.entry:
+ ; GFX12-LABEL: name: buffer_load_dword_32
+ ; GFX12: [[COPY:%[0-9]+]]:sgpr_32 = COPY $sgpr0
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:sgpr_32 = COPY $sgpr1
+ ; GFX12-NEXT: [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr2
+ ; GFX12-NEXT: [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr3
+ ; GFX12-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3
+ ; GFX12-NEXT: [[BUFFER_LOAD_DWORDX2_VBUFFER_OFFSET:%[0-9]+]]:vreg_64 = BUFFER_LOAD_DWORDX2_VBUFFER_OFFSET [[REG_SEQUENCE]], $sgpr_null, 4, 0, 0, implicit $exec :: (dereferenceable load (s64), align 1, addrspace 4)
+ ; GFX12-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[BUFFER_LOAD_DWORDX2_VBUFFER_OFFSET]].sub0
+ ; GFX12-NEXT: [[COPY5:%[0-9]+]]:vgpr_32 = COPY killed [[BUFFER_LOAD_DWORDX2_VBUFFER_OFFSET]].sub1
+ ; GFX12-NEXT: [[BUFFER_LOAD_DWORDX4_VBUFFER_OFFSET:%[0-9]+]]:vreg_128 = BUFFER_LOAD_DWORDX4_VBUFFER_OFFSET [[REG_SEQUENCE]], $sgpr_null, 16, 0, 0, implicit $exec :: (dereferenceable load (s128), align 1, addrspace 4)
+ ; GFX12-NEXT: [[COPY6:%[0-9]+]]:vreg_96 = COPY [[BUFFER_LOAD_DWORDX4_VBUFFER_OFFSET]].sub0_sub1_sub2
+ ; GFX12-NEXT: [[COPY7:%[0-9]+]]:vgpr_32 = COPY killed [[BUFFER_LOAD_DWORDX4_VBUFFER_OFFSET]].sub3
+ ; GFX12-NEXT: [[COPY8:%[0-9]+]]:vreg_64 = COPY [[COPY6]].sub0_sub1
+ ; GFX12-NEXT: [[COPY9:%[0-9]+]]:vgpr_32 = COPY killed [[COPY6]].sub2
+ ; GFX12-NEXT: [[COPY10:%[0-9]+]]:vgpr_32 = COPY [[COPY8]].sub0
+ ; GFX12-NEXT: [[COPY11:%[0-9]+]]:vgpr_32 = COPY killed [[COPY8]].sub1
+ ; GFX12-NEXT: [[BUFFER_LOAD_DWORDX3_VBUFFER_OFFSET:%[0-9]+]]:vreg_96 = BUFFER_LOAD_DWORDX3_VBUFFER_OFFSET [[REG_SEQUENCE]], $sgpr_null, 36, 0, 0, implicit $exec :: (dereferenceable load (s96), align 1, addrspace 4)
+ ; GFX12-NEXT: [[COPY12:%[0-9]+]]:vreg_64 = COPY [[BUFFER_LOAD_DWORDX3_VBUFFER_OFFSET]].sub0_sub1
+ ; GFX12-NEXT: [[COPY13:%[0-9]+]]:vgpr_32 = COPY killed [[BUFFER_LOAD_DWORDX3_VBUFFER_OFFSET]].sub2
+ ; GFX12-NEXT: [[COPY14:%[0-9]+]]:vgpr_32 = COPY [[COPY12]].sub0
+ ; GFX12-NEXT: [[COPY15:%[0-9]+]]:vgpr_32 = COPY killed [[COPY12]].sub1
+ %0:sgpr_32 = COPY $sgpr0
+ %1:sgpr_32 = COPY $sgpr1
+ %2:sgpr_32 = COPY $sgpr2
+ %3:sgpr_32 = COPY $sgpr3
+ %5:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3
+ %7:vgpr_32 = BUFFER_LOAD_DWORD_VBUFFER_OFFSET %5:sgpr_128, $sgpr_null, 4, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+ %8:vgpr_32 = BUFFER_LOAD_DWORD_VBUFFER_OFFSET %5:sgpr_128, $sgpr_null, 8, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+ %9:vgpr_32 = BUFFER_LOAD_DWORD_VBUFFER_OFFSET %5:sgpr_128, $sgpr_null, 16, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+ %10:vgpr_32 = BUFFER_LOAD_DWORD_VBUFFER_OFFSET %5:sgpr_128, $sgpr_null, 20, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+ %11:vgpr_32 = BUFFER_LOAD_DWORD_VBUFFER_OFFSET %5:sgpr_128, $sgpr_null, 24, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+ %12:vgpr_32 = BUFFER_LOAD_DWORD_VBUFFER_OFFSET %5:sgpr_128, $sgpr_null, 28, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+ %13:vgpr_32 = BUFFER_LOAD_DWORD_VBUFFER_OFFSET %5:sgpr_128, $sgpr_null, 36, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+ %14:vgpr_32 = BUFFER_LOAD_DWORD_VBUFFER_OFFSET %5:sgpr_128, $sgpr_null, 40, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+ %15:vgpr_32 = BUFFER_LOAD_DWORD_VBUFFER_OFFSET %5:sgpr_128, $sgpr_null, 44, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+...
+---
+
+#
+# buffer_store_dword
+#
+
+name: buffer_store_dword_xyz
+body: |
+ bb.0.entry:
+ liveins: $sgpr0, $sgpr1, $sgpr2, $sgpr3, $vgpr0, $vgpr1, $vgpr2, $vgpr3
+ ; GFX12-LABEL: name: buffer_store_dword_xyz
+ ; GFX12: liveins: $sgpr0, $sgpr1, $sgpr2, $sgpr3, $vgpr0, $vgpr1, $vgpr2, $vgpr3
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr3
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr2
+ ; GFX12-NEXT: [[COPY2:%[0-9]+]]:vgpr_32 = COPY $vgpr1
+ ; GFX12-NEXT: [[COPY3:%[0-9]+]]:vgpr_32 = COPY $vgpr0
+ ; GFX12-NEXT: [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr3
+ ; GFX12-NEXT: [[COPY5:%[0-9]+]]:sgpr_32 = COPY $sgpr2
+ ; GFX12-NEXT: [[COPY6:%[0-9]+]]:sgpr_32 = COPY $sgpr1
+ ; GFX12-NEXT: [[COPY7:%[0-9]+]]:sgpr_32 = COPY $sgpr0
+ ; GFX12-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY7]], %subreg.sub0, [[COPY6]], %subreg.sub1, [[COPY5]], %subreg.sub2, [[COPY4]], %subreg.sub3
+ ; GFX12-NEXT: [[REG_SEQUENCE1:%[0-9]+]]:vreg_96 = REG_SEQUENCE [[COPY3]], %subreg.sub0, [[COPY2]], %subreg.sub1, [[COPY1]], %subreg.sub2
+ ; GFX12-NEXT: [[REG_SEQUENCE2:%[0-9]+]]:vreg_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[REG_SEQUENCE1]], %subreg.sub1_sub2_sub3
+ ; GFX12-NEXT: BUFFER_STORE_DWORDX4_VBUFFER_OFFSET_exact killed [[REG_SEQUENCE2]], [[REG_SEQUENCE]], $sgpr_null, 4, 0, 0, implicit $exec :: (dereferenceable store (s128), align 1, addrspace 4)
+ %7:vgpr_32 = COPY $vgpr3
+ %6:vgpr_32 = COPY $vgpr2
+ %5:vgpr_32 = COPY $vgpr1
+ %4:vgpr_32 = COPY $vgpr0
+ %3:sgpr_32 = COPY $sgpr3
+ %2:sgpr_32 = COPY $sgpr2
+ %1:sgpr_32 = COPY $sgpr1
+ %0:sgpr_32 = COPY $sgpr0
+ %13:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3
+ %14:vreg_96 = REG_SEQUENCE %4:vgpr_32, %subreg.sub0, %5:vgpr_32, %subreg.sub1, %6:vgpr_32, %subreg.sub2
+ BUFFER_STORE_DWORD_VBUFFER_OFFSET_exact %7:vgpr_32, %13:sgpr_128, $sgpr_null, 4, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4)
+ BUFFER_STORE_DWORDX3_VBUFFER_OFFSET_exact %14:vreg_96, %13:sgpr_128, $sgpr_null, 8, 0, 0, implicit $exec :: (dereferenceable store (s96), align 1, addrspace 4)
+...
+---
+
+name: buffer_store_dwordx3_dword
+body: |
+ bb.0.entry:
+ liveins: $sgpr0, $sgpr1, $sgpr2, $sgpr3, $vgpr0, $vgpr1, $vgpr2, $vgpr3
+ ; GFX12-LABEL: name: buffer_store_dwordx3_dword
+ ; GFX12: liveins: $sgpr0, $sgpr1, $sgpr2, $sgpr3, $vgpr0, $vgpr1, $vgpr2, $vgpr3
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr3
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr2
+ ; GFX12-NEXT: [[COPY2:%[0-9]+]]:vgpr_32 = COPY $vgpr1
+ ; GFX12-NEXT: [[COPY3:%[0-9]+]]:vgpr_32 = COPY $vgpr0
+ ; GFX12-NEXT: [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr3
+ ; GFX12-NEXT: [[COPY5:%[0-9]+]]:sgpr_32 = COPY $sgpr2
+ ; GFX12-NEXT: [[COPY6:%[0-9]+]]:sgpr_32 = COPY $sgpr1
+ ; GFX12-NEXT: [[COPY7:%[0-9]+]]:sgpr_32 = COPY $sgpr0
+ ; GFX12-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY7]], %subreg.sub0, [[COPY6]], %subreg.sub1, [[COPY5]], %subreg.sub2, [[COPY4]], %subreg.sub3
+ ; GFX12-NEXT: [[REG_SEQUENCE1:%[0-9]+]]:vreg_96 = REG_SEQUENCE [[COPY3]], %subreg.sub0, [[COPY2]], %subreg.sub1, [[COPY1]], %subreg.sub2
+ ; GFX12-NEXT: [[REG_SEQUENCE2:%[0-9]+]]:vreg_128 = REG_SEQUENCE [[REG_SEQUENCE1]], %subreg.sub0_sub1_sub2, [[COPY]], %subreg.sub3
+ ; GFX12-NEXT: BUFFER_STORE_DWORDX4_VBUFFER_OFFSET_exact killed [[REG_SEQUENCE2]], [[REG_SEQUENCE]], $sgpr_null, 4, 0, 0, implicit $exec :: (dereferenceable store (s128), align 1, addrspace 4)
+ %7:vgpr_32 = COPY $vgpr3
+ %6:vgpr_32 = COPY $vgpr2
+ %5:vgpr_32 = COPY $vgpr1
+ %4:vgpr_32 = COPY $vgpr0
+ %3:sgpr_32 = COPY $sgpr3
+ %2:sgpr_32 = COPY $sgpr2
+ %1:sgpr_32 = COPY $sgpr1
+ %0:sgpr_32 = COPY $sgpr0
+ %13:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3
+ %14:vreg_96 = REG_SEQUENCE %4:vgpr_32, %subreg.sub0, %5:vgpr_32, %subreg.sub1, %6:vgpr_32, %subreg.sub2
+ BUFFER_STORE_DWORDX3_VBUFFER_OFFSET_exact %14:vreg_96, %13:sgpr_128, $sgpr_null, 4, 0, 0, implicit $exec :: (dereferenceable store (s96), align 1, addrspace 4)
+ BUFFER_STORE_DWORD_VBUFFER_OFFSET_exact %7:vgpr_32, %13:sgpr_128, $sgpr_null, 16, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4)
+...
+---
+
+name: buffer_store_dwordx2_dwordx2
+body: |
+ bb.0.entry:
+ liveins: $sgpr0, $sgpr1, $sgpr2, $sgpr3, $vgpr0, $vgpr1, $vgpr2, $vgpr3
+ ; GFX12-LABEL: name: buffer_store_dwordx2_dwordx2
+ ; GFX12: liveins: $sgpr0, $sgpr1, $sgpr2, $sgpr3, $vgpr0, $vgpr1, $vgpr2, $vgpr3
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr3
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr2
+ ; GFX12-NEXT: [[COPY2:%[0-9]+]]:vgpr_32 = COPY $vgpr1
+ ; GFX12-NEXT: [[COPY3:%[0-9]+]]:vgpr_32 = COPY $vgpr0
+ ; GFX12-NEXT: [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr3
+ ; GFX12-NEXT: [[COPY5:%[0-9]+]]:sgpr_32 = COPY $sgpr2
+ ; GFX12-NEXT: [[COPY6:%[0-9]+]]:sgpr_32 = COPY $sgpr1
+ ; GFX12-NEXT: [[COPY7:%[0-9]+]]:sgpr_32 = COPY $sgpr0
+ ; GFX12-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY7]], %subreg.sub0, [[COPY6]], %subreg.sub1, [[COPY5]], %subreg.sub2, [[COPY4]], %subreg.sub3
+ ; GFX12-NEXT: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[COPY3]], %subreg.sub0, [[COPY2]], %subreg.sub1
+ ; GFX12-NEXT: [[REG_SEQUENCE2:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[COPY1]], %subreg.sub0, [[COPY]], %subreg.sub1
+ ; GFX12-NEXT: [[REG_SEQUENCE3:%[0-9]+]]:vreg_128 = REG_SEQUENCE [[REG_SEQUENCE1]], %subreg.sub0_sub1, [[REG_SEQUENCE2]], %subreg.sub2_sub3
+ ; GFX12-NEXT: BUFFER_STORE_DWORDX4_VBUFFER_OFFSET_exact killed [[REG_SEQUENCE3]], [[REG_SEQUENCE]], $sgpr_null, 4, 0, 0, implicit $exec :: (dereferenceable store (s128), align 1, addrspace 4)
+ %7:vgpr_32 = COPY $vgpr3
+ %6:vgpr_32 = COPY $vgpr2
+ %5:vgpr_32 = COPY $vgpr1
+ %4:vgpr_32 = COPY $vgpr0
+ %3:sgpr_32 = COPY $sgpr3
+ %2:sgpr_32 = COPY $sgpr2
+ %1:sgpr_32 = COPY $sgpr1
+ %0:sgpr_32 = COPY $sgpr0
+ %13:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3
+ %14:vreg_64 = REG_SEQUENCE %4:vgpr_32, %subreg.sub0, %5:vgpr_32, %subreg.sub1
+ %15:vreg_64 = REG_SEQUENCE %6:vgpr_32, %subreg.sub0, %7:vgpr_32, %subreg.sub1
+ BUFFER_STORE_DWORDX2_VBUFFER_OFFSET_exact %14:vreg_64, %13:sgpr_128, $sgpr_null, 4, 0, 0, implicit $exec :: (dereferenceable store (s64), align 1, addrspace 4)
+ BUFFER_STORE_DWORDX2_VBUFFER_OFFSET_exact %15:vreg_64, %13:sgpr_128, $sgpr_null, 12, 0, 0, implicit $exec :: (dereferenceable store (s64), align 1, addrspace 4)
+...
+---
+
+name: buffer_store_dword_dwordx2
+body: |
+ bb.0.entry:
+ liveins: $sgpr0, $sgpr1, $sgpr2, $sgpr3, $vgpr0, $vgpr1, $vgpr2, $vgpr3
+ ; GFX12-LABEL: name: buffer_store_dword_dwordx2
+ ; GFX12: liveins: $sgpr0, $sgpr1, $sgpr2, $sgpr3, $vgpr0, $vgpr1, $vgpr2, $vgpr3
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr3
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr2
+ ; GFX12-NEXT: [[COPY2:%[0-9]+]]:vgpr_32 = COPY $vgpr1
+ ; GFX12-NEXT: [[COPY3:%[0-9]+]]:vgpr_32 = COPY $vgpr0
+ ; GFX12-NEXT: [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr3
+ ; GFX12-NEXT: [[COPY5:%[0-9]+]]:sgpr_32 = COPY $sgpr2
+ ; GFX12-NEXT: [[COPY6:%[0-9]+]]:sgpr_32 = COPY $sgpr1
+ ; GFX12-NEXT: [[COPY7:%[0-9]+]]:sgpr_32 = COPY $sgpr0
+ ; GFX12-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY7]], %subreg.sub0, [[COPY6]], %subreg.sub1, [[COPY5]], %subreg.sub2, [[COPY4]], %subreg.sub3
+ ; GFX12-NEXT: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[COPY3]], %subreg.sub0, [[COPY2]], %subreg.sub1
+ ; GFX12-NEXT: [[REG_SEQUENCE2:%[0-9]+]]:vreg_96 = REG_SEQUENCE [[COPY]], %subreg.sub0, %10:vreg_64, %subreg.sub1_sub2
+ ; GFX12-NEXT: BUFFER_STORE_DWORDX3_VBUFFER_OFFSET_exact killed [[REG_SEQUENCE2]], [[REG_SEQUENCE]], $sgpr_null, 4, 0, 0, implicit $exec :: (dereferenceable store (s96), align 1, addrspace 4)
+ %7:vgpr_32 = COPY $vgpr3
+ %6:vgpr_32 = COPY $vgpr2
+ %5:vgpr_32 = COPY $vgpr1
+ %4:vgpr_32 = COPY $vgpr0
+ %3:sgpr_32 = COPY $sgpr3
+ %2:sgpr_32 = COPY $sgpr2
+ %1:sgpr_32 = COPY $sgpr1
+ %0:sgpr_32 = COPY $sgpr0
+ %13:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3
+ %14:vreg_64 = REG_SEQUENCE %4:vgpr_32, %subreg.sub0, %5:vgpr_32, %subreg.sub1
+ BUFFER_STORE_DWORD_VBUFFER_OFFSET_exact %7:vgpr_32, %13:sgpr_128, $sgpr_null, 4, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4)
+ BUFFER_STORE_DWORDX2_VBUFFER_OFFSET_exact %15:vreg_64, %13:sgpr_128, $sgpr_null, 8, 0, 0, implicit $exec :: (dereferenceable store (s64), align 1, addrspace 4)
+...
+---
+
+name: buffer_store_dwordx2_dword
+body: |
+ bb.0.entry:
+ liveins: $sgpr0, $sgpr1, $sgpr2, $sgpr3, $vgpr0, $vgpr1, $vgpr2, $vgpr3
+ liveins: $sgpr0, $sgpr1, $sgpr2, $sgpr3, $vgpr0, $vgpr1, $vgpr2, $vgpr3
+ ; GFX12-LABEL: name: buffer_store_dwordx2_dword
+ ; GFX12: liveins: $sgpr0, $sgpr1, $sgpr2, $sgpr3, $vgpr0, $vgpr1, $vgpr2, $vgpr3, $sgpr0, $sgpr1, $sgpr2, $sgpr3, $vgpr0, $vgpr1, $vgpr2, $vgpr3
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr3
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr2
+ ; GFX12-NEXT: [[COPY2:%[0-9]+]]:vgpr_32 = COPY $vgpr1
+ ; GFX12-NEXT: [[COPY3:%[0-9]+]]:vgpr_32 = COPY $vgpr0
+ ; GFX12-NEXT: [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr3
+ ; GFX12-NEXT: [[COPY5:%[0-9]+]]:sgpr_32 = COPY $sgpr2
+ ; GFX12-NEXT: [[COPY6:%[0-9]+]]:sgpr_32 = COPY $sgpr1
+ ; GFX12-NEXT: [[COPY7:%[0-9]+]]:sgpr_32 = COPY $sgpr0
+ ; GFX12-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY7]], %subreg.sub0, [[COPY6]], %subreg.sub1, [[COPY5]], %subreg.sub2, [[COPY4]], %subreg.sub3
+ ; GFX12-NEXT: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[COPY3]], %subreg.sub0, [[COPY2]], %subreg.sub1
+ ; GFX12-NEXT: [[REG_SEQUENCE2:%[0-9]+]]:vreg_96 = REG_SEQUENCE [[REG_SEQUENCE1]], %subreg.sub0_sub1, [[COPY]], %subreg.sub2
+ ; GFX12-NEXT: BUFFER_STORE_DWORDX3_VBUFFER_OFFSET_exact killed [[REG_SEQUENCE2]], [[REG_SEQUENCE]], $sgpr_null, 4, 0, 0, implicit $exec :: (dereferenceable store (s96), align 1, addrspace 4)
+ %7:vgpr_32 = COPY $vgpr3
+ %6:vgpr_32 = COPY $vgpr2
+ %5:vgpr_32 = COPY $vgpr1
+ %4:vgpr_32 = COPY $vgpr0
+ %3:sgpr_32 = COPY $sgpr3
+ %2:sgpr_32 = COPY $sgpr2
+ %1:sgpr_32 = COPY $sgpr1
+ %0:sgpr_32 = COPY $sgpr0
+ %13:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3
+ %14:vreg_64 = REG_SEQUENCE %4:vgpr_32, %subreg.sub0, %5:vgpr_32, %subreg.sub1
+ BUFFER_STORE_DWORDX2_VBUFFER_OFFSET_exact %14:vreg_64, %13:sgpr_128, $sgpr_null, 4, 0, 0, implicit $exec :: (dereferenceable store (s64), align 1, addrspace 4)
+ BUFFER_STORE_DWORD_VBUFFER_OFFSET_exact %7:vgpr_32, %13:sgpr_128, $sgpr_null, 12, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4)
+...
+---
+
+name: buffer_store_dword_dword
+body: |
+ bb.0.entry:
+ liveins: $sgpr0, $sgpr1, $sgpr2, $sgpr3, $vgpr0, $vgpr1, $vgpr2, $vgpr3
+ ; GFX12-LABEL: name: buffer_store_dword_dword
+ ; GFX12: liveins: $sgpr0, $sgpr1, $sgpr2, $sgpr3, $vgpr0, $vgpr1, $vgpr2, $vgpr3
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr3
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr2
+ ; GFX12-NEXT: [[COPY2:%[0-9]+]]:vgpr_32 = COPY $vgpr1
+ ; GFX12-NEXT: [[COPY3:%[0-9]+]]:vgpr_32 = COPY $vgpr0
+ ; GFX12-NEXT: [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr3
+ ; GFX12-NEXT: [[COPY5:%[0-9]+]]:sgpr_32 = COPY $sgpr2
+ ; GFX12-NEXT: [[COPY6:%[0-9]+]]:sgpr_32 = COPY $sgpr1
+ ; GFX12-NEXT: [[COPY7:%[0-9]+]]:sgpr_32 = COPY $sgpr0
+ ; GFX12-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY7]], %subreg.sub0, [[COPY6]], %subreg.sub1, [[COPY5]], %subreg.sub2, [[COPY4]], %subreg.sub3
+ ; GFX12-NEXT: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[COPY1]], %subreg.sub0, [[COPY]], %subreg.sub1
+ ; GFX12-NEXT: BUFFER_STORE_DWORDX2_VBUFFER_OFFSET_exact killed [[REG_SEQUENCE1]], [[REG_SEQUENCE]], $sgpr_null, 4, 0, 0, implicit $exec :: (dereferenceable store (s64), align 1, addrspace 4)
+ %7:vgpr_32 = COPY $vgpr3
+ %6:vgpr_32 = COPY $vgpr2
+ %5:vgpr_32 = COPY $vgpr1
+ %4:vgpr_32 = COPY $vgpr0
+ %3:sgpr_32 = COPY $sgpr3
+ %2:sgpr_32 = COPY $sgpr2
+ %1:sgpr_32 = COPY $sgpr1
+ %0:sgpr_32 = COPY $sgpr0
+ %13:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3
+ BUFFER_STORE_DWORD_VBUFFER_OFFSET_exact %6:vgpr_32, %13:sgpr_128, $sgpr_null, 4, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4)
+ BUFFER_STORE_DWORD_VBUFFER_OFFSET_exact %7:vgpr_32, %13:sgpr_128, $sgpr_null, 8, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4)
+...
+---
+
+name: buffer_store_dword_32
+body: |
+ bb.0.entry:
+ liveins: $sgpr0, $sgpr1, $sgpr2, $sgpr3, $vgpr0, $vgpr1, $vgpr2, $vgpr3, $vgpr4, $vgpr5, $vgpr6, $vgpr7, $vgpr8
+ ; GFX12-LABEL: name: buffer_store_dword_32
+ ; GFX12: liveins: $sgpr0, $sgpr1, $sgpr2, $sgpr3, $vgpr0, $vgpr1, $vgpr2, $vgpr3, $vgpr4, $vgpr5, $vgpr6, $vgpr7, $vgpr8
+ ; GFX12-NEXT: {{ $}}
+ ; GFX12-NEXT: [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr8
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr7
+ ; GFX12-NEXT: [[COPY2:%[0-9]+]]:vgpr_32 = COPY $vgpr6
+ ; GFX12-NEXT: [[COPY3:%[0-9]+]]:vgpr_32 = COPY $vgpr5
+ ; GFX12-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY $vgpr4
+ ; GFX12-NEXT: [[COPY5:%[0-9]+]]:vgpr_32 = COPY $vgpr3
+ ; GFX12-NEXT: [[COPY6:%[0-9]+]]:vgpr_32 = COPY $vgpr2
+ ; GFX12-NEXT: [[COPY7:%[0-9]+]]:vgpr_32 = COPY $vgpr1
+ ; GFX12-NEXT: [[COPY8:%[0-9]+]]:vgpr_32 = COPY $vgpr0
+ ; GFX12-NEXT: [[COPY9:%[0-9]+]]:sgpr_32 = COPY $sgpr3
+ ; GFX12-NEXT: [[COPY10:%[0-9]+]]:sgpr_32 = COPY $sgpr2
+ ; GFX12-NEXT: [[COPY11:%[0-9]+]]:sgpr_32 = COPY $sgpr1
+ ; GFX12-NEXT: [[COPY12:%[0-9]+]]:sgpr_32 = COPY $sgpr0
+ ; GFX12-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY12]], %subreg.sub0, [[COPY11]], %subreg.sub1, [[COPY10]], %subreg.sub2, [[COPY9]], %subreg.sub3
+ ; GFX12-NEXT: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[COPY8]], %subreg.sub0, [[COPY7]], %subreg.sub1
+ ; GFX12-NEXT: BUFFER_STORE_DWORDX2_VBUFFER_OFFSET_exact killed [[REG_SEQUENCE1]], [[REG_SEQUENCE]], $sgpr_null, 4, 0, 0, implicit $exec :: (dereferenceable store (s64), align 1, addrspace 4)
+ ; GFX12-NEXT: [[REG_SEQUENCE2:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[COPY6]], %subreg.sub0, [[COPY5]], %subreg.sub1
+ ; GFX12-NEXT: [[REG_SEQUENCE3:%[0-9]+]]:vreg_96 = REG_SEQUENCE killed [[REG_SEQUENCE2]], %subreg.sub0_sub1, [[COPY4]], %subreg.sub2
+ ; GFX12-NEXT: [[REG_SEQUENCE4:%[0-9]+]]:vreg_128 = REG_SEQUENCE killed [[REG_SEQUENCE3]], %subreg.sub0_sub1_sub2, [[COPY3]], %subreg.sub3
+ ; GFX12-NEXT: BUFFER_STORE_DWORDX4_VBUFFER_OFFSET_exact killed [[REG_SEQUENCE4]], [[REG_SEQUENCE]], $sgpr_null, 16, 0, 0, implicit $exec :: (dereferenceable store (s128), align 1, addrspace 4)
+ ; GFX12-NEXT: [[REG_SEQUENCE5:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[COPY2]], %subreg.sub0, [[COPY1]], %subreg.sub1
+ ; GFX12-NEXT: [[REG_SEQUENCE6:%[0-9]+]]:vreg_96 = REG_SEQUENCE killed [[REG_SEQUENCE5]], %subreg.sub0_sub1, [[COPY]], %subreg.sub2
+ ; GFX12-NEXT: BUFFER_STORE_DWORDX3_VBUFFER_OFFSET_exact killed [[REG_SEQUENCE6]], [[REG_SEQUENCE]], $sgpr_null, 36, 0, 0, implicit $exec :: (dereferenceable store (s96), align 1, addrspace 4)
+ %12:vgpr_32 = COPY $vgpr8
+ %11:vgpr_32 = COPY $vgpr7
+ %10:vgpr_32 = COPY $vgpr6
+ %9:vgpr_32 = COPY $vgpr5
+ %8:vgpr_32 = COPY $vgpr4
+ %7:vgpr_32 = COPY $vgpr3
+ %6:vgpr_32 = COPY $vgpr2
+ %5:vgpr_32 = COPY $vgpr1
+ %4:vgpr_32 = COPY $vgpr0
+ %3:sgpr_32 = COPY $sgpr3
+ %2:sgpr_32 = COPY $sgpr2
+ %1:sgpr_32 = COPY $sgpr1
+ %0:sgpr_32 = COPY $sgpr0
+ %13:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3
+ BUFFER_STORE_DWORD_VBUFFER_OFFSET_exact %4:vgpr_32, %13:sgpr_128, $sgpr_null, 4, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4)
+ BUFFER_STORE_DWORD_VBUFFER_OFFSET_exact %5:vgpr_32, %13:sgpr_128, $sgpr_null, 8, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4)
+ BUFFER_STORE_DWORD_VBUFFER_OFFSET_exact %6:vgpr_32, %13:sgpr_128, $sgpr_null, 16, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4)
+ BUFFER_STORE_DWORD_VBUFFER_OFFSET_exact %7:vgpr_32, %13:sgpr_128, $sgpr_null, 20, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4)
+ BUFFER_STORE_DWORD_VBUFFER_OFFSET_exact %8:vgpr_32, %13:sgpr_128, $sgpr_null, 24, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4)
+ BUFFER_STORE_DWORD_VBUFFER_OFFSET_exact %9:vgpr_32, %13:sgpr_128, $sgpr_null, 28, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4)
+ BUFFER_STORE_DWORD_VBUFFER_OFFSET_exact %10:vgpr_32, %13:sgpr_128, $sgpr_null, 36, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4)
+ BUFFER_STORE_DWORD_VBUFFER_OFFSET_exact %11:vgpr_32, %13:sgpr_128, $sgpr_null, 40, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4)
+ BUFFER_STORE_DWORD_VBUFFER_OFFSET_exact %12:vgpr_32, %13:sgpr_128, $sgpr_null, 44, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4)
+...
+---
+
+name: buffer_load_dword_not_merged_swizzled_0
+body: |
+ bb.0.entry:
+ ; GFX12-LABEL: name: buffer_load_dword_not_merged_swizzled_0
+ ; GFX12: [[COPY:%[0-9]+]]:sgpr_32 = COPY $sgpr0
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:sgpr_32 = COPY $sgpr1
+ ; GFX12-NEXT: [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr2
+ ; GFX12-NEXT: [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr3
+ ; GFX12-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3
+ ; GFX12-NEXT: [[BUFFER_LOAD_DWORD_VBUFFER_OFFSET:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_DWORD_VBUFFER_OFFSET [[REG_SEQUENCE]], $sgpr_null, 4, 0, 1, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+ ; GFX12-NEXT: [[BUFFER_LOAD_DWORD_VBUFFER_OFFSET1:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_DWORD_VBUFFER_OFFSET [[REG_SEQUENCE]], $sgpr_null, 8, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+ %0:sgpr_32 = COPY $sgpr0
+ %1:sgpr_32 = COPY $sgpr1
+ %2:sgpr_32 = COPY $sgpr2
+ %3:sgpr_32 = COPY $sgpr3
+ %5:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3
+ %7:vgpr_32 = BUFFER_LOAD_DWORD_VBUFFER_OFFSET %5:sgpr_128, $sgpr_null, 4, 0, 1, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+ %8:vgpr_32 = BUFFER_LOAD_DWORD_VBUFFER_OFFSET %5:sgpr_128, $sgpr_null, 8, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+...
+---
+
+name: buffer_load_dword_not_merged_swizzled_1
+body: |
+ bb.0.entry:
+ ; GFX12-LABEL: name: buffer_load_dword_not_merged_swizzled_1
+ ; GFX12: [[COPY:%[0-9]+]]:sgpr_32 = COPY $sgpr0
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:sgpr_32 = COPY $sgpr1
+ ; GFX12-NEXT: [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr2
+ ; GFX12-NEXT: [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr3
+ ; GFX12-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3
+ ; GFX12-NEXT: [[BUFFER_LOAD_DWORD_VBUFFER_OFFSET:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_DWORD_VBUFFER_OFFSET [[REG_SEQUENCE]], $sgpr_null, 4, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+ ; GFX12-NEXT: [[BUFFER_LOAD_DWORD_VBUFFER_OFFSET1:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_DWORD_VBUFFER_OFFSET [[REG_SEQUENCE]], $sgpr_null, 8, 0, 1, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+ %0:sgpr_32 = COPY $sgpr0
+ %1:sgpr_32 = COPY $sgpr1
+ %2:sgpr_32 = COPY $sgpr2
+ %3:sgpr_32 = COPY $sgpr3
+ %5:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3
+ %7:vgpr_32 = BUFFER_LOAD_DWORD_VBUFFER_OFFSET %5:sgpr_128, $sgpr_null, 4, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+ %8:vgpr_32 = BUFFER_LOAD_DWORD_VBUFFER_OFFSET %5:sgpr_128, $sgpr_null, 8, 0, 1, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+...
+---
+
+name: buffer_load_dword_merge_across_swizzle
+body: |
+ bb.0.entry:
+ ; GFX12-LABEL: name: buffer_load_dword_merge_across_swizzle
+ ; GFX12: [[COPY:%[0-9]+]]:sgpr_32 = COPY $sgpr0
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:sgpr_32 = COPY $sgpr1
+ ; GFX12-NEXT: [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr2
+ ; GFX12-NEXT: [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr3
+ ; GFX12-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3
+ ; GFX12-NEXT: [[BUFFER_LOAD_DWORDX2_VBUFFER_OFFSET:%[0-9]+]]:vreg_64 = BUFFER_LOAD_DWORDX2_VBUFFER_OFFSET [[REG_SEQUENCE]], $sgpr_null, 4, 0, 0, implicit $exec :: (dereferenceable load (s64), align 1, addrspace 4)
+ ; GFX12-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[BUFFER_LOAD_DWORDX2_VBUFFER_OFFSET]].sub0
+ ; GFX12-NEXT: [[COPY5:%[0-9]+]]:vgpr_32 = COPY killed [[BUFFER_LOAD_DWORDX2_VBUFFER_OFFSET]].sub1
+ ; GFX12-NEXT: [[BUFFER_LOAD_DWORD_VBUFFER_OFFSET:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_DWORD_VBUFFER_OFFSET [[REG_SEQUENCE]], $sgpr_null, 12, 0, 1, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+ %0:sgpr_32 = COPY $sgpr0
+ %1:sgpr_32 = COPY $sgpr1
+ %2:sgpr_32 = COPY $sgpr2
+ %3:sgpr_32 = COPY $sgpr3
+ %4:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3
+ %5:vgpr_32 = BUFFER_LOAD_DWORD_VBUFFER_OFFSET %4:sgpr_128, $sgpr_null, 4, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+ %6:vgpr_32 = BUFFER_LOAD_DWORD_VBUFFER_OFFSET %4:sgpr_128, $sgpr_null, 12, 0, 1, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+ %7:vgpr_32 = BUFFER_LOAD_DWORD_VBUFFER_OFFSET %4:sgpr_128, $sgpr_null, 8, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+...
+---
+
+name: buffer_load_dword_not_merge_across_swizzled_store
+body: |
+ bb.0.entry:
+ ; GFX12-LABEL: name: buffer_load_dword_not_merge_across_swizzled_store
+ ; GFX12: [[COPY:%[0-9]+]]:sgpr_32 = COPY $sgpr0
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:sgpr_32 = COPY $sgpr1
+ ; GFX12-NEXT: [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr2
+ ; GFX12-NEXT: [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr3
+ ; GFX12-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY $vgpr0
+ ; GFX12-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3
+ ; GFX12-NEXT: [[BUFFER_LOAD_DWORD_VBUFFER_OFFSET:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_DWORD_VBUFFER_OFFSET [[REG_SEQUENCE]], $sgpr_null, 4, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+ ; GFX12-NEXT: BUFFER_STORE_DWORD_VBUFFER_OFFSET_exact [[COPY4]], [[REG_SEQUENCE]], $sgpr_null, 6, 0, 1, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4)
+ ; GFX12-NEXT: [[BUFFER_LOAD_DWORD_VBUFFER_OFFSET1:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_DWORD_VBUFFER_OFFSET [[REG_SEQUENCE]], $sgpr_null, 8, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+ %0:sgpr_32 = COPY $sgpr0
+ %1:sgpr_32 = COPY $sgpr1
+ %2:sgpr_32 = COPY $sgpr2
+ %3:sgpr_32 = COPY $sgpr3
+ %4:vgpr_32 = COPY $vgpr0
+ %5:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3
+ %6:vgpr_32 = BUFFER_LOAD_DWORD_VBUFFER_OFFSET %5:sgpr_128, $sgpr_null, 4, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+ BUFFER_STORE_DWORD_VBUFFER_OFFSET_exact %4:vgpr_32, %5:sgpr_128, $sgpr_null, 6, 0, 1, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4)
+ %7:vgpr_32 = BUFFER_LOAD_DWORD_VBUFFER_OFFSET %5:sgpr_128, $sgpr_null, 8, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+...
+---
+
+name: buffer_load_dword_merge_across_swizzled_store
+body: |
+ bb.0.entry:
+ ; GFX12-LABEL: name: buffer_load_dword_merge_across_swizzled_store
+ ; GFX12: [[COPY:%[0-9]+]]:sgpr_32 = COPY $sgpr0
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:sgpr_32 = COPY $sgpr1
+ ; GFX12-NEXT: [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr2
+ ; GFX12-NEXT: [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr3
+ ; GFX12-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY $vgpr0
+ ; GFX12-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3
+ ; GFX12-NEXT: [[BUFFER_LOAD_DWORDX2_VBUFFER_OFFSET:%[0-9]+]]:vreg_64 = BUFFER_LOAD_DWORDX2_VBUFFER_OFFSET [[REG_SEQUENCE]], $sgpr_null, 4, 0, 0, implicit $exec :: (dereferenceable load (s64), align 1, addrspace 4)
+ ; GFX12-NEXT: [[COPY5:%[0-9]+]]:vgpr_32 = COPY [[BUFFER_LOAD_DWORDX2_VBUFFER_OFFSET]].sub0
+ ; GFX12-NEXT: [[COPY6:%[0-9]+]]:vgpr_32 = COPY killed [[BUFFER_LOAD_DWORDX2_VBUFFER_OFFSET]].sub1
+ ; GFX12-NEXT: BUFFER_STORE_DWORD_VBUFFER_OFFSET_exact [[COPY4]], [[REG_SEQUENCE]], $sgpr_null, 12, 0, 1, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4)
+ %0:sgpr_32 = COPY $sgpr0
+ %1:sgpr_32 = COPY $sgpr1
+ %2:sgpr_32 = COPY $sgpr2
+ %3:sgpr_32 = COPY $sgpr3
+ %4:vgpr_32 = COPY $vgpr0
+ %5:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3
+ %6:vgpr_32 = BUFFER_LOAD_DWORD_VBUFFER_OFFSET %5:sgpr_128, $sgpr_null, 4, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+ BUFFER_STORE_DWORD_VBUFFER_OFFSET_exact %4:vgpr_32, %5:sgpr_128, $sgpr_null, 12, 0, 1, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4)
+ %7:vgpr_32 = BUFFER_LOAD_DWORD_VBUFFER_OFFSET %5:sgpr_128, $sgpr_null, 8, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+...
+---
+
+name: buffer_load_dword_dword_idxen
+body: |
+ bb.0.entry:
+ ; GFX12-LABEL: name: buffer_load_dword_dword_idxen
+ ; GFX12: [[COPY:%[0-9]+]]:sgpr_32 = COPY $sgpr0
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:sgpr_32 = COPY $sgpr1
+ ; GFX12-NEXT: [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr2
+ ; GFX12-NEXT: [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr3
+ ; GFX12-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY $vgpr0
+ ; GFX12-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3
+ ; GFX12-NEXT: [[BUFFER_LOAD_DWORDX2_VBUFFER_IDXEN:%[0-9]+]]:vreg_64 = BUFFER_LOAD_DWORDX2_VBUFFER_IDXEN [[COPY4]], [[REG_SEQUENCE]], $sgpr_null, 0, 0, 0, implicit $exec :: (dereferenceable load (s64), align 1, addrspace 4)
+ ; GFX12-NEXT: [[COPY5:%[0-9]+]]:vgpr_32 = COPY [[BUFFER_LOAD_DWORDX2_VBUFFER_IDXEN]].sub0
+ ; GFX12-NEXT: [[COPY6:%[0-9]+]]:vgpr_32 = COPY killed [[BUFFER_LOAD_DWORDX2_VBUFFER_IDXEN]].sub1
+ %0:sgpr_32 = COPY $sgpr0
+ %1:sgpr_32 = COPY $sgpr1
+ %2:sgpr_32 = COPY $sgpr2
+ %3:sgpr_32 = COPY $sgpr3
+ %4:vgpr_32 = COPY $vgpr0
+ %5:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3
+ %7:vgpr_32 = BUFFER_LOAD_DWORD_VBUFFER_IDXEN %4, %5:sgpr_128, $sgpr_null, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+ %8:vgpr_32 = BUFFER_LOAD_DWORD_VBUFFER_IDXEN %4, %5:sgpr_128, $sgpr_null, 4, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+...
+---
+
+name: buffer_load_dword_dwordx2_idxen
+body: |
+ bb.0.entry:
+ ; GFX12-LABEL: name: buffer_load_dword_dwordx2_idxen
+ ; GFX12: [[COPY:%[0-9]+]]:sgpr_32 = COPY $sgpr0
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:sgpr_32 = COPY $sgpr1
+ ; GFX12-NEXT: [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr2
+ ; GFX12-NEXT: [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr3
+ ; GFX12-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY $vgpr0
+ ; GFX12-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3
+ ; GFX12-NEXT: [[BUFFER_LOAD_DWORDX3_VBUFFER_IDXEN:%[0-9]+]]:vreg_96 = BUFFER_LOAD_DWORDX3_VBUFFER_IDXEN [[COPY4]], [[REG_SEQUENCE]], $sgpr_null, 0, 0, 0, implicit $exec :: (dereferenceable load (s64), align 1, addrspace 4)
+ ; GFX12-NEXT: [[COPY5:%[0-9]+]]:vgpr_32 = COPY [[BUFFER_LOAD_DWORDX3_VBUFFER_IDXEN]].sub0
+ ; GFX12-NEXT: [[COPY6:%[0-9]+]]:vreg_64 = COPY killed [[BUFFER_LOAD_DWORDX3_VBUFFER_IDXEN]].sub1_sub2
+ %0:sgpr_32 = COPY $sgpr0
+ %1:sgpr_32 = COPY $sgpr1
+ %2:sgpr_32 = COPY $sgpr2
+ %3:sgpr_32 = COPY $sgpr3
+ %4:vgpr_32 = COPY $vgpr0
+ %5:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3
+ %7:vgpr_32 = BUFFER_LOAD_DWORD_VBUFFER_IDXEN %4, %5:sgpr_128, $sgpr_null, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+ %8:vreg_64 = BUFFER_LOAD_DWORDX2_VBUFFER_IDXEN %4, %5:sgpr_128, $sgpr_null, 4, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+...
+---
+
+name: buffer_load_dwordx2_dwordx2_idxen
+body: |
+ bb.0.entry:
+ ; GFX12-LABEL: name: buffer_load_dwordx2_dwordx2_idxen
+ ; GFX12: [[COPY:%[0-9]+]]:sgpr_32 = COPY $sgpr0
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:sgpr_32 = COPY $sgpr1
+ ; GFX12-NEXT: [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr2
+ ; GFX12-NEXT: [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr3
+ ; GFX12-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY $vgpr0
+ ; GFX12-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3
+ ; GFX12-NEXT: [[BUFFER_LOAD_DWORDX4_VBUFFER_IDXEN:%[0-9]+]]:vreg_128 = BUFFER_LOAD_DWORDX4_VBUFFER_IDXEN [[COPY4]], [[REG_SEQUENCE]], $sgpr_null, 4, 0, 0, implicit $exec :: (dereferenceable load (s64), align 1, addrspace 4)
+ ; GFX12-NEXT: [[COPY5:%[0-9]+]]:vreg_64 = COPY [[BUFFER_LOAD_DWORDX4_VBUFFER_IDXEN]].sub0_sub1
+ ; GFX12-NEXT: [[COPY6:%[0-9]+]]:vreg_64 = COPY killed [[BUFFER_LOAD_DWORDX4_VBUFFER_IDXEN]].sub2_sub3
+ %0:sgpr_32 = COPY $sgpr0
+ %1:sgpr_32 = COPY $sgpr1
+ %2:sgpr_32 = COPY $sgpr2
+ %3:sgpr_32 = COPY $sgpr3
+ %4:vgpr_32 = COPY $vgpr0
+ %5:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3
+ %7:vreg_64 = BUFFER_LOAD_DWORDX2_VBUFFER_IDXEN %4, %5:sgpr_128, $sgpr_null, 4, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+ %8:vreg_64 = BUFFER_LOAD_DWORDX2_VBUFFER_IDXEN %4, %5:sgpr_128, $sgpr_null, 12, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+...
+---
+
+name: buffer_load_dword_dwordx3_idxen
+body: |
+ bb.0.entry:
+ ; GFX12-LABEL: name: buffer_load_dword_dwordx3_idxen
+ ; GFX12: [[COPY:%[0-9]+]]:sgpr_32 = COPY $sgpr0
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:sgpr_32 = COPY $sgpr1
+ ; GFX12-NEXT: [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr2
+ ; GFX12-NEXT: [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr3
+ ; GFX12-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY $vgpr0
+ ; GFX12-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3
+ ; GFX12-NEXT: [[BUFFER_LOAD_DWORDX4_VBUFFER_IDXEN:%[0-9]+]]:vreg_128 = BUFFER_LOAD_DWORDX4_VBUFFER_IDXEN [[COPY4]], [[REG_SEQUENCE]], $sgpr_null, 4, 0, 0, implicit $exec :: (dereferenceable load (s64), align 1, addrspace 4)
+ ; GFX12-NEXT: [[COPY5:%[0-9]+]]:vgpr_32 = COPY [[BUFFER_LOAD_DWORDX4_VBUFFER_IDXEN]].sub0
+ ; GFX12-NEXT: [[COPY6:%[0-9]+]]:vreg_96 = COPY killed [[BUFFER_LOAD_DWORDX4_VBUFFER_IDXEN]].sub1_sub2_sub3
+ %0:sgpr_32 = COPY $sgpr0
+ %1:sgpr_32 = COPY $sgpr1
+ %2:sgpr_32 = COPY $sgpr2
+ %3:sgpr_32 = COPY $sgpr3
+ %4:vgpr_32 = COPY $vgpr0
+ %5:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3
+ %7:vgpr_32 = BUFFER_LOAD_DWORD_VBUFFER_IDXEN %4, %5:sgpr_128, $sgpr_null, 4, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+ %8:vreg_96 = BUFFER_LOAD_DWORDX3_VBUFFER_IDXEN %4, %5:sgpr_128, $sgpr_null, 8, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+...
+---
+
+name: buffer_load_dword_dword_bothen
+body: |
+ bb.0.entry:
+ ; GFX12-LABEL: name: buffer_load_dword_dword_bothen
+ ; GFX12: [[COPY:%[0-9]+]]:sgpr_32 = COPY $sgpr0
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:sgpr_32 = COPY $sgpr1
+ ; GFX12-NEXT: [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr2
+ ; GFX12-NEXT: [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr3
+ ; GFX12-NEXT: [[COPY4:%[0-9]+]]:vreg_64 = COPY $vgpr0
+ ; GFX12-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3
+ ; GFX12-NEXT: [[BUFFER_LOAD_DWORDX2_VBUFFER_BOTHEN:%[0-9]+]]:vreg_64 = BUFFER_LOAD_DWORDX2_VBUFFER_BOTHEN [[COPY4]], [[REG_SEQUENCE]], $sgpr_null, 0, 0, 0, implicit $exec :: (dereferenceable load (s64), align 1, addrspace 4)
+ ; GFX12-NEXT: [[COPY5:%[0-9]+]]:vgpr_32 = COPY [[BUFFER_LOAD_DWORDX2_VBUFFER_BOTHEN]].sub0
+ ; GFX12-NEXT: [[COPY6:%[0-9]+]]:vgpr_32 = COPY killed [[BUFFER_LOAD_DWORDX2_VBUFFER_BOTHEN]].sub1
+ %0:sgpr_32 = COPY $sgpr0
+ %1:sgpr_32 = COPY $sgpr1
+ %2:sgpr_32 = COPY $sgpr2
+ %3:sgpr_32 = COPY $sgpr3
+ %4:vreg_64 = COPY $vgpr0
+ %5:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3
+ %7:vgpr_32 = BUFFER_LOAD_DWORD_VBUFFER_BOTHEN %4, %5:sgpr_128, $sgpr_null, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+ %8:vgpr_32 = BUFFER_LOAD_DWORD_VBUFFER_BOTHEN %4, %5:sgpr_128, $sgpr_null, 4, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+...
+---
+
+name: buffer_load_dword_dwordx2_bothen
+body: |
+ bb.0.entry:
+ ; GFX12-LABEL: name: buffer_load_dword_dwordx2_bothen
+ ; GFX12: [[COPY:%[0-9]+]]:sgpr_32 = COPY $sgpr0
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:sgpr_32 = COPY $sgpr1
+ ; GFX12-NEXT: [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr2
+ ; GFX12-NEXT: [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr3
+ ; GFX12-NEXT: [[COPY4:%[0-9]+]]:vreg_64 = COPY $vgpr0
+ ; GFX12-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3
+ ; GFX12-NEXT: [[BUFFER_LOAD_DWORDX3_VBUFFER_BOTHEN:%[0-9]+]]:vreg_96 = BUFFER_LOAD_DWORDX3_VBUFFER_BOTHEN [[COPY4]], [[REG_SEQUENCE]], $sgpr_null, 0, 0, 0, implicit $exec :: (dereferenceable load (s64), align 1, addrspace 4)
+ ; GFX12-NEXT: [[COPY5:%[0-9]+]]:vgpr_32 = COPY [[BUFFER_LOAD_DWORDX3_VBUFFER_BOTHEN]].sub0
+ ; GFX12-NEXT: [[COPY6:%[0-9]+]]:vreg_64 = COPY killed [[BUFFER_LOAD_DWORDX3_VBUFFER_BOTHEN]].sub1_sub2
+ %0:sgpr_32 = COPY $sgpr0
+ %1:sgpr_32 = COPY $sgpr1
+ %2:sgpr_32 = COPY $sgpr2
+ %3:sgpr_32 = COPY $sgpr3
+ %4:vreg_64 = COPY $vgpr0
+ %5:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3
+ %7:vgpr_32 = BUFFER_LOAD_DWORD_VBUFFER_BOTHEN %4, %5:sgpr_128, $sgpr_null, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+ %8:vreg_64 = BUFFER_LOAD_DWORDX2_VBUFFER_BOTHEN %4, %5:sgpr_128, $sgpr_null, 4, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+...
+---
+
+name: buffer_load_dwordx2_dwordx2_bothen
+body: |
+ bb.0.entry:
+ ; GFX12-LABEL: name: buffer_load_dwordx2_dwordx2_bothen
+ ; GFX12: [[COPY:%[0-9]+]]:sgpr_32 = COPY $sgpr0
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:sgpr_32 = COPY $sgpr1
+ ; GFX12-NEXT: [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr2
+ ; GFX12-NEXT: [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr3
+ ; GFX12-NEXT: [[COPY4:%[0-9]+]]:vreg_64 = COPY $vgpr0
+ ; GFX12-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3
+ ; GFX12-NEXT: [[BUFFER_LOAD_DWORDX4_VBUFFER_BOTHEN:%[0-9]+]]:vreg_128 = BUFFER_LOAD_DWORDX4_VBUFFER_BOTHEN [[COPY4]], [[REG_SEQUENCE]], $sgpr_null, 0, 0, 0, implicit $exec :: (dereferenceable load (s64), align 1, addrspace 4)
+ ; GFX12-NEXT: [[COPY5:%[0-9]+]]:vreg_64 = COPY [[BUFFER_LOAD_DWORDX4_VBUFFER_BOTHEN]].sub0_sub1
+ ; GFX12-NEXT: [[COPY6:%[0-9]+]]:vreg_64 = COPY killed [[BUFFER_LOAD_DWORDX4_VBUFFER_BOTHEN]].sub2_sub3
+ %0:sgpr_32 = COPY $sgpr0
+ %1:sgpr_32 = COPY $sgpr1
+ %2:sgpr_32 = COPY $sgpr2
+ %3:sgpr_32 = COPY $sgpr3
+ %4:vreg_64 = COPY $vgpr0
+ %5:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3
+ %7:vreg_64 = BUFFER_LOAD_DWORDX2_VBUFFER_BOTHEN %4, %5:sgpr_128, $sgpr_null, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+ %8:vreg_64 = BUFFER_LOAD_DWORDX2_VBUFFER_BOTHEN %4, %5:sgpr_128, $sgpr_null, 8, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+...
+---
+
+name: buffer_load_dword_dwordx3_bothen
+body: |
+ bb.0.entry:
+ ; GFX12-LABEL: name: buffer_load_dword_dwordx3_bothen
+ ; GFX12: [[COPY:%[0-9]+]]:sgpr_32 = COPY $sgpr0
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:sgpr_32 = COPY $sgpr1
+ ; GFX12-NEXT: [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr2
+ ; GFX12-NEXT: [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr3
+ ; GFX12-NEXT: [[COPY4:%[0-9]+]]:vreg_64 = COPY $vgpr0
+ ; GFX12-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3
+ ; GFX12-NEXT: [[BUFFER_LOAD_DWORDX4_VBUFFER_BOTHEN:%[0-9]+]]:vreg_128 = BUFFER_LOAD_DWORDX4_VBUFFER_BOTHEN [[COPY4]], [[REG_SEQUENCE]], $sgpr_null, 0, 0, 0, implicit $exec :: (dereferenceable load (s64), align 1, addrspace 4)
+ ; GFX12-NEXT: [[COPY5:%[0-9]+]]:vgpr_32 = COPY [[BUFFER_LOAD_DWORDX4_VBUFFER_BOTHEN]].sub0
+ ; GFX12-NEXT: [[COPY6:%[0-9]+]]:vreg_96 = COPY killed [[BUFFER_LOAD_DWORDX4_VBUFFER_BOTHEN]].sub1_sub2_sub3
+ %0:sgpr_32 = COPY $sgpr0
+ %1:sgpr_32 = COPY $sgpr1
+ %2:sgpr_32 = COPY $sgpr2
+ %3:sgpr_32 = COPY $sgpr3
+ %4:vreg_64 = COPY $vgpr0
+ %5:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3
+ %7:vgpr_32 = BUFFER_LOAD_DWORD_VBUFFER_BOTHEN %4, %5:sgpr_128, $sgpr_null, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+ %8:vreg_96 = BUFFER_LOAD_DWORDX3_VBUFFER_BOTHEN %4, %5:sgpr_128, $sgpr_null, 4, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+...
+---
+
+name: buffer_load_dword_dword_idxen_exact
+body: |
+ bb.0.entry:
+ ; GFX12-LABEL: name: buffer_load_dword_dword_idxen_exact
+ ; GFX12: [[COPY:%[0-9]+]]:sgpr_32 = COPY $sgpr0
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:sgpr_32 = COPY $sgpr1
+ ; GFX12-NEXT: [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr2
+ ; GFX12-NEXT: [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr3
+ ; GFX12-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY $vgpr0
+ ; GFX12-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3
+ ; GFX12-NEXT: [[BUFFER_LOAD_DWORDX2_VBUFFER_IDXEN_exact:%[0-9]+]]:vreg_64 = BUFFER_LOAD_DWORDX2_VBUFFER_IDXEN_exact [[COPY4]], [[REG_SEQUENCE]], $sgpr_null, 0, 0, 0, implicit $exec :: (dereferenceable load (s64), align 1, addrspace 4)
+ ; GFX12-NEXT: [[COPY5:%[0-9]+]]:vgpr_32 = COPY [[BUFFER_LOAD_DWORDX2_VBUFFER_IDXEN_exact]].sub0
+ ; GFX12-NEXT: [[COPY6:%[0-9]+]]:vgpr_32 = COPY killed [[BUFFER_LOAD_DWORDX2_VBUFFER_IDXEN_exact]].sub1
+ %0:sgpr_32 = COPY $sgpr0
+ %1:sgpr_32 = COPY $sgpr1
+ %2:sgpr_32 = COPY $sgpr2
+ %3:sgpr_32 = COPY $sgpr3
+ %4:vgpr_32 = COPY $vgpr0
+ %5:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3
+ %7:vgpr_32 = BUFFER_LOAD_DWORD_VBUFFER_IDXEN_exact %4, %5:sgpr_128, $sgpr_null, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+ %8:vgpr_32 = BUFFER_LOAD_DWORD_VBUFFER_IDXEN_exact %4, %5:sgpr_128, $sgpr_null, 4, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+...
+---
+
+name: buffer_load_dword_dwordx2_idxen_exact
+body: |
+ bb.0.entry:
+ ; GFX12-LABEL: name: buffer_load_dword_dwordx2_idxen_exact
+ ; GFX12: [[COPY:%[0-9]+]]:sgpr_32 = COPY $sgpr0
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:sgpr_32 = COPY $sgpr1
+ ; GFX12-NEXT: [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr2
+ ; GFX12-NEXT: [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr3
+ ; GFX12-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY $vgpr0
+ ; GFX12-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3
+ ; GFX12-NEXT: [[BUFFER_LOAD_DWORDX3_VBUFFER_IDXEN_exact:%[0-9]+]]:vreg_96 = BUFFER_LOAD_DWORDX3_VBUFFER_IDXEN_exact [[COPY4]], [[REG_SEQUENCE]], $sgpr_null, 0, 0, 0, implicit $exec :: (dereferenceable load (s64), align 1, addrspace 4)
+ ; GFX12-NEXT: [[COPY5:%[0-9]+]]:vgpr_32 = COPY [[BUFFER_LOAD_DWORDX3_VBUFFER_IDXEN_exact]].sub0
+ ; GFX12-NEXT: [[COPY6:%[0-9]+]]:vreg_64 = COPY killed [[BUFFER_LOAD_DWORDX3_VBUFFER_IDXEN_exact]].sub1_sub2
+ %0:sgpr_32 = COPY $sgpr0
+ %1:sgpr_32 = COPY $sgpr1
+ %2:sgpr_32 = COPY $sgpr2
+ %3:sgpr_32 = COPY $sgpr3
+ %4:vgpr_32 = COPY $vgpr0
+ %5:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3
+ %7:vgpr_32 = BUFFER_LOAD_DWORD_VBUFFER_IDXEN_exact %4, %5:sgpr_128, $sgpr_null, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+ %8:vreg_64 = BUFFER_LOAD_DWORDX2_VBUFFER_IDXEN_exact %4, %5:sgpr_128, $sgpr_null, 4, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+...
+---
+
+name: buffer_load_dwordx2_dwordx2_idxen_exact
+body: |
+ bb.0.entry:
+ ; GFX12-LABEL: name: buffer_load_dwordx2_dwordx2_idxen_exact
+ ; GFX12: [[COPY:%[0-9]+]]:sgpr_32 = COPY $sgpr0
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:sgpr_32 = COPY $sgpr1
+ ; GFX12-NEXT: [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr2
+ ; GFX12-NEXT: [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr3
+ ; GFX12-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY $vgpr0
+ ; GFX12-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3
+ ; GFX12-NEXT: [[BUFFER_LOAD_DWORDX4_VBUFFER_IDXEN_exact:%[0-9]+]]:vreg_128 = BUFFER_LOAD_DWORDX4_VBUFFER_IDXEN_exact [[COPY4]], [[REG_SEQUENCE]], $sgpr_null, 4, 0, 0, implicit $exec :: (dereferenceable load (s64), align 1, addrspace 4)
+ ; GFX12-NEXT: [[COPY5:%[0-9]+]]:vreg_64 = COPY [[BUFFER_LOAD_DWORDX4_VBUFFER_IDXEN_exact]].sub0_sub1
+ ; GFX12-NEXT: [[COPY6:%[0-9]+]]:vreg_64 = COPY killed [[BUFFER_LOAD_DWORDX4_VBUFFER_IDXEN_exact]].sub2_sub3
+ %0:sgpr_32 = COPY $sgpr0
+ %1:sgpr_32 = COPY $sgpr1
+ %2:sgpr_32 = COPY $sgpr2
+ %3:sgpr_32 = COPY $sgpr3
+ %4:vgpr_32 = COPY $vgpr0
+ %5:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3
+ %7:vreg_64 = BUFFER_LOAD_DWORDX2_VBUFFER_IDXEN_exact %4, %5:sgpr_128, $sgpr_null, 4, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+ %8:vreg_64 = BUFFER_LOAD_DWORDX2_VBUFFER_IDXEN_exact %4, %5:sgpr_128, $sgpr_null, 12, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+...
+---
+
+name: buffer_load_dword_dwordx3_idxen_exact
+body: |
+ bb.0.entry:
+ ; GFX12-LABEL: name: buffer_load_dword_dwordx3_idxen_exact
+ ; GFX12: [[COPY:%[0-9]+]]:sgpr_32 = COPY $sgpr0
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:sgpr_32 = COPY $sgpr1
+ ; GFX12-NEXT: [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr2
+ ; GFX12-NEXT: [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr3
+ ; GFX12-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY $vgpr0
+ ; GFX12-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3
+ ; GFX12-NEXT: [[BUFFER_LOAD_DWORDX4_VBUFFER_IDXEN_exact:%[0-9]+]]:vreg_128 = BUFFER_LOAD_DWORDX4_VBUFFER_IDXEN_exact [[COPY4]], [[REG_SEQUENCE]], $sgpr_null, 4, 0, 0, implicit $exec :: (dereferenceable load (s64), align 1, addrspace 4)
+ ; GFX12-NEXT: [[COPY5:%[0-9]+]]:vgpr_32 = COPY [[BUFFER_LOAD_DWORDX4_VBUFFER_IDXEN_exact]].sub0
+ ; GFX12-NEXT: [[COPY6:%[0-9]+]]:vreg_96 = COPY killed [[BUFFER_LOAD_DWORDX4_VBUFFER_IDXEN_exact]].sub1_sub2_sub3
+ %0:sgpr_32 = COPY $sgpr0
+ %1:sgpr_32 = COPY $sgpr1
+ %2:sgpr_32 = COPY $sgpr2
+ %3:sgpr_32 = COPY $sgpr3
+ %4:vgpr_32 = COPY $vgpr0
+ %5:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3
+ %7:vgpr_32 = BUFFER_LOAD_DWORD_VBUFFER_IDXEN_exact %4, %5:sgpr_128, $sgpr_null, 4, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+ %8:vreg_96 = BUFFER_LOAD_DWORDX3_VBUFFER_IDXEN_exact %4, %5:sgpr_128, $sgpr_null, 8, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+...
+---
+
+name: buffer_load_dword_dword_dword_idxen_exact
+body: |
+ bb.0.entry:
+ ; GFX12-LABEL: name: buffer_load_dword_dword_dword_idxen_exact
+ ; GFX12: [[COPY:%[0-9]+]]:sgpr_32 = COPY $sgpr0
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:sgpr_32 = COPY $sgpr1
+ ; GFX12-NEXT: [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr2
+ ; GFX12-NEXT: [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr3
+ ; GFX12-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY $vgpr0
+ ; GFX12-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3
+ ; GFX12-NEXT: [[BUFFER_LOAD_DWORDX3_VBUFFER_IDXEN_exact:%[0-9]+]]:vreg_96 = BUFFER_LOAD_DWORDX3_VBUFFER_IDXEN_exact [[COPY4]], [[REG_SEQUENCE]], $sgpr_null, 4, 0, 0, implicit $exec :: (dereferenceable load (s96), align 1, addrspace 4)
+ ; GFX12-NEXT: [[COPY5:%[0-9]+]]:vreg_64 = COPY [[BUFFER_LOAD_DWORDX3_VBUFFER_IDXEN_exact]].sub0_sub1
+ ; GFX12-NEXT: [[COPY6:%[0-9]+]]:vgpr_32 = COPY killed [[BUFFER_LOAD_DWORDX3_VBUFFER_IDXEN_exact]].sub2
+ ; GFX12-NEXT: [[COPY7:%[0-9]+]]:vgpr_32 = COPY [[COPY5]].sub0
+ ; GFX12-NEXT: [[COPY8:%[0-9]+]]:vgpr_32 = COPY killed [[COPY5]].sub1
+ %0:sgpr_32 = COPY $sgpr0
+ %1:sgpr_32 = COPY $sgpr1
+ %2:sgpr_32 = COPY $sgpr2
+ %3:sgpr_32 = COPY $sgpr3
+ %4:vgpr_32 = COPY $vgpr0
+ %5:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3
+ %7:vgpr_32 = BUFFER_LOAD_DWORD_VBUFFER_IDXEN_exact %4, %5:sgpr_128, $sgpr_null, 4, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+ %8:vgpr_32 = BUFFER_LOAD_DWORD_VBUFFER_IDXEN_exact %4, %5:sgpr_128, $sgpr_null, 8, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+ %9:vgpr_32 = BUFFER_LOAD_DWORD_VBUFFER_IDXEN_exact %4, %5:sgpr_128, $sgpr_null, 12, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+...
+---
+
+name: buffer_load_dword_dword_dword_idxen_exact_swizzled_0
+body: |
+ bb.0.entry:
+ ; GFX12-LABEL: name: buffer_load_dword_dword_dword_idxen_exact_swizzled_0
+ ; GFX12: [[COPY:%[0-9]+]]:sgpr_32 = COPY $sgpr0
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:sgpr_32 = COPY $sgpr1
+ ; GFX12-NEXT: [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr2
+ ; GFX12-NEXT: [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr3
+ ; GFX12-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY $vgpr0
+ ; GFX12-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3
+ ; GFX12-NEXT: [[BUFFER_LOAD_DWORD_VBUFFER_IDXEN_exact:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_DWORD_VBUFFER_IDXEN_exact [[COPY4]], [[REG_SEQUENCE]], $sgpr_null, 4, 0, 1, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+ ; GFX12-NEXT: [[BUFFER_LOAD_DWORDX2_VBUFFER_IDXEN_exact:%[0-9]+]]:vreg_64 = BUFFER_LOAD_DWORDX2_VBUFFER_IDXEN_exact [[COPY4]], [[REG_SEQUENCE]], $sgpr_null, 8, 0, 0, implicit $exec :: (dereferenceable load (s64), align 1, addrspace 4)
+ ; GFX12-NEXT: [[COPY5:%[0-9]+]]:vgpr_32 = COPY [[BUFFER_LOAD_DWORDX2_VBUFFER_IDXEN_exact]].sub0
+ ; GFX12-NEXT: [[COPY6:%[0-9]+]]:vgpr_32 = COPY killed [[BUFFER_LOAD_DWORDX2_VBUFFER_IDXEN_exact]].sub1
+ %0:sgpr_32 = COPY $sgpr0
+ %1:sgpr_32 = COPY $sgpr1
+ %2:sgpr_32 = COPY $sgpr2
+ %3:sgpr_32 = COPY $sgpr3
+ %4:vgpr_32 = COPY $vgpr0
+ %5:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3
+ %7:vgpr_32 = BUFFER_LOAD_DWORD_VBUFFER_IDXEN_exact %4, %5:sgpr_128, $sgpr_null, 4, 0, 1, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+ %8:vgpr_32 = BUFFER_LOAD_DWORD_VBUFFER_IDXEN_exact %4, %5:sgpr_128, $sgpr_null, 8, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+ %9:vgpr_32 = BUFFER_LOAD_DWORD_VBUFFER_IDXEN_exact %4, %5:sgpr_128, $sgpr_null, 12, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+...
+---
+
+name: buffer_load_dword_dword_bothen_exact
+body: |
+ bb.0.entry:
+ ; GFX12-LABEL: name: buffer_load_dword_dword_bothen_exact
+ ; GFX12: [[COPY:%[0-9]+]]:sgpr_32 = COPY $sgpr0
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:sgpr_32 = COPY $sgpr1
+ ; GFX12-NEXT: [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr2
+ ; GFX12-NEXT: [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr3
+ ; GFX12-NEXT: [[COPY4:%[0-9]+]]:vreg_64 = COPY $vgpr0
+ ; GFX12-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3
+ ; GFX12-NEXT: [[BUFFER_LOAD_DWORDX2_VBUFFER_BOTHEN_exact:%[0-9]+]]:vreg_64 = BUFFER_LOAD_DWORDX2_VBUFFER_BOTHEN_exact [[COPY4]], [[REG_SEQUENCE]], $sgpr_null, 0, 0, 0, implicit $exec :: (dereferenceable load (s64), align 1, addrspace 4)
+ ; GFX12-NEXT: [[COPY5:%[0-9]+]]:vgpr_32 = COPY [[BUFFER_LOAD_DWORDX2_VBUFFER_BOTHEN_exact]].sub0
+ ; GFX12-NEXT: [[COPY6:%[0-9]+]]:vgpr_32 = COPY killed [[BUFFER_LOAD_DWORDX2_VBUFFER_BOTHEN_exact]].sub1
+ %0:sgpr_32 = COPY $sgpr0
+ %1:sgpr_32 = COPY $sgpr1
+ %2:sgpr_32 = COPY $sgpr2
+ %3:sgpr_32 = COPY $sgpr3
+ %4:vreg_64 = COPY $vgpr0
+ %5:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3
+ %7:vgpr_32 = BUFFER_LOAD_DWORD_VBUFFER_BOTHEN_exact %4, %5:sgpr_128, $sgpr_null, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+ %8:vgpr_32 = BUFFER_LOAD_DWORD_VBUFFER_BOTHEN_exact %4, %5:sgpr_128, $sgpr_null, 4, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+...
+---
+
+name: buffer_load_dword_dwordx2_bothen_exact
+body: |
+ bb.0.entry:
+ ; GFX12-LABEL: name: buffer_load_dword_dwordx2_bothen_exact
+ ; GFX12: [[COPY:%[0-9]+]]:sgpr_32 = COPY $sgpr0
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:sgpr_32 = COPY $sgpr1
+ ; GFX12-NEXT: [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr2
+ ; GFX12-NEXT: [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr3
+ ; GFX12-NEXT: [[COPY4:%[0-9]+]]:vreg_64 = COPY $vgpr0
+ ; GFX12-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3
+ ; GFX12-NEXT: [[BUFFER_LOAD_DWORDX3_VBUFFER_BOTHEN_exact:%[0-9]+]]:vreg_96 = BUFFER_LOAD_DWORDX3_VBUFFER_BOTHEN_exact [[COPY4]], [[REG_SEQUENCE]], $sgpr_null, 0, 0, 0, implicit $exec :: (dereferenceable load (s64), align 1, addrspace 4)
+ ; GFX12-NEXT: [[COPY5:%[0-9]+]]:vgpr_32 = COPY [[BUFFER_LOAD_DWORDX3_VBUFFER_BOTHEN_exact]].sub0
+ ; GFX12-NEXT: [[COPY6:%[0-9]+]]:vreg_64 = COPY killed [[BUFFER_LOAD_DWORDX3_VBUFFER_BOTHEN_exact]].sub1_sub2
+ %0:sgpr_32 = COPY $sgpr0
+ %1:sgpr_32 = COPY $sgpr1
+ %2:sgpr_32 = COPY $sgpr2
+ %3:sgpr_32 = COPY $sgpr3
+ %4:vreg_64 = COPY $vgpr0
+ %5:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3
+ %7:vgpr_32 = BUFFER_LOAD_DWORD_VBUFFER_BOTHEN_exact %4, %5:sgpr_128, $sgpr_null, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+ %8:vreg_64 = BUFFER_LOAD_DWORDX2_VBUFFER_BOTHEN_exact %4, %5:sgpr_128, $sgpr_null, 4, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+...
+---
+
+name: buffer_load_dwordx2_dwordx2_bothen_exact
+body: |
+ bb.0.entry:
+
+ ; GFX12-LABEL: name: buffer_load_dwordx2_dwordx2_bothen_exact
+ ; GFX12: [[COPY:%[0-9]+]]:sgpr_32 = COPY $sgpr0
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:sgpr_32 = COPY $sgpr1
+ ; GFX12-NEXT: [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr2
+ ; GFX12-NEXT: [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr3
+ ; GFX12-NEXT: [[COPY4:%[0-9]+]]:vreg_64 = COPY $vgpr0
+ ; GFX12-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3
+ ; GFX12-NEXT: [[BUFFER_LOAD_DWORDX4_VBUFFER_BOTHEN_exact:%[0-9]+]]:vreg_128 = BUFFER_LOAD_DWORDX4_VBUFFER_BOTHEN_exact [[COPY4]], [[REG_SEQUENCE]], $sgpr_null, 0, 0, 0, implicit $exec :: (dereferenceable load (s64), align 1, addrspace 4)
+ ; GFX12-NEXT: [[COPY5:%[0-9]+]]:vreg_64 = COPY [[BUFFER_LOAD_DWORDX4_VBUFFER_BOTHEN_exact]].sub0_sub1
+ ; GFX12-NEXT: [[COPY6:%[0-9]+]]:vreg_64 = COPY killed [[BUFFER_LOAD_DWORDX4_VBUFFER_BOTHEN_exact]].sub2_sub3
+ %0:sgpr_32 = COPY $sgpr0
+ %1:sgpr_32 = COPY $sgpr1
+ %2:sgpr_32 = COPY $sgpr2
+ %3:sgpr_32 = COPY $sgpr3
+ %4:vreg_64 = COPY $vgpr0
+ %5:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3
+ %7:vreg_64 = BUFFER_LOAD_DWORDX2_VBUFFER_BOTHEN_exact %4, %5:sgpr_128, $sgpr_null, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+ %8:vreg_64 = BUFFER_LOAD_DWORDX2_VBUFFER_BOTHEN_exact %4, %5:sgpr_128, $sgpr_null, 8, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+...
+---
+
+name: buffer_load_dword_dwordx3_bothen_exact
+body: |
+ bb.0.entry:
+
+ ; GFX12-LABEL: name: buffer_load_dword_dwordx3_bothen_exact
+ ; GFX12: [[COPY:%[0-9]+]]:sgpr_32 = COPY $sgpr0
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:sgpr_32 = COPY $sgpr1
+ ; GFX12-NEXT: [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr2
+ ; GFX12-NEXT: [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr3
+ ; GFX12-NEXT: [[COPY4:%[0-9]+]]:vreg_64 = COPY $vgpr0
+ ; GFX12-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3
+ ; GFX12-NEXT: [[BUFFER_LOAD_DWORDX4_VBUFFER_BOTHEN_exact:%[0-9]+]]:vreg_128 = BUFFER_LOAD_DWORDX4_VBUFFER_BOTHEN_exact [[COPY4]], [[REG_SEQUENCE]], $sgpr_null, 0, 0, 0, implicit $exec :: (dereferenceable load (s64), align 1, addrspace 4)
+ ; GFX12-NEXT: [[COPY5:%[0-9]+]]:vgpr_32 = COPY [[BUFFER_LOAD_DWORDX4_VBUFFER_BOTHEN_exact]].sub0
+ ; GFX12-NEXT: [[COPY6:%[0-9]+]]:vreg_96 = COPY killed [[BUFFER_LOAD_DWORDX4_VBUFFER_BOTHEN_exact]].sub1_sub2_sub3
+ %0:sgpr_32 = COPY $sgpr0
+ %1:sgpr_32 = COPY $sgpr1
+ %2:sgpr_32 = COPY $sgpr2
+ %3:sgpr_32 = COPY $sgpr3
+ %4:vreg_64 = COPY $vgpr0
+ %5:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3
+ %7:vgpr_32 = BUFFER_LOAD_DWORD_VBUFFER_BOTHEN_exact %4, %5:sgpr_128, $sgpr_null, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+ %8:vreg_96 = BUFFER_LOAD_DWORDX3_VBUFFER_BOTHEN_exact %4, %5:sgpr_128, $sgpr_null, 4, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+...
+---
+
+name: buffer_load_dword_dword_dword_bothen_exact
+body: |
+ bb.0.entry:
+ ; GFX12-LABEL: name: buffer_load_dword_dword_dword_bothen_exact
+ ; GFX12: [[COPY:%[0-9]+]]:sgpr_32 = COPY $sgpr0
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:sgpr_32 = COPY $sgpr1
+ ; GFX12-NEXT: [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr2
+ ; GFX12-NEXT: [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr3
+ ; GFX12-NEXT: [[COPY4:%[0-9]+]]:vreg_64 = COPY $vgpr0
+ ; GFX12-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3
+ ; GFX12-NEXT: [[BUFFER_LOAD_DWORDX3_VBUFFER_BOTHEN_exact:%[0-9]+]]:vreg_96 = BUFFER_LOAD_DWORDX3_VBUFFER_BOTHEN_exact [[COPY4]], [[REG_SEQUENCE]], $sgpr_null, 4, 0, 0, implicit $exec :: (dereferenceable load (s96), align 1, addrspace 4)
+ ; GFX12-NEXT: [[COPY5:%[0-9]+]]:vreg_64 = COPY [[BUFFER_LOAD_DWORDX3_VBUFFER_BOTHEN_exact]].sub0_sub1
+ ; GFX12-NEXT: [[COPY6:%[0-9]+]]:vgpr_32 = COPY killed [[BUFFER_LOAD_DWORDX3_VBUFFER_BOTHEN_exact]].sub2
+ ; GFX12-NEXT: [[COPY7:%[0-9]+]]:vgpr_32 = COPY [[COPY5]].sub0
+ ; GFX12-NEXT: [[COPY8:%[0-9]+]]:vgpr_32 = COPY killed [[COPY5]].sub1
+ %0:sgpr_32 = COPY $sgpr0
+ %1:sgpr_32 = COPY $sgpr1
+ %2:sgpr_32 = COPY $sgpr2
+ %3:sgpr_32 = COPY $sgpr3
+ %4:vreg_64 = COPY $vgpr0
+ %5:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3
+ %7:vgpr_32 = BUFFER_LOAD_DWORD_VBUFFER_BOTHEN_exact %4, %5:sgpr_128, $sgpr_null, 4, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+ %8:vgpr_32 = BUFFER_LOAD_DWORD_VBUFFER_BOTHEN_exact %4, %5:sgpr_128, $sgpr_null, 8, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+ %9:vgpr_32 = BUFFER_LOAD_DWORD_VBUFFER_BOTHEN_exact %4, %5:sgpr_128, $sgpr_null, 12, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+...
+---
+
+name: buffer_load_dword_dword_dword_bothen_exact_swizzled_0
+body: |
+ bb.0.entry:
+ ; GFX12-LABEL: name: buffer_load_dword_dword_dword_bothen_exact_swizzled_0
+ ; GFX12: [[COPY:%[0-9]+]]:sgpr_32 = COPY $sgpr0
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:sgpr_32 = COPY $sgpr1
+ ; GFX12-NEXT: [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr2
+ ; GFX12-NEXT: [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr3
+ ; GFX12-NEXT: [[COPY4:%[0-9]+]]:vreg_64 = COPY $vgpr0
+ ; GFX12-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3
+ ; GFX12-NEXT: [[BUFFER_LOAD_DWORD_VBUFFER_BOTHEN_exact:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_DWORD_VBUFFER_BOTHEN_exact [[COPY4]], [[REG_SEQUENCE]], $sgpr_null, 4, 0, 1, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+ ; GFX12-NEXT: [[BUFFER_LOAD_DWORDX2_VBUFFER_BOTHEN_exact:%[0-9]+]]:vreg_64 = BUFFER_LOAD_DWORDX2_VBUFFER_BOTHEN_exact [[COPY4]], [[REG_SEQUENCE]], $sgpr_null, 8, 0, 0, implicit $exec :: (dereferenceable load (s64), align 1, addrspace 4)
+ ; GFX12-NEXT: [[COPY5:%[0-9]+]]:vgpr_32 = COPY [[BUFFER_LOAD_DWORDX2_VBUFFER_BOTHEN_exact]].sub0
+ ; GFX12-NEXT: [[COPY6:%[0-9]+]]:vgpr_32 = COPY killed [[BUFFER_LOAD_DWORDX2_VBUFFER_BOTHEN_exact]].sub1
+ %0:sgpr_32 = COPY $sgpr0
+ %1:sgpr_32 = COPY $sgpr1
+ %2:sgpr_32 = COPY $sgpr2
+ %3:sgpr_32 = COPY $sgpr3
+ %4:vreg_64 = COPY $vgpr0
+ %5:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3
+ %7:vgpr_32 = BUFFER_LOAD_DWORD_VBUFFER_BOTHEN_exact %4, %5:sgpr_128, $sgpr_null, 4, 0, 1, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+ %8:vgpr_32 = BUFFER_LOAD_DWORD_VBUFFER_BOTHEN_exact %4, %5:sgpr_128, $sgpr_null, 8, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+ %9:vgpr_32 = BUFFER_LOAD_DWORD_VBUFFER_BOTHEN_exact %4, %5:sgpr_128, $sgpr_null, 12, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+...
+---
+
+name: buffer_load_dwordx2_dwordx2_bothen_exact_
diff _vaddr
+body: |
+ bb.0.entry:
+ ; GFX12-LABEL: name: buffer_load_dwordx2_dwordx2_bothen_exact_
diff _vaddr
+ ; GFX12: [[COPY:%[0-9]+]]:sgpr_32 = COPY $sgpr0
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:sgpr_32 = COPY $sgpr1
+ ; GFX12-NEXT: [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr2
+ ; GFX12-NEXT: [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr3
+ ; GFX12-NEXT: [[COPY4:%[0-9]+]]:vreg_64 = COPY $vgpr0
+ ; GFX12-NEXT: [[COPY5:%[0-9]+]]:vreg_64 = COPY $vgpr1
+ ; GFX12-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3
+ ; GFX12-NEXT: [[BUFFER_LOAD_DWORDX2_VBUFFER_BOTHEN_exact:%[0-9]+]]:vreg_64 = BUFFER_LOAD_DWORDX2_VBUFFER_BOTHEN_exact [[COPY4]], [[REG_SEQUENCE]], $sgpr_null, 4, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+ ; GFX12-NEXT: [[BUFFER_LOAD_DWORDX2_VBUFFER_BOTHEN_exact1:%[0-9]+]]:vreg_64 = BUFFER_LOAD_DWORDX2_VBUFFER_BOTHEN_exact [[COPY5]], [[REG_SEQUENCE]], $sgpr_null, 12, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+ %0:sgpr_32 = COPY $sgpr0
+ %1:sgpr_32 = COPY $sgpr1
+ %2:sgpr_32 = COPY $sgpr2
+ %3:sgpr_32 = COPY $sgpr3
+ %4:vreg_64 = COPY $vgpr0
+ %5:vreg_64 = COPY $vgpr1
+ %6:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3
+ %8:vreg_64 = BUFFER_LOAD_DWORDX2_VBUFFER_BOTHEN_exact %4, %6:sgpr_128, $sgpr_null, 4, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+ %9:vreg_64 = BUFFER_LOAD_DWORDX2_VBUFFER_BOTHEN_exact %5, %6:sgpr_128, $sgpr_null, 12, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+...
+---
+
+name: buffer_load_dwordx2_dwordx2_bothen_exact_
diff _srsrc
+body: |
+ bb.0.entry:
+ ; GFX12-LABEL: name: buffer_load_dwordx2_dwordx2_bothen_exact_
diff _srsrc
+ ; GFX12: [[COPY:%[0-9]+]]:sgpr_32 = COPY $sgpr0
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:sgpr_32 = COPY $sgpr1
+ ; GFX12-NEXT: [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr2
+ ; GFX12-NEXT: [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr3
+ ; GFX12-NEXT: [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr4
+ ; GFX12-NEXT: [[COPY5:%[0-9]+]]:vreg_64 = COPY $vgpr0
+ ; GFX12-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3
+ ; GFX12-NEXT: [[REG_SEQUENCE1:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY1]], %subreg.sub0, [[COPY2]], %subreg.sub1, [[COPY3]], %subreg.sub2, [[COPY4]], %subreg.sub3
+ ; GFX12-NEXT: [[BUFFER_LOAD_DWORDX2_VBUFFER_BOTHEN_exact:%[0-9]+]]:vreg_64 = BUFFER_LOAD_DWORDX2_VBUFFER_BOTHEN_exact [[COPY5]], [[REG_SEQUENCE]], $sgpr_null, 4, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+ ; GFX12-NEXT: [[BUFFER_LOAD_DWORDX2_VBUFFER_BOTHEN_exact1:%[0-9]+]]:vreg_64 = BUFFER_LOAD_DWORDX2_VBUFFER_BOTHEN_exact [[COPY5]], [[REG_SEQUENCE1]], $sgpr_null, 12, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+ %0:sgpr_32 = COPY $sgpr0
+ %1:sgpr_32 = COPY $sgpr1
+ %2:sgpr_32 = COPY $sgpr2
+ %3:sgpr_32 = COPY $sgpr3
+ %4:sgpr_32 = COPY $sgpr4
+ %5:vreg_64 = COPY $vgpr0
+ %6:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3
+ %7:sgpr_128 = REG_SEQUENCE %1:sgpr_32, %subreg.sub0, %2:sgpr_32, %subreg.sub1, %3:sgpr_32, %subreg.sub2, %4:sgpr_32, %subreg.sub3
+ %8:vreg_64 = BUFFER_LOAD_DWORDX2_VBUFFER_BOTHEN_exact %5, %6:sgpr_128, $sgpr_null, 4, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+ %9:vreg_64 = BUFFER_LOAD_DWORDX2_VBUFFER_BOTHEN_exact %5, %7:sgpr_128, $sgpr_null, 12, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+...
+---
+
+name: buffer_load_dwordx2_dwordx2_idxen_exact_
diff _vaddr
+body: |
+ bb.0.entry:
+ ; GFX12-LABEL: name: buffer_load_dwordx2_dwordx2_idxen_exact_
diff _vaddr
+ ; GFX12: [[COPY:%[0-9]+]]:sgpr_32 = COPY $sgpr0
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:sgpr_32 = COPY $sgpr1
+ ; GFX12-NEXT: [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr2
+ ; GFX12-NEXT: [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr3
+ ; GFX12-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY $vgpr0
+ ; GFX12-NEXT: [[COPY5:%[0-9]+]]:vgpr_32 = COPY $vgpr1
+ ; GFX12-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3
+ ; GFX12-NEXT: [[BUFFER_LOAD_DWORDX2_VBUFFER_IDXEN_exact:%[0-9]+]]:vreg_64 = BUFFER_LOAD_DWORDX2_VBUFFER_IDXEN_exact [[COPY4]], [[REG_SEQUENCE]], $sgpr_null, 4, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+ ; GFX12-NEXT: [[BUFFER_LOAD_DWORDX2_VBUFFER_IDXEN_exact1:%[0-9]+]]:vreg_64 = BUFFER_LOAD_DWORDX2_VBUFFER_IDXEN_exact [[COPY5]], [[REG_SEQUENCE]], $sgpr_null, 12, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+ %0:sgpr_32 = COPY $sgpr0
+ %1:sgpr_32 = COPY $sgpr1
+ %2:sgpr_32 = COPY $sgpr2
+ %3:sgpr_32 = COPY $sgpr3
+ %4:vgpr_32 = COPY $vgpr0
+ %5:vgpr_32 = COPY $vgpr1
+ %6:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3
+ %8:vreg_64 = BUFFER_LOAD_DWORDX2_VBUFFER_IDXEN_exact %4, %6:sgpr_128, $sgpr_null, 4, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+ %9:vreg_64 = BUFFER_LOAD_DWORDX2_VBUFFER_IDXEN_exact %5, %6:sgpr_128, $sgpr_null, 12, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+...
+---
+
+name: buffer_load_dwordx2_dwordx2_idxen_exact_
diff _srsrc
+body: |
+ bb.0.entry:
+ ; GFX12-LABEL: name: buffer_load_dwordx2_dwordx2_idxen_exact_
diff _srsrc
+ ; GFX12: [[COPY:%[0-9]+]]:sgpr_32 = COPY $sgpr0
+ ; GFX12-NEXT: [[COPY1:%[0-9]+]]:sgpr_32 = COPY $sgpr1
+ ; GFX12-NEXT: [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr2
+ ; GFX12-NEXT: [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr3
+ ; GFX12-NEXT: [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr4
+ ; GFX12-NEXT: [[COPY5:%[0-9]+]]:vgpr_32 = COPY $vgpr0
+ ; GFX12-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3
+ ; GFX12-NEXT: [[REG_SEQUENCE1:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY1]], %subreg.sub0, [[COPY2]], %subreg.sub1, [[COPY3]], %subreg.sub2, [[COPY4]], %subreg.sub3
+ ; GFX12-NEXT: [[BUFFER_LOAD_DWORDX2_VBUFFER_IDXEN_exact:%[0-9]+]]:vreg_64 = BUFFER_LOAD_DWORDX2_VBUFFER_IDXEN_exact [[COPY5]], [[REG_SEQUENCE]], $sgpr_null, 4, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+ ; GFX12-NEXT: [[BUFFER_LOAD_DWORDX2_VBUFFER_IDXEN_exact1:%[0-9]+]]:vreg_64 = BUFFER_LOAD_DWORDX2_VBUFFER_IDXEN_exact [[COPY5]], [[REG_SEQUENCE1]], $sgpr_null, 12, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+ %0:sgpr_32 = COPY $sgpr0
+ %1:sgpr_32 = COPY $sgpr1
+ %2:sgpr_32 = COPY $sgpr2
+ %3:sgpr_32 = COPY $sgpr3
+ %4:sgpr_32 = COPY $sgpr4
+ %5:vgpr_32 = COPY $vgpr0
+ %6:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3
+ %7:sgpr_128 = REG_SEQUENCE %1:sgpr_32, %subreg.sub0, %2:sgpr_32, %subreg.sub1, %3:sgpr_32, %subreg.sub2, %4:sgpr_32, %subreg.sub3
+ %8:vreg_64 = BUFFER_LOAD_DWORDX2_VBUFFER_IDXEN_exact %5, %6:sgpr_128, $sgpr_null, 4, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+ %9:vreg_64 = BUFFER_LOAD_DWORDX2_VBUFFER_IDXEN_exact %5, %7:sgpr_128, $sgpr_null, 12, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+...
diff --git a/llvm/test/CodeGen/AMDGPU/merge-buffer.mir b/llvm/test/CodeGen/AMDGPU/merge-buffer.mir
new file mode 100644
index 00000000000000..1c6d429d20eac2
--- /dev/null
+++ b/llvm/test/CodeGen/AMDGPU/merge-buffer.mir
@@ -0,0 +1,1130 @@
+# NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py UTC_ARGS: --version 2
+# RUN: llc -mtriple=amdgcn -mcpu=gfx900 -verify-machineinstrs -run-pass si-load-store-opt -o - %s | FileCheck -check-prefixes=GCN %s
+# RUN: llc -mtriple=amdgcn -mcpu=gfx1010 -verify-machineinstrs -run-pass si-load-store-opt -o - %s | FileCheck -check-prefixes=GCN %s
+# RUN: llc -mtriple=amdgcn -mcpu=gfx1100 -verify-machineinstrs -run-pass si-load-store-opt -o - %s | FileCheck -check-prefixes=GCN %s
+
+---
+name: buffer_load_dword_dwordx3
+body: |
+ bb.0.entry:
+ ; GCN-LABEL: name: buffer_load_dword_dwordx3
+ ; GCN: [[COPY:%[0-9]+]]:sgpr_32 = COPY $sgpr0
+ ; GCN-NEXT: [[COPY1:%[0-9]+]]:sgpr_32 = COPY $sgpr1
+ ; GCN-NEXT: [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr2
+ ; GCN-NEXT: [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr3
+ ; GCN-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3
+ ; GCN-NEXT: [[BUFFER_LOAD_DWORDX4_OFFSET:%[0-9]+]]:vreg_128 = BUFFER_LOAD_DWORDX4_OFFSET [[REG_SEQUENCE]], 0, 4, 0, 0, implicit $exec :: (dereferenceable load (s128), align 1, addrspace 4)
+ ; GCN-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[BUFFER_LOAD_DWORDX4_OFFSET]].sub0
+ ; GCN-NEXT: [[COPY5:%[0-9]+]]:vreg_96 = COPY killed [[BUFFER_LOAD_DWORDX4_OFFSET]].sub1_sub2_sub3
+ %0:sgpr_32 = COPY $sgpr0
+ %1:sgpr_32 = COPY $sgpr1
+ %2:sgpr_32 = COPY $sgpr2
+ %3:sgpr_32 = COPY $sgpr3
+ %5:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3
+ %7:vgpr_32 = BUFFER_LOAD_DWORD_OFFSET %5:sgpr_128, 0, 4, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+ %8:vreg_96 = BUFFER_LOAD_DWORDX3_OFFSET %5:sgpr_128, 0, 8, 0, 0, implicit $exec :: (dereferenceable load (s96), align 1, addrspace 4)
+...
+---
+
+name: buffer_load_dwordx3_dword
+body: |
+ bb.0.entry:
+ ; GCN-LABEL: name: buffer_load_dwordx3_dword
+ ; GCN: [[COPY:%[0-9]+]]:sgpr_32 = COPY $sgpr0
+ ; GCN-NEXT: [[COPY1:%[0-9]+]]:sgpr_32 = COPY $sgpr1
+ ; GCN-NEXT: [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr2
+ ; GCN-NEXT: [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr3
+ ; GCN-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3
+ ; GCN-NEXT: [[BUFFER_LOAD_DWORDX4_OFFSET:%[0-9]+]]:vreg_128 = BUFFER_LOAD_DWORDX4_OFFSET [[REG_SEQUENCE]], 0, 4, 0, 0, implicit $exec :: (dereferenceable load (s128), align 1, addrspace 4)
+ ; GCN-NEXT: [[COPY4:%[0-9]+]]:vreg_96 = COPY [[BUFFER_LOAD_DWORDX4_OFFSET]].sub0_sub1_sub2
+ ; GCN-NEXT: [[COPY5:%[0-9]+]]:vgpr_32 = COPY killed [[BUFFER_LOAD_DWORDX4_OFFSET]].sub3
+ %0:sgpr_32 = COPY $sgpr0
+ %1:sgpr_32 = COPY $sgpr1
+ %2:sgpr_32 = COPY $sgpr2
+ %3:sgpr_32 = COPY $sgpr3
+ %5:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3
+ %7:vreg_96 = BUFFER_LOAD_DWORDX3_OFFSET %5:sgpr_128, 0, 4, 0, 0, implicit $exec :: (dereferenceable load (s96), align 1, addrspace 4)
+ %8:vgpr_32 = BUFFER_LOAD_DWORD_OFFSET %5:sgpr_128, 0, 16, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+...
+---
+
+name: buffer_load_dwordx2_dwordx2
+body: |
+ bb.0.entry:
+ ; GCN-LABEL: name: buffer_load_dwordx2_dwordx2
+ ; GCN: [[COPY:%[0-9]+]]:sgpr_32 = COPY $sgpr0
+ ; GCN-NEXT: [[COPY1:%[0-9]+]]:sgpr_32 = COPY $sgpr1
+ ; GCN-NEXT: [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr2
+ ; GCN-NEXT: [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr3
+ ; GCN-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3
+ ; GCN-NEXT: [[BUFFER_LOAD_DWORDX4_OFFSET:%[0-9]+]]:vreg_128 = BUFFER_LOAD_DWORDX4_OFFSET [[REG_SEQUENCE]], 0, 4, 0, 0, implicit $exec :: (dereferenceable load (s128), align 1, addrspace 4)
+ ; GCN-NEXT: [[COPY4:%[0-9]+]]:vreg_64 = COPY [[BUFFER_LOAD_DWORDX4_OFFSET]].sub0_sub1
+ ; GCN-NEXT: [[COPY5:%[0-9]+]]:vreg_64 = COPY killed [[BUFFER_LOAD_DWORDX4_OFFSET]].sub2_sub3
+ %0:sgpr_32 = COPY $sgpr0
+ %1:sgpr_32 = COPY $sgpr1
+ %2:sgpr_32 = COPY $sgpr2
+ %3:sgpr_32 = COPY $sgpr3
+ %5:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3
+ %7:vreg_64 = BUFFER_LOAD_DWORDX2_OFFSET %5:sgpr_128, 0, 4, 0, 0, implicit $exec :: (dereferenceable load (s64), align 1, addrspace 4)
+ %8:vreg_64 = BUFFER_LOAD_DWORDX2_OFFSET %5:sgpr_128, 0, 12, 0, 0, implicit $exec :: (dereferenceable load (s64), align 1, addrspace 4)
+...
+---
+
+name: buffer_load_dword_dwordx2
+body: |
+ bb.0.entry:
+ ; GCN-LABEL: name: buffer_load_dword_dwordx2
+ ; GCN: [[COPY:%[0-9]+]]:sgpr_32 = COPY $sgpr0
+ ; GCN-NEXT: [[COPY1:%[0-9]+]]:sgpr_32 = COPY $sgpr1
+ ; GCN-NEXT: [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr2
+ ; GCN-NEXT: [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr3
+ ; GCN-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3
+ ; GCN-NEXT: [[BUFFER_LOAD_DWORDX3_OFFSET:%[0-9]+]]:vreg_96 = BUFFER_LOAD_DWORDX3_OFFSET [[REG_SEQUENCE]], 0, 4, 0, 0, implicit $exec :: (dereferenceable load (s96), align 1, addrspace 4)
+ ; GCN-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[BUFFER_LOAD_DWORDX3_OFFSET]].sub0
+ ; GCN-NEXT: [[COPY5:%[0-9]+]]:vreg_64 = COPY killed [[BUFFER_LOAD_DWORDX3_OFFSET]].sub1_sub2
+ %0:sgpr_32 = COPY $sgpr0
+ %1:sgpr_32 = COPY $sgpr1
+ %2:sgpr_32 = COPY $sgpr2
+ %3:sgpr_32 = COPY $sgpr3
+ %5:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3
+ %7:vgpr_32 = BUFFER_LOAD_DWORD_OFFSET %5:sgpr_128, 0, 4, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+ %8:vreg_64 = BUFFER_LOAD_DWORDX2_OFFSET %5:sgpr_128, 0, 8, 0, 0, implicit $exec :: (dereferenceable load (s64), align 1, addrspace 4)
+...
+---
+
+name: buffer_load_dwordx2_dword
+body: |
+ bb.0.entry:
+ ; GCN-LABEL: name: buffer_load_dwordx2_dword
+ ; GCN: [[COPY:%[0-9]+]]:sgpr_32 = COPY $sgpr0
+ ; GCN-NEXT: [[COPY1:%[0-9]+]]:sgpr_32 = COPY $sgpr1
+ ; GCN-NEXT: [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr2
+ ; GCN-NEXT: [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr3
+ ; GCN-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3
+ ; GCN-NEXT: [[BUFFER_LOAD_DWORDX3_OFFSET:%[0-9]+]]:vreg_96 = BUFFER_LOAD_DWORDX3_OFFSET [[REG_SEQUENCE]], 0, 4, 0, 0, implicit $exec :: (dereferenceable load (s96), align 1, addrspace 4)
+ ; GCN-NEXT: [[COPY4:%[0-9]+]]:vreg_64 = COPY [[BUFFER_LOAD_DWORDX3_OFFSET]].sub0_sub1
+ ; GCN-NEXT: [[COPY5:%[0-9]+]]:vgpr_32 = COPY killed [[BUFFER_LOAD_DWORDX3_OFFSET]].sub2
+ %0:sgpr_32 = COPY $sgpr0
+ %1:sgpr_32 = COPY $sgpr1
+ %2:sgpr_32 = COPY $sgpr2
+ %3:sgpr_32 = COPY $sgpr3
+ %5:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3
+ %7:vreg_64 = BUFFER_LOAD_DWORDX2_OFFSET %5:sgpr_128, 0, 4, 0, 0, implicit $exec :: (dereferenceable load (s64), align 1, addrspace 4)
+ %8:vgpr_32 = BUFFER_LOAD_DWORD_OFFSET %5:sgpr_128, 0, 12, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+...
+---
+
+
+name: buffer_load_dword_dword
+body: |
+ bb.0.entry:
+ ; GCN-LABEL: name: buffer_load_dword_dword
+ ; GCN: [[COPY:%[0-9]+]]:sgpr_32 = COPY $sgpr0
+ ; GCN-NEXT: [[COPY1:%[0-9]+]]:sgpr_32 = COPY $sgpr1
+ ; GCN-NEXT: [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr2
+ ; GCN-NEXT: [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr3
+ ; GCN-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3
+ ; GCN-NEXT: [[BUFFER_LOAD_DWORDX2_OFFSET:%[0-9]+]]:vreg_64 = BUFFER_LOAD_DWORDX2_OFFSET [[REG_SEQUENCE]], 0, 4, 0, 0, implicit $exec :: (dereferenceable load (s64), align 1, addrspace 4)
+ ; GCN-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[BUFFER_LOAD_DWORDX2_OFFSET]].sub0
+ ; GCN-NEXT: [[COPY5:%[0-9]+]]:vgpr_32 = COPY killed [[BUFFER_LOAD_DWORDX2_OFFSET]].sub1
+ %0:sgpr_32 = COPY $sgpr0
+ %1:sgpr_32 = COPY $sgpr1
+ %2:sgpr_32 = COPY $sgpr2
+ %3:sgpr_32 = COPY $sgpr3
+ %5:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3
+ %7:vgpr_32 = BUFFER_LOAD_DWORD_OFFSET %5:sgpr_128, 0, 4, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+ %8:vgpr_32 = BUFFER_LOAD_DWORD_OFFSET %5:sgpr_128, 0, 8, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+...
+---
+
+name: buffer_load_dword_32
+body: |
+ bb.0.entry:
+ ; GCN-LABEL: name: buffer_load_dword_32
+ ; GCN: [[COPY:%[0-9]+]]:sgpr_32 = COPY $sgpr0
+ ; GCN-NEXT: [[COPY1:%[0-9]+]]:sgpr_32 = COPY $sgpr1
+ ; GCN-NEXT: [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr2
+ ; GCN-NEXT: [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr3
+ ; GCN-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3
+ ; GCN-NEXT: [[BUFFER_LOAD_DWORDX2_OFFSET:%[0-9]+]]:vreg_64 = BUFFER_LOAD_DWORDX2_OFFSET [[REG_SEQUENCE]], 0, 4, 0, 0, implicit $exec :: (dereferenceable load (s64), align 1, addrspace 4)
+ ; GCN-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[BUFFER_LOAD_DWORDX2_OFFSET]].sub0
+ ; GCN-NEXT: [[COPY5:%[0-9]+]]:vgpr_32 = COPY killed [[BUFFER_LOAD_DWORDX2_OFFSET]].sub1
+ ; GCN-NEXT: [[BUFFER_LOAD_DWORDX4_OFFSET:%[0-9]+]]:vreg_128 = BUFFER_LOAD_DWORDX4_OFFSET [[REG_SEQUENCE]], 0, 16, 0, 0, implicit $exec :: (dereferenceable load (s128), align 1, addrspace 4)
+ ; GCN-NEXT: [[COPY6:%[0-9]+]]:vreg_96 = COPY [[BUFFER_LOAD_DWORDX4_OFFSET]].sub0_sub1_sub2
+ ; GCN-NEXT: [[COPY7:%[0-9]+]]:vgpr_32 = COPY killed [[BUFFER_LOAD_DWORDX4_OFFSET]].sub3
+ ; GCN-NEXT: [[COPY8:%[0-9]+]]:vreg_64 = COPY [[COPY6]].sub0_sub1
+ ; GCN-NEXT: [[COPY9:%[0-9]+]]:vgpr_32 = COPY killed [[COPY6]].sub2
+ ; GCN-NEXT: [[COPY10:%[0-9]+]]:vgpr_32 = COPY [[COPY8]].sub0
+ ; GCN-NEXT: [[COPY11:%[0-9]+]]:vgpr_32 = COPY killed [[COPY8]].sub1
+ ; GCN-NEXT: [[BUFFER_LOAD_DWORDX3_OFFSET:%[0-9]+]]:vreg_96 = BUFFER_LOAD_DWORDX3_OFFSET [[REG_SEQUENCE]], 0, 36, 0, 0, implicit $exec :: (dereferenceable load (s96), align 1, addrspace 4)
+ ; GCN-NEXT: [[COPY12:%[0-9]+]]:vreg_64 = COPY [[BUFFER_LOAD_DWORDX3_OFFSET]].sub0_sub1
+ ; GCN-NEXT: [[COPY13:%[0-9]+]]:vgpr_32 = COPY killed [[BUFFER_LOAD_DWORDX3_OFFSET]].sub2
+ ; GCN-NEXT: [[COPY14:%[0-9]+]]:vgpr_32 = COPY [[COPY12]].sub0
+ ; GCN-NEXT: [[COPY15:%[0-9]+]]:vgpr_32 = COPY killed [[COPY12]].sub1
+ %0:sgpr_32 = COPY $sgpr0
+ %1:sgpr_32 = COPY $sgpr1
+ %2:sgpr_32 = COPY $sgpr2
+ %3:sgpr_32 = COPY $sgpr3
+ %5:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3
+ %7:vgpr_32 = BUFFER_LOAD_DWORD_OFFSET %5:sgpr_128, 0, 4, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+ %8:vgpr_32 = BUFFER_LOAD_DWORD_OFFSET %5:sgpr_128, 0, 8, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+ %9:vgpr_32 = BUFFER_LOAD_DWORD_OFFSET %5:sgpr_128, 0, 16, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+ %10:vgpr_32 = BUFFER_LOAD_DWORD_OFFSET %5:sgpr_128, 0, 20, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+ %11:vgpr_32 = BUFFER_LOAD_DWORD_OFFSET %5:sgpr_128, 0, 24, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+ %12:vgpr_32 = BUFFER_LOAD_DWORD_OFFSET %5:sgpr_128, 0, 28, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+ %13:vgpr_32 = BUFFER_LOAD_DWORD_OFFSET %5:sgpr_128, 0, 36, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+ %14:vgpr_32 = BUFFER_LOAD_DWORD_OFFSET %5:sgpr_128, 0, 40, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+ %15:vgpr_32 = BUFFER_LOAD_DWORD_OFFSET %5:sgpr_128, 0, 44, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+...
+---
+
+#
+# buffer_store_dword
+#
+
+name: buffer_store_dword_xyz
+body: |
+ bb.0.entry:
+ liveins: $sgpr0, $sgpr1, $sgpr2, $sgpr3, $vgpr0, $vgpr1, $vgpr2, $vgpr3
+ ; GCN-LABEL: name: buffer_store_dword_xyz
+ ; GCN: liveins: $sgpr0, $sgpr1, $sgpr2, $sgpr3, $vgpr0, $vgpr1, $vgpr2, $vgpr3
+ ; GCN-NEXT: {{ $}}
+ ; GCN-NEXT: [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr3
+ ; GCN-NEXT: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr2
+ ; GCN-NEXT: [[COPY2:%[0-9]+]]:vgpr_32 = COPY $vgpr1
+ ; GCN-NEXT: [[COPY3:%[0-9]+]]:vgpr_32 = COPY $vgpr0
+ ; GCN-NEXT: [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr3
+ ; GCN-NEXT: [[COPY5:%[0-9]+]]:sgpr_32 = COPY $sgpr2
+ ; GCN-NEXT: [[COPY6:%[0-9]+]]:sgpr_32 = COPY $sgpr1
+ ; GCN-NEXT: [[COPY7:%[0-9]+]]:sgpr_32 = COPY $sgpr0
+ ; GCN-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY7]], %subreg.sub0, [[COPY6]], %subreg.sub1, [[COPY5]], %subreg.sub2, [[COPY4]], %subreg.sub3
+ ; GCN-NEXT: [[REG_SEQUENCE1:%[0-9]+]]:vreg_96 = REG_SEQUENCE [[COPY3]], %subreg.sub0, [[COPY2]], %subreg.sub1, [[COPY1]], %subreg.sub2
+ ; GCN-NEXT: [[REG_SEQUENCE2:%[0-9]+]]:vreg_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[REG_SEQUENCE1]], %subreg.sub1_sub2_sub3
+ ; GCN-NEXT: BUFFER_STORE_DWORDX4_OFFSET_exact killed [[REG_SEQUENCE2]], [[REG_SEQUENCE]], 0, 4, 0, 0, implicit $exec :: (dereferenceable store (s128), align 1, addrspace 4)
+ %7:vgpr_32 = COPY $vgpr3
+ %6:vgpr_32 = COPY $vgpr2
+ %5:vgpr_32 = COPY $vgpr1
+ %4:vgpr_32 = COPY $vgpr0
+ %3:sgpr_32 = COPY $sgpr3
+ %2:sgpr_32 = COPY $sgpr2
+ %1:sgpr_32 = COPY $sgpr1
+ %0:sgpr_32 = COPY $sgpr0
+ %13:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3
+ %14:vreg_96 = REG_SEQUENCE %4:vgpr_32, %subreg.sub0, %5:vgpr_32, %subreg.sub1, %6:vgpr_32, %subreg.sub2
+ BUFFER_STORE_DWORD_OFFSET_exact %7:vgpr_32, %13:sgpr_128, 0, 4, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4)
+ BUFFER_STORE_DWORDX3_OFFSET_exact %14:vreg_96, %13:sgpr_128, 0, 8, 0, 0, implicit $exec :: (dereferenceable store (s96), align 1, addrspace 4)
+...
+---
+
+name: buffer_store_dwordx3_dword
+body: |
+ bb.0.entry:
+ liveins: $sgpr0, $sgpr1, $sgpr2, $sgpr3, $vgpr0, $vgpr1, $vgpr2, $vgpr3
+ ; GCN-LABEL: name: buffer_store_dwordx3_dword
+ ; GCN: liveins: $sgpr0, $sgpr1, $sgpr2, $sgpr3, $vgpr0, $vgpr1, $vgpr2, $vgpr3
+ ; GCN-NEXT: {{ $}}
+ ; GCN-NEXT: [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr3
+ ; GCN-NEXT: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr2
+ ; GCN-NEXT: [[COPY2:%[0-9]+]]:vgpr_32 = COPY $vgpr1
+ ; GCN-NEXT: [[COPY3:%[0-9]+]]:vgpr_32 = COPY $vgpr0
+ ; GCN-NEXT: [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr3
+ ; GCN-NEXT: [[COPY5:%[0-9]+]]:sgpr_32 = COPY $sgpr2
+ ; GCN-NEXT: [[COPY6:%[0-9]+]]:sgpr_32 = COPY $sgpr1
+ ; GCN-NEXT: [[COPY7:%[0-9]+]]:sgpr_32 = COPY $sgpr0
+ ; GCN-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY7]], %subreg.sub0, [[COPY6]], %subreg.sub1, [[COPY5]], %subreg.sub2, [[COPY4]], %subreg.sub3
+ ; GCN-NEXT: [[REG_SEQUENCE1:%[0-9]+]]:vreg_96 = REG_SEQUENCE [[COPY3]], %subreg.sub0, [[COPY2]], %subreg.sub1, [[COPY1]], %subreg.sub2
+ ; GCN-NEXT: [[REG_SEQUENCE2:%[0-9]+]]:vreg_128 = REG_SEQUENCE [[REG_SEQUENCE1]], %subreg.sub0_sub1_sub2, [[COPY]], %subreg.sub3
+ ; GCN-NEXT: BUFFER_STORE_DWORDX4_OFFSET_exact killed [[REG_SEQUENCE2]], [[REG_SEQUENCE]], 0, 4, 0, 0, implicit $exec :: (dereferenceable store (s128), align 1, addrspace 4)
+ %7:vgpr_32 = COPY $vgpr3
+ %6:vgpr_32 = COPY $vgpr2
+ %5:vgpr_32 = COPY $vgpr1
+ %4:vgpr_32 = COPY $vgpr0
+ %3:sgpr_32 = COPY $sgpr3
+ %2:sgpr_32 = COPY $sgpr2
+ %1:sgpr_32 = COPY $sgpr1
+ %0:sgpr_32 = COPY $sgpr0
+ %13:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3
+ %14:vreg_96 = REG_SEQUENCE %4:vgpr_32, %subreg.sub0, %5:vgpr_32, %subreg.sub1, %6:vgpr_32, %subreg.sub2
+ BUFFER_STORE_DWORDX3_OFFSET_exact %14:vreg_96, %13:sgpr_128, 0, 4, 0, 0, implicit $exec :: (dereferenceable store (s96), align 1, addrspace 4)
+ BUFFER_STORE_DWORD_OFFSET_exact %7:vgpr_32, %13:sgpr_128, 0, 16, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4)
+...
+---
+
+name: buffer_store_dwordx2_dwordx2
+body: |
+ bb.0.entry:
+ liveins: $sgpr0, $sgpr1, $sgpr2, $sgpr3, $vgpr0, $vgpr1, $vgpr2, $vgpr3
+ ; GCN-LABEL: name: buffer_store_dwordx2_dwordx2
+ ; GCN: liveins: $sgpr0, $sgpr1, $sgpr2, $sgpr3, $vgpr0, $vgpr1, $vgpr2, $vgpr3
+ ; GCN-NEXT: {{ $}}
+ ; GCN-NEXT: [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr3
+ ; GCN-NEXT: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr2
+ ; GCN-NEXT: [[COPY2:%[0-9]+]]:vgpr_32 = COPY $vgpr1
+ ; GCN-NEXT: [[COPY3:%[0-9]+]]:vgpr_32 = COPY $vgpr0
+ ; GCN-NEXT: [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr3
+ ; GCN-NEXT: [[COPY5:%[0-9]+]]:sgpr_32 = COPY $sgpr2
+ ; GCN-NEXT: [[COPY6:%[0-9]+]]:sgpr_32 = COPY $sgpr1
+ ; GCN-NEXT: [[COPY7:%[0-9]+]]:sgpr_32 = COPY $sgpr0
+ ; GCN-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY7]], %subreg.sub0, [[COPY6]], %subreg.sub1, [[COPY5]], %subreg.sub2, [[COPY4]], %subreg.sub3
+ ; GCN-NEXT: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[COPY3]], %subreg.sub0, [[COPY2]], %subreg.sub1
+ ; GCN-NEXT: [[REG_SEQUENCE2:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[COPY1]], %subreg.sub0, [[COPY]], %subreg.sub1
+ ; GCN-NEXT: [[REG_SEQUENCE3:%[0-9]+]]:vreg_128 = REG_SEQUENCE [[REG_SEQUENCE1]], %subreg.sub0_sub1, [[REG_SEQUENCE2]], %subreg.sub2_sub3
+ ; GCN-NEXT: BUFFER_STORE_DWORDX4_OFFSET_exact killed [[REG_SEQUENCE3]], [[REG_SEQUENCE]], 0, 4, 0, 0, implicit $exec :: (dereferenceable store (s128), align 1, addrspace 4)
+ %7:vgpr_32 = COPY $vgpr3
+ %6:vgpr_32 = COPY $vgpr2
+ %5:vgpr_32 = COPY $vgpr1
+ %4:vgpr_32 = COPY $vgpr0
+ %3:sgpr_32 = COPY $sgpr3
+ %2:sgpr_32 = COPY $sgpr2
+ %1:sgpr_32 = COPY $sgpr1
+ %0:sgpr_32 = COPY $sgpr0
+ %13:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3
+ %14:vreg_64 = REG_SEQUENCE %4:vgpr_32, %subreg.sub0, %5:vgpr_32, %subreg.sub1
+ %15:vreg_64 = REG_SEQUENCE %6:vgpr_32, %subreg.sub0, %7:vgpr_32, %subreg.sub1
+ BUFFER_STORE_DWORDX2_OFFSET_exact %14:vreg_64, %13:sgpr_128, 0, 4, 0, 0, implicit $exec :: (dereferenceable store (s64), align 1, addrspace 4)
+ BUFFER_STORE_DWORDX2_OFFSET_exact %15:vreg_64, %13:sgpr_128, 0, 12, 0, 0, implicit $exec :: (dereferenceable store (s64), align 1, addrspace 4)
+...
+---
+
+name: buffer_store_dword_dwordx2
+body: |
+ bb.0.entry:
+ liveins: $sgpr0, $sgpr1, $sgpr2, $sgpr3, $vgpr0, $vgpr1, $vgpr2, $vgpr3
+ ; GCN-LABEL: name: buffer_store_dword_dwordx2
+ ; GCN: liveins: $sgpr0, $sgpr1, $sgpr2, $sgpr3, $vgpr0, $vgpr1, $vgpr2, $vgpr3
+ ; GCN-NEXT: {{ $}}
+ ; GCN-NEXT: [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr3
+ ; GCN-NEXT: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr2
+ ; GCN-NEXT: [[COPY2:%[0-9]+]]:vgpr_32 = COPY $vgpr1
+ ; GCN-NEXT: [[COPY3:%[0-9]+]]:vgpr_32 = COPY $vgpr0
+ ; GCN-NEXT: [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr3
+ ; GCN-NEXT: [[COPY5:%[0-9]+]]:sgpr_32 = COPY $sgpr2
+ ; GCN-NEXT: [[COPY6:%[0-9]+]]:sgpr_32 = COPY $sgpr1
+ ; GCN-NEXT: [[COPY7:%[0-9]+]]:sgpr_32 = COPY $sgpr0
+ ; GCN-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY7]], %subreg.sub0, [[COPY6]], %subreg.sub1, [[COPY5]], %subreg.sub2, [[COPY4]], %subreg.sub3
+ ; GCN-NEXT: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[COPY3]], %subreg.sub0, [[COPY2]], %subreg.sub1
+ ; GCN-NEXT: [[REG_SEQUENCE2:%[0-9]+]]:vreg_96 = REG_SEQUENCE [[COPY]], %subreg.sub0, %10:vreg_64, %subreg.sub1_sub2
+ ; GCN-NEXT: BUFFER_STORE_DWORDX3_OFFSET_exact killed [[REG_SEQUENCE2]], [[REG_SEQUENCE]], 0, 4, 0, 0, implicit $exec :: (dereferenceable store (s96), align 1, addrspace 4)
+ %7:vgpr_32 = COPY $vgpr3
+ %6:vgpr_32 = COPY $vgpr2
+ %5:vgpr_32 = COPY $vgpr1
+ %4:vgpr_32 = COPY $vgpr0
+ %3:sgpr_32 = COPY $sgpr3
+ %2:sgpr_32 = COPY $sgpr2
+ %1:sgpr_32 = COPY $sgpr1
+ %0:sgpr_32 = COPY $sgpr0
+ %13:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3
+ %14:vreg_64 = REG_SEQUENCE %4:vgpr_32, %subreg.sub0, %5:vgpr_32, %subreg.sub1
+ BUFFER_STORE_DWORD_OFFSET_exact %7:vgpr_32, %13:sgpr_128, 0, 4, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4)
+ BUFFER_STORE_DWORDX2_OFFSET_exact %15:vreg_64, %13:sgpr_128, 0, 8, 0, 0, implicit $exec :: (dereferenceable store (s64), align 1, addrspace 4)
+...
+---
+
+name: buffer_store_dwordx2_dword
+body: |
+ bb.0.entry:
+ liveins: $sgpr0, $sgpr1, $sgpr2, $sgpr3, $vgpr0, $vgpr1, $vgpr2, $vgpr3
+ liveins: $sgpr0, $sgpr1, $sgpr2, $sgpr3, $vgpr0, $vgpr1, $vgpr2, $vgpr3
+ ; GCN-LABEL: name: buffer_store_dwordx2_dword
+ ; GCN: liveins: $sgpr0, $sgpr1, $sgpr2, $sgpr3, $vgpr0, $vgpr1, $vgpr2, $vgpr3, $sgpr0, $sgpr1, $sgpr2, $sgpr3, $vgpr0, $vgpr1, $vgpr2, $vgpr3
+ ; GCN-NEXT: {{ $}}
+ ; GCN-NEXT: [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr3
+ ; GCN-NEXT: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr2
+ ; GCN-NEXT: [[COPY2:%[0-9]+]]:vgpr_32 = COPY $vgpr1
+ ; GCN-NEXT: [[COPY3:%[0-9]+]]:vgpr_32 = COPY $vgpr0
+ ; GCN-NEXT: [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr3
+ ; GCN-NEXT: [[COPY5:%[0-9]+]]:sgpr_32 = COPY $sgpr2
+ ; GCN-NEXT: [[COPY6:%[0-9]+]]:sgpr_32 = COPY $sgpr1
+ ; GCN-NEXT: [[COPY7:%[0-9]+]]:sgpr_32 = COPY $sgpr0
+ ; GCN-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY7]], %subreg.sub0, [[COPY6]], %subreg.sub1, [[COPY5]], %subreg.sub2, [[COPY4]], %subreg.sub3
+ ; GCN-NEXT: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[COPY3]], %subreg.sub0, [[COPY2]], %subreg.sub1
+ ; GCN-NEXT: [[REG_SEQUENCE2:%[0-9]+]]:vreg_96 = REG_SEQUENCE [[REG_SEQUENCE1]], %subreg.sub0_sub1, [[COPY]], %subreg.sub2
+ ; GCN-NEXT: BUFFER_STORE_DWORDX3_OFFSET_exact killed [[REG_SEQUENCE2]], [[REG_SEQUENCE]], 0, 4, 0, 0, implicit $exec :: (dereferenceable store (s96), align 1, addrspace 4)
+ %7:vgpr_32 = COPY $vgpr3
+ %6:vgpr_32 = COPY $vgpr2
+ %5:vgpr_32 = COPY $vgpr1
+ %4:vgpr_32 = COPY $vgpr0
+ %3:sgpr_32 = COPY $sgpr3
+ %2:sgpr_32 = COPY $sgpr2
+ %1:sgpr_32 = COPY $sgpr1
+ %0:sgpr_32 = COPY $sgpr0
+ %13:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3
+ %14:vreg_64 = REG_SEQUENCE %4:vgpr_32, %subreg.sub0, %5:vgpr_32, %subreg.sub1
+ BUFFER_STORE_DWORDX2_OFFSET_exact %14:vreg_64, %13:sgpr_128, 0, 4, 0, 0, implicit $exec :: (dereferenceable store (s64), align 1, addrspace 4)
+ BUFFER_STORE_DWORD_OFFSET_exact %7:vgpr_32, %13:sgpr_128, 0, 12, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4)
+...
+---
+
+name: buffer_store_dword_dword
+body: |
+ bb.0.entry:
+ liveins: $sgpr0, $sgpr1, $sgpr2, $sgpr3, $vgpr0, $vgpr1, $vgpr2, $vgpr3
+ ; GCN-LABEL: name: buffer_store_dword_dword
+ ; GCN: liveins: $sgpr0, $sgpr1, $sgpr2, $sgpr3, $vgpr0, $vgpr1, $vgpr2, $vgpr3
+ ; GCN-NEXT: {{ $}}
+ ; GCN-NEXT: [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr3
+ ; GCN-NEXT: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr2
+ ; GCN-NEXT: [[COPY2:%[0-9]+]]:vgpr_32 = COPY $vgpr1
+ ; GCN-NEXT: [[COPY3:%[0-9]+]]:vgpr_32 = COPY $vgpr0
+ ; GCN-NEXT: [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr3
+ ; GCN-NEXT: [[COPY5:%[0-9]+]]:sgpr_32 = COPY $sgpr2
+ ; GCN-NEXT: [[COPY6:%[0-9]+]]:sgpr_32 = COPY $sgpr1
+ ; GCN-NEXT: [[COPY7:%[0-9]+]]:sgpr_32 = COPY $sgpr0
+ ; GCN-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY7]], %subreg.sub0, [[COPY6]], %subreg.sub1, [[COPY5]], %subreg.sub2, [[COPY4]], %subreg.sub3
+ ; GCN-NEXT: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[COPY1]], %subreg.sub0, [[COPY]], %subreg.sub1
+ ; GCN-NEXT: BUFFER_STORE_DWORDX2_OFFSET_exact killed [[REG_SEQUENCE1]], [[REG_SEQUENCE]], 0, 4, 0, 0, implicit $exec :: (dereferenceable store (s64), align 1, addrspace 4)
+ %7:vgpr_32 = COPY $vgpr3
+ %6:vgpr_32 = COPY $vgpr2
+ %5:vgpr_32 = COPY $vgpr1
+ %4:vgpr_32 = COPY $vgpr0
+ %3:sgpr_32 = COPY $sgpr3
+ %2:sgpr_32 = COPY $sgpr2
+ %1:sgpr_32 = COPY $sgpr1
+ %0:sgpr_32 = COPY $sgpr0
+ %13:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3
+ BUFFER_STORE_DWORD_OFFSET_exact %6:vgpr_32, %13:sgpr_128, 0, 4, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4)
+ BUFFER_STORE_DWORD_OFFSET_exact %7:vgpr_32, %13:sgpr_128, 0, 8, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4)
+...
+---
+
+name: buffer_store_dword_32
+body: |
+ bb.0.entry:
+ liveins: $sgpr0, $sgpr1, $sgpr2, $sgpr3, $vgpr0, $vgpr1, $vgpr2, $vgpr3, $vgpr4, $vgpr5, $vgpr6, $vgpr7, $vgpr8
+ ; GCN-LABEL: name: buffer_store_dword_32
+ ; GCN: liveins: $sgpr0, $sgpr1, $sgpr2, $sgpr3, $vgpr0, $vgpr1, $vgpr2, $vgpr3, $vgpr4, $vgpr5, $vgpr6, $vgpr7, $vgpr8
+ ; GCN-NEXT: {{ $}}
+ ; GCN-NEXT: [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr8
+ ; GCN-NEXT: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr7
+ ; GCN-NEXT: [[COPY2:%[0-9]+]]:vgpr_32 = COPY $vgpr6
+ ; GCN-NEXT: [[COPY3:%[0-9]+]]:vgpr_32 = COPY $vgpr5
+ ; GCN-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY $vgpr4
+ ; GCN-NEXT: [[COPY5:%[0-9]+]]:vgpr_32 = COPY $vgpr3
+ ; GCN-NEXT: [[COPY6:%[0-9]+]]:vgpr_32 = COPY $vgpr2
+ ; GCN-NEXT: [[COPY7:%[0-9]+]]:vgpr_32 = COPY $vgpr1
+ ; GCN-NEXT: [[COPY8:%[0-9]+]]:vgpr_32 = COPY $vgpr0
+ ; GCN-NEXT: [[COPY9:%[0-9]+]]:sgpr_32 = COPY $sgpr3
+ ; GCN-NEXT: [[COPY10:%[0-9]+]]:sgpr_32 = COPY $sgpr2
+ ; GCN-NEXT: [[COPY11:%[0-9]+]]:sgpr_32 = COPY $sgpr1
+ ; GCN-NEXT: [[COPY12:%[0-9]+]]:sgpr_32 = COPY $sgpr0
+ ; GCN-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY12]], %subreg.sub0, [[COPY11]], %subreg.sub1, [[COPY10]], %subreg.sub2, [[COPY9]], %subreg.sub3
+ ; GCN-NEXT: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[COPY8]], %subreg.sub0, [[COPY7]], %subreg.sub1
+ ; GCN-NEXT: BUFFER_STORE_DWORDX2_OFFSET_exact killed [[REG_SEQUENCE1]], [[REG_SEQUENCE]], 0, 4, 0, 0, implicit $exec :: (dereferenceable store (s64), align 1, addrspace 4)
+ ; GCN-NEXT: [[REG_SEQUENCE2:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[COPY6]], %subreg.sub0, [[COPY5]], %subreg.sub1
+ ; GCN-NEXT: [[REG_SEQUENCE3:%[0-9]+]]:vreg_96 = REG_SEQUENCE killed [[REG_SEQUENCE2]], %subreg.sub0_sub1, [[COPY4]], %subreg.sub2
+ ; GCN-NEXT: [[REG_SEQUENCE4:%[0-9]+]]:vreg_128 = REG_SEQUENCE killed [[REG_SEQUENCE3]], %subreg.sub0_sub1_sub2, [[COPY3]], %subreg.sub3
+ ; GCN-NEXT: BUFFER_STORE_DWORDX4_OFFSET_exact killed [[REG_SEQUENCE4]], [[REG_SEQUENCE]], 0, 16, 0, 0, implicit $exec :: (dereferenceable store (s128), align 1, addrspace 4)
+ ; GCN-NEXT: [[REG_SEQUENCE5:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[COPY2]], %subreg.sub0, [[COPY1]], %subreg.sub1
+ ; GCN-NEXT: [[REG_SEQUENCE6:%[0-9]+]]:vreg_96 = REG_SEQUENCE killed [[REG_SEQUENCE5]], %subreg.sub0_sub1, [[COPY]], %subreg.sub2
+ ; GCN-NEXT: BUFFER_STORE_DWORDX3_OFFSET_exact killed [[REG_SEQUENCE6]], [[REG_SEQUENCE]], 0, 36, 0, 0, implicit $exec :: (dereferenceable store (s96), align 1, addrspace 4)
+ %12:vgpr_32 = COPY $vgpr8
+ %11:vgpr_32 = COPY $vgpr7
+ %10:vgpr_32 = COPY $vgpr6
+ %9:vgpr_32 = COPY $vgpr5
+ %8:vgpr_32 = COPY $vgpr4
+ %7:vgpr_32 = COPY $vgpr3
+ %6:vgpr_32 = COPY $vgpr2
+ %5:vgpr_32 = COPY $vgpr1
+ %4:vgpr_32 = COPY $vgpr0
+ %3:sgpr_32 = COPY $sgpr3
+ %2:sgpr_32 = COPY $sgpr2
+ %1:sgpr_32 = COPY $sgpr1
+ %0:sgpr_32 = COPY $sgpr0
+ %13:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3
+ BUFFER_STORE_DWORD_OFFSET_exact %4:vgpr_32, %13:sgpr_128, 0, 4, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4)
+ BUFFER_STORE_DWORD_OFFSET_exact %5:vgpr_32, %13:sgpr_128, 0, 8, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4)
+ BUFFER_STORE_DWORD_OFFSET_exact %6:vgpr_32, %13:sgpr_128, 0, 16, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4)
+ BUFFER_STORE_DWORD_OFFSET_exact %7:vgpr_32, %13:sgpr_128, 0, 20, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4)
+ BUFFER_STORE_DWORD_OFFSET_exact %8:vgpr_32, %13:sgpr_128, 0, 24, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4)
+ BUFFER_STORE_DWORD_OFFSET_exact %9:vgpr_32, %13:sgpr_128, 0, 28, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4)
+ BUFFER_STORE_DWORD_OFFSET_exact %10:vgpr_32, %13:sgpr_128, 0, 36, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4)
+ BUFFER_STORE_DWORD_OFFSET_exact %11:vgpr_32, %13:sgpr_128, 0, 40, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4)
+ BUFFER_STORE_DWORD_OFFSET_exact %12:vgpr_32, %13:sgpr_128, 0, 44, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4)
+...
+---
+
+name: buffer_load_dword_not_merged_swizzled_0
+body: |
+ bb.0.entry:
+ ; GCN-LABEL: name: buffer_load_dword_not_merged_swizzled_0
+ ; GCN: [[COPY:%[0-9]+]]:sgpr_32 = COPY $sgpr0
+ ; GCN-NEXT: [[COPY1:%[0-9]+]]:sgpr_32 = COPY $sgpr1
+ ; GCN-NEXT: [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr2
+ ; GCN-NEXT: [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr3
+ ; GCN-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3
+ ; GCN-NEXT: [[BUFFER_LOAD_DWORD_OFFSET:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_DWORD_OFFSET [[REG_SEQUENCE]], 0, 4, 0, 1, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+ ; GCN-NEXT: [[BUFFER_LOAD_DWORD_OFFSET1:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_DWORD_OFFSET [[REG_SEQUENCE]], 0, 8, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+ %0:sgpr_32 = COPY $sgpr0
+ %1:sgpr_32 = COPY $sgpr1
+ %2:sgpr_32 = COPY $sgpr2
+ %3:sgpr_32 = COPY $sgpr3
+ %5:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3
+ %7:vgpr_32 = BUFFER_LOAD_DWORD_OFFSET %5:sgpr_128, 0, 4, 0, 1, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+ %8:vgpr_32 = BUFFER_LOAD_DWORD_OFFSET %5:sgpr_128, 0, 8, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+...
+---
+
+name: buffer_load_dword_not_merged_swizzled_1
+body: |
+ bb.0.entry:
+ ; GCN-LABEL: name: buffer_load_dword_not_merged_swizzled_1
+ ; GCN: [[COPY:%[0-9]+]]:sgpr_32 = COPY $sgpr0
+ ; GCN-NEXT: [[COPY1:%[0-9]+]]:sgpr_32 = COPY $sgpr1
+ ; GCN-NEXT: [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr2
+ ; GCN-NEXT: [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr3
+ ; GCN-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3
+ ; GCN-NEXT: [[BUFFER_LOAD_DWORD_OFFSET:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_DWORD_OFFSET [[REG_SEQUENCE]], 0, 4, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+ ; GCN-NEXT: [[BUFFER_LOAD_DWORD_OFFSET1:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_DWORD_OFFSET [[REG_SEQUENCE]], 0, 8, 0, 1, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+ %0:sgpr_32 = COPY $sgpr0
+ %1:sgpr_32 = COPY $sgpr1
+ %2:sgpr_32 = COPY $sgpr2
+ %3:sgpr_32 = COPY $sgpr3
+ %5:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3
+ %7:vgpr_32 = BUFFER_LOAD_DWORD_OFFSET %5:sgpr_128, 0, 4, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+ %8:vgpr_32 = BUFFER_LOAD_DWORD_OFFSET %5:sgpr_128, 0, 8, 0, 1, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+...
+---
+
+name: buffer_load_dword_merge_across_swizzle
+body: |
+ bb.0.entry:
+ ; GCN-LABEL: name: buffer_load_dword_merge_across_swizzle
+ ; GCN: [[COPY:%[0-9]+]]:sgpr_32 = COPY $sgpr0
+ ; GCN-NEXT: [[COPY1:%[0-9]+]]:sgpr_32 = COPY $sgpr1
+ ; GCN-NEXT: [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr2
+ ; GCN-NEXT: [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr3
+ ; GCN-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3
+ ; GCN-NEXT: [[BUFFER_LOAD_DWORDX2_OFFSET:%[0-9]+]]:vreg_64 = BUFFER_LOAD_DWORDX2_OFFSET [[REG_SEQUENCE]], 0, 4, 0, 0, implicit $exec :: (dereferenceable load (s64), align 1, addrspace 4)
+ ; GCN-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[BUFFER_LOAD_DWORDX2_OFFSET]].sub0
+ ; GCN-NEXT: [[COPY5:%[0-9]+]]:vgpr_32 = COPY killed [[BUFFER_LOAD_DWORDX2_OFFSET]].sub1
+ ; GCN-NEXT: [[BUFFER_LOAD_DWORD_OFFSET:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_DWORD_OFFSET [[REG_SEQUENCE]], 0, 12, 0, 1, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+ %0:sgpr_32 = COPY $sgpr0
+ %1:sgpr_32 = COPY $sgpr1
+ %2:sgpr_32 = COPY $sgpr2
+ %3:sgpr_32 = COPY $sgpr3
+ %4:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3
+ %5:vgpr_32 = BUFFER_LOAD_DWORD_OFFSET %4:sgpr_128, 0, 4, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+ %6:vgpr_32 = BUFFER_LOAD_DWORD_OFFSET %4:sgpr_128, 0, 12, 0, 1, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+ %7:vgpr_32 = BUFFER_LOAD_DWORD_OFFSET %4:sgpr_128, 0, 8, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+...
+---
+
+name: buffer_load_dword_merge_across_swizzled_store
+body: |
+ bb.0.entry:
+ ; GCN-LABEL: name: buffer_load_dword_merge_across_swizzled_store
+ ; GCN: [[COPY:%[0-9]+]]:sgpr_32 = COPY $sgpr0
+ ; GCN-NEXT: [[COPY1:%[0-9]+]]:sgpr_32 = COPY $sgpr1
+ ; GCN-NEXT: [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr2
+ ; GCN-NEXT: [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr3
+ ; GCN-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY $vgpr0
+ ; GCN-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3
+ ; GCN-NEXT: [[BUFFER_LOAD_DWORD_OFFSET:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_DWORD_OFFSET [[REG_SEQUENCE]], 0, 4, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+ ; GCN-NEXT: BUFFER_STORE_DWORD_OFFSET_exact [[COPY4]], [[REG_SEQUENCE]], 0, 6, 0, 1, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4)
+ ; GCN-NEXT: [[BUFFER_LOAD_DWORD_OFFSET1:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_DWORD_OFFSET [[REG_SEQUENCE]], 0, 8, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+ %0:sgpr_32 = COPY $sgpr0
+ %1:sgpr_32 = COPY $sgpr1
+ %2:sgpr_32 = COPY $sgpr2
+ %3:sgpr_32 = COPY $sgpr3
+ %4:vgpr_32 = COPY $vgpr0
+ %5:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3
+ %6:vgpr_32 = BUFFER_LOAD_DWORD_OFFSET %5:sgpr_128, 0, 4, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+ BUFFER_STORE_DWORD_OFFSET_exact %4:vgpr_32, %5:sgpr_128, 0, 6, 0, 1, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4)
+ %7:vgpr_32 = BUFFER_LOAD_DWORD_OFFSET %5:sgpr_128, 0, 8, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+...
+---
+
+name: buffer_load_dword_dword_idxen
+body: |
+ bb.0.entry:
+ ; GCN-LABEL: name: buffer_load_dword_dword_idxen
+ ; GCN: [[COPY:%[0-9]+]]:sgpr_32 = COPY $sgpr0
+ ; GCN-NEXT: [[COPY1:%[0-9]+]]:sgpr_32 = COPY $sgpr1
+ ; GCN-NEXT: [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr2
+ ; GCN-NEXT: [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr3
+ ; GCN-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY $vgpr0
+ ; GCN-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3
+ ; GCN-NEXT: [[BUFFER_LOAD_DWORDX2_IDXEN:%[0-9]+]]:vreg_64 = BUFFER_LOAD_DWORDX2_IDXEN [[COPY4]], [[REG_SEQUENCE]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s64), align 1, addrspace 4)
+ ; GCN-NEXT: [[COPY5:%[0-9]+]]:vgpr_32 = COPY [[BUFFER_LOAD_DWORDX2_IDXEN]].sub0
+ ; GCN-NEXT: [[COPY6:%[0-9]+]]:vgpr_32 = COPY killed [[BUFFER_LOAD_DWORDX2_IDXEN]].sub1
+ %0:sgpr_32 = COPY $sgpr0
+ %1:sgpr_32 = COPY $sgpr1
+ %2:sgpr_32 = COPY $sgpr2
+ %3:sgpr_32 = COPY $sgpr3
+ %4:vgpr_32 = COPY $vgpr0
+ %5:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3
+ %7:vgpr_32 = BUFFER_LOAD_DWORD_IDXEN %4, %5:sgpr_128, 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+ %8:vgpr_32 = BUFFER_LOAD_DWORD_IDXEN %4, %5:sgpr_128, 0, 4, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+...
+---
+
+name: buffer_load_dword_dwordx2_idxen
+body: |
+ bb.0.entry:
+ ; GCN-LABEL: name: buffer_load_dword_dwordx2_idxen
+ ; GCN: [[COPY:%[0-9]+]]:sgpr_32 = COPY $sgpr0
+ ; GCN-NEXT: [[COPY1:%[0-9]+]]:sgpr_32 = COPY $sgpr1
+ ; GCN-NEXT: [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr2
+ ; GCN-NEXT: [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr3
+ ; GCN-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY $vgpr0
+ ; GCN-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3
+ ; GCN-NEXT: [[BUFFER_LOAD_DWORDX3_IDXEN:%[0-9]+]]:vreg_96 = BUFFER_LOAD_DWORDX3_IDXEN [[COPY4]], [[REG_SEQUENCE]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s64), align 1, addrspace 4)
+ ; GCN-NEXT: [[COPY5:%[0-9]+]]:vgpr_32 = COPY [[BUFFER_LOAD_DWORDX3_IDXEN]].sub0
+ ; GCN-NEXT: [[COPY6:%[0-9]+]]:vreg_64 = COPY killed [[BUFFER_LOAD_DWORDX3_IDXEN]].sub1_sub2
+ %0:sgpr_32 = COPY $sgpr0
+ %1:sgpr_32 = COPY $sgpr1
+ %2:sgpr_32 = COPY $sgpr2
+ %3:sgpr_32 = COPY $sgpr3
+ %4:vgpr_32 = COPY $vgpr0
+ %5:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3
+ %7:vgpr_32 = BUFFER_LOAD_DWORD_IDXEN %4, %5:sgpr_128, 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+ %8:vreg_64 = BUFFER_LOAD_DWORDX2_IDXEN %4, %5:sgpr_128, 0, 4, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+...
+---
+
+name: buffer_load_dwordx2_dwordx2_idxen
+body: |
+ bb.0.entry:
+ ; GCN-LABEL: name: buffer_load_dwordx2_dwordx2_idxen
+ ; GCN: [[COPY:%[0-9]+]]:sgpr_32 = COPY $sgpr0
+ ; GCN-NEXT: [[COPY1:%[0-9]+]]:sgpr_32 = COPY $sgpr1
+ ; GCN-NEXT: [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr2
+ ; GCN-NEXT: [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr3
+ ; GCN-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY $vgpr0
+ ; GCN-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3
+ ; GCN-NEXT: [[BUFFER_LOAD_DWORDX4_IDXEN:%[0-9]+]]:vreg_128 = BUFFER_LOAD_DWORDX4_IDXEN [[COPY4]], [[REG_SEQUENCE]], 0, 4, 0, 0, implicit $exec :: (dereferenceable load (s64), align 1, addrspace 4)
+ ; GCN-NEXT: [[COPY5:%[0-9]+]]:vreg_64 = COPY [[BUFFER_LOAD_DWORDX4_IDXEN]].sub0_sub1
+ ; GCN-NEXT: [[COPY6:%[0-9]+]]:vreg_64 = COPY killed [[BUFFER_LOAD_DWORDX4_IDXEN]].sub2_sub3
+ %0:sgpr_32 = COPY $sgpr0
+ %1:sgpr_32 = COPY $sgpr1
+ %2:sgpr_32 = COPY $sgpr2
+ %3:sgpr_32 = COPY $sgpr3
+ %4:vgpr_32 = COPY $vgpr0
+ %5:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3
+ %7:vreg_64 = BUFFER_LOAD_DWORDX2_IDXEN %4, %5:sgpr_128, 0, 4, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+ %8:vreg_64 = BUFFER_LOAD_DWORDX2_IDXEN %4, %5:sgpr_128, 0, 12, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+...
+---
+
+name: buffer_load_dword_dwordx3_idxen
+body: |
+ bb.0.entry:
+ ; GCN-LABEL: name: buffer_load_dword_dwordx3_idxen
+ ; GCN: [[COPY:%[0-9]+]]:sgpr_32 = COPY $sgpr0
+ ; GCN-NEXT: [[COPY1:%[0-9]+]]:sgpr_32 = COPY $sgpr1
+ ; GCN-NEXT: [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr2
+ ; GCN-NEXT: [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr3
+ ; GCN-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY $vgpr0
+ ; GCN-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3
+ ; GCN-NEXT: [[BUFFER_LOAD_DWORDX4_IDXEN:%[0-9]+]]:vreg_128 = BUFFER_LOAD_DWORDX4_IDXEN [[COPY4]], [[REG_SEQUENCE]], 0, 4, 0, 0, implicit $exec :: (dereferenceable load (s64), align 1, addrspace 4)
+ ; GCN-NEXT: [[COPY5:%[0-9]+]]:vgpr_32 = COPY [[BUFFER_LOAD_DWORDX4_IDXEN]].sub0
+ ; GCN-NEXT: [[COPY6:%[0-9]+]]:vreg_96 = COPY killed [[BUFFER_LOAD_DWORDX4_IDXEN]].sub1_sub2_sub3
+ %0:sgpr_32 = COPY $sgpr0
+ %1:sgpr_32 = COPY $sgpr1
+ %2:sgpr_32 = COPY $sgpr2
+ %3:sgpr_32 = COPY $sgpr3
+ %4:vgpr_32 = COPY $vgpr0
+ %5:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3
+ %7:vgpr_32 = BUFFER_LOAD_DWORD_IDXEN %4, %5:sgpr_128, 0, 4, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+ %8:vreg_96 = BUFFER_LOAD_DWORDX3_IDXEN %4, %5:sgpr_128, 0, 8, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+...
+---
+
+name: buffer_load_dword_dword_bothen
+body: |
+ bb.0.entry:
+ ; GCN-LABEL: name: buffer_load_dword_dword_bothen
+ ; GCN: [[COPY:%[0-9]+]]:sgpr_32 = COPY $sgpr0
+ ; GCN-NEXT: [[COPY1:%[0-9]+]]:sgpr_32 = COPY $sgpr1
+ ; GCN-NEXT: [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr2
+ ; GCN-NEXT: [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr3
+ ; GCN-NEXT: [[COPY4:%[0-9]+]]:vreg_64 = COPY $vgpr0
+ ; GCN-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3
+ ; GCN-NEXT: [[BUFFER_LOAD_DWORDX2_BOTHEN:%[0-9]+]]:vreg_64 = BUFFER_LOAD_DWORDX2_BOTHEN [[COPY4]], [[REG_SEQUENCE]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s64), align 1, addrspace 4)
+ ; GCN-NEXT: [[COPY5:%[0-9]+]]:vgpr_32 = COPY [[BUFFER_LOAD_DWORDX2_BOTHEN]].sub0
+ ; GCN-NEXT: [[COPY6:%[0-9]+]]:vgpr_32 = COPY killed [[BUFFER_LOAD_DWORDX2_BOTHEN]].sub1
+ %0:sgpr_32 = COPY $sgpr0
+ %1:sgpr_32 = COPY $sgpr1
+ %2:sgpr_32 = COPY $sgpr2
+ %3:sgpr_32 = COPY $sgpr3
+ %4:vreg_64 = COPY $vgpr0
+ %5:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3
+ %7:vgpr_32 = BUFFER_LOAD_DWORD_BOTHEN %4, %5:sgpr_128, 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+ %8:vgpr_32 = BUFFER_LOAD_DWORD_BOTHEN %4, %5:sgpr_128, 0, 4, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+...
+---
+
+name: buffer_load_dword_dwordx2_bothen
+body: |
+ bb.0.entry:
+ ; GCN-LABEL: name: buffer_load_dword_dwordx2_bothen
+ ; GCN: [[COPY:%[0-9]+]]:sgpr_32 = COPY $sgpr0
+ ; GCN-NEXT: [[COPY1:%[0-9]+]]:sgpr_32 = COPY $sgpr1
+ ; GCN-NEXT: [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr2
+ ; GCN-NEXT: [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr3
+ ; GCN-NEXT: [[COPY4:%[0-9]+]]:vreg_64 = COPY $vgpr0
+ ; GCN-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3
+ ; GCN-NEXT: [[BUFFER_LOAD_DWORDX3_BOTHEN:%[0-9]+]]:vreg_96 = BUFFER_LOAD_DWORDX3_BOTHEN [[COPY4]], [[REG_SEQUENCE]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s64), align 1, addrspace 4)
+ ; GCN-NEXT: [[COPY5:%[0-9]+]]:vgpr_32 = COPY [[BUFFER_LOAD_DWORDX3_BOTHEN]].sub0
+ ; GCN-NEXT: [[COPY6:%[0-9]+]]:vreg_64 = COPY killed [[BUFFER_LOAD_DWORDX3_BOTHEN]].sub1_sub2
+ %0:sgpr_32 = COPY $sgpr0
+ %1:sgpr_32 = COPY $sgpr1
+ %2:sgpr_32 = COPY $sgpr2
+ %3:sgpr_32 = COPY $sgpr3
+ %4:vreg_64 = COPY $vgpr0
+ %5:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3
+ %7:vgpr_32 = BUFFER_LOAD_DWORD_BOTHEN %4, %5:sgpr_128, 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+ %8:vreg_64 = BUFFER_LOAD_DWORDX2_BOTHEN %4, %5:sgpr_128, 0, 4, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+...
+---
+
+name: buffer_load_dwordx2_dwordx2_bothen
+body: |
+ bb.0.entry:
+ ; GCN-LABEL: name: buffer_load_dwordx2_dwordx2_bothen
+ ; GCN: [[COPY:%[0-9]+]]:sgpr_32 = COPY $sgpr0
+ ; GCN-NEXT: [[COPY1:%[0-9]+]]:sgpr_32 = COPY $sgpr1
+ ; GCN-NEXT: [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr2
+ ; GCN-NEXT: [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr3
+ ; GCN-NEXT: [[COPY4:%[0-9]+]]:vreg_64 = COPY $vgpr0
+ ; GCN-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3
+ ; GCN-NEXT: [[BUFFER_LOAD_DWORDX4_BOTHEN:%[0-9]+]]:vreg_128 = BUFFER_LOAD_DWORDX4_BOTHEN [[COPY4]], [[REG_SEQUENCE]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s64), align 1, addrspace 4)
+ ; GCN-NEXT: [[COPY5:%[0-9]+]]:vreg_64 = COPY [[BUFFER_LOAD_DWORDX4_BOTHEN]].sub0_sub1
+ ; GCN-NEXT: [[COPY6:%[0-9]+]]:vreg_64 = COPY killed [[BUFFER_LOAD_DWORDX4_BOTHEN]].sub2_sub3
+ %0:sgpr_32 = COPY $sgpr0
+ %1:sgpr_32 = COPY $sgpr1
+ %2:sgpr_32 = COPY $sgpr2
+ %3:sgpr_32 = COPY $sgpr3
+ %4:vreg_64 = COPY $vgpr0
+ %5:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3
+ %7:vreg_64 = BUFFER_LOAD_DWORDX2_BOTHEN %4, %5:sgpr_128, 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+ %8:vreg_64 = BUFFER_LOAD_DWORDX2_BOTHEN %4, %5:sgpr_128, 0, 8, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+...
+---
+
+name: buffer_load_dword_dwordx3_bothen
+body: |
+ bb.0.entry:
+ ; GCN-LABEL: name: buffer_load_dword_dwordx3_bothen
+ ; GCN: [[COPY:%[0-9]+]]:sgpr_32 = COPY $sgpr0
+ ; GCN-NEXT: [[COPY1:%[0-9]+]]:sgpr_32 = COPY $sgpr1
+ ; GCN-NEXT: [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr2
+ ; GCN-NEXT: [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr3
+ ; GCN-NEXT: [[COPY4:%[0-9]+]]:vreg_64 = COPY $vgpr0
+ ; GCN-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3
+ ; GCN-NEXT: [[BUFFER_LOAD_DWORDX4_BOTHEN:%[0-9]+]]:vreg_128 = BUFFER_LOAD_DWORDX4_BOTHEN [[COPY4]], [[REG_SEQUENCE]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s64), align 1, addrspace 4)
+ ; GCN-NEXT: [[COPY5:%[0-9]+]]:vgpr_32 = COPY [[BUFFER_LOAD_DWORDX4_BOTHEN]].sub0
+ ; GCN-NEXT: [[COPY6:%[0-9]+]]:vreg_96 = COPY killed [[BUFFER_LOAD_DWORDX4_BOTHEN]].sub1_sub2_sub3
+ %0:sgpr_32 = COPY $sgpr0
+ %1:sgpr_32 = COPY $sgpr1
+ %2:sgpr_32 = COPY $sgpr2
+ %3:sgpr_32 = COPY $sgpr3
+ %4:vreg_64 = COPY $vgpr0
+ %5:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3
+ %7:vgpr_32 = BUFFER_LOAD_DWORD_BOTHEN %4, %5:sgpr_128, 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+ %8:vreg_96 = BUFFER_LOAD_DWORDX3_BOTHEN %4, %5:sgpr_128, 0, 4, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+...
+---
+
+name: buffer_load_dword_dword_idxen_exact
+body: |
+ bb.0.entry:
+ ; GCN-LABEL: name: buffer_load_dword_dword_idxen_exact
+ ; GCN: [[COPY:%[0-9]+]]:sgpr_32 = COPY $sgpr0
+ ; GCN-NEXT: [[COPY1:%[0-9]+]]:sgpr_32 = COPY $sgpr1
+ ; GCN-NEXT: [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr2
+ ; GCN-NEXT: [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr3
+ ; GCN-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY $vgpr0
+ ; GCN-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3
+ ; GCN-NEXT: [[BUFFER_LOAD_DWORDX2_IDXEN_exact:%[0-9]+]]:vreg_64 = BUFFER_LOAD_DWORDX2_IDXEN_exact [[COPY4]], [[REG_SEQUENCE]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s64), align 1, addrspace 4)
+ ; GCN-NEXT: [[COPY5:%[0-9]+]]:vgpr_32 = COPY [[BUFFER_LOAD_DWORDX2_IDXEN_exact]].sub0
+ ; GCN-NEXT: [[COPY6:%[0-9]+]]:vgpr_32 = COPY killed [[BUFFER_LOAD_DWORDX2_IDXEN_exact]].sub1
+ %0:sgpr_32 = COPY $sgpr0
+ %1:sgpr_32 = COPY $sgpr1
+ %2:sgpr_32 = COPY $sgpr2
+ %3:sgpr_32 = COPY $sgpr3
+ %4:vgpr_32 = COPY $vgpr0
+ %5:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3
+ %7:vgpr_32 = BUFFER_LOAD_DWORD_IDXEN_exact %4, %5:sgpr_128, 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+ %8:vgpr_32 = BUFFER_LOAD_DWORD_IDXEN_exact %4, %5:sgpr_128, 0, 4, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+...
+---
+
+name: buffer_load_dword_dwordx2_idxen_exact
+body: |
+ bb.0.entry:
+ ; GCN-LABEL: name: buffer_load_dword_dwordx2_idxen_exact
+ ; GCN: [[COPY:%[0-9]+]]:sgpr_32 = COPY $sgpr0
+ ; GCN-NEXT: [[COPY1:%[0-9]+]]:sgpr_32 = COPY $sgpr1
+ ; GCN-NEXT: [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr2
+ ; GCN-NEXT: [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr3
+ ; GCN-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY $vgpr0
+ ; GCN-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3
+ ; GCN-NEXT: [[BUFFER_LOAD_DWORDX3_IDXEN_exact:%[0-9]+]]:vreg_96 = BUFFER_LOAD_DWORDX3_IDXEN_exact [[COPY4]], [[REG_SEQUENCE]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s64), align 1, addrspace 4)
+ ; GCN-NEXT: [[COPY5:%[0-9]+]]:vgpr_32 = COPY [[BUFFER_LOAD_DWORDX3_IDXEN_exact]].sub0
+ ; GCN-NEXT: [[COPY6:%[0-9]+]]:vreg_64 = COPY killed [[BUFFER_LOAD_DWORDX3_IDXEN_exact]].sub1_sub2
+ %0:sgpr_32 = COPY $sgpr0
+ %1:sgpr_32 = COPY $sgpr1
+ %2:sgpr_32 = COPY $sgpr2
+ %3:sgpr_32 = COPY $sgpr3
+ %4:vgpr_32 = COPY $vgpr0
+ %5:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3
+ %7:vgpr_32 = BUFFER_LOAD_DWORD_IDXEN_exact %4, %5:sgpr_128, 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+ %8:vreg_64 = BUFFER_LOAD_DWORDX2_IDXEN_exact %4, %5:sgpr_128, 0, 4, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+...
+---
+
+name: buffer_load_dwordx2_dwordx2_idxen_exact
+body: |
+ bb.0.entry:
+ ; GCN-LABEL: name: buffer_load_dwordx2_dwordx2_idxen_exact
+ ; GCN: [[COPY:%[0-9]+]]:sgpr_32 = COPY $sgpr0
+ ; GCN-NEXT: [[COPY1:%[0-9]+]]:sgpr_32 = COPY $sgpr1
+ ; GCN-NEXT: [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr2
+ ; GCN-NEXT: [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr3
+ ; GCN-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY $vgpr0
+ ; GCN-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3
+ ; GCN-NEXT: [[BUFFER_LOAD_DWORDX4_IDXEN_exact:%[0-9]+]]:vreg_128 = BUFFER_LOAD_DWORDX4_IDXEN_exact [[COPY4]], [[REG_SEQUENCE]], 0, 4, 0, 0, implicit $exec :: (dereferenceable load (s64), align 1, addrspace 4)
+ ; GCN-NEXT: [[COPY5:%[0-9]+]]:vreg_64 = COPY [[BUFFER_LOAD_DWORDX4_IDXEN_exact]].sub0_sub1
+ ; GCN-NEXT: [[COPY6:%[0-9]+]]:vreg_64 = COPY killed [[BUFFER_LOAD_DWORDX4_IDXEN_exact]].sub2_sub3
+ %0:sgpr_32 = COPY $sgpr0
+ %1:sgpr_32 = COPY $sgpr1
+ %2:sgpr_32 = COPY $sgpr2
+ %3:sgpr_32 = COPY $sgpr3
+ %4:vgpr_32 = COPY $vgpr0
+ %5:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3
+ %7:vreg_64 = BUFFER_LOAD_DWORDX2_IDXEN_exact %4, %5:sgpr_128, 0, 4, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+ %8:vreg_64 = BUFFER_LOAD_DWORDX2_IDXEN_exact %4, %5:sgpr_128, 0, 12, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+...
+---
+
+name: buffer_load_dword_dwordx3_idxen_exact
+body: |
+ bb.0.entry:
+ ; GCN-LABEL: name: buffer_load_dword_dwordx3_idxen_exact
+ ; GCN: [[COPY:%[0-9]+]]:sgpr_32 = COPY $sgpr0
+ ; GCN-NEXT: [[COPY1:%[0-9]+]]:sgpr_32 = COPY $sgpr1
+ ; GCN-NEXT: [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr2
+ ; GCN-NEXT: [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr3
+ ; GCN-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY $vgpr0
+ ; GCN-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3
+ ; GCN-NEXT: [[BUFFER_LOAD_DWORDX4_IDXEN_exact:%[0-9]+]]:vreg_128 = BUFFER_LOAD_DWORDX4_IDXEN_exact [[COPY4]], [[REG_SEQUENCE]], 0, 4, 0, 0, implicit $exec :: (dereferenceable load (s64), align 1, addrspace 4)
+ ; GCN-NEXT: [[COPY5:%[0-9]+]]:vgpr_32 = COPY [[BUFFER_LOAD_DWORDX4_IDXEN_exact]].sub0
+ ; GCN-NEXT: [[COPY6:%[0-9]+]]:vreg_96 = COPY killed [[BUFFER_LOAD_DWORDX4_IDXEN_exact]].sub1_sub2_sub3
+ %0:sgpr_32 = COPY $sgpr0
+ %1:sgpr_32 = COPY $sgpr1
+ %2:sgpr_32 = COPY $sgpr2
+ %3:sgpr_32 = COPY $sgpr3
+ %4:vgpr_32 = COPY $vgpr0
+ %5:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3
+ %7:vgpr_32 = BUFFER_LOAD_DWORD_IDXEN_exact %4, %5:sgpr_128, 0, 4, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+ %8:vreg_96 = BUFFER_LOAD_DWORDX3_IDXEN_exact %4, %5:sgpr_128, 0, 8, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+...
+---
+
+name: buffer_load_dword_dword_dword_idxen_exact
+body: |
+ bb.0.entry:
+ ; GCN-LABEL: name: buffer_load_dword_dword_dword_idxen_exact
+ ; GCN: [[COPY:%[0-9]+]]:sgpr_32 = COPY $sgpr0
+ ; GCN-NEXT: [[COPY1:%[0-9]+]]:sgpr_32 = COPY $sgpr1
+ ; GCN-NEXT: [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr2
+ ; GCN-NEXT: [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr3
+ ; GCN-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY $vgpr0
+ ; GCN-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3
+ ; GCN-NEXT: [[BUFFER_LOAD_DWORDX3_IDXEN_exact:%[0-9]+]]:vreg_96 = BUFFER_LOAD_DWORDX3_IDXEN_exact [[COPY4]], [[REG_SEQUENCE]], 0, 4, 0, 0, implicit $exec :: (dereferenceable load (s96), align 1, addrspace 4)
+ ; GCN-NEXT: [[COPY5:%[0-9]+]]:vreg_64 = COPY [[BUFFER_LOAD_DWORDX3_IDXEN_exact]].sub0_sub1
+ ; GCN-NEXT: [[COPY6:%[0-9]+]]:vgpr_32 = COPY killed [[BUFFER_LOAD_DWORDX3_IDXEN_exact]].sub2
+ ; GCN-NEXT: [[COPY7:%[0-9]+]]:vgpr_32 = COPY [[COPY5]].sub0
+ ; GCN-NEXT: [[COPY8:%[0-9]+]]:vgpr_32 = COPY killed [[COPY5]].sub1
+ %0:sgpr_32 = COPY $sgpr0
+ %1:sgpr_32 = COPY $sgpr1
+ %2:sgpr_32 = COPY $sgpr2
+ %3:sgpr_32 = COPY $sgpr3
+ %4:vgpr_32 = COPY $vgpr0
+ %5:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3
+ %7:vgpr_32 = BUFFER_LOAD_DWORD_IDXEN_exact %4, %5:sgpr_128, 0, 4, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+ %8:vgpr_32 = BUFFER_LOAD_DWORD_IDXEN_exact %4, %5:sgpr_128, 0, 8, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+ %9:vgpr_32 = BUFFER_LOAD_DWORD_IDXEN_exact %4, %5:sgpr_128, 0, 12, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+...
+---
+
+name: buffer_load_dword_dword_dword_idxen_exact_swizzled_0
+body: |
+ bb.0.entry:
+ ; GCN-LABEL: name: buffer_load_dword_dword_dword_idxen_exact_swizzled_0
+ ; GCN: [[COPY:%[0-9]+]]:sgpr_32 = COPY $sgpr0
+ ; GCN-NEXT: [[COPY1:%[0-9]+]]:sgpr_32 = COPY $sgpr1
+ ; GCN-NEXT: [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr2
+ ; GCN-NEXT: [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr3
+ ; GCN-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY $vgpr0
+ ; GCN-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3
+ ; GCN-NEXT: [[BUFFER_LOAD_DWORD_IDXEN_exact:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_DWORD_IDXEN_exact [[COPY4]], [[REG_SEQUENCE]], 0, 4, 0, 1, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+ ; GCN-NEXT: [[BUFFER_LOAD_DWORDX2_IDXEN_exact:%[0-9]+]]:vreg_64 = BUFFER_LOAD_DWORDX2_IDXEN_exact [[COPY4]], [[REG_SEQUENCE]], 0, 8, 0, 0, implicit $exec :: (dereferenceable load (s64), align 1, addrspace 4)
+ ; GCN-NEXT: [[COPY5:%[0-9]+]]:vgpr_32 = COPY [[BUFFER_LOAD_DWORDX2_IDXEN_exact]].sub0
+ ; GCN-NEXT: [[COPY6:%[0-9]+]]:vgpr_32 = COPY killed [[BUFFER_LOAD_DWORDX2_IDXEN_exact]].sub1
+ %0:sgpr_32 = COPY $sgpr0
+ %1:sgpr_32 = COPY $sgpr1
+ %2:sgpr_32 = COPY $sgpr2
+ %3:sgpr_32 = COPY $sgpr3
+ %4:vgpr_32 = COPY $vgpr0
+ %5:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3
+ %7:vgpr_32 = BUFFER_LOAD_DWORD_IDXEN_exact %4, %5:sgpr_128, 0, 4, 0, 1, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+ %8:vgpr_32 = BUFFER_LOAD_DWORD_IDXEN_exact %4, %5:sgpr_128, 0, 8, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+ %9:vgpr_32 = BUFFER_LOAD_DWORD_IDXEN_exact %4, %5:sgpr_128, 0, 12, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+...
+---
+
+name: buffer_load_dword_dword_bothen_exact
+body: |
+ bb.0.entry:
+ ; GCN-LABEL: name: buffer_load_dword_dword_bothen_exact
+ ; GCN: [[COPY:%[0-9]+]]:sgpr_32 = COPY $sgpr0
+ ; GCN-NEXT: [[COPY1:%[0-9]+]]:sgpr_32 = COPY $sgpr1
+ ; GCN-NEXT: [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr2
+ ; GCN-NEXT: [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr3
+ ; GCN-NEXT: [[COPY4:%[0-9]+]]:vreg_64 = COPY $vgpr0
+ ; GCN-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3
+ ; GCN-NEXT: [[BUFFER_LOAD_DWORDX2_BOTHEN_exact:%[0-9]+]]:vreg_64 = BUFFER_LOAD_DWORDX2_BOTHEN_exact [[COPY4]], [[REG_SEQUENCE]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s64), align 1, addrspace 4)
+ ; GCN-NEXT: [[COPY5:%[0-9]+]]:vgpr_32 = COPY [[BUFFER_LOAD_DWORDX2_BOTHEN_exact]].sub0
+ ; GCN-NEXT: [[COPY6:%[0-9]+]]:vgpr_32 = COPY killed [[BUFFER_LOAD_DWORDX2_BOTHEN_exact]].sub1
+ %0:sgpr_32 = COPY $sgpr0
+ %1:sgpr_32 = COPY $sgpr1
+ %2:sgpr_32 = COPY $sgpr2
+ %3:sgpr_32 = COPY $sgpr3
+ %4:vreg_64 = COPY $vgpr0
+ %5:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3
+ %7:vgpr_32 = BUFFER_LOAD_DWORD_BOTHEN_exact %4, %5:sgpr_128, 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+ %8:vgpr_32 = BUFFER_LOAD_DWORD_BOTHEN_exact %4, %5:sgpr_128, 0, 4, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+...
+---
+
+name: buffer_load_dword_dwordx2_bothen_exact
+body: |
+ bb.0.entry:
+ ; GCN-LABEL: name: buffer_load_dword_dwordx2_bothen_exact
+ ; GCN: [[COPY:%[0-9]+]]:sgpr_32 = COPY $sgpr0
+ ; GCN-NEXT: [[COPY1:%[0-9]+]]:sgpr_32 = COPY $sgpr1
+ ; GCN-NEXT: [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr2
+ ; GCN-NEXT: [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr3
+ ; GCN-NEXT: [[COPY4:%[0-9]+]]:vreg_64 = COPY $vgpr0
+ ; GCN-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3
+ ; GCN-NEXT: [[BUFFER_LOAD_DWORDX3_BOTHEN_exact:%[0-9]+]]:vreg_96 = BUFFER_LOAD_DWORDX3_BOTHEN_exact [[COPY4]], [[REG_SEQUENCE]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s64), align 1, addrspace 4)
+ ; GCN-NEXT: [[COPY5:%[0-9]+]]:vgpr_32 = COPY [[BUFFER_LOAD_DWORDX3_BOTHEN_exact]].sub0
+ ; GCN-NEXT: [[COPY6:%[0-9]+]]:vreg_64 = COPY killed [[BUFFER_LOAD_DWORDX3_BOTHEN_exact]].sub1_sub2
+ %0:sgpr_32 = COPY $sgpr0
+ %1:sgpr_32 = COPY $sgpr1
+ %2:sgpr_32 = COPY $sgpr2
+ %3:sgpr_32 = COPY $sgpr3
+ %4:vreg_64 = COPY $vgpr0
+ %5:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3
+ %7:vgpr_32 = BUFFER_LOAD_DWORD_BOTHEN_exact %4, %5:sgpr_128, 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+ %8:vreg_64 = BUFFER_LOAD_DWORDX2_BOTHEN_exact %4, %5:sgpr_128, 0, 4, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+...
+---
+
+name: buffer_load_dwordx2_dwordx2_bothen_exact
+body: |
+ bb.0.entry:
+
+ ; GCN-LABEL: name: buffer_load_dwordx2_dwordx2_bothen_exact
+ ; GCN: [[COPY:%[0-9]+]]:sgpr_32 = COPY $sgpr0
+ ; GCN-NEXT: [[COPY1:%[0-9]+]]:sgpr_32 = COPY $sgpr1
+ ; GCN-NEXT: [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr2
+ ; GCN-NEXT: [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr3
+ ; GCN-NEXT: [[COPY4:%[0-9]+]]:vreg_64 = COPY $vgpr0
+ ; GCN-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3
+ ; GCN-NEXT: [[BUFFER_LOAD_DWORDX4_BOTHEN_exact:%[0-9]+]]:vreg_128 = BUFFER_LOAD_DWORDX4_BOTHEN_exact [[COPY4]], [[REG_SEQUENCE]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s64), align 1, addrspace 4)
+ ; GCN-NEXT: [[COPY5:%[0-9]+]]:vreg_64 = COPY [[BUFFER_LOAD_DWORDX4_BOTHEN_exact]].sub0_sub1
+ ; GCN-NEXT: [[COPY6:%[0-9]+]]:vreg_64 = COPY killed [[BUFFER_LOAD_DWORDX4_BOTHEN_exact]].sub2_sub3
+ %0:sgpr_32 = COPY $sgpr0
+ %1:sgpr_32 = COPY $sgpr1
+ %2:sgpr_32 = COPY $sgpr2
+ %3:sgpr_32 = COPY $sgpr3
+ %4:vreg_64 = COPY $vgpr0
+ %5:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3
+ %7:vreg_64 = BUFFER_LOAD_DWORDX2_BOTHEN_exact %4, %5:sgpr_128, 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+ %8:vreg_64 = BUFFER_LOAD_DWORDX2_BOTHEN_exact %4, %5:sgpr_128, 0, 8, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+...
+---
+
+name: buffer_load_dword_dwordx3_bothen_exact
+body: |
+ bb.0.entry:
+
+ ; GCN-LABEL: name: buffer_load_dword_dwordx3_bothen_exact
+ ; GCN: [[COPY:%[0-9]+]]:sgpr_32 = COPY $sgpr0
+ ; GCN-NEXT: [[COPY1:%[0-9]+]]:sgpr_32 = COPY $sgpr1
+ ; GCN-NEXT: [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr2
+ ; GCN-NEXT: [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr3
+ ; GCN-NEXT: [[COPY4:%[0-9]+]]:vreg_64 = COPY $vgpr0
+ ; GCN-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3
+ ; GCN-NEXT: [[BUFFER_LOAD_DWORDX4_BOTHEN_exact:%[0-9]+]]:vreg_128 = BUFFER_LOAD_DWORDX4_BOTHEN_exact [[COPY4]], [[REG_SEQUENCE]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s64), align 1, addrspace 4)
+ ; GCN-NEXT: [[COPY5:%[0-9]+]]:vgpr_32 = COPY [[BUFFER_LOAD_DWORDX4_BOTHEN_exact]].sub0
+ ; GCN-NEXT: [[COPY6:%[0-9]+]]:vreg_96 = COPY killed [[BUFFER_LOAD_DWORDX4_BOTHEN_exact]].sub1_sub2_sub3
+ %0:sgpr_32 = COPY $sgpr0
+ %1:sgpr_32 = COPY $sgpr1
+ %2:sgpr_32 = COPY $sgpr2
+ %3:sgpr_32 = COPY $sgpr3
+ %4:vreg_64 = COPY $vgpr0
+ %5:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3
+ %7:vgpr_32 = BUFFER_LOAD_DWORD_BOTHEN_exact %4, %5:sgpr_128, 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+ %8:vreg_96 = BUFFER_LOAD_DWORDX3_BOTHEN_exact %4, %5:sgpr_128, 0, 4, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+...
+---
+
+name: buffer_load_dword_dword_dword_bothen_exact
+body: |
+ bb.0.entry:
+ ; GCN-LABEL: name: buffer_load_dword_dword_dword_bothen_exact
+ ; GCN: [[COPY:%[0-9]+]]:sgpr_32 = COPY $sgpr0
+ ; GCN-NEXT: [[COPY1:%[0-9]+]]:sgpr_32 = COPY $sgpr1
+ ; GCN-NEXT: [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr2
+ ; GCN-NEXT: [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr3
+ ; GCN-NEXT: [[COPY4:%[0-9]+]]:vreg_64 = COPY $vgpr0
+ ; GCN-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3
+ ; GCN-NEXT: [[BUFFER_LOAD_DWORDX3_BOTHEN_exact:%[0-9]+]]:vreg_96 = BUFFER_LOAD_DWORDX3_BOTHEN_exact [[COPY4]], [[REG_SEQUENCE]], 0, 4, 0, 0, implicit $exec :: (dereferenceable load (s96), align 1, addrspace 4)
+ ; GCN-NEXT: [[COPY5:%[0-9]+]]:vreg_64 = COPY [[BUFFER_LOAD_DWORDX3_BOTHEN_exact]].sub0_sub1
+ ; GCN-NEXT: [[COPY6:%[0-9]+]]:vgpr_32 = COPY killed [[BUFFER_LOAD_DWORDX3_BOTHEN_exact]].sub2
+ ; GCN-NEXT: [[COPY7:%[0-9]+]]:vgpr_32 = COPY [[COPY5]].sub0
+ ; GCN-NEXT: [[COPY8:%[0-9]+]]:vgpr_32 = COPY killed [[COPY5]].sub1
+ %0:sgpr_32 = COPY $sgpr0
+ %1:sgpr_32 = COPY $sgpr1
+ %2:sgpr_32 = COPY $sgpr2
+ %3:sgpr_32 = COPY $sgpr3
+ %4:vreg_64 = COPY $vgpr0
+ %5:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3
+ %7:vgpr_32 = BUFFER_LOAD_DWORD_BOTHEN_exact %4, %5:sgpr_128, 0, 4, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+ %8:vgpr_32 = BUFFER_LOAD_DWORD_BOTHEN_exact %4, %5:sgpr_128, 0, 8, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+ %9:vgpr_32 = BUFFER_LOAD_DWORD_BOTHEN_exact %4, %5:sgpr_128, 0, 12, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+...
+---
+
+name: buffer_load_dword_dword_dword_bothen_exact_swizzled_0
+body: |
+ bb.0.entry:
+ ; GCN-LABEL: name: buffer_load_dword_dword_dword_bothen_exact_swizzled_0
+ ; GCN: [[COPY:%[0-9]+]]:sgpr_32 = COPY $sgpr0
+ ; GCN-NEXT: [[COPY1:%[0-9]+]]:sgpr_32 = COPY $sgpr1
+ ; GCN-NEXT: [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr2
+ ; GCN-NEXT: [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr3
+ ; GCN-NEXT: [[COPY4:%[0-9]+]]:vreg_64 = COPY $vgpr0
+ ; GCN-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3
+ ; GCN-NEXT: [[BUFFER_LOAD_DWORD_BOTHEN_exact:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_DWORD_BOTHEN_exact [[COPY4]], [[REG_SEQUENCE]], 0, 4, 0, 1, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+ ; GCN-NEXT: [[BUFFER_LOAD_DWORDX2_BOTHEN_exact:%[0-9]+]]:vreg_64 = BUFFER_LOAD_DWORDX2_BOTHEN_exact [[COPY4]], [[REG_SEQUENCE]], 0, 8, 0, 0, implicit $exec :: (dereferenceable load (s64), align 1, addrspace 4)
+ ; GCN-NEXT: [[COPY5:%[0-9]+]]:vgpr_32 = COPY [[BUFFER_LOAD_DWORDX2_BOTHEN_exact]].sub0
+ ; GCN-NEXT: [[COPY6:%[0-9]+]]:vgpr_32 = COPY killed [[BUFFER_LOAD_DWORDX2_BOTHEN_exact]].sub1
+ %0:sgpr_32 = COPY $sgpr0
+ %1:sgpr_32 = COPY $sgpr1
+ %2:sgpr_32 = COPY $sgpr2
+ %3:sgpr_32 = COPY $sgpr3
+ %4:vreg_64 = COPY $vgpr0
+ %5:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3
+ %7:vgpr_32 = BUFFER_LOAD_DWORD_BOTHEN_exact %4, %5:sgpr_128, 0, 4, 0, 1, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+ %8:vgpr_32 = BUFFER_LOAD_DWORD_BOTHEN_exact %4, %5:sgpr_128, 0, 8, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+ %9:vgpr_32 = BUFFER_LOAD_DWORD_BOTHEN_exact %4, %5:sgpr_128, 0, 12, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+...
+---
+
+name: buffer_load_dwordx2_dwordx2_bothen_exact_
diff _vaddr
+body: |
+ bb.0.entry:
+ ; GCN-LABEL: name: buffer_load_dwordx2_dwordx2_bothen_exact_
diff _vaddr
+ ; GCN: [[COPY:%[0-9]+]]:sgpr_32 = COPY $sgpr0
+ ; GCN-NEXT: [[COPY1:%[0-9]+]]:sgpr_32 = COPY $sgpr1
+ ; GCN-NEXT: [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr2
+ ; GCN-NEXT: [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr3
+ ; GCN-NEXT: [[COPY4:%[0-9]+]]:vreg_64 = COPY $vgpr0
+ ; GCN-NEXT: [[COPY5:%[0-9]+]]:vreg_64 = COPY $vgpr1
+ ; GCN-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3
+ ; GCN-NEXT: [[BUFFER_LOAD_DWORDX2_BOTHEN_exact:%[0-9]+]]:vreg_64 = BUFFER_LOAD_DWORDX2_BOTHEN_exact [[COPY4]], [[REG_SEQUENCE]], 0, 4, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+ ; GCN-NEXT: [[BUFFER_LOAD_DWORDX2_BOTHEN_exact1:%[0-9]+]]:vreg_64 = BUFFER_LOAD_DWORDX2_BOTHEN_exact [[COPY5]], [[REG_SEQUENCE]], 0, 12, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+ %0:sgpr_32 = COPY $sgpr0
+ %1:sgpr_32 = COPY $sgpr1
+ %2:sgpr_32 = COPY $sgpr2
+ %3:sgpr_32 = COPY $sgpr3
+ %4:vreg_64 = COPY $vgpr0
+ %5:vreg_64 = COPY $vgpr1
+ %6:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3
+ %8:vreg_64 = BUFFER_LOAD_DWORDX2_BOTHEN_exact %4, %6:sgpr_128, 0, 4, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+ %9:vreg_64 = BUFFER_LOAD_DWORDX2_BOTHEN_exact %5, %6:sgpr_128, 0, 12, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+...
+---
+
+name: buffer_load_dwordx2_dwordx2_bothen_exact_
diff _srsrc
+body: |
+ bb.0.entry:
+ ; GCN-LABEL: name: buffer_load_dwordx2_dwordx2_bothen_exact_
diff _srsrc
+ ; GCN: [[COPY:%[0-9]+]]:sgpr_32 = COPY $sgpr0
+ ; GCN-NEXT: [[COPY1:%[0-9]+]]:sgpr_32 = COPY $sgpr1
+ ; GCN-NEXT: [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr2
+ ; GCN-NEXT: [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr3
+ ; GCN-NEXT: [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr4
+ ; GCN-NEXT: [[COPY5:%[0-9]+]]:vreg_64 = COPY $vgpr0
+ ; GCN-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3
+ ; GCN-NEXT: [[REG_SEQUENCE1:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY1]], %subreg.sub0, [[COPY2]], %subreg.sub1, [[COPY3]], %subreg.sub2, [[COPY4]], %subreg.sub3
+ ; GCN-NEXT: [[BUFFER_LOAD_DWORDX2_BOTHEN_exact:%[0-9]+]]:vreg_64 = BUFFER_LOAD_DWORDX2_BOTHEN_exact [[COPY5]], [[REG_SEQUENCE]], 0, 4, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+ ; GCN-NEXT: [[BUFFER_LOAD_DWORDX2_BOTHEN_exact1:%[0-9]+]]:vreg_64 = BUFFER_LOAD_DWORDX2_BOTHEN_exact [[COPY5]], [[REG_SEQUENCE1]], 0, 12, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+ %0:sgpr_32 = COPY $sgpr0
+ %1:sgpr_32 = COPY $sgpr1
+ %2:sgpr_32 = COPY $sgpr2
+ %3:sgpr_32 = COPY $sgpr3
+ %4:sgpr_32 = COPY $sgpr4
+ %5:vreg_64 = COPY $vgpr0
+ %6:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3
+ %7:sgpr_128 = REG_SEQUENCE %1:sgpr_32, %subreg.sub0, %2:sgpr_32, %subreg.sub1, %3:sgpr_32, %subreg.sub2, %4:sgpr_32, %subreg.sub3
+ %8:vreg_64 = BUFFER_LOAD_DWORDX2_BOTHEN_exact %5, %6:sgpr_128, 0, 4, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+ %9:vreg_64 = BUFFER_LOAD_DWORDX2_BOTHEN_exact %5, %7:sgpr_128, 0, 12, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+...
+---
+
+name: buffer_load_dwordx2_dwordx2_idxen_exact_
diff _vaddr
+body: |
+ bb.0.entry:
+ ; GCN-LABEL: name: buffer_load_dwordx2_dwordx2_idxen_exact_
diff _vaddr
+ ; GCN: [[COPY:%[0-9]+]]:sgpr_32 = COPY $sgpr0
+ ; GCN-NEXT: [[COPY1:%[0-9]+]]:sgpr_32 = COPY $sgpr1
+ ; GCN-NEXT: [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr2
+ ; GCN-NEXT: [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr3
+ ; GCN-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY $vgpr0
+ ; GCN-NEXT: [[COPY5:%[0-9]+]]:vgpr_32 = COPY $vgpr1
+ ; GCN-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3
+ ; GCN-NEXT: [[BUFFER_LOAD_DWORDX2_IDXEN_exact:%[0-9]+]]:vreg_64 = BUFFER_LOAD_DWORDX2_IDXEN_exact [[COPY4]], [[REG_SEQUENCE]], 0, 4, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+ ; GCN-NEXT: [[BUFFER_LOAD_DWORDX2_IDXEN_exact1:%[0-9]+]]:vreg_64 = BUFFER_LOAD_DWORDX2_IDXEN_exact [[COPY5]], [[REG_SEQUENCE]], 0, 12, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+ %0:sgpr_32 = COPY $sgpr0
+ %1:sgpr_32 = COPY $sgpr1
+ %2:sgpr_32 = COPY $sgpr2
+ %3:sgpr_32 = COPY $sgpr3
+ %4:vgpr_32 = COPY $vgpr0
+ %5:vgpr_32 = COPY $vgpr1
+ %6:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3
+ %8:vreg_64 = BUFFER_LOAD_DWORDX2_IDXEN_exact %4, %6:sgpr_128, 0, 4, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+ %9:vreg_64 = BUFFER_LOAD_DWORDX2_IDXEN_exact %5, %6:sgpr_128, 0, 12, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+...
+---
+
+name: buffer_load_dwordx2_dwordx2_idxen_exact_
diff _srsrc
+body: |
+ bb.0.entry:
+ ; GCN-LABEL: name: buffer_load_dwordx2_dwordx2_idxen_exact_
diff _srsrc
+ ; GCN: [[COPY:%[0-9]+]]:sgpr_32 = COPY $sgpr0
+ ; GCN-NEXT: [[COPY1:%[0-9]+]]:sgpr_32 = COPY $sgpr1
+ ; GCN-NEXT: [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr2
+ ; GCN-NEXT: [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr3
+ ; GCN-NEXT: [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr4
+ ; GCN-NEXT: [[COPY5:%[0-9]+]]:vgpr_32 = COPY $vgpr0
+ ; GCN-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3
+ ; GCN-NEXT: [[REG_SEQUENCE1:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY1]], %subreg.sub0, [[COPY2]], %subreg.sub1, [[COPY3]], %subreg.sub2, [[COPY4]], %subreg.sub3
+ ; GCN-NEXT: [[BUFFER_LOAD_DWORDX2_IDXEN_exact:%[0-9]+]]:vreg_64 = BUFFER_LOAD_DWORDX2_IDXEN_exact [[COPY5]], [[REG_SEQUENCE]], 0, 4, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+ ; GCN-NEXT: [[BUFFER_LOAD_DWORDX2_IDXEN_exact1:%[0-9]+]]:vreg_64 = BUFFER_LOAD_DWORDX2_IDXEN_exact [[COPY5]], [[REG_SEQUENCE1]], 0, 12, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+ %0:sgpr_32 = COPY $sgpr0
+ %1:sgpr_32 = COPY $sgpr1
+ %2:sgpr_32 = COPY $sgpr2
+ %3:sgpr_32 = COPY $sgpr3
+ %4:sgpr_32 = COPY $sgpr4
+ %5:vgpr_32 = COPY $vgpr0
+ %6:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3
+ %7:sgpr_128 = REG_SEQUENCE %1:sgpr_32, %subreg.sub0, %2:sgpr_32, %subreg.sub1, %3:sgpr_32, %subreg.sub2, %4:sgpr_32, %subreg.sub3
+ %8:vreg_64 = BUFFER_LOAD_DWORDX2_IDXEN_exact %5, %6:sgpr_128, 0, 4, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+ %9:vreg_64 = BUFFER_LOAD_DWORDX2_IDXEN_exact %5, %7:sgpr_128, 0, 12, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+...
diff --git a/llvm/test/CodeGen/AMDGPU/merge-tbuffer.mir b/llvm/test/CodeGen/AMDGPU/merge-tbuffer.mir
index c86b5adec372d6..9766b427b43258 100644
--- a/llvm/test/CodeGen/AMDGPU/merge-tbuffer.mir
+++ b/llvm/test/CodeGen/AMDGPU/merge-tbuffer.mir
@@ -7,9 +7,37 @@
# GFX9 tests
#
+---
name: gfx9_tbuffer_load_x_xyz
body: |
bb.0.entry:
+ ; GFX9-LABEL: name: gfx9_tbuffer_load_x_xyz
+ ; GFX9: [[COPY:%[0-9]+]]:sgpr_32 = COPY $sgpr0
+ ; GFX9-NEXT: [[COPY1:%[0-9]+]]:sgpr_32 = COPY $sgpr1
+ ; GFX9-NEXT: [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr2
+ ; GFX9-NEXT: [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr3
+ ; GFX9-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3
+ ; GFX9-NEXT: [[TBUFFER_LOAD_FORMAT_XYZW_OFFSET:%[0-9]+]]:vreg_128 = TBUFFER_LOAD_FORMAT_XYZW_OFFSET [[REG_SEQUENCE]], 0, 4, 126, 0, 0, implicit $exec :: (dereferenceable load (s128), align 1, addrspace 4)
+ ; GFX9-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[TBUFFER_LOAD_FORMAT_XYZW_OFFSET]].sub0
+ ; GFX9-NEXT: [[COPY5:%[0-9]+]]:vreg_96 = COPY killed [[TBUFFER_LOAD_FORMAT_XYZW_OFFSET]].sub1_sub2_sub3
+ ;
+ ; GFX10-LABEL: name: gfx9_tbuffer_load_x_xyz
+ ; GFX10: [[COPY:%[0-9]+]]:sgpr_32 = COPY $sgpr0
+ ; GFX10-NEXT: [[COPY1:%[0-9]+]]:sgpr_32 = COPY $sgpr1
+ ; GFX10-NEXT: [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr2
+ ; GFX10-NEXT: [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr3
+ ; GFX10-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3
+ ; GFX10-NEXT: [[TBUFFER_LOAD_FORMAT_X_OFFSET:%[0-9]+]]:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET [[REG_SEQUENCE]], 0, 4, 116, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+ ; GFX10-NEXT: [[TBUFFER_LOAD_FORMAT_XYZ_OFFSET:%[0-9]+]]:vreg_96 = TBUFFER_LOAD_FORMAT_XYZ_OFFSET [[REG_SEQUENCE]], 0, 8, 125, 0, 0, implicit $exec :: (dereferenceable load (s96), align 1, addrspace 4)
+ ;
+ ; GFX11-LABEL: name: gfx9_tbuffer_load_x_xyz
+ ; GFX11: [[COPY:%[0-9]+]]:sgpr_32 = COPY $sgpr0
+ ; GFX11-NEXT: [[COPY1:%[0-9]+]]:sgpr_32 = COPY $sgpr1
+ ; GFX11-NEXT: [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr2
+ ; GFX11-NEXT: [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr3
+ ; GFX11-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3
+ ; GFX11-NEXT: [[TBUFFER_LOAD_FORMAT_X_OFFSET:%[0-9]+]]:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET [[REG_SEQUENCE]], 0, 4, 116, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
+ ; GFX11-NEXT: [[TBUFFER_LOAD_FORMAT_XYZ_OFFSET:%[0-9]+]]:vreg_96 = TBUFFER_LOAD_FORMAT_XYZ_OFFSET [[REG_SEQUENCE]], 0, 8, 125, 0, 0, implicit $exec :: (dereferenceable load (s96), align 1, addrspace 4)
%0:sgpr_32 = COPY $sgpr0
%1:sgpr_32 = COPY $sgpr1
%2:sgpr_32 = COPY $sgpr2
More information about the llvm-commits
mailing list