merge-image-load-gfx12.mir | Explore in Territory

# RUN: llc -mtriple=amdgcn -mcpu=gfx1200 -verify-machineinstrs -run-pass si-load-store-opt -o - %s | FileCheck -check-prefix=GFX12 %s

---
# GFX12-LABEL: name: image_load_merged_v1v3
# GFX12: %{{[0-9]+}}:vreg_128 = IMAGE_LOAD_V4_V2_gfx12 %6, %7, %3, 15, 1, 0, 0, 0, 0, 0, implicit $exec, implicit $exec :: (dereferenceable load (s128), align 4, addrspace 4)
# GFX12: %{{[0-9]+}}:vgpr_32 = COPY %10.sub0
# GFX12: %{{[0-9]+}}:vreg_96 = COPY killed %10.sub1_sub2_sub3

name: image_load_merged_v1v3
body:             |
  bb.0.entry:
    %0:sgpr_64 = COPY $sgpr0_sgpr1
    %1:sreg_64_xexec = S_LOAD_DWORDX2_IMM %0, 36, 0
    %2:sgpr_128 = COPY $sgpr96_sgpr97_sgpr98_sgpr99
    %3:sgpr_256 = S_LOAD_DWORDX8_IMM %1, 208, 0
    %4:vgpr_32 = COPY %2.sub3
    %5:vreg_64 = BUFFER_LOAD_DWORDX2_OFFSET %2:sgpr_128, 0, 0, 0, 0, implicit $exec :: (dereferenceable invariant load 16)
    %6:vgpr_32 = COPY %5.sub0
    %7:vgpr_32 = COPY %5.sub1
    %8:vgpr_32 = IMAGE_LOAD_V1_V2_gfx12 %6:vgpr_32, %7:vgpr_32, %3:sgpr_256, 1, 1, 0, 0, 0, 0, 0, implicit $exec :: (dereferenceable load 4, addrspace 4)
    %9:vreg_96 = IMAGE_LOAD_V3_V2_gfx12 %6:vgpr_32, %7:vgpr_32, %3:sgpr_256, 14, 1, 0, 0, 0, 0, 0, implicit $exec :: (dereferenceable load 12, align 16, addrspace 4)
...
---
# GFX12-LABEL: name: image_load_merged_v1v3_reversed
# GFX12: %{{[0-9]+}}:vreg_128 = IMAGE_LOAD_V4_V2_gfx12 %6, %7, %3, 15, 1, 0, 0, 0, 0, 0, implicit $exec, implicit $exec :: (dereferenceable load (s128), addrspace 4)
# GFX12: %{{[0-9]+}}:vgpr_32 = COPY %10.sub3
# GFX12: %{{[0-9]+}}:vreg_96 = COPY killed %10.sub0_sub1_sub2

name: image_load_merged_v1v3_reversed
body:             |
  bb.0.entry:
    %0:sgpr_64 = COPY $sgpr0_sgpr1
    %1:sreg_64_xexec = S_LOAD_DWORDX2_IMM %0, 36, 0
    %2:sgpr_128 = COPY $sgpr96_sgpr97_sgpr98_sgpr99
    %3:sgpr_256 = S_LOAD_DWORDX8_IMM %1, 208, 0
    %4:vgpr_32 = COPY %2.sub3
    %5:vreg_64 = BUFFER_LOAD_DWORDX2_OFFSET %2:sgpr_128, 0, 0, 0, 0, implicit $exec :: (dereferenceable invariant load 16)
    %6:vgpr_32 = COPY %5.sub0
    %7:vgpr_32 = COPY %5.sub1
    %8:vgpr_32 = IMAGE_LOAD_V1_V2_gfx12 %6:vgpr_32, %7:vgpr_32, %3:sgpr_256, 8, 1, 0, 0, 0, 0, 0, implicit $exec :: (dereferenceable load 4, addrspace 4)
    %9:vreg_96 = IMAGE_LOAD_V3_V2_gfx12 %6:vgpr_32, %7:vgpr_32, %3:sgpr_256, 7, 1, 0, 0, 0, 0, 0, implicit $exec :: (dereferenceable load 12, align 16, addrspace 4)
...
---

# GFX12-LABEL: name: image_load_merged_v2v2
# GFX12: %{{[0-9]+}}:vreg_128 = IMAGE_LOAD_V4_V2_gfx12 %6, %7, %3, 15, 1, 0, 0, 0, 0, 0, implicit $exec, implicit $exec :: (dereferenceable load (s128), addrspace 4)
# GFX12: %{{[0-9]+}}:vreg_64 = COPY %10.sub0_sub1
# GFX12: %{{[0-9]+}}:vreg_64 = COPY killed %10.sub2_sub3

name: image_load_merged_v2v2
body:             |
  bb.0.entry:
    %0:sgpr_64 = COPY $sgpr0_sgpr1
    %1:sreg_64_xexec = S_LOAD_DWORDX2_IMM %0, 36, 0
    %2:sgpr_128 = COPY $sgpr96_sgpr97_sgpr98_sgpr99
    %3:sgpr_256 = S_LOAD_DWORDX8_IMM %1, 208, 0
    %4:vgpr_32 = COPY %2.sub3
    %5:vreg_64 = BUFFER_LOAD_DWORDX2_OFFSET %2:sgpr_128, 0, 0, 0, 0, implicit $exec :: (dereferenceable invariant load 16)
    %6:vgpr_32 = COPY %5.sub0
    %7:vgpr_32 = COPY %5.sub1
    %8:vreg_64 = IMAGE_LOAD_V2_V2_gfx12 %6:vgpr_32, %7:vgpr_32, %3:sgpr_256, 3, 1, 0, 0, 0, 0, 0, implicit $exec :: (dereferenceable load 8, align 16, addrspace 4)
    %9:vreg_64 = IMAGE_LOAD_V2_V2_gfx12 %6:vgpr_32, %7:vgpr_32, %3:sgpr_256, 12, 1, 0, 0, 0, 0, 0, implicit $exec :: (dereferenceable load 8, align 16, addrspace 4)
...
---

# GFX12-LABEL: name: image_load_merged_v2v2_reversed
# GFX12: %{{[0-9]+}}:vreg_128 = IMAGE_LOAD_V4_V2_gfx12 %6, %7, %3, 15, 1, 0, 0, 0, 0, 0, implicit $exec, implicit $exec :: (dereferenceable load (s128), addrspace 4)
# GFX12: %{{[0-9]+}}:vreg_64 = COPY %10.sub2_sub3
# GFX12: %{{[0-9]+}}:vreg_64 = COPY killed %10.sub0_sub1

name: image_load_merged_v2v2_reversed
body:             |
  bb.0.entry:
    %0:sgpr_64 = COPY $sgpr0_sgpr1
    %1:sreg_64_xexec = S_LOAD_DWORDX2_IMM %0, 36, 0
    %2:sgpr_128 = COPY $sgpr96_sgpr97_sgpr98_sgpr99
    %3:sgpr_256 = S_LOAD_DWORDX8_IMM %1, 208, 0
    %4:vgpr_32 = COPY %2.sub3
    %5:vreg_64 = BUFFER_LOAD_DWORDX2_OFFSET %2:sgpr_128, 0, 0, 0, 0, implicit $exec :: (dereferenceable invariant load 16)
    %6:vgpr_32 = COPY %5.sub0
    %7:vgpr_32 = COPY %5.sub1
    %8:vreg_64 = IMAGE_LOAD_V2_V2_gfx12 %6:vgpr_32, %7:vgpr_32, %3:sgpr_256, 12, 1, 0, 0, 0, 0, 0, implicit $exec :: (dereferenceable load 8, align 16, addrspace 4)
    %9:vreg_64 = IMAGE_LOAD_V2_V2_gfx12 %6:vgpr_32, %7:vgpr_32, %3:sgpr_256, 3, 1, 0, 0, 0, 0, 0, implicit $exec :: (dereferenceable load 8, align 16, addrspace 4)
...
---

# GFX12-LABEL: name: image_load_merged_v3v1
# GFX12: %{{[0-9]+}}:vreg_128 = IMAGE_LOAD_V4_V2_gfx12 %6, %7, %3, 15, 1, 0, 0, 0, 0, 0, implicit $exec, implicit $exec :: (dereferenceable load (s128), addrspace 4)
# GFX12: %{{[0-9]+}}:vreg_96 = COPY %10.sub0_sub1_sub2
# GFX12: %{{[0-9]+}}:vgpr_32 = COPY killed %10.sub3

name: image_load_merged_v3v1
body:             |
  bb.0.entry:
    %0:sgpr_64 = COPY $sgpr0_sgpr1
    %1:sreg_64_xexec = S_LOAD_DWORDX2_IMM %0, 36, 0
    %2:sgpr_128 = COPY $sgpr96_sgpr97_sgpr98_sgpr99
    %3:sgpr_256 = S_LOAD_DWORDX8_IMM %1, 208, 0
    %4:vgpr_32 = COPY %2.sub3
    %5:vreg_64 = BUFFER_LOAD_DWORDX2_OFFSET %2:sgpr_128, 0, 0, 0, 0, implicit $exec :: (dereferenceable invariant load 16)
    %6:vgpr_32 = COPY %5.sub0
    %7:vgpr_32 = COPY %5.sub1
    %8:vreg_96 = IMAGE_LOAD_V3_V2_gfx12 %6:vgpr_32, %7:vgpr_32, %3:sgpr_256, 7, 1, 0, 0, 0, 0, 0, implicit $exec :: (dereferenceable load 12, align 16, addrspace 4)
    %9:vgpr_32 = IMAGE_LOAD_V1_V2_gfx12 %6:vgpr_32, %7:vgpr_32, %3:sgpr_256, 8, 1, 0, 0, 0, 0, 0, implicit $exec :: (dereferenceable load 4, addrspace 4)
...
---

# GFX12-LABEL: name: image_load_merged_v3v1_reversed
# GFX12: %{{[0-9]+}}:vreg_128 = IMAGE_LOAD_V4_V2_gfx12 %6, %7, %3, 15, 1, 0, 0, 0, 0, 0, implicit $exec, implicit $exec :: (dereferenceable load (s128), align 4, addrspace 4)
# GFX12: %{{[0-9]+}}:vreg_96 = COPY %10.sub1_sub2_sub3
# GFX12: %{{[0-9]+}}:vgpr_32 = COPY killed %10.sub0

name: image_load_merged_v3v1_reversed
body:             |
  bb.0.entry:
    %0:sgpr_64 = COPY $sgpr0_sgpr1
    %1:sreg_64_xexec = S_LOAD_DWORDX2_IMM %0, 36, 0
    %2:sgpr_128 = COPY $sgpr96_sgpr97_sgpr98_sgpr99
    %3:sgpr_256 = S_LOAD_DWORDX8_IMM %1, 208, 0
    %4:vgpr_32 = COPY %2.sub3
    %5:vreg_64 = BUFFER_LOAD_DWORDX2_OFFSET %2:sgpr_128, 0, 0, 0, 0, implicit $exec :: (dereferenceable invariant load 16)
    %6:vgpr_32 = COPY %5.sub0
    %7:vgpr_32 = COPY %5.sub1
    %8:vreg_96 = IMAGE_LOAD_V3_V2_gfx12 %6:vgpr_32, %7:vgpr_32, %3:sgpr_256, 14, 1, 0, 0, 0, 0, 0, implicit $exec :: (dereferenceable load 12, align 16, addrspace 4)
    %9:vgpr_32 = IMAGE_LOAD_V1_V2_gfx12 %6:vgpr_32, %7:vgpr_32, %3:sgpr_256, 1, 1, 0, 0, 0, 0, 0, implicit $exec :: (dereferenceable load 4, addrspace 4)
...
---

# GFX12-LABEL: name: image_load_divided_merged
# GFX12: %{{[0-9]+}}:vreg_128 = IMAGE_LOAD_V4_V2_gfx12 %6, %7, %3, 15, 1, 0, 0, 0, 0, 0, implicit $exec, implicit $exec :: (dereferenceable load (s128), addrspace 4)

name: image_load_divided_merged
body:             |
  bb.0.entry:
    %0:sgpr_64 = COPY $sgpr0_sgpr1
    %1:sreg_64_xexec = S_LOAD_DWORDX2_IMM %0, 36, 0
    %2:sgpr_128 = COPY $sgpr96_sgpr97_sgpr98_sgpr99
    %3:sgpr_256 = S_LOAD_DWORDX8_IMM %1, 208, 0
    %4:vgpr_32 = COPY %2.sub3
    %5:vreg_64 = BUFFER_LOAD_DWORDX2_OFFSET %2:sgpr_128, 0, 0, 0, 0, implicit $exec :: (dereferenceable invariant load 16)
    %6:vgpr_32 = COPY %5.sub0
    %7:vgpr_32 = COPY %5.sub1
    %8:vgpr_32 = IMAGE_LOAD_V1_V2_gfx12 %6:vgpr_32, %7:vgpr_32, %3:sgpr_256, 8, 1, 0, 0, 0, 0, 0, implicit $exec :: (dereferenceable load 4, addrspace 4)
    %9:vreg_64 = BUFFER_LOAD_DWORDX2_OFFSET %2:sgpr_128, 0, 0, 0, 0, implicit $exec :: (dereferenceable invariant load 16)
    %10:vgpr_32 = COPY %9.sub0
    %11:vgpr_32 = COPY %9.sub1
    %12:vreg_128 = BUFFER_LOAD_DWORDX4_OFFSET %2:sgpr_128, 0, 0, 0, 0, implicit $exec :: (dereferenceable invariant load 16)
    %13:vreg_96 = IMAGE_LOAD_V3_V2_gfx12 %10:vgpr_32, %11:vgpr_32, %3:sgpr_256, 7, 1, 0, 0, 0, 0, 0, implicit $exec :: (dereferenceable load 12, align 16, addrspace 4)
    %14:vreg_128 = BUFFER_LOAD_DWORDX4_OFFSET %2:sgpr_128, 0, 0, 0, 0, implicit $exec :: (dereferenceable invariant load 16)
    %15:vreg_96 = IMAGE_LOAD_V3_V2_gfx12 %6:vgpr_32, %7:vgpr_32, %3:sgpr_256, 7, 1, 0, 0, 0, 0, 0, implicit $exec :: (dereferenceable load 12, align 16, addrspace 4)
...
---

# GFX12-LABEL: name: image_load_divided_not_merged
# GFX12: %{{[0-9]+}}:vgpr_32 = IMAGE_LOAD_V1_V2_gfx12 %6, %7, %3, 8, 1, 0, 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), addrspace 4)
# GFX12: %{{[0-9]+}}:vreg_96 = IMAGE_LOAD_V3_V2_gfx12 %6, %7, %3, 7, 1, 0, 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s96), align 16, addrspace 4)

name: image_load_divided_not_merged
body:             |
  bb.0.entry:
    %0:sgpr_64 = COPY $sgpr0_sgpr1
    %1:sreg_64_xexec = S_LOAD_DWORDX2_IMM %0, 36, 0
    %2:sgpr_128 = COPY $sgpr96_sgpr97_sgpr98_sgpr99
    %3:sgpr_256 = S_LOAD_DWORDX8_IMM %1, 208, 0
    %4:vreg_128 = COPY %2
    %5:vreg_64 = BUFFER_LOAD_DWORDX2_OFFSET %2:sgpr_128, 0, 0, 0, 0, implicit $exec :: (dereferenceable invariant load 16)
    %6:vgpr_32 = COPY %5.sub0
    %7:vgpr_32 = COPY %5.sub1
    %8:vgpr_32 = IMAGE_LOAD_V1_V2_gfx12 %6:vgpr_32, %7:vgpr_32, %3:sgpr_256, 8, 1, 0, 0, 0, 0, 0, implicit $exec :: (dereferenceable load 4, addrspace 4)
    IMAGE_STORE_V4_V2_gfx12 %4:vreg_128, %6:vgpr_32, %7:vgpr_32, %3:sgpr_256, 15, 1, 0, 0, 0, 0, 0, implicit $exec :: (store 16)
    %9:vreg_96 = IMAGE_LOAD_V3_V2_gfx12 %6:vgpr_32, %7:vgpr_32, %3:sgpr_256, 7, 1, 0, 0, 0, 0, 0, implicit $exec :: (dereferenceable load 12, align 16, addrspace 4)
...
---

# GFX12-LABEL: name: image_load_dmask_overlapped_not_merged
# GFX12: %{{[0-9]+}}:vgpr_32 = IMAGE_LOAD_V1_V2_gfx12 %6, %7, %3, 4, 1, 0, 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), addrspace 4)
# GFX12: %{{[0-9]+}}:vreg_96 = IMAGE_LOAD_V3_V2_gfx12 %6, %7, %3, 7, 1, 0, 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s96), align 16, addrspace 4)

name: image_load_dmask_overlapped_not_merged
body:             |
  bb.0.entry:
    %0:sgpr_64 = COPY $sgpr0_sgpr1
    %1:sreg_64_xexec = S_LOAD_DWORDX2_IMM %0, 36, 0
    %2:sgpr_128 = COPY $sgpr96_sgpr97_sgpr98_sgpr99
    %3:sgpr_256 = S_LOAD_DWORDX8_IMM %1, 208, 0
    %4:vgpr_32 = COPY %2.sub3
    %5:vreg_64 = BUFFER_LOAD_DWORDX2_OFFSET %2:sgpr_128, 0, 0, 0, 0, implicit $exec :: (dereferenceable invariant load 16)
    %6:vgpr_32 = COPY %5.sub0
    %7:vgpr_32 = COPY %5.sub1
    %8:vgpr_32 = IMAGE_LOAD_V1_V2_gfx12 %6:vgpr_32, %7:vgpr_32, %3:sgpr_256, 4, 1, 0, 0, 0, 0, 0, implicit $exec :: (dereferenceable load 4, addrspace 4)
    %9:vreg_96 = IMAGE_LOAD_V3_V2_gfx12 %6:vgpr_32, %7:vgpr_32, %3:sgpr_256, 7, 1, 0, 0, 0, 0, 0, implicit $exec :: (dereferenceable load 12, align 16, addrspace 4)
...
---

# GFX12-LABEL: name: image_load_dmask_not_disjoint_not_merged
# GFX12: %{{[0-9]+}}:vgpr_32 = IMAGE_LOAD_V1_V2_gfx12 %6, %7, %3, 4, 1, 0, 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), addrspace 4)
# GFX12: %{{[0-9]+}}:vreg_96 = IMAGE_LOAD_V3_V2_gfx12 %6, %7, %3, 11, 1, 0, 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s96), align 16, addrspace 4)

name: image_load_dmask_not_disjoint_not_merged
body:             |
  bb.0.entry:
    %0:sgpr_64 = COPY $sgpr0_sgpr1
    %1:sreg_64_xexec = S_LOAD_DWORDX2_IMM %0, 36, 0
    %2:sgpr_128 = COPY $sgpr96_sgpr97_sgpr98_sgpr99
    %3:sgpr_256 = S_LOAD_DWORDX8_IMM %1, 208, 0
    %4:vgpr_32 = COPY %2.sub3
    %5:vreg_64 = BUFFER_LOAD_DWORDX2_OFFSET %2:sgpr_128, 0, 0, 0, 0, implicit $exec :: (dereferenceable invariant load 16)
    %6:vgpr_32 = COPY %5.sub0
    %7:vgpr_32 = COPY %5.sub1
    %8:vgpr_32 = IMAGE_LOAD_V1_V2_gfx12 %6:vgpr_32, %7:vgpr_32, %3:sgpr_256, 4, 1, 0, 0, 0, 0, 0, implicit $exec :: (dereferenceable load 4, addrspace 4)
    %9:vreg_96 = IMAGE_LOAD_V3_V2_gfx12 %6:vgpr_32, %7:vgpr_32, %3:sgpr_256, 11, 1, 0, 0, 0, 0, 0, implicit $exec :: (dereferenceable load 12, align 16, addrspace 4)
...
---

# GFX12-LABEL: name: image_load_not_merged_0
# GFX12: %{{[0-9]+}}:vgpr_32 = IMAGE_LOAD_V1_V2_gfx12 %6, %7, %3, 8, 1, 0, 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), addrspace 4)
# GFX12: %{{[0-9]+}}:vreg_96 = IMAGE_LOAD_V3_V2_gfx12 %9, %10, %3, 7, 1, 0, 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s96), align 16, addrspace 4)

name: image_load_not_merged_0
body:             |
  bb.0.entry:
    %0:sgpr_64 = COPY $sgpr0_sgpr1
    %1:sreg_64_xexec = S_LOAD_DWORDX2_IMM %0, 36, 0
    %2:sgpr_128 = COPY $sgpr96_sgpr97_sgpr98_sgpr99
    %3:sgpr_256 = S_LOAD_DWORDX8_IMM %1, 208, 0
    %4:vgpr_32 = COPY %2.sub3
    %5:vreg_64 = BUFFER_LOAD_DWORDX2_OFFSET %2, 0, 0, 0, 0, implicit $exec :: (dereferenceable invariant load 16)
    %6:vgpr_32 = COPY %5.sub0
    %7:vgpr_32 = COPY %5.sub1
    %8:vreg_64 = BUFFER_LOAD_DWORDX2_OFFSET %2, 1, 0, 0, 0, implicit $exec :: (dereferenceable invariant load 16)
    %9:vgpr_32 = COPY %8.sub0
    %10:vgpr_32 = COPY %8.sub1
    %11:vgpr_32 = IMAGE_LOAD_V1_V2_gfx12 %6, %7, %3, 8, 1, 0, 0, 0, 0, 0, implicit $exec :: (dereferenceable load 4, addrspace 4)
    %12:vreg_96 = IMAGE_LOAD_V3_V2_gfx12 %9, %10, %3, 7, 1, 0, 0, 0, 0, 0, implicit $exec :: (dereferenceable load 12, align 16, addrspace 4)
...
---

# GFX12-LABEL: name: image_load_not_merged_1
# GFX12: %{{[0-9]+}}:vgpr_32 = IMAGE_LOAD_V1_V2_gfx12 %7, %8, %3, 8, 1, 0, 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), addrspace 4)
# GFX12: %{{[0-9]+}}:vreg_96 = IMAGE_LOAD_V3_V2_gfx12 %7, %8, %4, 7, 1, 0, 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s96), align 16, addrspace 4)

name: image_load_not_merged_1
body:             |
  bb.0.entry:
    %0:sgpr_64 = COPY $sgpr0_sgpr1
    %1:sreg_64_xexec = S_LOAD_DWORDX2_IMM %0, 36, 0
    %2:sgpr_128 = COPY $sgpr96_sgpr97_sgpr98_sgpr99
    %3:sgpr_256 = S_LOAD_DWORDX8_IMM %1, 208, 0
    %4:sgpr_256 = S_LOAD_DWORDX8_IMM %1, 208, 0
    %5:vgpr_32 = COPY %2.sub3
    %6:vreg_64 = BUFFER_LOAD_DWORDX2_OFFSET %2, 0, 0, 0, 0, implicit $exec :: (dereferenceable invariant load 16)
    %7:vgpr_32 = COPY %6.sub0
    %8:vgpr_32 = COPY %6.sub1
    %9:vgpr_32 = IMAGE_LOAD_V1_V2_gfx12 %7, %8, %3, 8, 1, 0, 0, 0, 0, 0, implicit $exec :: (dereferenceable load 4, addrspace 4)
    %10:vreg_96 = IMAGE_LOAD_V3_V2_gfx12 %7, %8, %4, 7, 1, 0, 0, 0, 0, 0, implicit $exec :: (dereferenceable load 12, align 16, addrspace 4)
...
---

# GFX12-LABEL: name: image_load_not_merged_4
# GFX12: %{{[0-9]+}}:vgpr_32 = IMAGE_LOAD_V1_V2_gfx12 %6, %7, %3, 8, 1, 0, 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), addrspace 4)
# GFX12: %{{[0-9]+}}:vreg_96 = IMAGE_LOAD_V3_V2_gfx12 %6, %7, %3, 7, 1, 1, 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s96), align 16, addrspace 4)

name: image_load_not_merged_4
body:             |
  bb.0.entry:
    %0:sgpr_64 = COPY $sgpr0_sgpr1
    %1:sreg_64_xexec = S_LOAD_DWORDX2_IMM %0, 36, 0
    %2:sgpr_128 = COPY $sgpr96_sgpr97_sgpr98_sgpr99
    %3:sgpr_256 = S_LOAD_DWORDX8_IMM %1, 208, 0
    %4:vgpr_32 = COPY %2.sub3
    %5:vreg_64 = BUFFER_LOAD_DWORDX2_OFFSET %2, 0, 0, 0, 0, implicit $exec :: (dereferenceable invariant load 16)
    %6:vgpr_32 = COPY %5.sub0
    %7:vgpr_32 = COPY %5.sub1
    %8:vgpr_32 = IMAGE_LOAD_V1_V2_gfx12 %6, %7, %3, 8, 1, 0, 0, 0, 0, 0, implicit $exec :: (dereferenceable load 4, addrspace 4)
    %9:vreg_96 = IMAGE_LOAD_V3_V2_gfx12 %6, %7, %3, 7, 1, 1, 0, 0, 0, 0, implicit $exec :: (dereferenceable load 12, align 16, addrspace 4)
...
---

# GFX12-LABEL: name: image_load_not_merged_5
# GFX12: %{{[0-9]+}}:vgpr_32 = IMAGE_LOAD_V1_V2_gfx12 %6, %7, %3, 8, 1, 0, 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), addrspace 4)
# GFX12: %{{[0-9]+}}:vreg_96 = IMAGE_LOAD_V3_V2_gfx12 %6, %7, %3, 7, 1, 0, 1, 0, 0, 0, implicit $exec :: (dereferenceable load (s96), align 16, addrspace 4)

name: image_load_not_merged_5
body:             |
  bb.0.entry:
    %0:sgpr_64 = COPY $sgpr0_sgpr1
    %1:sreg_64_xexec = S_LOAD_DWORDX2_IMM %0, 36, 0
    %2:sgpr_128 = COPY $sgpr96_sgpr97_sgpr98_sgpr99
    %3:sgpr_256 = S_LOAD_DWORDX8_IMM %1, 208, 0
    %4:vgpr_32 = COPY %2.sub3
    %5:vreg_64 = BUFFER_LOAD_DWORDX2_OFFSET %2, 0, 0, 0, 0, implicit $exec :: (dereferenceable invariant load 16)
    %6:vgpr_32 = COPY %5.sub0
    %7:vgpr_32 = COPY %5.sub1
    %8:vgpr_32 = IMAGE_LOAD_V1_V2_gfx12 %6, %7, %3, 8, 1, 0, 0, 0, 0, 0, implicit $exec :: (dereferenceable load 4, addrspace 4)
    %9:vreg_96 = IMAGE_LOAD_V3_V2_gfx12 %6, %7, %3, 7, 1, 0, 1, 0, 0, 0, implicit $exec :: (dereferenceable load 12, align 16, addrspace 4)
...
---

# GFX12-LABEL: name: image_load_not_merged_6
# GFX12: %{{[0-9]+}}:vgpr_32 = IMAGE_LOAD_V1_V2_gfx12 %6, %7, %3, 8, 1, 0, 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), addrspace 4)
# GFX12: %{{[0-9]+}}:vreg_96 = IMAGE_LOAD_V3_V2_gfx12 %6, %7, %3, 7, 1, 1, 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s96), align 16, addrspace 4)

name: image_load_not_merged_6
body:             |
  bb.0.entry:
    %0:sgpr_64 = COPY $sgpr0_sgpr1
    %1:sreg_64_xexec = S_LOAD_DWORDX2_IMM %0, 36, 0
    %2:sgpr_128 = COPY $sgpr96_sgpr97_sgpr98_sgpr99
    %3:sgpr_256 = S_LOAD_DWORDX8_IMM %1, 208, 0
    %4:vgpr_32 = COPY %2.sub3
    %5:vreg_64 = BUFFER_LOAD_DWORDX2_OFFSET %2, 0, 0, 0, 0, implicit $exec :: (dereferenceable invariant load 16)
    %6:vgpr_32 = COPY %5.sub0
    %7:vgpr_32 = COPY %5.sub1
    %8:vgpr_32 = IMAGE_LOAD_V1_V2_gfx12 %6, %7, %3, 8, 1, 0, 0, 0, 0, 0, implicit $exec :: (dereferenceable load 4, addrspace 4)
    %9:vreg_96 = IMAGE_LOAD_V3_V2_gfx12 %6, %7, %3, 7, 1, 1, 0, 0, 0, 0, implicit $exec :: (dereferenceable load 12, align 16, addrspace 4)
...
---

# GFX12-LABEL: name: image_load_not_merged_7
# GFX12: %{{[0-9]+}}:vgpr_32 = IMAGE_LOAD_V1_V2_gfx12 %6, %7, %3, 8, 1, 0, 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), addrspace 4)
# GFX12: %{{[0-9]+}}:vreg_96 = IMAGE_LOAD_V3_V2_gfx12 %6, %7, %3, 7, 1, 0, 1, 0, 0, 0, implicit $exec :: (dereferenceable load (s96), align 16, addrspace 4)

name: image_load_not_merged_7
body:             |
  bb.0.entry:
    %0:sgpr_64 = COPY $sgpr0_sgpr1
    %1:sreg_64_xexec = S_LOAD_DWORDX2_IMM %0, 36, 0
    %2:sgpr_128 = COPY $sgpr96_sgpr97_sgpr98_sgpr99
    %3:sgpr_256 = S_LOAD_DWORDX8_IMM %1, 208, 0
    %4:vgpr_32 = COPY %2.sub3
    %5:vreg_64 = BUFFER_LOAD_DWORDX2_OFFSET %2, 0, 0, 0, 0, implicit $exec :: (dereferenceable invariant load 16)
    %6:vgpr_32 = COPY %5.sub0
    %7:vgpr_32 = COPY %5.sub1
    %8:vgpr_32 = IMAGE_LOAD_V1_V2_gfx12 %6, %7, %3, 8, 1, 0, 0, 0, 0, 0, implicit $exec :: (dereferenceable load 4, addrspace 4)
    %9:vreg_96 = IMAGE_LOAD_V3_V2_gfx12 %6, %7, %3, 7, 1, 0, 1, 0, 0, 0, implicit $exec :: (dereferenceable load 12, align 16, addrspace 4)
...
---

# GFX12-LABEL: name: image_load_not_merged_8
# GFX12: %{{[0-9]+}}:vgpr_32 = IMAGE_LOAD_V1_V1_gfx12 %6, %3, 8, 1, 0, 0, 1, 0, 0, implicit $exec :: (dereferenceable load (s64), align 16, addrspace 4)
# GFX12: %{{[0-9]+}}:vreg_96 = IMAGE_LOAD_V3_V2_gfx12 %6, %7, %3, 7, 1, 0, 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s96), align 16, addrspace 4)

name: image_load_not_merged_8
body:             |
  bb.0.entry:
    %0:sgpr_64 = COPY $sgpr0_sgpr1
    %1:sreg_64_xexec = S_LOAD_DWORDX2_IMM %0, 36, 0
    %2:sgpr_128 = COPY $sgpr96_sgpr97_sgpr98_sgpr99
    %3:sgpr_256 = S_LOAD_DWORDX8_IMM %1, 208, 0
    %4:vgpr_32 = COPY %2.sub3
    %5:vreg_64 = BUFFER_LOAD_DWORDX2_OFFSET %2, 0, 0, 0, 0, implicit $exec :: (dereferenceable invariant load 16)
    %6:vgpr_32 = COPY %5.sub0
    %7:vgpr_32 = COPY %5.sub1
    %8:vgpr_32 = IMAGE_LOAD_V1_V1_gfx12 %6, %3, 8, 1, 0, 0, 1, 0, 0, implicit $exec :: (dereferenceable load 8, align 16, addrspace 4)
    %9:vreg_96 = IMAGE_LOAD_V3_V2_gfx12 %6, %7, %3, 7, 1, 0, 0, 0, 0, 0, implicit $exec :: (dereferenceable load 12, align 16, addrspace 4)
...
---

# GFX12-LABEL: name: image_load_not_merged_9
# GFX12: %{{[0-9]+}}:vreg_64 = IMAGE_LOAD_V2_V2_gfx12 %6, %7, %3, 8, 1, 0, 0, 0, 1, 0, implicit $exec :: (dereferenceable load (s32), addrspace 4)
# GFX12: %{{[0-9]+}}:vreg_96 = IMAGE_LOAD_V3_V2_gfx12 %6, %7, %3, 7, 1, 0, 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s96), align 16, addrspace 4)

name: image_load_not_merged_9
body:             |
  bb.0.entry:
    %0:sgpr_64 = COPY $sgpr0_sgpr1
    %1:sreg_64_xexec = S_LOAD_DWORDX2_IMM %0, 36, 0
    %2:sgpr_128 = COPY $sgpr96_sgpr97_sgpr98_sgpr99
    %3:sgpr_256 = S_LOAD_DWORDX8_IMM %1, 208, 0
    %4:vgpr_32 = COPY %2.sub3
    %5:vreg_64 = BUFFER_LOAD_DWORDX2_OFFSET %2, 0, 0, 0, 0, implicit $exec :: (dereferenceable invariant load 16)
    %6:vgpr_32 = COPY %5.sub0
    %7:vgpr_32 = COPY %5.sub1
    %8:vreg_64 = IMAGE_LOAD_V2_V2_gfx12 %6, %7, %3, 8, 1, 0, 0, 0, 1, 0, implicit $exec :: (dereferenceable load 4, addrspace 4)
    %9:vreg_96 = IMAGE_LOAD_V3_V2_gfx12 %6, %7, %3, 7, 1, 0, 0, 0, 0, 0, implicit $exec :: (dereferenceable load 12, align 16, addrspace 4)
...
---

# GFX12-LABEL: name: image_load_not_merged_11
# GFX12: %{{[0-9]+}}:vgpr_32 = IMAGE_LOAD_V1_V2_gfx12 %6, %7, %3, 8, 1, 0, 0, 0, 0, 1, implicit $exec :: (dereferenceable load (s32), addrspace 4)
# GFX12: %{{[0-9]+}}:vreg_96 = IMAGE_LOAD_V3_V2_gfx12 %6, %7, %3, 7, 1, 0, 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s96), align 16, addrspace 4)

name: image_load_not_merged_11
body:             |
  bb.0.entry:
    %0:sgpr_64 = COPY $sgpr0_sgpr1
    %1:sreg_64_xexec = S_LOAD_DWORDX2_IMM %0, 36, 0
    %2:sgpr_128 = COPY $sgpr96_sgpr97_sgpr98_sgpr99
    %3:sgpr_256 = S_LOAD_DWORDX8_IMM %1, 208, 0
    %4:vgpr_32 = COPY %2.sub3
    %5:vreg_64 = BUFFER_LOAD_DWORDX2_OFFSET %2, 0, 0, 0, 0, implicit $exec :: (dereferenceable invariant load 16)
    %6:vgpr_32 = COPY %5.sub0
    %7:vgpr_32 = COPY %5.sub1
    %8:vgpr_32 = IMAGE_LOAD_V1_V2_gfx12 %6, %7, %3, 8, 1, 0, 0, 0, 0, 1, implicit $exec :: (dereferenceable load 4, addrspace 4)
    %9:vreg_96 = IMAGE_LOAD_V3_V2_gfx12 %6, %7, %3, 7, 1, 0, 0, 0, 0, 0, implicit $exec :: (dereferenceable load 12, align 16, addrspace 4)
...
---

# GFX12-LABEL: name: image_load_mip_merged_v1v3
# GFX12: %{{[0-9]+}}:vreg_128 = IMAGE_LOAD_MIP_V4_V3_gfx12 %6, %7, %8, %3, 15, 1, 0, 0, 0, 0, 0, implicit $exec, implicit $exec :: (dereferenceable load (s128), align 4, addrspace 4)
# GFX12: %{{[0-9]+}}:vgpr_32 = COPY %11.sub0
# GFX12: %{{[0-9]+}}:vreg_96 = COPY killed %11.sub1_sub2_sub3

name: image_load_mip_merged_v1v3
body:             |
  bb.0.entry:
    %0:sgpr_64 = COPY $sgpr0_sgpr1
    %1:sreg_64_xexec = S_LOAD_DWORDX2_IMM %0, 36, 0
    %2:sgpr_128 = COPY $sgpr96_sgpr97_sgpr98_sgpr99
    %3:sgpr_256 = S_LOAD_DWORDX8_IMM %1, 208, 0
    %4:vgpr_32 = COPY %2.sub3
    %5:vreg_96 = BUFFER_LOAD_DWORDX3_OFFSET %2:sgpr_128, 0, 0, 0, 0, implicit $exec :: (dereferenceable invariant load 16)
    %6:vgpr_32 = COPY %5.sub0
    %7:vgpr_32 = COPY %5.sub1
    %8:vgpr_32 = COPY %5.sub2
    %9:vgpr_32 = IMAGE_LOAD_MIP_V1_V3_gfx12 %6:vgpr_32, %7:vgpr_32, %8:vgpr_32, %3:sgpr_256, 1, 1, 0, 0, 0, 0, 0, implicit $exec :: (dereferenceable load 4, addrspace 4)
    %10:vreg_96 = IMAGE_LOAD_MIP_V3_V3_gfx12 %6:vgpr_32, %7:vgpr_32, %8:vgpr_32, %3:sgpr_256, 14, 1, 0, 0, 0, 0, 0, implicit $exec :: (dereferenceable load 12, align 16, addrspace 4)
...
---

# GFX12-LABEL: name: image_load_mip_pck_merged_v1v3
# GFX12: %{{[0-9]+}}:vreg_128 = IMAGE_LOAD_MIP_PCK_V4_V3_gfx12 %6, %7, %8, %3, 15, 1, 0, 0, 0, 0, implicit $exec, implicit $exec :: (dereferenceable load (s128), align 4, addrspace 4)
# GFX12: %{{[0-9]+}}:vgpr_32 = COPY %11.sub0
# GFX12: %{{[0-9]+}}:vreg_96 = COPY killed %11.sub1_sub2_sub3

name: image_load_mip_pck_merged_v1v3
body:             |
  bb.0.entry:
    %0:sgpr_64 = COPY $sgpr0_sgpr1
    %1:sreg_64_xexec = S_LOAD_DWORDX2_IMM %0, 36, 0
    %2:sgpr_128 = COPY $sgpr96_sgpr97_sgpr98_sgpr99
    %3:sgpr_256 = S_LOAD_DWORDX8_IMM %1, 208, 0
    %4:vgpr_32 = COPY %2.sub3
    %5:vreg_96 = BUFFER_LOAD_DWORDX3_OFFSET %2:sgpr_128, 0, 0, 0, 0, implicit $exec :: (dereferenceable invariant load 16)
    %6:vgpr_32 = COPY %5.sub0
    %7:vgpr_32 = COPY %5.sub1
    %8:vgpr_32 = COPY %5.sub2
    %9:vgpr_32 = IMAGE_LOAD_MIP_PCK_V1_V3_gfx12 %6:vgpr_32, %7:vgpr_32, %8:vgpr_32, %3:sgpr_256, 1, 1, 0, 0, 0, 0, implicit $exec :: (dereferenceable load 4, addrspace 4)
    %10:vreg_96 = IMAGE_LOAD_MIP_PCK_V3_V3_gfx12 %6:vgpr_32, %7:vgpr_32, %8:vgpr_32, %3:sgpr_256, 14, 1, 0, 0, 0, 0, implicit $exec :: (dereferenceable load 12, align 16, addrspace 4)
...
---

# GFX12-LABEL: name: image_load_mip_pck_sgn_merged_v1v3
# GFX12: %{{[0-9]+}}:vreg_128 = IMAGE_LOAD_MIP_PCK_SGN_V4_V3_gfx12 %6, %7, %8, %3, 15, 1, 0, 0, 0, 0, implicit $exec, implicit $exec :: (dereferenceable load (s128), align 4, addrspace 4)
# GFX12: %{{[0-9]+}}:vgpr_32 = COPY %11.sub0
# GFX12: %{{[0-9]+}}:vreg_96 = COPY killed %11.sub1_sub2_sub3

name: image_load_mip_pck_sgn_merged_v1v3
body:             |
  bb.0.entry:
    %0:sgpr_64 = COPY $sgpr0_sgpr1
    %1:sreg_64_xexec = S_LOAD_DWORDX2_IMM %0, 36, 0
    %2:sgpr_128 = COPY $sgpr96_sgpr97_sgpr98_sgpr99
    %3:sgpr_256 = S_LOAD_DWORDX8_IMM %1, 208, 0
    %4:vgpr_32 = COPY %2.sub3
    %5:vreg_96 = BUFFER_LOAD_DWORDX3_OFFSET %2:sgpr_128, 0, 0, 0, 0, implicit $exec :: (dereferenceable invariant load 16)
    %6:vgpr_32 = COPY %5.sub0
    %7:vgpr_32 = COPY %5.sub1
    %8:vgpr_32 = COPY %5.sub2
    %9:vgpr_32 = IMAGE_LOAD_MIP_PCK_SGN_V1_V3_gfx12 %6:vgpr_32, %7:vgpr_32, %8:vgpr_32, %3:sgpr_256, 1, 1, 0, 0, 0, 0, implicit $exec :: (dereferenceable load 4, addrspace 4)
    %10:vreg_96 = IMAGE_LOAD_MIP_PCK_SGN_V3_V3_gfx12 %6:vgpr_32, %7:vgpr_32, %8:vgpr_32, %3:sgpr_256, 14, 1, 0, 0, 0, 0, implicit $exec :: (dereferenceable load 12, align 16, addrspace 4)
...
---

# GFX12-LABEL: name: image_load_pck_merged_v1v3
# GFX12: %{{[0-9]+}}:vreg_128 = IMAGE_LOAD_PCK_V4_V2_gfx12 %6, %7, %3, 15, 1, 0, 0, 0, 0, implicit $exec, implicit $exec :: (dereferenceable load (s128), align 4, addrspace 4)
# GFX12: %{{[0-9]+}}:vgpr_32 = COPY %10.sub0
# GFX12: %{{[0-9]+}}:vreg_96 = COPY killed %10.sub1_sub2_sub3

name: image_load_pck_merged_v1v3
body:             |
  bb.0.entry:
    %0:sgpr_64 = COPY $sgpr0_sgpr1
    %1:sreg_64_xexec = S_LOAD_DWORDX2_IMM %0, 36, 0
    %2:sgpr_128 = COPY $sgpr96_sgpr97_sgpr98_sgpr99
    %3:sgpr_256 = S_LOAD_DWORDX8_IMM %1, 208, 0
    %4:vgpr_32 = COPY %2.sub3
    %5:vreg_64 = BUFFER_LOAD_DWORDX2_OFFSET %2:sgpr_128, 0, 0, 0, 0, implicit $exec :: (dereferenceable invariant load 16)
    %6:vgpr_32 = COPY %5.sub0
    %7:vgpr_32 = COPY %5.sub1
    %8:vgpr_32 = IMAGE_LOAD_PCK_V1_V2_gfx12 %6:vgpr_32, %7:vgpr_32, %3:sgpr_256, 1, 1, 0, 0, 0, 0, implicit $exec :: (dereferenceable load 4, addrspace 4)
    %9:vreg_96 = IMAGE_LOAD_PCK_V3_V2_gfx12 %6:vgpr_32, %7:vgpr_32, %3:sgpr_256, 14, 1, 0, 0, 0, 0, implicit $exec :: (dereferenceable load 12, align 16, addrspace 4)
...
---

# GFX12-LABEL: name: image_load_pck_sgn_merged_v1v3
# GFX12: %{{[0-9]+}}:vreg_128 = IMAGE_LOAD_PCK_SGN_V4_V2_gfx12 %6, %7, %3, 15, 1, 0, 0, 0, 0, implicit $exec, implicit $exec :: (dereferenceable load (s128), align 4, addrspace 4)
# GFX12: %{{[0-9]+}}:vgpr_32 = COPY %10.sub0
# GFX12: %{{[0-9]+}}:vreg_96 = COPY killed %10.sub1_sub2_sub3

name: image_load_pck_sgn_merged_v1v3
body:             |
  bb.0.entry:
    %0:sgpr_64 = COPY $sgpr0_sgpr1
    %1:sreg_64_xexec = S_LOAD_DWORDX2_IMM %0, 36, 0
    %2:sgpr_128 = COPY $sgpr96_sgpr97_sgpr98_sgpr99
    %3:sgpr_256 = S_LOAD_DWORDX8_IMM %1, 208, 0
    %4:vgpr_32 = COPY %2.sub3
    %5:vreg_64 = BUFFER_LOAD_DWORDX2_OFFSET %2:sgpr_128, 0, 0, 0, 0, implicit $exec :: (dereferenceable invariant load 16)
    %6:vgpr_32 = COPY %5.sub0
    %7:vgpr_32 = COPY %5.sub1
    %8:vgpr_32 = IMAGE_LOAD_PCK_SGN_V1_V2_gfx12 %6:vgpr_32, %7:vgpr_32, %3:sgpr_256, 1, 1, 0, 0, 0, 0, implicit $exec :: (dereferenceable load 4, addrspace 4)
    %9:vreg_96 = IMAGE_LOAD_PCK_SGN_V3_V2_gfx12 %6:vgpr_32, %7:vgpr_32, %3:sgpr_256, 14, 1, 0, 0, 0, 0, implicit $exec :: (dereferenceable load 12, align 16, addrspace 4)
...
---
llvm/llvm/test/CodeGen/AMDGPU/merge-image-load-gfx12.mir
