| # NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py |
| # RUN: llc -amdgpu-global-isel-new-legality -march=amdgcn -mcpu=hawaii -run-pass=instruction-select -verify-machineinstrs -global-isel-abort=0 -o - %s | FileCheck -check-prefix=GFX7 %s |
| # RUN: llc -amdgpu-global-isel-new-legality -march=amdgcn -mcpu=fiji -run-pass=instruction-select -verify-machineinstrs -global-isel-abort=0 -o - %s | FileCheck -check-prefix=GFX8 %s |
| # RUN: llc -amdgpu-global-isel-new-legality -march=amdgcn -mcpu=gfx900 -run-pass=instruction-select -verify-machineinstrs -global-isel-abort=0 -o - %s | FileCheck -check-prefix=GFX9 %s |
| # RUN: llc -amdgpu-global-isel-new-legality -march=amdgcn -mcpu=gfx1010 -run-pass=instruction-select -verify-machineinstrs -global-isel-abort=0 -o - %s | FileCheck -check-prefix=GFX10 %s |
| # RUN: llc -amdgpu-global-isel-new-legality -march=amdgcn -mcpu=gfx1100 -run-pass=instruction-select -verify-machineinstrs -global-isel-abort=0 -o - %s | FileCheck -check-prefix=GFX11 %s |
| |
| |
| --- |
| |
| name: load_flat_s32_from_4 |
| legalized: true |
| regBankSelected: true |
| tracksRegLiveness: true |
| |
| body: | |
| bb.0: |
| liveins: $vgpr0_vgpr1 |
| |
| ; GFX7-LABEL: name: load_flat_s32_from_4 |
| ; GFX7: liveins: $vgpr0_vgpr1 |
| ; GFX7-NEXT: {{ $}} |
| ; GFX7-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1 |
| ; GFX7-NEXT: [[FLAT_LOAD_DWORD:%[0-9]+]]:vgpr_32 = FLAT_LOAD_DWORD [[COPY]], 0, 0, implicit $exec, implicit $flat_scr :: (load (s32)) |
| ; GFX7-NEXT: $vgpr0 = COPY [[FLAT_LOAD_DWORD]] |
| ; GFX8-LABEL: name: load_flat_s32_from_4 |
| ; GFX8: liveins: $vgpr0_vgpr1 |
| ; GFX8-NEXT: {{ $}} |
| ; GFX8-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1 |
| ; GFX8-NEXT: [[FLAT_LOAD_DWORD:%[0-9]+]]:vgpr_32 = FLAT_LOAD_DWORD [[COPY]], 0, 0, implicit $exec, implicit $flat_scr :: (load (s32)) |
| ; GFX8-NEXT: $vgpr0 = COPY [[FLAT_LOAD_DWORD]] |
| ; GFX9-LABEL: name: load_flat_s32_from_4 |
| ; GFX9: liveins: $vgpr0_vgpr1 |
| ; GFX9-NEXT: {{ $}} |
| ; GFX9-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1 |
| ; GFX9-NEXT: [[FLAT_LOAD_DWORD:%[0-9]+]]:vgpr_32 = FLAT_LOAD_DWORD [[COPY]], 0, 0, implicit $exec, implicit $flat_scr :: (load (s32)) |
| ; GFX9-NEXT: $vgpr0 = COPY [[FLAT_LOAD_DWORD]] |
| ; GFX10-LABEL: name: load_flat_s32_from_4 |
| ; GFX10: liveins: $vgpr0_vgpr1 |
| ; GFX10-NEXT: {{ $}} |
| ; GFX10-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1 |
| ; GFX10-NEXT: [[FLAT_LOAD_DWORD:%[0-9]+]]:vgpr_32 = FLAT_LOAD_DWORD [[COPY]], 0, 0, implicit $exec, implicit $flat_scr :: (load (s32)) |
| ; GFX10-NEXT: $vgpr0 = COPY [[FLAT_LOAD_DWORD]] |
| ; GFX11-LABEL: name: load_flat_s32_from_4 |
| ; GFX11: liveins: $vgpr0_vgpr1 |
| ; GFX11-NEXT: {{ $}} |
| ; GFX11-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1 |
| ; GFX11-NEXT: [[FLAT_LOAD_DWORD:%[0-9]+]]:vgpr_32 = FLAT_LOAD_DWORD [[COPY]], 0, 0, implicit $exec, implicit $flat_scr :: (load (s32)) |
| ; GFX11-NEXT: $vgpr0 = COPY [[FLAT_LOAD_DWORD]] |
| %0:vgpr(p1) = COPY $vgpr0_vgpr1 |
| %1:vgpr(s32) = G_LOAD %0 :: (load (s32), align 4, addrspace 0) |
| $vgpr0 = COPY %1 |
| |
| ... |
| |
| --- |
| |
| name: load_flat_s32_from_2 |
| legalized: true |
| regBankSelected: true |
| tracksRegLiveness: true |
| |
| body: | |
| bb.0: |
| liveins: $vgpr0_vgpr1 |
| |
| ; GFX7-LABEL: name: load_flat_s32_from_2 |
| ; GFX7: liveins: $vgpr0_vgpr1 |
| ; GFX7-NEXT: {{ $}} |
| ; GFX7-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1 |
| ; GFX7-NEXT: [[FLAT_LOAD_USHORT:%[0-9]+]]:vgpr_32 = FLAT_LOAD_USHORT [[COPY]], 0, 0, implicit $exec, implicit $flat_scr :: (load (s16)) |
| ; GFX7-NEXT: $vgpr0 = COPY [[FLAT_LOAD_USHORT]] |
| ; GFX8-LABEL: name: load_flat_s32_from_2 |
| ; GFX8: liveins: $vgpr0_vgpr1 |
| ; GFX8-NEXT: {{ $}} |
| ; GFX8-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1 |
| ; GFX8-NEXT: [[FLAT_LOAD_USHORT:%[0-9]+]]:vgpr_32 = FLAT_LOAD_USHORT [[COPY]], 0, 0, implicit $exec, implicit $flat_scr :: (load (s16)) |
| ; GFX8-NEXT: $vgpr0 = COPY [[FLAT_LOAD_USHORT]] |
| ; GFX9-LABEL: name: load_flat_s32_from_2 |
| ; GFX9: liveins: $vgpr0_vgpr1 |
| ; GFX9-NEXT: {{ $}} |
| ; GFX9-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1 |
| ; GFX9-NEXT: [[FLAT_LOAD_USHORT:%[0-9]+]]:vgpr_32 = FLAT_LOAD_USHORT [[COPY]], 0, 0, implicit $exec, implicit $flat_scr :: (load (s16)) |
| ; GFX9-NEXT: $vgpr0 = COPY [[FLAT_LOAD_USHORT]] |
| ; GFX10-LABEL: name: load_flat_s32_from_2 |
| ; GFX10: liveins: $vgpr0_vgpr1 |
| ; GFX10-NEXT: {{ $}} |
| ; GFX10-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1 |
| ; GFX10-NEXT: [[FLAT_LOAD_USHORT:%[0-9]+]]:vgpr_32 = FLAT_LOAD_USHORT [[COPY]], 0, 0, implicit $exec, implicit $flat_scr :: (load (s16)) |
| ; GFX10-NEXT: $vgpr0 = COPY [[FLAT_LOAD_USHORT]] |
| ; GFX11-LABEL: name: load_flat_s32_from_2 |
| ; GFX11: liveins: $vgpr0_vgpr1 |
| ; GFX11-NEXT: {{ $}} |
| ; GFX11-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1 |
| ; GFX11-NEXT: [[FLAT_LOAD_USHORT:%[0-9]+]]:vgpr_32 = FLAT_LOAD_USHORT [[COPY]], 0, 0, implicit $exec, implicit $flat_scr :: (load (s16)) |
| ; GFX11-NEXT: $vgpr0 = COPY [[FLAT_LOAD_USHORT]] |
| %0:vgpr(p1) = COPY $vgpr0_vgpr1 |
| %1:vgpr(s32) = G_LOAD %0 :: (load (s16), align 2, addrspace 0) |
| $vgpr0 = COPY %1 |
| |
| ... |
| |
| --- |
| |
| name: load_flat_s32_from_1 |
| legalized: true |
| regBankSelected: true |
| tracksRegLiveness: true |
| |
| body: | |
| bb.0: |
| liveins: $vgpr0_vgpr1 |
| |
| ; GFX7-LABEL: name: load_flat_s32_from_1 |
| ; GFX7: liveins: $vgpr0_vgpr1 |
| ; GFX7-NEXT: {{ $}} |
| ; GFX7-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1 |
| ; GFX7-NEXT: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[COPY]], 0, 0, implicit $exec, implicit $flat_scr :: (load (s8)) |
| ; GFX7-NEXT: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]] |
| ; GFX8-LABEL: name: load_flat_s32_from_1 |
| ; GFX8: liveins: $vgpr0_vgpr1 |
| ; GFX8-NEXT: {{ $}} |
| ; GFX8-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1 |
| ; GFX8-NEXT: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[COPY]], 0, 0, implicit $exec, implicit $flat_scr :: (load (s8)) |
| ; GFX8-NEXT: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]] |
| ; GFX9-LABEL: name: load_flat_s32_from_1 |
| ; GFX9: liveins: $vgpr0_vgpr1 |
| ; GFX9-NEXT: {{ $}} |
| ; GFX9-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1 |
| ; GFX9-NEXT: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[COPY]], 0, 0, implicit $exec, implicit $flat_scr :: (load (s8)) |
| ; GFX9-NEXT: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]] |
| ; GFX10-LABEL: name: load_flat_s32_from_1 |
| ; GFX10: liveins: $vgpr0_vgpr1 |
| ; GFX10-NEXT: {{ $}} |
| ; GFX10-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1 |
| ; GFX10-NEXT: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[COPY]], 0, 0, implicit $exec, implicit $flat_scr :: (load (s8)) |
| ; GFX10-NEXT: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]] |
| ; GFX11-LABEL: name: load_flat_s32_from_1 |
| ; GFX11: liveins: $vgpr0_vgpr1 |
| ; GFX11-NEXT: {{ $}} |
| ; GFX11-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1 |
| ; GFX11-NEXT: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[COPY]], 0, 0, implicit $exec, implicit $flat_scr :: (load (s8)) |
| ; GFX11-NEXT: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]] |
| %0:vgpr(p1) = COPY $vgpr0_vgpr1 |
| %1:vgpr(s32) = G_LOAD %0 :: (load (s8), align 1, addrspace 0) |
| $vgpr0 = COPY %1 |
| |
| ... |
| |
| --- |
| |
| name: load_flat_v2s32 |
| legalized: true |
| regBankSelected: true |
| |
| |
| body: | |
| bb.0: |
| liveins: $vgpr0_vgpr1 |
| |
| ; GFX7-LABEL: name: load_flat_v2s32 |
| ; GFX7: liveins: $vgpr0_vgpr1 |
| ; GFX7-NEXT: {{ $}} |
| ; GFX7-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1 |
| ; GFX7-NEXT: [[FLAT_LOAD_DWORDX2_:%[0-9]+]]:vreg_64 = FLAT_LOAD_DWORDX2 [[COPY]], 0, 0, implicit $exec, implicit $flat_scr :: (load (<2 x s32>)) |
| ; GFX7-NEXT: $vgpr0_vgpr1 = COPY [[FLAT_LOAD_DWORDX2_]] |
| ; GFX8-LABEL: name: load_flat_v2s32 |
| ; GFX8: liveins: $vgpr0_vgpr1 |
| ; GFX8-NEXT: {{ $}} |
| ; GFX8-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1 |
| ; GFX8-NEXT: [[FLAT_LOAD_DWORDX2_:%[0-9]+]]:vreg_64 = FLAT_LOAD_DWORDX2 [[COPY]], 0, 0, implicit $exec, implicit $flat_scr :: (load (<2 x s32>)) |
| ; GFX8-NEXT: $vgpr0_vgpr1 = COPY [[FLAT_LOAD_DWORDX2_]] |
| ; GFX9-LABEL: name: load_flat_v2s32 |
| ; GFX9: liveins: $vgpr0_vgpr1 |
| ; GFX9-NEXT: {{ $}} |
| ; GFX9-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1 |
| ; GFX9-NEXT: [[FLAT_LOAD_DWORDX2_:%[0-9]+]]:vreg_64 = FLAT_LOAD_DWORDX2 [[COPY]], 0, 0, implicit $exec, implicit $flat_scr :: (load (<2 x s32>)) |
| ; GFX9-NEXT: $vgpr0_vgpr1 = COPY [[FLAT_LOAD_DWORDX2_]] |
| ; GFX10-LABEL: name: load_flat_v2s32 |
| ; GFX10: liveins: $vgpr0_vgpr1 |
| ; GFX10-NEXT: {{ $}} |
| ; GFX10-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1 |
| ; GFX10-NEXT: [[FLAT_LOAD_DWORDX2_:%[0-9]+]]:vreg_64 = FLAT_LOAD_DWORDX2 [[COPY]], 0, 0, implicit $exec, implicit $flat_scr :: (load (<2 x s32>)) |
| ; GFX10-NEXT: $vgpr0_vgpr1 = COPY [[FLAT_LOAD_DWORDX2_]] |
| ; GFX11-LABEL: name: load_flat_v2s32 |
| ; GFX11: liveins: $vgpr0_vgpr1 |
| ; GFX11-NEXT: {{ $}} |
| ; GFX11-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1 |
| ; GFX11-NEXT: [[FLAT_LOAD_DWORDX2_:%[0-9]+]]:vreg_64 = FLAT_LOAD_DWORDX2 [[COPY]], 0, 0, implicit $exec, implicit $flat_scr :: (load (<2 x s32>)) |
| ; GFX11-NEXT: $vgpr0_vgpr1 = COPY [[FLAT_LOAD_DWORDX2_]] |
| %0:vgpr(p1) = COPY $vgpr0_vgpr1 |
| %1:vgpr(<2 x s32>) = G_LOAD %0 :: (load (<2 x s32>), align 8, addrspace 0) |
| $vgpr0_vgpr1 = COPY %1 |
| |
| ... |
| |
| --- |
| |
| name: load_flat_v3s32 |
| legalized: true |
| regBankSelected: true |
| tracksRegLiveness: true |
| |
| body: | |
| bb.0: |
| liveins: $vgpr0_vgpr1 |
| |
| ; GFX7-LABEL: name: load_flat_v3s32 |
| ; GFX7: liveins: $vgpr0_vgpr1 |
| ; GFX7-NEXT: {{ $}} |
| ; GFX7-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1 |
| ; GFX7-NEXT: [[FLAT_LOAD_DWORDX3_:%[0-9]+]]:vreg_96 = FLAT_LOAD_DWORDX3 [[COPY]], 0, 0, implicit $exec, implicit $flat_scr :: (load (<3 x s32>), align 4) |
| ; GFX7-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[FLAT_LOAD_DWORDX3_]] |
| ; GFX8-LABEL: name: load_flat_v3s32 |
| ; GFX8: liveins: $vgpr0_vgpr1 |
| ; GFX8-NEXT: {{ $}} |
| ; GFX8-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1 |
| ; GFX8-NEXT: [[FLAT_LOAD_DWORDX3_:%[0-9]+]]:vreg_96 = FLAT_LOAD_DWORDX3 [[COPY]], 0, 0, implicit $exec, implicit $flat_scr :: (load (<3 x s32>), align 4) |
| ; GFX8-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[FLAT_LOAD_DWORDX3_]] |
| ; GFX9-LABEL: name: load_flat_v3s32 |
| ; GFX9: liveins: $vgpr0_vgpr1 |
| ; GFX9-NEXT: {{ $}} |
| ; GFX9-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1 |
| ; GFX9-NEXT: [[FLAT_LOAD_DWORDX3_:%[0-9]+]]:vreg_96 = FLAT_LOAD_DWORDX3 [[COPY]], 0, 0, implicit $exec, implicit $flat_scr :: (load (<3 x s32>), align 4) |
| ; GFX9-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[FLAT_LOAD_DWORDX3_]] |
| ; GFX10-LABEL: name: load_flat_v3s32 |
| ; GFX10: liveins: $vgpr0_vgpr1 |
| ; GFX10-NEXT: {{ $}} |
| ; GFX10-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1 |
| ; GFX10-NEXT: [[FLAT_LOAD_DWORDX3_:%[0-9]+]]:vreg_96 = FLAT_LOAD_DWORDX3 [[COPY]], 0, 0, implicit $exec, implicit $flat_scr :: (load (<3 x s32>), align 4) |
| ; GFX10-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[FLAT_LOAD_DWORDX3_]] |
| ; GFX11-LABEL: name: load_flat_v3s32 |
| ; GFX11: liveins: $vgpr0_vgpr1 |
| ; GFX11-NEXT: {{ $}} |
| ; GFX11-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1 |
| ; GFX11-NEXT: [[FLAT_LOAD_DWORDX3_:%[0-9]+]]:vreg_96 = FLAT_LOAD_DWORDX3 [[COPY]], 0, 0, implicit $exec, implicit $flat_scr :: (load (<3 x s32>), align 4) |
| ; GFX11-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[FLAT_LOAD_DWORDX3_]] |
| %0:vgpr(p1) = COPY $vgpr0_vgpr1 |
| %1:vgpr(<3 x s32>) = G_LOAD %0 :: (load (<3 x s32>), align 4, addrspace 0) |
| $vgpr0_vgpr1_vgpr2 = COPY %1 |
| |
| ... |
| |
| --- |
| |
| name: load_flat_v4s32 |
| legalized: true |
| regBankSelected: true |
| tracksRegLiveness: true |
| |
| body: | |
| bb.0: |
| liveins: $vgpr0_vgpr1 |
| |
| ; GFX7-LABEL: name: load_flat_v4s32 |
| ; GFX7: liveins: $vgpr0_vgpr1 |
| ; GFX7-NEXT: {{ $}} |
| ; GFX7-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1 |
| ; GFX7-NEXT: [[FLAT_LOAD_DWORDX4_:%[0-9]+]]:vreg_128 = FLAT_LOAD_DWORDX4 [[COPY]], 0, 0, implicit $exec, implicit $flat_scr :: (load (<4 x s32>), align 4) |
| ; GFX7-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[FLAT_LOAD_DWORDX4_]] |
| ; GFX8-LABEL: name: load_flat_v4s32 |
| ; GFX8: liveins: $vgpr0_vgpr1 |
| ; GFX8-NEXT: {{ $}} |
| ; GFX8-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1 |
| ; GFX8-NEXT: [[FLAT_LOAD_DWORDX4_:%[0-9]+]]:vreg_128 = FLAT_LOAD_DWORDX4 [[COPY]], 0, 0, implicit $exec, implicit $flat_scr :: (load (<4 x s32>), align 4) |
| ; GFX8-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[FLAT_LOAD_DWORDX4_]] |
| ; GFX9-LABEL: name: load_flat_v4s32 |
| ; GFX9: liveins: $vgpr0_vgpr1 |
| ; GFX9-NEXT: {{ $}} |
| ; GFX9-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1 |
| ; GFX9-NEXT: [[FLAT_LOAD_DWORDX4_:%[0-9]+]]:vreg_128 = FLAT_LOAD_DWORDX4 [[COPY]], 0, 0, implicit $exec, implicit $flat_scr :: (load (<4 x s32>), align 4) |
| ; GFX9-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[FLAT_LOAD_DWORDX4_]] |
| ; GFX10-LABEL: name: load_flat_v4s32 |
| ; GFX10: liveins: $vgpr0_vgpr1 |
| ; GFX10-NEXT: {{ $}} |
| ; GFX10-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1 |
| ; GFX10-NEXT: [[FLAT_LOAD_DWORDX4_:%[0-9]+]]:vreg_128 = FLAT_LOAD_DWORDX4 [[COPY]], 0, 0, implicit $exec, implicit $flat_scr :: (load (<4 x s32>), align 4) |
| ; GFX10-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[FLAT_LOAD_DWORDX4_]] |
| ; GFX11-LABEL: name: load_flat_v4s32 |
| ; GFX11: liveins: $vgpr0_vgpr1 |
| ; GFX11-NEXT: {{ $}} |
| ; GFX11-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1 |
| ; GFX11-NEXT: [[FLAT_LOAD_DWORDX4_:%[0-9]+]]:vreg_128 = FLAT_LOAD_DWORDX4 [[COPY]], 0, 0, implicit $exec, implicit $flat_scr :: (load (<4 x s32>), align 4) |
| ; GFX11-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[FLAT_LOAD_DWORDX4_]] |
| %0:vgpr(p1) = COPY $vgpr0_vgpr1 |
| %1:vgpr(<4 x s32>) = G_LOAD %0 :: (load (<4 x s32>), align 4, addrspace 0) |
| $vgpr0_vgpr1_vgpr2_vgpr3 = COPY %1 |
| |
| ... |
| |
| --- |
| |
| name: load_flat_s64 |
| legalized: true |
| regBankSelected: true |
| tracksRegLiveness: true |
| |
| body: | |
| bb.0: |
| liveins: $vgpr0_vgpr1 |
| |
| ; GFX7-LABEL: name: load_flat_s64 |
| ; GFX7: liveins: $vgpr0_vgpr1 |
| ; GFX7-NEXT: {{ $}} |
| ; GFX7-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1 |
| ; GFX7-NEXT: [[FLAT_LOAD_DWORDX2_:%[0-9]+]]:vreg_64 = FLAT_LOAD_DWORDX2 [[COPY]], 0, 0, implicit $exec, implicit $flat_scr :: (load (s64)) |
| ; GFX7-NEXT: $vgpr0_vgpr1 = COPY [[FLAT_LOAD_DWORDX2_]] |
| ; GFX8-LABEL: name: load_flat_s64 |
| ; GFX8: liveins: $vgpr0_vgpr1 |
| ; GFX8-NEXT: {{ $}} |
| ; GFX8-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1 |
| ; GFX8-NEXT: [[FLAT_LOAD_DWORDX2_:%[0-9]+]]:vreg_64 = FLAT_LOAD_DWORDX2 [[COPY]], 0, 0, implicit $exec, implicit $flat_scr :: (load (s64)) |
| ; GFX8-NEXT: $vgpr0_vgpr1 = COPY [[FLAT_LOAD_DWORDX2_]] |
| ; GFX9-LABEL: name: load_flat_s64 |
| ; GFX9: liveins: $vgpr0_vgpr1 |
| ; GFX9-NEXT: {{ $}} |
| ; GFX9-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1 |
| ; GFX9-NEXT: [[FLAT_LOAD_DWORDX2_:%[0-9]+]]:vreg_64 = FLAT_LOAD_DWORDX2 [[COPY]], 0, 0, implicit $exec, implicit $flat_scr :: (load (s64)) |
| ; GFX9-NEXT: $vgpr0_vgpr1 = COPY [[FLAT_LOAD_DWORDX2_]] |
| ; GFX10-LABEL: name: load_flat_s64 |
| ; GFX10: liveins: $vgpr0_vgpr1 |
| ; GFX10-NEXT: {{ $}} |
| ; GFX10-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1 |
| ; GFX10-NEXT: [[FLAT_LOAD_DWORDX2_:%[0-9]+]]:vreg_64 = FLAT_LOAD_DWORDX2 [[COPY]], 0, 0, implicit $exec, implicit $flat_scr :: (load (s64)) |
| ; GFX10-NEXT: $vgpr0_vgpr1 = COPY [[FLAT_LOAD_DWORDX2_]] |
| ; GFX11-LABEL: name: load_flat_s64 |
| ; GFX11: liveins: $vgpr0_vgpr1 |
| ; GFX11-NEXT: {{ $}} |
| ; GFX11-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1 |
| ; GFX11-NEXT: [[FLAT_LOAD_DWORDX2_:%[0-9]+]]:vreg_64 = FLAT_LOAD_DWORDX2 [[COPY]], 0, 0, implicit $exec, implicit $flat_scr :: (load (s64)) |
| ; GFX11-NEXT: $vgpr0_vgpr1 = COPY [[FLAT_LOAD_DWORDX2_]] |
| %0:vgpr(p1) = COPY $vgpr0_vgpr1 |
| %1:vgpr(s64) = G_LOAD %0 :: (load (s64), align 8, addrspace 0) |
| $vgpr0_vgpr1 = COPY %1 |
| |
| ... |
| |
| --- |
| |
| name: load_flat_v2s64 |
| legalized: true |
| regBankSelected: true |
| tracksRegLiveness: true |
| |
| body: | |
| bb.0: |
| liveins: $vgpr0_vgpr1 |
| |
| ; GFX7-LABEL: name: load_flat_v2s64 |
| ; GFX7: liveins: $vgpr0_vgpr1 |
| ; GFX7-NEXT: {{ $}} |
| ; GFX7-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1 |
| ; GFX7-NEXT: [[FLAT_LOAD_DWORDX4_:%[0-9]+]]:vreg_128 = FLAT_LOAD_DWORDX4 [[COPY]], 0, 0, implicit $exec, implicit $flat_scr :: (load (<2 x s64>), align 4) |
| ; GFX7-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[FLAT_LOAD_DWORDX4_]] |
| ; GFX8-LABEL: name: load_flat_v2s64 |
| ; GFX8: liveins: $vgpr0_vgpr1 |
| ; GFX8-NEXT: {{ $}} |
| ; GFX8-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1 |
| ; GFX8-NEXT: [[FLAT_LOAD_DWORDX4_:%[0-9]+]]:vreg_128 = FLAT_LOAD_DWORDX4 [[COPY]], 0, 0, implicit $exec, implicit $flat_scr :: (load (<2 x s64>), align 4) |
| ; GFX8-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[FLAT_LOAD_DWORDX4_]] |
| ; GFX9-LABEL: name: load_flat_v2s64 |
| ; GFX9: liveins: $vgpr0_vgpr1 |
| ; GFX9-NEXT: {{ $}} |
| ; GFX9-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1 |
| ; GFX9-NEXT: [[FLAT_LOAD_DWORDX4_:%[0-9]+]]:vreg_128 = FLAT_LOAD_DWORDX4 [[COPY]], 0, 0, implicit $exec, implicit $flat_scr :: (load (<2 x s64>), align 4) |
| ; GFX9-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[FLAT_LOAD_DWORDX4_]] |
| ; GFX10-LABEL: name: load_flat_v2s64 |
| ; GFX10: liveins: $vgpr0_vgpr1 |
| ; GFX10-NEXT: {{ $}} |
| ; GFX10-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1 |
| ; GFX10-NEXT: [[FLAT_LOAD_DWORDX4_:%[0-9]+]]:vreg_128 = FLAT_LOAD_DWORDX4 [[COPY]], 0, 0, implicit $exec, implicit $flat_scr :: (load (<2 x s64>), align 4) |
| ; GFX10-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[FLAT_LOAD_DWORDX4_]] |
| ; GFX11-LABEL: name: load_flat_v2s64 |
| ; GFX11: liveins: $vgpr0_vgpr1 |
| ; GFX11-NEXT: {{ $}} |
| ; GFX11-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1 |
| ; GFX11-NEXT: [[FLAT_LOAD_DWORDX4_:%[0-9]+]]:vreg_128 = FLAT_LOAD_DWORDX4 [[COPY]], 0, 0, implicit $exec, implicit $flat_scr :: (load (<2 x s64>), align 4) |
| ; GFX11-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[FLAT_LOAD_DWORDX4_]] |
| %0:vgpr(p1) = COPY $vgpr0_vgpr1 |
| %1:vgpr(<2 x s64>) = G_LOAD %0 :: (load (<2 x s64>), align 4, addrspace 0) |
| $vgpr0_vgpr1_vgpr2_vgpr3 = COPY %1 |
| |
| ... |
| |
| --- |
| |
| name: load_flat_v2p1 |
| legalized: true |
| regBankSelected: true |
| tracksRegLiveness: true |
| |
| body: | |
| bb.0: |
| liveins: $vgpr0_vgpr1 |
| |
| ; GFX7-LABEL: name: load_flat_v2p1 |
| ; GFX7: liveins: $vgpr0_vgpr1 |
| ; GFX7-NEXT: {{ $}} |
| ; GFX7-NEXT: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1 |
| ; GFX7-NEXT: [[LOAD:%[0-9]+]]:vreg_128(<2 x p1>) = G_LOAD [[COPY]](p1) :: (load (<2 x p1>), align 4) |
| ; GFX7-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<2 x p1>) |
| ; GFX8-LABEL: name: load_flat_v2p1 |
| ; GFX8: liveins: $vgpr0_vgpr1 |
| ; GFX8-NEXT: {{ $}} |
| ; GFX8-NEXT: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1 |
| ; GFX8-NEXT: [[LOAD:%[0-9]+]]:vreg_128(<2 x p1>) = G_LOAD [[COPY]](p1) :: (load (<2 x p1>), align 4) |
| ; GFX8-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<2 x p1>) |
| ; GFX9-LABEL: name: load_flat_v2p1 |
| ; GFX9: liveins: $vgpr0_vgpr1 |
| ; GFX9-NEXT: {{ $}} |
| ; GFX9-NEXT: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1 |
| ; GFX9-NEXT: [[LOAD:%[0-9]+]]:vreg_128(<2 x p1>) = G_LOAD [[COPY]](p1) :: (load (<2 x p1>), align 4) |
| ; GFX9-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<2 x p1>) |
| ; GFX10-LABEL: name: load_flat_v2p1 |
| ; GFX10: liveins: $vgpr0_vgpr1 |
| ; GFX10-NEXT: {{ $}} |
| ; GFX10-NEXT: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1 |
| ; GFX10-NEXT: [[LOAD:%[0-9]+]]:vreg_128(<2 x p1>) = G_LOAD [[COPY]](p1) :: (load (<2 x p1>), align 4) |
| ; GFX10-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<2 x p1>) |
| ; GFX11-LABEL: name: load_flat_v2p1 |
| ; GFX11: liveins: $vgpr0_vgpr1 |
| ; GFX11-NEXT: {{ $}} |
| ; GFX11-NEXT: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1 |
| ; GFX11-NEXT: [[LOAD:%[0-9]+]]:vreg_128(<2 x p1>) = G_LOAD [[COPY]](p1) :: (load (<2 x p1>), align 4) |
| ; GFX11-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<2 x p1>) |
| %0:vgpr(p1) = COPY $vgpr0_vgpr1 |
| %1:vgpr(<2 x p1>) = G_LOAD %0 :: (load (<2 x p1>), align 4, addrspace 0) |
| $vgpr0_vgpr1_vgpr2_vgpr3 = COPY %1 |
| |
| ... |
| |
| --- |
| |
| name: load_flat_s96 |
| legalized: true |
| regBankSelected: true |
| tracksRegLiveness: true |
| |
| body: | |
| bb.0: |
| liveins: $vgpr0_vgpr1 |
| |
| ; GFX7-LABEL: name: load_flat_s96 |
| ; GFX7: liveins: $vgpr0_vgpr1 |
| ; GFX7-NEXT: {{ $}} |
| ; GFX7-NEXT: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1 |
| ; GFX7-NEXT: [[LOAD:%[0-9]+]]:vreg_96(s96) = G_LOAD [[COPY]](p1) :: (load (s96), align 4) |
| ; GFX7-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[LOAD]](s96) |
| ; GFX8-LABEL: name: load_flat_s96 |
| ; GFX8: liveins: $vgpr0_vgpr1 |
| ; GFX8-NEXT: {{ $}} |
| ; GFX8-NEXT: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1 |
| ; GFX8-NEXT: [[LOAD:%[0-9]+]]:vreg_96(s96) = G_LOAD [[COPY]](p1) :: (load (s96), align 4) |
| ; GFX8-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[LOAD]](s96) |
| ; GFX9-LABEL: name: load_flat_s96 |
| ; GFX9: liveins: $vgpr0_vgpr1 |
| ; GFX9-NEXT: {{ $}} |
| ; GFX9-NEXT: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1 |
| ; GFX9-NEXT: [[LOAD:%[0-9]+]]:vreg_96(s96) = G_LOAD [[COPY]](p1) :: (load (s96), align 4) |
| ; GFX9-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[LOAD]](s96) |
| ; GFX10-LABEL: name: load_flat_s96 |
| ; GFX10: liveins: $vgpr0_vgpr1 |
| ; GFX10-NEXT: {{ $}} |
| ; GFX10-NEXT: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1 |
| ; GFX10-NEXT: [[LOAD:%[0-9]+]]:vreg_96(s96) = G_LOAD [[COPY]](p1) :: (load (s96), align 4) |
| ; GFX10-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[LOAD]](s96) |
| ; GFX11-LABEL: name: load_flat_s96 |
| ; GFX11: liveins: $vgpr0_vgpr1 |
| ; GFX11-NEXT: {{ $}} |
| ; GFX11-NEXT: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1 |
| ; GFX11-NEXT: [[LOAD:%[0-9]+]]:vreg_96(s96) = G_LOAD [[COPY]](p1) :: (load (s96), align 4) |
| ; GFX11-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[LOAD]](s96) |
| %0:vgpr(p1) = COPY $vgpr0_vgpr1 |
| %1:vgpr(s96) = G_LOAD %0 :: (load (s96), align 4, addrspace 0) |
| $vgpr0_vgpr1_vgpr2 = COPY %1 |
| |
| ... |
| |
| --- |
| |
| name: load_flat_s128 |
| legalized: true |
| regBankSelected: true |
| tracksRegLiveness: true |
| |
| body: | |
| bb.0: |
| liveins: $vgpr0_vgpr1 |
| |
| ; GFX7-LABEL: name: load_flat_s128 |
| ; GFX7: liveins: $vgpr0_vgpr1 |
| ; GFX7-NEXT: {{ $}} |
| ; GFX7-NEXT: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1 |
| ; GFX7-NEXT: [[LOAD:%[0-9]+]]:vreg_128(s128) = G_LOAD [[COPY]](p1) :: (load (s128), align 4) |
| ; GFX7-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](s128) |
| ; GFX8-LABEL: name: load_flat_s128 |
| ; GFX8: liveins: $vgpr0_vgpr1 |
| ; GFX8-NEXT: {{ $}} |
| ; GFX8-NEXT: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1 |
| ; GFX8-NEXT: [[LOAD:%[0-9]+]]:vreg_128(s128) = G_LOAD [[COPY]](p1) :: (load (s128), align 4) |
| ; GFX8-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](s128) |
| ; GFX9-LABEL: name: load_flat_s128 |
| ; GFX9: liveins: $vgpr0_vgpr1 |
| ; GFX9-NEXT: {{ $}} |
| ; GFX9-NEXT: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1 |
| ; GFX9-NEXT: [[LOAD:%[0-9]+]]:vreg_128(s128) = G_LOAD [[COPY]](p1) :: (load (s128), align 4) |
| ; GFX9-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](s128) |
| ; GFX10-LABEL: name: load_flat_s128 |
| ; GFX10: liveins: $vgpr0_vgpr1 |
| ; GFX10-NEXT: {{ $}} |
| ; GFX10-NEXT: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1 |
| ; GFX10-NEXT: [[LOAD:%[0-9]+]]:vreg_128(s128) = G_LOAD [[COPY]](p1) :: (load (s128), align 4) |
| ; GFX10-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](s128) |
| ; GFX11-LABEL: name: load_flat_s128 |
| ; GFX11: liveins: $vgpr0_vgpr1 |
| ; GFX11-NEXT: {{ $}} |
| ; GFX11-NEXT: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1 |
| ; GFX11-NEXT: [[LOAD:%[0-9]+]]:vreg_128(s128) = G_LOAD [[COPY]](p1) :: (load (s128), align 4) |
| ; GFX11-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](s128) |
| %0:vgpr(p1) = COPY $vgpr0_vgpr1 |
| %1:vgpr(s128) = G_LOAD %0 :: (load (s128), align 4, addrspace 0) |
| $vgpr0_vgpr1_vgpr2_vgpr3 = COPY %1 |
| |
| ... |
| |
| --- |
| |
| name: load_flat_p3_from_4 |
| legalized: true |
| regBankSelected: true |
| tracksRegLiveness: true |
| |
| body: | |
| bb.0: |
| liveins: $vgpr0_vgpr1 |
| |
| ; GFX7-LABEL: name: load_flat_p3_from_4 |
| ; GFX7: liveins: $vgpr0_vgpr1 |
| ; GFX7-NEXT: {{ $}} |
| ; GFX7-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1 |
| ; GFX7-NEXT: [[FLAT_LOAD_DWORD:%[0-9]+]]:vgpr_32 = FLAT_LOAD_DWORD [[COPY]], 0, 0, implicit $exec, implicit $flat_scr :: (load (p3)) |
| ; GFX7-NEXT: $vgpr0 = COPY [[FLAT_LOAD_DWORD]] |
| ; GFX8-LABEL: name: load_flat_p3_from_4 |
| ; GFX8: liveins: $vgpr0_vgpr1 |
| ; GFX8-NEXT: {{ $}} |
| ; GFX8-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1 |
| ; GFX8-NEXT: [[FLAT_LOAD_DWORD:%[0-9]+]]:vgpr_32 = FLAT_LOAD_DWORD [[COPY]], 0, 0, implicit $exec, implicit $flat_scr :: (load (p3)) |
| ; GFX8-NEXT: $vgpr0 = COPY [[FLAT_LOAD_DWORD]] |
| ; GFX9-LABEL: name: load_flat_p3_from_4 |
| ; GFX9: liveins: $vgpr0_vgpr1 |
| ; GFX9-NEXT: {{ $}} |
| ; GFX9-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1 |
| ; GFX9-NEXT: [[FLAT_LOAD_DWORD:%[0-9]+]]:vgpr_32 = FLAT_LOAD_DWORD [[COPY]], 0, 0, implicit $exec, implicit $flat_scr :: (load (p3)) |
| ; GFX9-NEXT: $vgpr0 = COPY [[FLAT_LOAD_DWORD]] |
| ; GFX10-LABEL: name: load_flat_p3_from_4 |
| ; GFX10: liveins: $vgpr0_vgpr1 |
| ; GFX10-NEXT: {{ $}} |
| ; GFX10-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1 |
| ; GFX10-NEXT: [[FLAT_LOAD_DWORD:%[0-9]+]]:vgpr_32 = FLAT_LOAD_DWORD [[COPY]], 0, 0, implicit $exec, implicit $flat_scr :: (load (p3)) |
| ; GFX10-NEXT: $vgpr0 = COPY [[FLAT_LOAD_DWORD]] |
| ; GFX11-LABEL: name: load_flat_p3_from_4 |
| ; GFX11: liveins: $vgpr0_vgpr1 |
| ; GFX11-NEXT: {{ $}} |
| ; GFX11-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1 |
| ; GFX11-NEXT: [[FLAT_LOAD_DWORD:%[0-9]+]]:vgpr_32 = FLAT_LOAD_DWORD [[COPY]], 0, 0, implicit $exec, implicit $flat_scr :: (load (p3)) |
| ; GFX11-NEXT: $vgpr0 = COPY [[FLAT_LOAD_DWORD]] |
| %0:vgpr(p1) = COPY $vgpr0_vgpr1 |
| %1:vgpr(p3) = G_LOAD %0 :: (load (p3), align 4, addrspace 0) |
| $vgpr0 = COPY %1 |
| |
| ... |
| |
| --- |
| |
| name: load_flat_p1_from_8 |
| legalized: true |
| regBankSelected: true |
| tracksRegLiveness: true |
| |
| body: | |
| bb.0: |
| liveins: $vgpr0_vgpr1 |
| |
| ; GFX7-LABEL: name: load_flat_p1_from_8 |
| ; GFX7: liveins: $vgpr0_vgpr1 |
| ; GFX7-NEXT: {{ $}} |
| ; GFX7-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1 |
| ; GFX7-NEXT: [[FLAT_LOAD_DWORDX2_:%[0-9]+]]:vreg_64 = FLAT_LOAD_DWORDX2 [[COPY]], 0, 0, implicit $exec, implicit $flat_scr :: (load (p1)) |
| ; GFX7-NEXT: $vgpr0_vgpr1 = COPY [[FLAT_LOAD_DWORDX2_]] |
| ; GFX8-LABEL: name: load_flat_p1_from_8 |
| ; GFX8: liveins: $vgpr0_vgpr1 |
| ; GFX8-NEXT: {{ $}} |
| ; GFX8-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1 |
| ; GFX8-NEXT: [[FLAT_LOAD_DWORDX2_:%[0-9]+]]:vreg_64 = FLAT_LOAD_DWORDX2 [[COPY]], 0, 0, implicit $exec, implicit $flat_scr :: (load (p1)) |
| ; GFX8-NEXT: $vgpr0_vgpr1 = COPY [[FLAT_LOAD_DWORDX2_]] |
| ; GFX9-LABEL: name: load_flat_p1_from_8 |
| ; GFX9: liveins: $vgpr0_vgpr1 |
| ; GFX9-NEXT: {{ $}} |
| ; GFX9-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1 |
| ; GFX9-NEXT: [[FLAT_LOAD_DWORDX2_:%[0-9]+]]:vreg_64 = FLAT_LOAD_DWORDX2 [[COPY]], 0, 0, implicit $exec, implicit $flat_scr :: (load (p1)) |
| ; GFX9-NEXT: $vgpr0_vgpr1 = COPY [[FLAT_LOAD_DWORDX2_]] |
| ; GFX10-LABEL: name: load_flat_p1_from_8 |
| ; GFX10: liveins: $vgpr0_vgpr1 |
| ; GFX10-NEXT: {{ $}} |
| ; GFX10-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1 |
| ; GFX10-NEXT: [[FLAT_LOAD_DWORDX2_:%[0-9]+]]:vreg_64 = FLAT_LOAD_DWORDX2 [[COPY]], 0, 0, implicit $exec, implicit $flat_scr :: (load (p1)) |
| ; GFX10-NEXT: $vgpr0_vgpr1 = COPY [[FLAT_LOAD_DWORDX2_]] |
| ; GFX11-LABEL: name: load_flat_p1_from_8 |
| ; GFX11: liveins: $vgpr0_vgpr1 |
| ; GFX11-NEXT: {{ $}} |
| ; GFX11-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1 |
| ; GFX11-NEXT: [[FLAT_LOAD_DWORDX2_:%[0-9]+]]:vreg_64 = FLAT_LOAD_DWORDX2 [[COPY]], 0, 0, implicit $exec, implicit $flat_scr :: (load (p1)) |
| ; GFX11-NEXT: $vgpr0_vgpr1 = COPY [[FLAT_LOAD_DWORDX2_]] |
| %0:vgpr(p1) = COPY $vgpr0_vgpr1 |
| %1:vgpr(p1) = G_LOAD %0 :: (load (p1), align 8, addrspace 0) |
| $vgpr0_vgpr1 = COPY %1 |
| |
| ... |
| |
| --- |
| |
| name: load_flat_p999_from_8 |
| legalized: true |
| regBankSelected: true |
| tracksRegLiveness: true |
| |
| body: | |
| bb.0: |
| liveins: $vgpr0_vgpr1 |
| |
| ; GFX7-LABEL: name: load_flat_p999_from_8 |
| ; GFX7: liveins: $vgpr0_vgpr1 |
| ; GFX7-NEXT: {{ $}} |
| ; GFX7-NEXT: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1 |
| ; GFX7-NEXT: [[LOAD:%[0-9]+]]:vreg_64(p999) = G_LOAD [[COPY]](p1) :: (load (p999)) |
| ; GFX7-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p999) |
| ; GFX8-LABEL: name: load_flat_p999_from_8 |
| ; GFX8: liveins: $vgpr0_vgpr1 |
| ; GFX8-NEXT: {{ $}} |
| ; GFX8-NEXT: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1 |
| ; GFX8-NEXT: [[LOAD:%[0-9]+]]:vreg_64(p999) = G_LOAD [[COPY]](p1) :: (load (p999)) |
| ; GFX8-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p999) |
| ; GFX9-LABEL: name: load_flat_p999_from_8 |
| ; GFX9: liveins: $vgpr0_vgpr1 |
| ; GFX9-NEXT: {{ $}} |
| ; GFX9-NEXT: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1 |
| ; GFX9-NEXT: [[LOAD:%[0-9]+]]:vreg_64(p999) = G_LOAD [[COPY]](p1) :: (load (p999)) |
| ; GFX9-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p999) |
| ; GFX10-LABEL: name: load_flat_p999_from_8 |
| ; GFX10: liveins: $vgpr0_vgpr1 |
| ; GFX10-NEXT: {{ $}} |
| ; GFX10-NEXT: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1 |
| ; GFX10-NEXT: [[LOAD:%[0-9]+]]:vreg_64(p999) = G_LOAD [[COPY]](p1) :: (load (p999)) |
| ; GFX10-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p999) |
| ; GFX11-LABEL: name: load_flat_p999_from_8 |
| ; GFX11: liveins: $vgpr0_vgpr1 |
| ; GFX11-NEXT: {{ $}} |
| ; GFX11-NEXT: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1 |
| ; GFX11-NEXT: [[LOAD:%[0-9]+]]:vreg_64(p999) = G_LOAD [[COPY]](p1) :: (load (p999)) |
| ; GFX11-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](p999) |
| %0:vgpr(p1) = COPY $vgpr0_vgpr1 |
| %1:vgpr(p999) = G_LOAD %0 :: (load (p999), align 8, addrspace 0) |
| $vgpr0_vgpr1 = COPY %1 |
| |
| ... |
| |
| --- |
| |
| name: load_flat_v2p3 |
| legalized: true |
| regBankSelected: true |
| |
| |
| body: | |
| bb.0: |
| liveins: $vgpr0_vgpr1 |
| |
| ; GFX7-LABEL: name: load_flat_v2p3 |
| ; GFX7: liveins: $vgpr0_vgpr1 |
| ; GFX7-NEXT: {{ $}} |
| ; GFX7-NEXT: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1 |
| ; GFX7-NEXT: [[LOAD:%[0-9]+]]:vreg_64(<2 x p3>) = G_LOAD [[COPY]](p1) :: (load (<2 x p3>)) |
| ; GFX7-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x p3>) |
| ; GFX8-LABEL: name: load_flat_v2p3 |
| ; GFX8: liveins: $vgpr0_vgpr1 |
| ; GFX8-NEXT: {{ $}} |
| ; GFX8-NEXT: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1 |
| ; GFX8-NEXT: [[LOAD:%[0-9]+]]:vreg_64(<2 x p3>) = G_LOAD [[COPY]](p1) :: (load (<2 x p3>)) |
| ; GFX8-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x p3>) |
| ; GFX9-LABEL: name: load_flat_v2p3 |
| ; GFX9: liveins: $vgpr0_vgpr1 |
| ; GFX9-NEXT: {{ $}} |
| ; GFX9-NEXT: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1 |
| ; GFX9-NEXT: [[LOAD:%[0-9]+]]:vreg_64(<2 x p3>) = G_LOAD [[COPY]](p1) :: (load (<2 x p3>)) |
| ; GFX9-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x p3>) |
| ; GFX10-LABEL: name: load_flat_v2p3 |
| ; GFX10: liveins: $vgpr0_vgpr1 |
| ; GFX10-NEXT: {{ $}} |
| ; GFX10-NEXT: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1 |
| ; GFX10-NEXT: [[LOAD:%[0-9]+]]:vreg_64(<2 x p3>) = G_LOAD [[COPY]](p1) :: (load (<2 x p3>)) |
| ; GFX10-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x p3>) |
| ; GFX11-LABEL: name: load_flat_v2p3 |
| ; GFX11: liveins: $vgpr0_vgpr1 |
| ; GFX11-NEXT: {{ $}} |
| ; GFX11-NEXT: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1 |
| ; GFX11-NEXT: [[LOAD:%[0-9]+]]:vreg_64(<2 x p3>) = G_LOAD [[COPY]](p1) :: (load (<2 x p3>)) |
| ; GFX11-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x p3>) |
| %0:vgpr(p1) = COPY $vgpr0_vgpr1 |
| %1:vgpr(<2 x p3>) = G_LOAD %0 :: (load (<2 x p3>), align 8, addrspace 0) |
| $vgpr0_vgpr1 = COPY %1 |
| |
| ... |
| |
| --- |
| |
| name: load_flat_v2s16 |
| legalized: true |
| regBankSelected: true |
| tracksRegLiveness: true |
| |
| body: | |
| bb.0: |
| liveins: $vgpr0_vgpr1 |
| |
| ; GFX7-LABEL: name: load_flat_v2s16 |
| ; GFX7: liveins: $vgpr0_vgpr1 |
| ; GFX7-NEXT: {{ $}} |
| ; GFX7-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1 |
| ; GFX7-NEXT: [[FLAT_LOAD_DWORD:%[0-9]+]]:vgpr_32 = FLAT_LOAD_DWORD [[COPY]], 0, 0, implicit $exec, implicit $flat_scr :: (load (<2 x s16>)) |
| ; GFX7-NEXT: $vgpr0 = COPY [[FLAT_LOAD_DWORD]] |
| ; GFX8-LABEL: name: load_flat_v2s16 |
| ; GFX8: liveins: $vgpr0_vgpr1 |
| ; GFX8-NEXT: {{ $}} |
| ; GFX8-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1 |
| ; GFX8-NEXT: [[FLAT_LOAD_DWORD:%[0-9]+]]:vgpr_32 = FLAT_LOAD_DWORD [[COPY]], 0, 0, implicit $exec, implicit $flat_scr :: (load (<2 x s16>)) |
| ; GFX8-NEXT: $vgpr0 = COPY [[FLAT_LOAD_DWORD]] |
| ; GFX9-LABEL: name: load_flat_v2s16 |
| ; GFX9: liveins: $vgpr0_vgpr1 |
| ; GFX9-NEXT: {{ $}} |
| ; GFX9-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1 |
| ; GFX9-NEXT: [[FLAT_LOAD_DWORD:%[0-9]+]]:vgpr_32 = FLAT_LOAD_DWORD [[COPY]], 0, 0, implicit $exec, implicit $flat_scr :: (load (<2 x s16>)) |
| ; GFX9-NEXT: $vgpr0 = COPY [[FLAT_LOAD_DWORD]] |
| ; GFX10-LABEL: name: load_flat_v2s16 |
| ; GFX10: liveins: $vgpr0_vgpr1 |
| ; GFX10-NEXT: {{ $}} |
| ; GFX10-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1 |
| ; GFX10-NEXT: [[FLAT_LOAD_DWORD:%[0-9]+]]:vgpr_32 = FLAT_LOAD_DWORD [[COPY]], 0, 0, implicit $exec, implicit $flat_scr :: (load (<2 x s16>)) |
| ; GFX10-NEXT: $vgpr0 = COPY [[FLAT_LOAD_DWORD]] |
| ; GFX11-LABEL: name: load_flat_v2s16 |
| ; GFX11: liveins: $vgpr0_vgpr1 |
| ; GFX11-NEXT: {{ $}} |
| ; GFX11-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1 |
| ; GFX11-NEXT: [[FLAT_LOAD_DWORD:%[0-9]+]]:vgpr_32 = FLAT_LOAD_DWORD [[COPY]], 0, 0, implicit $exec, implicit $flat_scr :: (load (<2 x s16>)) |
| ; GFX11-NEXT: $vgpr0 = COPY [[FLAT_LOAD_DWORD]] |
| %0:vgpr(p1) = COPY $vgpr0_vgpr1 |
| %1:vgpr(<2 x s16>) = G_LOAD %0 :: (load (<2 x s16>), align 4, addrspace 0) |
| $vgpr0 = COPY %1 |
| |
| ... |
| |
| --- |
| |
| name: load_flat_v4s16 |
| legalized: true |
| regBankSelected: true |
| tracksRegLiveness: true |
| |
| body: | |
| bb.0: |
| liveins: $vgpr0_vgpr1 |
| |
| ; GFX7-LABEL: name: load_flat_v4s16 |
| ; GFX7: liveins: $vgpr0_vgpr1 |
| ; GFX7-NEXT: {{ $}} |
| ; GFX7-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1 |
| ; GFX7-NEXT: [[FLAT_LOAD_DWORDX2_:%[0-9]+]]:vreg_64 = FLAT_LOAD_DWORDX2 [[COPY]], 0, 0, implicit $exec, implicit $flat_scr :: (load (<4 x s16>)) |
| ; GFX7-NEXT: $vgpr0_vgpr1 = COPY [[FLAT_LOAD_DWORDX2_]] |
| ; GFX8-LABEL: name: load_flat_v4s16 |
| ; GFX8: liveins: $vgpr0_vgpr1 |
| ; GFX8-NEXT: {{ $}} |
| ; GFX8-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1 |
| ; GFX8-NEXT: [[FLAT_LOAD_DWORDX2_:%[0-9]+]]:vreg_64 = FLAT_LOAD_DWORDX2 [[COPY]], 0, 0, implicit $exec, implicit $flat_scr :: (load (<4 x s16>)) |
| ; GFX8-NEXT: $vgpr0_vgpr1 = COPY [[FLAT_LOAD_DWORDX2_]] |
| ; GFX9-LABEL: name: load_flat_v4s16 |
| ; GFX9: liveins: $vgpr0_vgpr1 |
| ; GFX9-NEXT: {{ $}} |
| ; GFX9-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1 |
| ; GFX9-NEXT: [[FLAT_LOAD_DWORDX2_:%[0-9]+]]:vreg_64 = FLAT_LOAD_DWORDX2 [[COPY]], 0, 0, implicit $exec, implicit $flat_scr :: (load (<4 x s16>)) |
| ; GFX9-NEXT: $vgpr0_vgpr1 = COPY [[FLAT_LOAD_DWORDX2_]] |
| ; GFX10-LABEL: name: load_flat_v4s16 |
| ; GFX10: liveins: $vgpr0_vgpr1 |
| ; GFX10-NEXT: {{ $}} |
| ; GFX10-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1 |
| ; GFX10-NEXT: [[FLAT_LOAD_DWORDX2_:%[0-9]+]]:vreg_64 = FLAT_LOAD_DWORDX2 [[COPY]], 0, 0, implicit $exec, implicit $flat_scr :: (load (<4 x s16>)) |
| ; GFX10-NEXT: $vgpr0_vgpr1 = COPY [[FLAT_LOAD_DWORDX2_]] |
| ; GFX11-LABEL: name: load_flat_v4s16 |
| ; GFX11: liveins: $vgpr0_vgpr1 |
| ; GFX11-NEXT: {{ $}} |
| ; GFX11-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1 |
| ; GFX11-NEXT: [[FLAT_LOAD_DWORDX2_:%[0-9]+]]:vreg_64 = FLAT_LOAD_DWORDX2 [[COPY]], 0, 0, implicit $exec, implicit $flat_scr :: (load (<4 x s16>)) |
| ; GFX11-NEXT: $vgpr0_vgpr1 = COPY [[FLAT_LOAD_DWORDX2_]] |
| %0:vgpr(p1) = COPY $vgpr0_vgpr1 |
| %1:vgpr(<4 x s16>) = G_LOAD %0 :: (load (<4 x s16>), align 8, addrspace 0) |
| $vgpr0_vgpr1 = COPY %1 |
| |
| ... |
| |
| --- |
| |
| name: load_flat_v6s16 |
| legalized: true |
| regBankSelected: true |
| tracksRegLiveness: true |
| |
| body: | |
| bb.0: |
| liveins: $vgpr0_vgpr1 |
| |
| ; GFX7-LABEL: name: load_flat_v6s16 |
| ; GFX7: liveins: $vgpr0_vgpr1 |
| ; GFX7-NEXT: {{ $}} |
| ; GFX7-NEXT: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1 |
| ; GFX7-NEXT: [[LOAD:%[0-9]+]]:vreg_96(<6 x s16>) = G_LOAD [[COPY]](p1) :: (load (<6 x s16>), align 4) |
| ; GFX7-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[LOAD]](<6 x s16>) |
| ; GFX8-LABEL: name: load_flat_v6s16 |
| ; GFX8: liveins: $vgpr0_vgpr1 |
| ; GFX8-NEXT: {{ $}} |
| ; GFX8-NEXT: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1 |
| ; GFX8-NEXT: [[LOAD:%[0-9]+]]:vreg_96(<6 x s16>) = G_LOAD [[COPY]](p1) :: (load (<6 x s16>), align 4) |
| ; GFX8-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[LOAD]](<6 x s16>) |
| ; GFX9-LABEL: name: load_flat_v6s16 |
| ; GFX9: liveins: $vgpr0_vgpr1 |
| ; GFX9-NEXT: {{ $}} |
| ; GFX9-NEXT: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1 |
| ; GFX9-NEXT: [[LOAD:%[0-9]+]]:vreg_96(<6 x s16>) = G_LOAD [[COPY]](p1) :: (load (<6 x s16>), align 4) |
| ; GFX9-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[LOAD]](<6 x s16>) |
| ; GFX10-LABEL: name: load_flat_v6s16 |
| ; GFX10: liveins: $vgpr0_vgpr1 |
| ; GFX10-NEXT: {{ $}} |
| ; GFX10-NEXT: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1 |
| ; GFX10-NEXT: [[LOAD:%[0-9]+]]:vreg_96(<6 x s16>) = G_LOAD [[COPY]](p1) :: (load (<6 x s16>), align 4) |
| ; GFX10-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[LOAD]](<6 x s16>) |
| ; GFX11-LABEL: name: load_flat_v6s16 |
| ; GFX11: liveins: $vgpr0_vgpr1 |
| ; GFX11-NEXT: {{ $}} |
| ; GFX11-NEXT: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1 |
| ; GFX11-NEXT: [[LOAD:%[0-9]+]]:vreg_96(<6 x s16>) = G_LOAD [[COPY]](p1) :: (load (<6 x s16>), align 4) |
| ; GFX11-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[LOAD]](<6 x s16>) |
| %0:vgpr(p1) = COPY $vgpr0_vgpr1 |
| %1:vgpr(<6 x s16>) = G_LOAD %0 :: (load (<6 x s16>), align 4, addrspace 0) |
| $vgpr0_vgpr1_vgpr2 = COPY %1 |
| |
| ... |
| |
| --- |
| |
| name: load_flat_v8s16 |
| legalized: true |
| regBankSelected: true |
| tracksRegLiveness: true |
| |
| body: | |
| bb.0: |
| liveins: $vgpr0_vgpr1 |
| |
| ; GFX7-LABEL: name: load_flat_v8s16 |
| ; GFX7: liveins: $vgpr0_vgpr1 |
| ; GFX7-NEXT: {{ $}} |
| ; GFX7-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1 |
| ; GFX7-NEXT: [[FLAT_LOAD_DWORDX4_:%[0-9]+]]:vreg_128 = FLAT_LOAD_DWORDX4 [[COPY]], 0, 0, implicit $exec, implicit $flat_scr :: (load (<8 x s16>), align 4) |
| ; GFX7-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[FLAT_LOAD_DWORDX4_]] |
| ; GFX8-LABEL: name: load_flat_v8s16 |
| ; GFX8: liveins: $vgpr0_vgpr1 |
| ; GFX8-NEXT: {{ $}} |
| ; GFX8-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1 |
| ; GFX8-NEXT: [[FLAT_LOAD_DWORDX4_:%[0-9]+]]:vreg_128 = FLAT_LOAD_DWORDX4 [[COPY]], 0, 0, implicit $exec, implicit $flat_scr :: (load (<8 x s16>), align 4) |
| ; GFX8-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[FLAT_LOAD_DWORDX4_]] |
| ; GFX9-LABEL: name: load_flat_v8s16 |
| ; GFX9: liveins: $vgpr0_vgpr1 |
| ; GFX9-NEXT: {{ $}} |
| ; GFX9-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1 |
| ; GFX9-NEXT: [[FLAT_LOAD_DWORDX4_:%[0-9]+]]:vreg_128 = FLAT_LOAD_DWORDX4 [[COPY]], 0, 0, implicit $exec, implicit $flat_scr :: (load (<8 x s16>), align 4) |
| ; GFX9-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[FLAT_LOAD_DWORDX4_]] |
| ; GFX10-LABEL: name: load_flat_v8s16 |
| ; GFX10: liveins: $vgpr0_vgpr1 |
| ; GFX10-NEXT: {{ $}} |
| ; GFX10-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1 |
| ; GFX10-NEXT: [[FLAT_LOAD_DWORDX4_:%[0-9]+]]:vreg_128 = FLAT_LOAD_DWORDX4 [[COPY]], 0, 0, implicit $exec, implicit $flat_scr :: (load (<8 x s16>), align 4) |
| ; GFX10-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[FLAT_LOAD_DWORDX4_]] |
| ; GFX11-LABEL: name: load_flat_v8s16 |
| ; GFX11: liveins: $vgpr0_vgpr1 |
| ; GFX11-NEXT: {{ $}} |
| ; GFX11-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1 |
| ; GFX11-NEXT: [[FLAT_LOAD_DWORDX4_:%[0-9]+]]:vreg_128 = FLAT_LOAD_DWORDX4 [[COPY]], 0, 0, implicit $exec, implicit $flat_scr :: (load (<8 x s16>), align 4) |
| ; GFX11-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[FLAT_LOAD_DWORDX4_]] |
| %0:vgpr(p1) = COPY $vgpr0_vgpr1 |
| %1:vgpr(<8 x s16>) = G_LOAD %0 :: (load (<8 x s16>), align 4, addrspace 0) |
| $vgpr0_vgpr1_vgpr2_vgpr3 = COPY %1 |
| |
| ... |
| |
| ################################################################################ |
| ### Stress addressing modes |
| ################################################################################ |
| |
| --- |
| |
| name: load_flat_s32_from_1_gep_2047 |
| legalized: true |
| regBankSelected: true |
| tracksRegLiveness: true |
| |
| body: | |
| bb.0: |
| liveins: $vgpr0_vgpr1 |
| |
| ; GFX7-LABEL: name: load_flat_s32_from_1_gep_2047 |
| ; GFX7: liveins: $vgpr0_vgpr1 |
| ; GFX7-NEXT: {{ $}} |
| ; GFX7-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1 |
| ; GFX7-NEXT: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 2047, implicit $exec |
| ; GFX7-NEXT: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 0, implicit $exec |
| ; GFX7-NEXT: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1 |
| ; GFX7-NEXT: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0 |
| ; GFX7-NEXT: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0 |
| ; GFX7-NEXT: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1 |
| ; GFX7-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1 |
| ; GFX7-NEXT: [[V_ADD_CO_U32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_CO_U32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADD_CO_U32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec |
| ; GFX7-NEXT: [[V_ADDC_U32_e64_:%[0-9]+]]:vgpr_32, dead [[V_ADDC_U32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_CO_U32_e64_1]], 0, implicit $exec |
| ; GFX7-NEXT: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_CO_U32_e64_]], %subreg.sub0, [[V_ADDC_U32_e64_]], %subreg.sub1 |
| ; GFX7-NEXT: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, implicit $exec, implicit $flat_scr :: (load (s8)) |
| ; GFX7-NEXT: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]] |
| ; GFX8-LABEL: name: load_flat_s32_from_1_gep_2047 |
| ; GFX8: liveins: $vgpr0_vgpr1 |
| ; GFX8-NEXT: {{ $}} |
| ; GFX8-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1 |
| ; GFX8-NEXT: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 2047, implicit $exec |
| ; GFX8-NEXT: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 0, implicit $exec |
| ; GFX8-NEXT: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1 |
| ; GFX8-NEXT: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0 |
| ; GFX8-NEXT: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0 |
| ; GFX8-NEXT: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1 |
| ; GFX8-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1 |
| ; GFX8-NEXT: [[V_ADD_CO_U32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_CO_U32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADD_CO_U32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec |
| ; GFX8-NEXT: [[V_ADDC_U32_e64_:%[0-9]+]]:vgpr_32, dead [[V_ADDC_U32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_CO_U32_e64_1]], 0, implicit $exec |
| ; GFX8-NEXT: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_CO_U32_e64_]], %subreg.sub0, [[V_ADDC_U32_e64_]], %subreg.sub1 |
| ; GFX8-NEXT: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, implicit $exec, implicit $flat_scr :: (load (s8)) |
| ; GFX8-NEXT: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]] |
| ; GFX9-LABEL: name: load_flat_s32_from_1_gep_2047 |
| ; GFX9: liveins: $vgpr0_vgpr1 |
| ; GFX9-NEXT: {{ $}} |
| ; GFX9-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1 |
| ; GFX9-NEXT: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[COPY]], 2047, 0, implicit $exec, implicit $flat_scr :: (load (s8)) |
| ; GFX9-NEXT: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]] |
| ; GFX10-LABEL: name: load_flat_s32_from_1_gep_2047 |
| ; GFX10: liveins: $vgpr0_vgpr1 |
| ; GFX10-NEXT: {{ $}} |
| ; GFX10-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1 |
| ; GFX10-NEXT: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 2047, implicit $exec |
| ; GFX10-NEXT: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 0, implicit $exec |
| ; GFX10-NEXT: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1 |
| ; GFX10-NEXT: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0 |
| ; GFX10-NEXT: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0 |
| ; GFX10-NEXT: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1 |
| ; GFX10-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1 |
| ; GFX10-NEXT: [[V_ADD_CO_U32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_CO_U32_e64_1:%[0-9]+]]:sreg_32_xm0_xexec = V_ADD_CO_U32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec |
| ; GFX10-NEXT: [[V_ADDC_U32_e64_:%[0-9]+]]:vgpr_32, dead [[V_ADDC_U32_e64_1:%[0-9]+]]:sreg_32_xm0_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_CO_U32_e64_1]], 0, implicit $exec |
| ; GFX10-NEXT: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_CO_U32_e64_]], %subreg.sub0, [[V_ADDC_U32_e64_]], %subreg.sub1 |
| ; GFX10-NEXT: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, implicit $exec, implicit $flat_scr :: (load (s8)) |
| ; GFX10-NEXT: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]] |
| ; GFX11-LABEL: name: load_flat_s32_from_1_gep_2047 |
| ; GFX11: liveins: $vgpr0_vgpr1 |
| ; GFX11-NEXT: {{ $}} |
| ; GFX11-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1 |
| ; GFX11-NEXT: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[COPY]], 2047, 0, implicit $exec, implicit $flat_scr :: (load (s8)) |
| ; GFX11-NEXT: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]] |
| %0:vgpr(p1) = COPY $vgpr0_vgpr1 |
| %1:vgpr(s64) = G_CONSTANT i64 2047 |
| %2:vgpr(p1) = G_PTR_ADD %0, %1 |
| %3:vgpr(s32) = G_LOAD %2 :: (load (s8), align 1, addrspace 0) |
| $vgpr0 = COPY %3 |
| |
| ... |
| |
| --- |
| |
| name: load_flat_s32_from_1_gep_2048 |
| legalized: true |
| regBankSelected: true |
| tracksRegLiveness: true |
| |
| body: | |
| bb.0: |
| liveins: $vgpr0_vgpr1 |
| |
| ; GFX7-LABEL: name: load_flat_s32_from_1_gep_2048 |
| ; GFX7: liveins: $vgpr0_vgpr1 |
| ; GFX7-NEXT: {{ $}} |
| ; GFX7-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1 |
| ; GFX7-NEXT: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 2048, implicit $exec |
| ; GFX7-NEXT: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 0, implicit $exec |
| ; GFX7-NEXT: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1 |
| ; GFX7-NEXT: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0 |
| ; GFX7-NEXT: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0 |
| ; GFX7-NEXT: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1 |
| ; GFX7-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1 |
| ; GFX7-NEXT: [[V_ADD_CO_U32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_CO_U32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADD_CO_U32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec |
| ; GFX7-NEXT: [[V_ADDC_U32_e64_:%[0-9]+]]:vgpr_32, dead [[V_ADDC_U32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_CO_U32_e64_1]], 0, implicit $exec |
| ; GFX7-NEXT: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_CO_U32_e64_]], %subreg.sub0, [[V_ADDC_U32_e64_]], %subreg.sub1 |
| ; GFX7-NEXT: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, implicit $exec, implicit $flat_scr :: (load (s8)) |
| ; GFX7-NEXT: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]] |
| ; GFX8-LABEL: name: load_flat_s32_from_1_gep_2048 |
| ; GFX8: liveins: $vgpr0_vgpr1 |
| ; GFX8-NEXT: {{ $}} |
| ; GFX8-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1 |
| ; GFX8-NEXT: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 2048, implicit $exec |
| ; GFX8-NEXT: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 0, implicit $exec |
| ; GFX8-NEXT: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1 |
| ; GFX8-NEXT: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0 |
| ; GFX8-NEXT: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0 |
| ; GFX8-NEXT: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1 |
| ; GFX8-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1 |
| ; GFX8-NEXT: [[V_ADD_CO_U32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_CO_U32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADD_CO_U32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec |
| ; GFX8-NEXT: [[V_ADDC_U32_e64_:%[0-9]+]]:vgpr_32, dead [[V_ADDC_U32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_CO_U32_e64_1]], 0, implicit $exec |
| ; GFX8-NEXT: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_CO_U32_e64_]], %subreg.sub0, [[V_ADDC_U32_e64_]], %subreg.sub1 |
| ; GFX8-NEXT: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, implicit $exec, implicit $flat_scr :: (load (s8)) |
| ; GFX8-NEXT: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]] |
| ; GFX9-LABEL: name: load_flat_s32_from_1_gep_2048 |
| ; GFX9: liveins: $vgpr0_vgpr1 |
| ; GFX9-NEXT: {{ $}} |
| ; GFX9-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1 |
| ; GFX9-NEXT: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[COPY]], 2048, 0, implicit $exec, implicit $flat_scr :: (load (s8)) |
| ; GFX9-NEXT: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]] |
| ; GFX10-LABEL: name: load_flat_s32_from_1_gep_2048 |
| ; GFX10: liveins: $vgpr0_vgpr1 |
| ; GFX10-NEXT: {{ $}} |
| ; GFX10-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1 |
| ; GFX10-NEXT: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 2048, implicit $exec |
| ; GFX10-NEXT: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 0, implicit $exec |
| ; GFX10-NEXT: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1 |
| ; GFX10-NEXT: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0 |
| ; GFX10-NEXT: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0 |
| ; GFX10-NEXT: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1 |
| ; GFX10-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1 |
| ; GFX10-NEXT: [[V_ADD_CO_U32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_CO_U32_e64_1:%[0-9]+]]:sreg_32_xm0_xexec = V_ADD_CO_U32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec |
| ; GFX10-NEXT: [[V_ADDC_U32_e64_:%[0-9]+]]:vgpr_32, dead [[V_ADDC_U32_e64_1:%[0-9]+]]:sreg_32_xm0_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_CO_U32_e64_1]], 0, implicit $exec |
| ; GFX10-NEXT: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_CO_U32_e64_]], %subreg.sub0, [[V_ADDC_U32_e64_]], %subreg.sub1 |
| ; GFX10-NEXT: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, implicit $exec, implicit $flat_scr :: (load (s8)) |
| ; GFX10-NEXT: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]] |
| ; GFX11-LABEL: name: load_flat_s32_from_1_gep_2048 |
| ; GFX11: liveins: $vgpr0_vgpr1 |
| ; GFX11-NEXT: {{ $}} |
| ; GFX11-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1 |
| ; GFX11-NEXT: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[COPY]], 2048, 0, implicit $exec, implicit $flat_scr :: (load (s8)) |
| ; GFX11-NEXT: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]] |
| %0:vgpr(p1) = COPY $vgpr0_vgpr1 |
| %1:vgpr(s64) = G_CONSTANT i64 2048 |
| %2:vgpr(p1) = G_PTR_ADD %0, %1 |
| %3:vgpr(s32) = G_LOAD %2 :: (load (s8), align 1, addrspace 0) |
| $vgpr0 = COPY %3 |
| |
| ... |
| |
| --- |
| |
| name: load_flat_s32_from_1_gep_m2047 |
| legalized: true |
| regBankSelected: true |
| tracksRegLiveness: true |
| |
| body: | |
| bb.0: |
| liveins: $vgpr0_vgpr1 |
| |
| ; GFX7-LABEL: name: load_flat_s32_from_1_gep_m2047 |
| ; GFX7: liveins: $vgpr0_vgpr1 |
| ; GFX7-NEXT: {{ $}} |
| ; GFX7-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1 |
| ; GFX7-NEXT: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 4294965249, implicit $exec |
| ; GFX7-NEXT: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 -1, implicit $exec |
| ; GFX7-NEXT: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1 |
| ; GFX7-NEXT: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0 |
| ; GFX7-NEXT: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0 |
| ; GFX7-NEXT: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1 |
| ; GFX7-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1 |
| ; GFX7-NEXT: [[V_ADD_CO_U32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_CO_U32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADD_CO_U32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec |
| ; GFX7-NEXT: [[V_ADDC_U32_e64_:%[0-9]+]]:vgpr_32, dead [[V_ADDC_U32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_CO_U32_e64_1]], 0, implicit $exec |
| ; GFX7-NEXT: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_CO_U32_e64_]], %subreg.sub0, [[V_ADDC_U32_e64_]], %subreg.sub1 |
| ; GFX7-NEXT: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, implicit $exec, implicit $flat_scr :: (load (s8)) |
| ; GFX7-NEXT: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]] |
| ; GFX8-LABEL: name: load_flat_s32_from_1_gep_m2047 |
| ; GFX8: liveins: $vgpr0_vgpr1 |
| ; GFX8-NEXT: {{ $}} |
| ; GFX8-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1 |
| ; GFX8-NEXT: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 4294965249, implicit $exec |
| ; GFX8-NEXT: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 -1, implicit $exec |
| ; GFX8-NEXT: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1 |
| ; GFX8-NEXT: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0 |
| ; GFX8-NEXT: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0 |
| ; GFX8-NEXT: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1 |
| ; GFX8-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1 |
| ; GFX8-NEXT: [[V_ADD_CO_U32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_CO_U32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADD_CO_U32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec |
| ; GFX8-NEXT: [[V_ADDC_U32_e64_:%[0-9]+]]:vgpr_32, dead [[V_ADDC_U32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_CO_U32_e64_1]], 0, implicit $exec |
| ; GFX8-NEXT: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_CO_U32_e64_]], %subreg.sub0, [[V_ADDC_U32_e64_]], %subreg.sub1 |
| ; GFX8-NEXT: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, implicit $exec, implicit $flat_scr :: (load (s8)) |
| ; GFX8-NEXT: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]] |
| ; GFX9-LABEL: name: load_flat_s32_from_1_gep_m2047 |
| ; GFX9: liveins: $vgpr0_vgpr1 |
| ; GFX9-NEXT: {{ $}} |
| ; GFX9-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1 |
| ; GFX9-NEXT: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 4294965249, implicit $exec |
| ; GFX9-NEXT: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 -1, implicit $exec |
| ; GFX9-NEXT: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1 |
| ; GFX9-NEXT: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0 |
| ; GFX9-NEXT: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0 |
| ; GFX9-NEXT: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1 |
| ; GFX9-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1 |
| ; GFX9-NEXT: [[V_ADD_CO_U32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_CO_U32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADD_CO_U32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec |
| ; GFX9-NEXT: [[V_ADDC_U32_e64_:%[0-9]+]]:vgpr_32, dead [[V_ADDC_U32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_CO_U32_e64_1]], 0, implicit $exec |
| ; GFX9-NEXT: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_CO_U32_e64_]], %subreg.sub0, [[V_ADDC_U32_e64_]], %subreg.sub1 |
| ; GFX9-NEXT: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, implicit $exec, implicit $flat_scr :: (load (s8)) |
| ; GFX9-NEXT: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]] |
| ; GFX10-LABEL: name: load_flat_s32_from_1_gep_m2047 |
| ; GFX10: liveins: $vgpr0_vgpr1 |
| ; GFX10-NEXT: {{ $}} |
| ; GFX10-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1 |
| ; GFX10-NEXT: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 4294965249, implicit $exec |
| ; GFX10-NEXT: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 -1, implicit $exec |
| ; GFX10-NEXT: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1 |
| ; GFX10-NEXT: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0 |
| ; GFX10-NEXT: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0 |
| ; GFX10-NEXT: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1 |
| ; GFX10-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1 |
| ; GFX10-NEXT: [[V_ADD_CO_U32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_CO_U32_e64_1:%[0-9]+]]:sreg_32_xm0_xexec = V_ADD_CO_U32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec |
| ; GFX10-NEXT: [[V_ADDC_U32_e64_:%[0-9]+]]:vgpr_32, dead [[V_ADDC_U32_e64_1:%[0-9]+]]:sreg_32_xm0_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_CO_U32_e64_1]], 0, implicit $exec |
| ; GFX10-NEXT: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_CO_U32_e64_]], %subreg.sub0, [[V_ADDC_U32_e64_]], %subreg.sub1 |
| ; GFX10-NEXT: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, implicit $exec, implicit $flat_scr :: (load (s8)) |
| ; GFX10-NEXT: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]] |
| ; GFX11-LABEL: name: load_flat_s32_from_1_gep_m2047 |
| ; GFX11: liveins: $vgpr0_vgpr1 |
| ; GFX11-NEXT: {{ $}} |
| ; GFX11-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1 |
| ; GFX11-NEXT: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 4294965249, implicit $exec |
| ; GFX11-NEXT: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 -1, implicit $exec |
| ; GFX11-NEXT: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1 |
| ; GFX11-NEXT: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0 |
| ; GFX11-NEXT: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0 |
| ; GFX11-NEXT: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1 |
| ; GFX11-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1 |
| ; GFX11-NEXT: [[V_ADD_CO_U32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_CO_U32_e64_1:%[0-9]+]]:sreg_32_xm0_xexec = V_ADD_CO_U32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec |
| ; GFX11-NEXT: [[V_ADDC_U32_e64_:%[0-9]+]]:vgpr_32, dead [[V_ADDC_U32_e64_1:%[0-9]+]]:sreg_32_xm0_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_CO_U32_e64_1]], 0, implicit $exec |
| ; GFX11-NEXT: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_CO_U32_e64_]], %subreg.sub0, [[V_ADDC_U32_e64_]], %subreg.sub1 |
| ; GFX11-NEXT: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, implicit $exec, implicit $flat_scr :: (load (s8)) |
| ; GFX11-NEXT: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]] |
| %0:vgpr(p1) = COPY $vgpr0_vgpr1 |
| %1:vgpr(s64) = G_CONSTANT i64 -2047 |
| %2:vgpr(p1) = G_PTR_ADD %0, %1 |
| %3:vgpr(s32) = G_LOAD %2 :: (load (s8), align 1, addrspace 0) |
| $vgpr0 = COPY %3 |
| |
| ... |
| |
| --- |
| |
| name: load_flat_s32_from_1_gep_m2048 |
| legalized: true |
| regBankSelected: true |
| tracksRegLiveness: true |
| |
| body: | |
| bb.0: |
| liveins: $vgpr0_vgpr1 |
| |
| ; GFX7-LABEL: name: load_flat_s32_from_1_gep_m2048 |
| ; GFX7: liveins: $vgpr0_vgpr1 |
| ; GFX7-NEXT: {{ $}} |
| ; GFX7-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1 |
| ; GFX7-NEXT: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 4294965248, implicit $exec |
| ; GFX7-NEXT: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 -1, implicit $exec |
| ; GFX7-NEXT: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1 |
| ; GFX7-NEXT: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0 |
| ; GFX7-NEXT: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0 |
| ; GFX7-NEXT: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1 |
| ; GFX7-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1 |
| ; GFX7-NEXT: [[V_ADD_CO_U32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_CO_U32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADD_CO_U32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec |
| ; GFX7-NEXT: [[V_ADDC_U32_e64_:%[0-9]+]]:vgpr_32, dead [[V_ADDC_U32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_CO_U32_e64_1]], 0, implicit $exec |
| ; GFX7-NEXT: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_CO_U32_e64_]], %subreg.sub0, [[V_ADDC_U32_e64_]], %subreg.sub1 |
| ; GFX7-NEXT: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, implicit $exec, implicit $flat_scr :: (load (s8)) |
| ; GFX7-NEXT: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]] |
| ; GFX8-LABEL: name: load_flat_s32_from_1_gep_m2048 |
| ; GFX8: liveins: $vgpr0_vgpr1 |
| ; GFX8-NEXT: {{ $}} |
| ; GFX8-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1 |
| ; GFX8-NEXT: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 4294965248, implicit $exec |
| ; GFX8-NEXT: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 -1, implicit $exec |
| ; GFX8-NEXT: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1 |
| ; GFX8-NEXT: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0 |
| ; GFX8-NEXT: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0 |
| ; GFX8-NEXT: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1 |
| ; GFX8-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1 |
| ; GFX8-NEXT: [[V_ADD_CO_U32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_CO_U32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADD_CO_U32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec |
| ; GFX8-NEXT: [[V_ADDC_U32_e64_:%[0-9]+]]:vgpr_32, dead [[V_ADDC_U32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_CO_U32_e64_1]], 0, implicit $exec |
| ; GFX8-NEXT: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_CO_U32_e64_]], %subreg.sub0, [[V_ADDC_U32_e64_]], %subreg.sub1 |
| ; GFX8-NEXT: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, implicit $exec, implicit $flat_scr :: (load (s8)) |
| ; GFX8-NEXT: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]] |
| ; GFX9-LABEL: name: load_flat_s32_from_1_gep_m2048 |
| ; GFX9: liveins: $vgpr0_vgpr1 |
| ; GFX9-NEXT: {{ $}} |
| ; GFX9-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1 |
| ; GFX9-NEXT: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 4294965248, implicit $exec |
| ; GFX9-NEXT: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 -1, implicit $exec |
| ; GFX9-NEXT: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1 |
| ; GFX9-NEXT: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0 |
| ; GFX9-NEXT: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0 |
| ; GFX9-NEXT: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1 |
| ; GFX9-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1 |
| ; GFX9-NEXT: [[V_ADD_CO_U32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_CO_U32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADD_CO_U32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec |
| ; GFX9-NEXT: [[V_ADDC_U32_e64_:%[0-9]+]]:vgpr_32, dead [[V_ADDC_U32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_CO_U32_e64_1]], 0, implicit $exec |
| ; GFX9-NEXT: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_CO_U32_e64_]], %subreg.sub0, [[V_ADDC_U32_e64_]], %subreg.sub1 |
| ; GFX9-NEXT: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, implicit $exec, implicit $flat_scr :: (load (s8)) |
| ; GFX9-NEXT: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]] |
| ; GFX10-LABEL: name: load_flat_s32_from_1_gep_m2048 |
| ; GFX10: liveins: $vgpr0_vgpr1 |
| ; GFX10-NEXT: {{ $}} |
| ; GFX10-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1 |
| ; GFX10-NEXT: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 4294965248, implicit $exec |
| ; GFX10-NEXT: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 -1, implicit $exec |
| ; GFX10-NEXT: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1 |
| ; GFX10-NEXT: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0 |
| ; GFX10-NEXT: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0 |
| ; GFX10-NEXT: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1 |
| ; GFX10-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1 |
| ; GFX10-NEXT: [[V_ADD_CO_U32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_CO_U32_e64_1:%[0-9]+]]:sreg_32_xm0_xexec = V_ADD_CO_U32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec |
| ; GFX10-NEXT: [[V_ADDC_U32_e64_:%[0-9]+]]:vgpr_32, dead [[V_ADDC_U32_e64_1:%[0-9]+]]:sreg_32_xm0_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_CO_U32_e64_1]], 0, implicit $exec |
| ; GFX10-NEXT: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_CO_U32_e64_]], %subreg.sub0, [[V_ADDC_U32_e64_]], %subreg.sub1 |
| ; GFX10-NEXT: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, implicit $exec, implicit $flat_scr :: (load (s8)) |
| ; GFX10-NEXT: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]] |
| ; GFX11-LABEL: name: load_flat_s32_from_1_gep_m2048 |
| ; GFX11: liveins: $vgpr0_vgpr1 |
| ; GFX11-NEXT: {{ $}} |
| ; GFX11-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1 |
| ; GFX11-NEXT: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 4294965248, implicit $exec |
| ; GFX11-NEXT: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 -1, implicit $exec |
| ; GFX11-NEXT: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1 |
| ; GFX11-NEXT: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0 |
| ; GFX11-NEXT: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0 |
| ; GFX11-NEXT: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1 |
| ; GFX11-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1 |
| ; GFX11-NEXT: [[V_ADD_CO_U32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_CO_U32_e64_1:%[0-9]+]]:sreg_32_xm0_xexec = V_ADD_CO_U32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec |
| ; GFX11-NEXT: [[V_ADDC_U32_e64_:%[0-9]+]]:vgpr_32, dead [[V_ADDC_U32_e64_1:%[0-9]+]]:sreg_32_xm0_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_CO_U32_e64_1]], 0, implicit $exec |
| ; GFX11-NEXT: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_CO_U32_e64_]], %subreg.sub0, [[V_ADDC_U32_e64_]], %subreg.sub1 |
| ; GFX11-NEXT: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, implicit $exec, implicit $flat_scr :: (load (s8)) |
| ; GFX11-NEXT: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]] |
| %0:vgpr(p1) = COPY $vgpr0_vgpr1 |
| %1:vgpr(s64) = G_CONSTANT i64 -2048 |
| %2:vgpr(p1) = G_PTR_ADD %0, %1 |
| %3:vgpr(s32) = G_LOAD %2 :: (load (s8), align 1, addrspace 0) |
| $vgpr0 = COPY %3 |
| |
| ... |
| |
| --- |
| |
| name: load_flat_s32_from_1_gep_4095 |
| legalized: true |
| regBankSelected: true |
| tracksRegLiveness: true |
| |
| body: | |
| bb.0: |
| liveins: $vgpr0_vgpr1 |
| |
| ; GFX7-LABEL: name: load_flat_s32_from_1_gep_4095 |
| ; GFX7: liveins: $vgpr0_vgpr1 |
| ; GFX7-NEXT: {{ $}} |
| ; GFX7-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1 |
| ; GFX7-NEXT: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 4095, implicit $exec |
| ; GFX7-NEXT: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 0, implicit $exec |
| ; GFX7-NEXT: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1 |
| ; GFX7-NEXT: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0 |
| ; GFX7-NEXT: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0 |
| ; GFX7-NEXT: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1 |
| ; GFX7-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1 |
| ; GFX7-NEXT: [[V_ADD_CO_U32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_CO_U32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADD_CO_U32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec |
| ; GFX7-NEXT: [[V_ADDC_U32_e64_:%[0-9]+]]:vgpr_32, dead [[V_ADDC_U32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_CO_U32_e64_1]], 0, implicit $exec |
| ; GFX7-NEXT: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_CO_U32_e64_]], %subreg.sub0, [[V_ADDC_U32_e64_]], %subreg.sub1 |
| ; GFX7-NEXT: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, implicit $exec, implicit $flat_scr :: (load (s8)) |
| ; GFX7-NEXT: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]] |
| ; GFX8-LABEL: name: load_flat_s32_from_1_gep_4095 |
| ; GFX8: liveins: $vgpr0_vgpr1 |
| ; GFX8-NEXT: {{ $}} |
| ; GFX8-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1 |
| ; GFX8-NEXT: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 4095, implicit $exec |
| ; GFX8-NEXT: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 0, implicit $exec |
| ; GFX8-NEXT: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1 |
| ; GFX8-NEXT: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0 |
| ; GFX8-NEXT: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0 |
| ; GFX8-NEXT: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1 |
| ; GFX8-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1 |
| ; GFX8-NEXT: [[V_ADD_CO_U32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_CO_U32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADD_CO_U32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec |
| ; GFX8-NEXT: [[V_ADDC_U32_e64_:%[0-9]+]]:vgpr_32, dead [[V_ADDC_U32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_CO_U32_e64_1]], 0, implicit $exec |
| ; GFX8-NEXT: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_CO_U32_e64_]], %subreg.sub0, [[V_ADDC_U32_e64_]], %subreg.sub1 |
| ; GFX8-NEXT: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, implicit $exec, implicit $flat_scr :: (load (s8)) |
| ; GFX8-NEXT: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]] |
| ; GFX9-LABEL: name: load_flat_s32_from_1_gep_4095 |
| ; GFX9: liveins: $vgpr0_vgpr1 |
| ; GFX9-NEXT: {{ $}} |
| ; GFX9-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1 |
| ; GFX9-NEXT: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[COPY]], 4095, 0, implicit $exec, implicit $flat_scr :: (load (s8)) |
| ; GFX9-NEXT: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]] |
| ; GFX10-LABEL: name: load_flat_s32_from_1_gep_4095 |
| ; GFX10: liveins: $vgpr0_vgpr1 |
| ; GFX10-NEXT: {{ $}} |
| ; GFX10-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1 |
| ; GFX10-NEXT: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 4095, implicit $exec |
| ; GFX10-NEXT: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 0, implicit $exec |
| ; GFX10-NEXT: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1 |
| ; GFX10-NEXT: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0 |
| ; GFX10-NEXT: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0 |
| ; GFX10-NEXT: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1 |
| ; GFX10-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1 |
| ; GFX10-NEXT: [[V_ADD_CO_U32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_CO_U32_e64_1:%[0-9]+]]:sreg_32_xm0_xexec = V_ADD_CO_U32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec |
| ; GFX10-NEXT: [[V_ADDC_U32_e64_:%[0-9]+]]:vgpr_32, dead [[V_ADDC_U32_e64_1:%[0-9]+]]:sreg_32_xm0_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_CO_U32_e64_1]], 0, implicit $exec |
| ; GFX10-NEXT: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_CO_U32_e64_]], %subreg.sub0, [[V_ADDC_U32_e64_]], %subreg.sub1 |
| ; GFX10-NEXT: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, implicit $exec, implicit $flat_scr :: (load (s8)) |
| ; GFX10-NEXT: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]] |
| ; GFX11-LABEL: name: load_flat_s32_from_1_gep_4095 |
| ; GFX11: liveins: $vgpr0_vgpr1 |
| ; GFX11-NEXT: {{ $}} |
| ; GFX11-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1 |
| ; GFX11-NEXT: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[COPY]], 4095, 0, implicit $exec, implicit $flat_scr :: (load (s8)) |
| ; GFX11-NEXT: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]] |
| %0:vgpr(p1) = COPY $vgpr0_vgpr1 |
| %1:vgpr(s64) = G_CONSTANT i64 4095 |
| %2:vgpr(p1) = G_PTR_ADD %0, %1 |
| %3:vgpr(s32) = G_LOAD %2 :: (load (s8), align 1, addrspace 0) |
| $vgpr0 = COPY %3 |
| |
| ... |
| |
| --- |
| |
| name: load_flat_s32_from_1_gep_4096 |
| legalized: true |
| regBankSelected: true |
| tracksRegLiveness: true |
| |
| body: | |
| bb.0: |
| liveins: $vgpr0_vgpr1 |
| |
| ; GFX7-LABEL: name: load_flat_s32_from_1_gep_4096 |
| ; GFX7: liveins: $vgpr0_vgpr1 |
| ; GFX7-NEXT: {{ $}} |
| ; GFX7-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1 |
| ; GFX7-NEXT: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 4096, implicit $exec |
| ; GFX7-NEXT: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 0, implicit $exec |
| ; GFX7-NEXT: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1 |
| ; GFX7-NEXT: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0 |
| ; GFX7-NEXT: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0 |
| ; GFX7-NEXT: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1 |
| ; GFX7-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1 |
| ; GFX7-NEXT: [[V_ADD_CO_U32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_CO_U32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADD_CO_U32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec |
| ; GFX7-NEXT: [[V_ADDC_U32_e64_:%[0-9]+]]:vgpr_32, dead [[V_ADDC_U32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_CO_U32_e64_1]], 0, implicit $exec |
| ; GFX7-NEXT: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_CO_U32_e64_]], %subreg.sub0, [[V_ADDC_U32_e64_]], %subreg.sub1 |
| ; GFX7-NEXT: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, implicit $exec, implicit $flat_scr :: (load (s8)) |
| ; GFX7-NEXT: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]] |
| ; GFX8-LABEL: name: load_flat_s32_from_1_gep_4096 |
| ; GFX8: liveins: $vgpr0_vgpr1 |
| ; GFX8-NEXT: {{ $}} |
| ; GFX8-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1 |
| ; GFX8-NEXT: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 4096, implicit $exec |
| ; GFX8-NEXT: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 0, implicit $exec |
| ; GFX8-NEXT: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1 |
| ; GFX8-NEXT: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0 |
| ; GFX8-NEXT: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0 |
| ; GFX8-NEXT: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1 |
| ; GFX8-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1 |
| ; GFX8-NEXT: [[V_ADD_CO_U32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_CO_U32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADD_CO_U32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec |
| ; GFX8-NEXT: [[V_ADDC_U32_e64_:%[0-9]+]]:vgpr_32, dead [[V_ADDC_U32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_CO_U32_e64_1]], 0, implicit $exec |
| ; GFX8-NEXT: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_CO_U32_e64_]], %subreg.sub0, [[V_ADDC_U32_e64_]], %subreg.sub1 |
| ; GFX8-NEXT: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, implicit $exec, implicit $flat_scr :: (load (s8)) |
| ; GFX8-NEXT: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]] |
| ; GFX9-LABEL: name: load_flat_s32_from_1_gep_4096 |
| ; GFX9: liveins: $vgpr0_vgpr1 |
| ; GFX9-NEXT: {{ $}} |
| ; GFX9-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1 |
| ; GFX9-NEXT: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 4096, implicit $exec |
| ; GFX9-NEXT: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 0, implicit $exec |
| ; GFX9-NEXT: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1 |
| ; GFX9-NEXT: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0 |
| ; GFX9-NEXT: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0 |
| ; GFX9-NEXT: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1 |
| ; GFX9-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1 |
| ; GFX9-NEXT: [[V_ADD_CO_U32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_CO_U32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADD_CO_U32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec |
| ; GFX9-NEXT: [[V_ADDC_U32_e64_:%[0-9]+]]:vgpr_32, dead [[V_ADDC_U32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_CO_U32_e64_1]], 0, implicit $exec |
| ; GFX9-NEXT: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_CO_U32_e64_]], %subreg.sub0, [[V_ADDC_U32_e64_]], %subreg.sub1 |
| ; GFX9-NEXT: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, implicit $exec, implicit $flat_scr :: (load (s8)) |
| ; GFX9-NEXT: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]] |
| ; GFX10-LABEL: name: load_flat_s32_from_1_gep_4096 |
| ; GFX10: liveins: $vgpr0_vgpr1 |
| ; GFX10-NEXT: {{ $}} |
| ; GFX10-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1 |
| ; GFX10-NEXT: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 4096, implicit $exec |
| ; GFX10-NEXT: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 0, implicit $exec |
| ; GFX10-NEXT: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1 |
| ; GFX10-NEXT: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0 |
| ; GFX10-NEXT: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0 |
| ; GFX10-NEXT: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1 |
| ; GFX10-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1 |
| ; GFX10-NEXT: [[V_ADD_CO_U32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_CO_U32_e64_1:%[0-9]+]]:sreg_32_xm0_xexec = V_ADD_CO_U32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec |
| ; GFX10-NEXT: [[V_ADDC_U32_e64_:%[0-9]+]]:vgpr_32, dead [[V_ADDC_U32_e64_1:%[0-9]+]]:sreg_32_xm0_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_CO_U32_e64_1]], 0, implicit $exec |
| ; GFX10-NEXT: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_CO_U32_e64_]], %subreg.sub0, [[V_ADDC_U32_e64_]], %subreg.sub1 |
| ; GFX10-NEXT: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, implicit $exec, implicit $flat_scr :: (load (s8)) |
| ; GFX10-NEXT: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]] |
| ; GFX11-LABEL: name: load_flat_s32_from_1_gep_4096 |
| ; GFX11: liveins: $vgpr0_vgpr1 |
| ; GFX11-NEXT: {{ $}} |
| ; GFX11-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1 |
| ; GFX11-NEXT: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 4096, implicit $exec |
| ; GFX11-NEXT: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 0, implicit $exec |
| ; GFX11-NEXT: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1 |
| ; GFX11-NEXT: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0 |
| ; GFX11-NEXT: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0 |
| ; GFX11-NEXT: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1 |
| ; GFX11-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1 |
| ; GFX11-NEXT: [[V_ADD_CO_U32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_CO_U32_e64_1:%[0-9]+]]:sreg_32_xm0_xexec = V_ADD_CO_U32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec |
| ; GFX11-NEXT: [[V_ADDC_U32_e64_:%[0-9]+]]:vgpr_32, dead [[V_ADDC_U32_e64_1:%[0-9]+]]:sreg_32_xm0_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_CO_U32_e64_1]], 0, implicit $exec |
| ; GFX11-NEXT: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_CO_U32_e64_]], %subreg.sub0, [[V_ADDC_U32_e64_]], %subreg.sub1 |
| ; GFX11-NEXT: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, implicit $exec, implicit $flat_scr :: (load (s8)) |
| ; GFX11-NEXT: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]] |
| %0:vgpr(p1) = COPY $vgpr0_vgpr1 |
| %1:vgpr(s64) = G_CONSTANT i64 4096 |
| %2:vgpr(p1) = G_PTR_ADD %0, %1 |
| %3:vgpr(s32) = G_LOAD %2 :: (load (s8), align 1, addrspace 0) |
| $vgpr0 = COPY %3 |
| |
| ... |
| |
| --- |
| |
| name: load_flat_s32_from_1_gep_m4095 |
| legalized: true |
| regBankSelected: true |
| tracksRegLiveness: true |
| |
| body: | |
| bb.0: |
| liveins: $vgpr0_vgpr1 |
| |
| ; GFX7-LABEL: name: load_flat_s32_from_1_gep_m4095 |
| ; GFX7: liveins: $vgpr0_vgpr1 |
| ; GFX7-NEXT: {{ $}} |
| ; GFX7-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1 |
| ; GFX7-NEXT: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 4294963201, implicit $exec |
| ; GFX7-NEXT: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 -1, implicit $exec |
| ; GFX7-NEXT: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1 |
| ; GFX7-NEXT: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0 |
| ; GFX7-NEXT: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0 |
| ; GFX7-NEXT: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1 |
| ; GFX7-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1 |
| ; GFX7-NEXT: [[V_ADD_CO_U32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_CO_U32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADD_CO_U32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec |
| ; GFX7-NEXT: [[V_ADDC_U32_e64_:%[0-9]+]]:vgpr_32, dead [[V_ADDC_U32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_CO_U32_e64_1]], 0, implicit $exec |
| ; GFX7-NEXT: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_CO_U32_e64_]], %subreg.sub0, [[V_ADDC_U32_e64_]], %subreg.sub1 |
| ; GFX7-NEXT: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, implicit $exec, implicit $flat_scr :: (load (s8)) |
| ; GFX7-NEXT: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]] |
| ; GFX8-LABEL: name: load_flat_s32_from_1_gep_m4095 |
| ; GFX8: liveins: $vgpr0_vgpr1 |
| ; GFX8-NEXT: {{ $}} |
| ; GFX8-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1 |
| ; GFX8-NEXT: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 4294963201, implicit $exec |
| ; GFX8-NEXT: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 -1, implicit $exec |
| ; GFX8-NEXT: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1 |
| ; GFX8-NEXT: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0 |
| ; GFX8-NEXT: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0 |
| ; GFX8-NEXT: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1 |
| ; GFX8-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1 |
| ; GFX8-NEXT: [[V_ADD_CO_U32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_CO_U32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADD_CO_U32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec |
| ; GFX8-NEXT: [[V_ADDC_U32_e64_:%[0-9]+]]:vgpr_32, dead [[V_ADDC_U32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_CO_U32_e64_1]], 0, implicit $exec |
| ; GFX8-NEXT: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_CO_U32_e64_]], %subreg.sub0, [[V_ADDC_U32_e64_]], %subreg.sub1 |
| ; GFX8-NEXT: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, implicit $exec, implicit $flat_scr :: (load (s8)) |
| ; GFX8-NEXT: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]] |
| ; GFX9-LABEL: name: load_flat_s32_from_1_gep_m4095 |
| ; GFX9: liveins: $vgpr0_vgpr1 |
| ; GFX9-NEXT: {{ $}} |
| ; GFX9-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1 |
| ; GFX9-NEXT: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 4294963201, implicit $exec |
| ; GFX9-NEXT: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 -1, implicit $exec |
| ; GFX9-NEXT: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1 |
| ; GFX9-NEXT: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0 |
| ; GFX9-NEXT: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0 |
| ; GFX9-NEXT: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1 |
| ; GFX9-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1 |
| ; GFX9-NEXT: [[V_ADD_CO_U32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_CO_U32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADD_CO_U32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec |
| ; GFX9-NEXT: [[V_ADDC_U32_e64_:%[0-9]+]]:vgpr_32, dead [[V_ADDC_U32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_CO_U32_e64_1]], 0, implicit $exec |
| ; GFX9-NEXT: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_CO_U32_e64_]], %subreg.sub0, [[V_ADDC_U32_e64_]], %subreg.sub1 |
| ; GFX9-NEXT: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, implicit $exec, implicit $flat_scr :: (load (s8)) |
| ; GFX9-NEXT: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]] |
| ; GFX10-LABEL: name: load_flat_s32_from_1_gep_m4095 |
| ; GFX10: liveins: $vgpr0_vgpr1 |
| ; GFX10-NEXT: {{ $}} |
| ; GFX10-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1 |
| ; GFX10-NEXT: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 4294963201, implicit $exec |
| ; GFX10-NEXT: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 -1, implicit $exec |
| ; GFX10-NEXT: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1 |
| ; GFX10-NEXT: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0 |
| ; GFX10-NEXT: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0 |
| ; GFX10-NEXT: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1 |
| ; GFX10-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1 |
| ; GFX10-NEXT: [[V_ADD_CO_U32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_CO_U32_e64_1:%[0-9]+]]:sreg_32_xm0_xexec = V_ADD_CO_U32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec |
| ; GFX10-NEXT: [[V_ADDC_U32_e64_:%[0-9]+]]:vgpr_32, dead [[V_ADDC_U32_e64_1:%[0-9]+]]:sreg_32_xm0_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_CO_U32_e64_1]], 0, implicit $exec |
| ; GFX10-NEXT: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_CO_U32_e64_]], %subreg.sub0, [[V_ADDC_U32_e64_]], %subreg.sub1 |
| ; GFX10-NEXT: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, implicit $exec, implicit $flat_scr :: (load (s8)) |
| ; GFX10-NEXT: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]] |
| ; GFX11-LABEL: name: load_flat_s32_from_1_gep_m4095 |
| ; GFX11: liveins: $vgpr0_vgpr1 |
| ; GFX11-NEXT: {{ $}} |
| ; GFX11-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1 |
| ; GFX11-NEXT: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 4294963201, implicit $exec |
| ; GFX11-NEXT: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 -1, implicit $exec |
| ; GFX11-NEXT: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1 |
| ; GFX11-NEXT: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0 |
| ; GFX11-NEXT: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0 |
| ; GFX11-NEXT: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1 |
| ; GFX11-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1 |
| ; GFX11-NEXT: [[V_ADD_CO_U32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_CO_U32_e64_1:%[0-9]+]]:sreg_32_xm0_xexec = V_ADD_CO_U32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec |
| ; GFX11-NEXT: [[V_ADDC_U32_e64_:%[0-9]+]]:vgpr_32, dead [[V_ADDC_U32_e64_1:%[0-9]+]]:sreg_32_xm0_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_CO_U32_e64_1]], 0, implicit $exec |
| ; GFX11-NEXT: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_CO_U32_e64_]], %subreg.sub0, [[V_ADDC_U32_e64_]], %subreg.sub1 |
| ; GFX11-NEXT: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, implicit $exec, implicit $flat_scr :: (load (s8)) |
| ; GFX11-NEXT: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]] |
| %0:vgpr(p1) = COPY $vgpr0_vgpr1 |
| %1:vgpr(s64) = G_CONSTANT i64 -4095 |
| %2:vgpr(p1) = G_PTR_ADD %0, %1 |
| %3:vgpr(s32) = G_LOAD %2 :: (load (s8), align 1, addrspace 0) |
| $vgpr0 = COPY %3 |
| |
| ... |
| |
| --- |
| |
| name: load_flat_s32_from_1_gep_m4096 |
| legalized: true |
| regBankSelected: true |
| tracksRegLiveness: true |
| |
| body: | |
| bb.0: |
| liveins: $vgpr0_vgpr1 |
| |
| ; GFX7-LABEL: name: load_flat_s32_from_1_gep_m4096 |
| ; GFX7: liveins: $vgpr0_vgpr1 |
| ; GFX7-NEXT: {{ $}} |
| ; GFX7-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1 |
| ; GFX7-NEXT: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 4294963200, implicit $exec |
| ; GFX7-NEXT: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 -1, implicit $exec |
| ; GFX7-NEXT: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1 |
| ; GFX7-NEXT: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0 |
| ; GFX7-NEXT: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0 |
| ; GFX7-NEXT: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1 |
| ; GFX7-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1 |
| ; GFX7-NEXT: [[V_ADD_CO_U32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_CO_U32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADD_CO_U32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec |
| ; GFX7-NEXT: [[V_ADDC_U32_e64_:%[0-9]+]]:vgpr_32, dead [[V_ADDC_U32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_CO_U32_e64_1]], 0, implicit $exec |
| ; GFX7-NEXT: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_CO_U32_e64_]], %subreg.sub0, [[V_ADDC_U32_e64_]], %subreg.sub1 |
| ; GFX7-NEXT: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, implicit $exec, implicit $flat_scr :: (load (s8)) |
| ; GFX7-NEXT: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]] |
| ; GFX8-LABEL: name: load_flat_s32_from_1_gep_m4096 |
| ; GFX8: liveins: $vgpr0_vgpr1 |
| ; GFX8-NEXT: {{ $}} |
| ; GFX8-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1 |
| ; GFX8-NEXT: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 4294963200, implicit $exec |
| ; GFX8-NEXT: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 -1, implicit $exec |
| ; GFX8-NEXT: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1 |
| ; GFX8-NEXT: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0 |
| ; GFX8-NEXT: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0 |
| ; GFX8-NEXT: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1 |
| ; GFX8-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1 |
| ; GFX8-NEXT: [[V_ADD_CO_U32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_CO_U32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADD_CO_U32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec |
| ; GFX8-NEXT: [[V_ADDC_U32_e64_:%[0-9]+]]:vgpr_32, dead [[V_ADDC_U32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_CO_U32_e64_1]], 0, implicit $exec |
| ; GFX8-NEXT: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_CO_U32_e64_]], %subreg.sub0, [[V_ADDC_U32_e64_]], %subreg.sub1 |
| ; GFX8-NEXT: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, implicit $exec, implicit $flat_scr :: (load (s8)) |
| ; GFX8-NEXT: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]] |
| ; GFX9-LABEL: name: load_flat_s32_from_1_gep_m4096 |
| ; GFX9: liveins: $vgpr0_vgpr1 |
| ; GFX9-NEXT: {{ $}} |
| ; GFX9-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1 |
| ; GFX9-NEXT: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 4294963200, implicit $exec |
| ; GFX9-NEXT: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 -1, implicit $exec |
| ; GFX9-NEXT: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1 |
| ; GFX9-NEXT: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0 |
| ; GFX9-NEXT: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0 |
| ; GFX9-NEXT: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1 |
| ; GFX9-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1 |
| ; GFX9-NEXT: [[V_ADD_CO_U32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_CO_U32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADD_CO_U32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec |
| ; GFX9-NEXT: [[V_ADDC_U32_e64_:%[0-9]+]]:vgpr_32, dead [[V_ADDC_U32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_CO_U32_e64_1]], 0, implicit $exec |
| ; GFX9-NEXT: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_CO_U32_e64_]], %subreg.sub0, [[V_ADDC_U32_e64_]], %subreg.sub1 |
| ; GFX9-NEXT: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, implicit $exec, implicit $flat_scr :: (load (s8)) |
| ; GFX9-NEXT: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]] |
| ; GFX10-LABEL: name: load_flat_s32_from_1_gep_m4096 |
| ; GFX10: liveins: $vgpr0_vgpr1 |
| ; GFX10-NEXT: {{ $}} |
| ; GFX10-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1 |
| ; GFX10-NEXT: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 4294963200, implicit $exec |
| ; GFX10-NEXT: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 -1, implicit $exec |
| ; GFX10-NEXT: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1 |
| ; GFX10-NEXT: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0 |
| ; GFX10-NEXT: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0 |
| ; GFX10-NEXT: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1 |
| ; GFX10-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1 |
| ; GFX10-NEXT: [[V_ADD_CO_U32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_CO_U32_e64_1:%[0-9]+]]:sreg_32_xm0_xexec = V_ADD_CO_U32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec |
| ; GFX10-NEXT: [[V_ADDC_U32_e64_:%[0-9]+]]:vgpr_32, dead [[V_ADDC_U32_e64_1:%[0-9]+]]:sreg_32_xm0_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_CO_U32_e64_1]], 0, implicit $exec |
| ; GFX10-NEXT: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_CO_U32_e64_]], %subreg.sub0, [[V_ADDC_U32_e64_]], %subreg.sub1 |
| ; GFX10-NEXT: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, implicit $exec, implicit $flat_scr :: (load (s8)) |
| ; GFX10-NEXT: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]] |
| ; GFX11-LABEL: name: load_flat_s32_from_1_gep_m4096 |
| ; GFX11: liveins: $vgpr0_vgpr1 |
| ; GFX11-NEXT: {{ $}} |
| ; GFX11-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1 |
| ; GFX11-NEXT: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 4294963200, implicit $exec |
| ; GFX11-NEXT: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 -1, implicit $exec |
| ; GFX11-NEXT: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1 |
| ; GFX11-NEXT: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0 |
| ; GFX11-NEXT: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0 |
| ; GFX11-NEXT: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1 |
| ; GFX11-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1 |
| ; GFX11-NEXT: [[V_ADD_CO_U32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_CO_U32_e64_1:%[0-9]+]]:sreg_32_xm0_xexec = V_ADD_CO_U32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec |
| ; GFX11-NEXT: [[V_ADDC_U32_e64_:%[0-9]+]]:vgpr_32, dead [[V_ADDC_U32_e64_1:%[0-9]+]]:sreg_32_xm0_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_CO_U32_e64_1]], 0, implicit $exec |
| ; GFX11-NEXT: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_CO_U32_e64_]], %subreg.sub0, [[V_ADDC_U32_e64_]], %subreg.sub1 |
| ; GFX11-NEXT: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, implicit $exec, implicit $flat_scr :: (load (s8)) |
| ; GFX11-NEXT: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]] |
| %0:vgpr(p1) = COPY $vgpr0_vgpr1 |
| %1:vgpr(s64) = G_CONSTANT i64 -4096 |
| %2:vgpr(p1) = G_PTR_ADD %0, %1 |
| %3:vgpr(s32) = G_LOAD %2 :: (load (s8), align 1, addrspace 0) |
| $vgpr0 = COPY %3 |
| |
| ... |
| |
| --- |
| |
| name: load_flat_s32_from_1_gep_8191 |
| legalized: true |
| regBankSelected: true |
| tracksRegLiveness: true |
| |
| body: | |
| bb.0: |
| liveins: $vgpr0_vgpr1 |
| |
| ; GFX7-LABEL: name: load_flat_s32_from_1_gep_8191 |
| ; GFX7: liveins: $vgpr0_vgpr1 |
| ; GFX7-NEXT: {{ $}} |
| ; GFX7-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1 |
| ; GFX7-NEXT: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 8191, implicit $exec |
| ; GFX7-NEXT: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 0, implicit $exec |
| ; GFX7-NEXT: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1 |
| ; GFX7-NEXT: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0 |
| ; GFX7-NEXT: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0 |
| ; GFX7-NEXT: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1 |
| ; GFX7-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1 |
| ; GFX7-NEXT: [[V_ADD_CO_U32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_CO_U32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADD_CO_U32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec |
| ; GFX7-NEXT: [[V_ADDC_U32_e64_:%[0-9]+]]:vgpr_32, dead [[V_ADDC_U32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_CO_U32_e64_1]], 0, implicit $exec |
| ; GFX7-NEXT: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_CO_U32_e64_]], %subreg.sub0, [[V_ADDC_U32_e64_]], %subreg.sub1 |
| ; GFX7-NEXT: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, implicit $exec, implicit $flat_scr :: (load (s8)) |
| ; GFX7-NEXT: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]] |
| ; GFX8-LABEL: name: load_flat_s32_from_1_gep_8191 |
| ; GFX8: liveins: $vgpr0_vgpr1 |
| ; GFX8-NEXT: {{ $}} |
| ; GFX8-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1 |
| ; GFX8-NEXT: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 8191, implicit $exec |
| ; GFX8-NEXT: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 0, implicit $exec |
| ; GFX8-NEXT: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1 |
| ; GFX8-NEXT: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0 |
| ; GFX8-NEXT: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0 |
| ; GFX8-NEXT: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1 |
| ; GFX8-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1 |
| ; GFX8-NEXT: [[V_ADD_CO_U32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_CO_U32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADD_CO_U32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec |
| ; GFX8-NEXT: [[V_ADDC_U32_e64_:%[0-9]+]]:vgpr_32, dead [[V_ADDC_U32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_CO_U32_e64_1]], 0, implicit $exec |
| ; GFX8-NEXT: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_CO_U32_e64_]], %subreg.sub0, [[V_ADDC_U32_e64_]], %subreg.sub1 |
| ; GFX8-NEXT: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, implicit $exec, implicit $flat_scr :: (load (s8)) |
| ; GFX8-NEXT: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]] |
| ; GFX9-LABEL: name: load_flat_s32_from_1_gep_8191 |
| ; GFX9: liveins: $vgpr0_vgpr1 |
| ; GFX9-NEXT: {{ $}} |
| ; GFX9-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1 |
| ; GFX9-NEXT: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 8191, implicit $exec |
| ; GFX9-NEXT: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 0, implicit $exec |
| ; GFX9-NEXT: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1 |
| ; GFX9-NEXT: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0 |
| ; GFX9-NEXT: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0 |
| ; GFX9-NEXT: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1 |
| ; GFX9-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1 |
| ; GFX9-NEXT: [[V_ADD_CO_U32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_CO_U32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADD_CO_U32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec |
| ; GFX9-NEXT: [[V_ADDC_U32_e64_:%[0-9]+]]:vgpr_32, dead [[V_ADDC_U32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_CO_U32_e64_1]], 0, implicit $exec |
| ; GFX9-NEXT: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_CO_U32_e64_]], %subreg.sub0, [[V_ADDC_U32_e64_]], %subreg.sub1 |
| ; GFX9-NEXT: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, implicit $exec, implicit $flat_scr :: (load (s8)) |
| ; GFX9-NEXT: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]] |
| ; GFX10-LABEL: name: load_flat_s32_from_1_gep_8191 |
| ; GFX10: liveins: $vgpr0_vgpr1 |
| ; GFX10-NEXT: {{ $}} |
| ; GFX10-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1 |
| ; GFX10-NEXT: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 8191, implicit $exec |
| ; GFX10-NEXT: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 0, implicit $exec |
| ; GFX10-NEXT: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1 |
| ; GFX10-NEXT: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0 |
| ; GFX10-NEXT: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0 |
| ; GFX10-NEXT: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1 |
| ; GFX10-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1 |
| ; GFX10-NEXT: [[V_ADD_CO_U32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_CO_U32_e64_1:%[0-9]+]]:sreg_32_xm0_xexec = V_ADD_CO_U32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec |
| ; GFX10-NEXT: [[V_ADDC_U32_e64_:%[0-9]+]]:vgpr_32, dead [[V_ADDC_U32_e64_1:%[0-9]+]]:sreg_32_xm0_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_CO_U32_e64_1]], 0, implicit $exec |
| ; GFX10-NEXT: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_CO_U32_e64_]], %subreg.sub0, [[V_ADDC_U32_e64_]], %subreg.sub1 |
| ; GFX10-NEXT: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, implicit $exec, implicit $flat_scr :: (load (s8)) |
| ; GFX10-NEXT: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]] |
| ; GFX11-LABEL: name: load_flat_s32_from_1_gep_8191 |
| ; GFX11: liveins: $vgpr0_vgpr1 |
| ; GFX11-NEXT: {{ $}} |
| ; GFX11-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1 |
| ; GFX11-NEXT: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 8191, implicit $exec |
| ; GFX11-NEXT: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 0, implicit $exec |
| ; GFX11-NEXT: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1 |
| ; GFX11-NEXT: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0 |
| ; GFX11-NEXT: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0 |
| ; GFX11-NEXT: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1 |
| ; GFX11-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1 |
| ; GFX11-NEXT: [[V_ADD_CO_U32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_CO_U32_e64_1:%[0-9]+]]:sreg_32_xm0_xexec = V_ADD_CO_U32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec |
| ; GFX11-NEXT: [[V_ADDC_U32_e64_:%[0-9]+]]:vgpr_32, dead [[V_ADDC_U32_e64_1:%[0-9]+]]:sreg_32_xm0_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_CO_U32_e64_1]], 0, implicit $exec |
| ; GFX11-NEXT: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_CO_U32_e64_]], %subreg.sub0, [[V_ADDC_U32_e64_]], %subreg.sub1 |
| ; GFX11-NEXT: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, implicit $exec, implicit $flat_scr :: (load (s8)) |
| ; GFX11-NEXT: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]] |
| %0:vgpr(p1) = COPY $vgpr0_vgpr1 |
| %1:vgpr(s64) = G_CONSTANT i64 8191 |
| %2:vgpr(p1) = G_PTR_ADD %0, %1 |
| %3:vgpr(s32) = G_LOAD %2 :: (load (s8), align 1, addrspace 0) |
| $vgpr0 = COPY %3 |
| |
| ... |
| |
| --- |
| |
| name: load_flat_s32_from_1_gep_8192 |
| legalized: true |
| regBankSelected: true |
| tracksRegLiveness: true |
| |
| body: | |
| bb.0: |
| liveins: $vgpr0_vgpr1 |
| |
| ; GFX7-LABEL: name: load_flat_s32_from_1_gep_8192 |
| ; GFX7: liveins: $vgpr0_vgpr1 |
| ; GFX7-NEXT: {{ $}} |
| ; GFX7-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1 |
| ; GFX7-NEXT: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 8192, implicit $exec |
| ; GFX7-NEXT: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 0, implicit $exec |
| ; GFX7-NEXT: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1 |
| ; GFX7-NEXT: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0 |
| ; GFX7-NEXT: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0 |
| ; GFX7-NEXT: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1 |
| ; GFX7-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1 |
| ; GFX7-NEXT: [[V_ADD_CO_U32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_CO_U32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADD_CO_U32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec |
| ; GFX7-NEXT: [[V_ADDC_U32_e64_:%[0-9]+]]:vgpr_32, dead [[V_ADDC_U32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_CO_U32_e64_1]], 0, implicit $exec |
| ; GFX7-NEXT: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_CO_U32_e64_]], %subreg.sub0, [[V_ADDC_U32_e64_]], %subreg.sub1 |
| ; GFX7-NEXT: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, implicit $exec, implicit $flat_scr :: (load (s8)) |
| ; GFX7-NEXT: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]] |
| ; GFX8-LABEL: name: load_flat_s32_from_1_gep_8192 |
| ; GFX8: liveins: $vgpr0_vgpr1 |
| ; GFX8-NEXT: {{ $}} |
| ; GFX8-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1 |
| ; GFX8-NEXT: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 8192, implicit $exec |
| ; GFX8-NEXT: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 0, implicit $exec |
| ; GFX8-NEXT: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1 |
| ; GFX8-NEXT: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0 |
| ; GFX8-NEXT: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0 |
| ; GFX8-NEXT: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1 |
| ; GFX8-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1 |
| ; GFX8-NEXT: [[V_ADD_CO_U32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_CO_U32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADD_CO_U32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec |
| ; GFX8-NEXT: [[V_ADDC_U32_e64_:%[0-9]+]]:vgpr_32, dead [[V_ADDC_U32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_CO_U32_e64_1]], 0, implicit $exec |
| ; GFX8-NEXT: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_CO_U32_e64_]], %subreg.sub0, [[V_ADDC_U32_e64_]], %subreg.sub1 |
| ; GFX8-NEXT: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, implicit $exec, implicit $flat_scr :: (load (s8)) |
| ; GFX8-NEXT: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]] |
| ; GFX9-LABEL: name: load_flat_s32_from_1_gep_8192 |
| ; GFX9: liveins: $vgpr0_vgpr1 |
| ; GFX9-NEXT: {{ $}} |
| ; GFX9-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1 |
| ; GFX9-NEXT: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 8192, implicit $exec |
| ; GFX9-NEXT: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 0, implicit $exec |
| ; GFX9-NEXT: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1 |
| ; GFX9-NEXT: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0 |
| ; GFX9-NEXT: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0 |
| ; GFX9-NEXT: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1 |
| ; GFX9-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1 |
| ; GFX9-NEXT: [[V_ADD_CO_U32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_CO_U32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADD_CO_U32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec |
| ; GFX9-NEXT: [[V_ADDC_U32_e64_:%[0-9]+]]:vgpr_32, dead [[V_ADDC_U32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_CO_U32_e64_1]], 0, implicit $exec |
| ; GFX9-NEXT: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_CO_U32_e64_]], %subreg.sub0, [[V_ADDC_U32_e64_]], %subreg.sub1 |
| ; GFX9-NEXT: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, implicit $exec, implicit $flat_scr :: (load (s8)) |
| ; GFX9-NEXT: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]] |
| ; GFX10-LABEL: name: load_flat_s32_from_1_gep_8192 |
| ; GFX10: liveins: $vgpr0_vgpr1 |
| ; GFX10-NEXT: {{ $}} |
| ; GFX10-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1 |
| ; GFX10-NEXT: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 8192, implicit $exec |
| ; GFX10-NEXT: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 0, implicit $exec |
| ; GFX10-NEXT: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1 |
| ; GFX10-NEXT: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0 |
| ; GFX10-NEXT: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0 |
| ; GFX10-NEXT: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1 |
| ; GFX10-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1 |
| ; GFX10-NEXT: [[V_ADD_CO_U32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_CO_U32_e64_1:%[0-9]+]]:sreg_32_xm0_xexec = V_ADD_CO_U32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec |
| ; GFX10-NEXT: [[V_ADDC_U32_e64_:%[0-9]+]]:vgpr_32, dead [[V_ADDC_U32_e64_1:%[0-9]+]]:sreg_32_xm0_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_CO_U32_e64_1]], 0, implicit $exec |
| ; GFX10-NEXT: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_CO_U32_e64_]], %subreg.sub0, [[V_ADDC_U32_e64_]], %subreg.sub1 |
| ; GFX10-NEXT: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, implicit $exec, implicit $flat_scr :: (load (s8)) |
| ; GFX10-NEXT: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]] |
| ; GFX11-LABEL: name: load_flat_s32_from_1_gep_8192 |
| ; GFX11: liveins: $vgpr0_vgpr1 |
| ; GFX11-NEXT: {{ $}} |
| ; GFX11-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1 |
| ; GFX11-NEXT: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 8192, implicit $exec |
| ; GFX11-NEXT: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 0, implicit $exec |
| ; GFX11-NEXT: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1 |
| ; GFX11-NEXT: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0 |
| ; GFX11-NEXT: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0 |
| ; GFX11-NEXT: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1 |
| ; GFX11-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1 |
| ; GFX11-NEXT: [[V_ADD_CO_U32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_CO_U32_e64_1:%[0-9]+]]:sreg_32_xm0_xexec = V_ADD_CO_U32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec |
| ; GFX11-NEXT: [[V_ADDC_U32_e64_:%[0-9]+]]:vgpr_32, dead [[V_ADDC_U32_e64_1:%[0-9]+]]:sreg_32_xm0_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_CO_U32_e64_1]], 0, implicit $exec |
| ; GFX11-NEXT: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_CO_U32_e64_]], %subreg.sub0, [[V_ADDC_U32_e64_]], %subreg.sub1 |
| ; GFX11-NEXT: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, implicit $exec, implicit $flat_scr :: (load (s8)) |
| ; GFX11-NEXT: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]] |
| %0:vgpr(p1) = COPY $vgpr0_vgpr1 |
| %1:vgpr(s64) = G_CONSTANT i64 8192 |
| %2:vgpr(p1) = G_PTR_ADD %0, %1 |
| %3:vgpr(s32) = G_LOAD %2 :: (load (s8), align 1, addrspace 0) |
| $vgpr0 = COPY %3 |
| |
| ... |
| |
| --- |
| |
| name: load_flat_s32_from_1_gep_m8191 |
| legalized: true |
| regBankSelected: true |
| tracksRegLiveness: true |
| |
| body: | |
| bb.0: |
| liveins: $vgpr0_vgpr1 |
| |
| ; GFX7-LABEL: name: load_flat_s32_from_1_gep_m8191 |
| ; GFX7: liveins: $vgpr0_vgpr1 |
| ; GFX7-NEXT: {{ $}} |
| ; GFX7-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1 |
| ; GFX7-NEXT: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 4294959105, implicit $exec |
| ; GFX7-NEXT: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 -1, implicit $exec |
| ; GFX7-NEXT: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1 |
| ; GFX7-NEXT: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0 |
| ; GFX7-NEXT: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0 |
| ; GFX7-NEXT: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1 |
| ; GFX7-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1 |
| ; GFX7-NEXT: [[V_ADD_CO_U32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_CO_U32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADD_CO_U32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec |
| ; GFX7-NEXT: [[V_ADDC_U32_e64_:%[0-9]+]]:vgpr_32, dead [[V_ADDC_U32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_CO_U32_e64_1]], 0, implicit $exec |
| ; GFX7-NEXT: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_CO_U32_e64_]], %subreg.sub0, [[V_ADDC_U32_e64_]], %subreg.sub1 |
| ; GFX7-NEXT: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, implicit $exec, implicit $flat_scr :: (load (s8)) |
| ; GFX7-NEXT: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]] |
| ; GFX8-LABEL: name: load_flat_s32_from_1_gep_m8191 |
| ; GFX8: liveins: $vgpr0_vgpr1 |
| ; GFX8-NEXT: {{ $}} |
| ; GFX8-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1 |
| ; GFX8-NEXT: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 4294959105, implicit $exec |
| ; GFX8-NEXT: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 -1, implicit $exec |
| ; GFX8-NEXT: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1 |
| ; GFX8-NEXT: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0 |
| ; GFX8-NEXT: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0 |
| ; GFX8-NEXT: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1 |
| ; GFX8-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1 |
| ; GFX8-NEXT: [[V_ADD_CO_U32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_CO_U32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADD_CO_U32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec |
| ; GFX8-NEXT: [[V_ADDC_U32_e64_:%[0-9]+]]:vgpr_32, dead [[V_ADDC_U32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_CO_U32_e64_1]], 0, implicit $exec |
| ; GFX8-NEXT: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_CO_U32_e64_]], %subreg.sub0, [[V_ADDC_U32_e64_]], %subreg.sub1 |
| ; GFX8-NEXT: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, implicit $exec, implicit $flat_scr :: (load (s8)) |
| ; GFX8-NEXT: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]] |
| ; GFX9-LABEL: name: load_flat_s32_from_1_gep_m8191 |
| ; GFX9: liveins: $vgpr0_vgpr1 |
| ; GFX9-NEXT: {{ $}} |
| ; GFX9-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1 |
| ; GFX9-NEXT: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 4294959105, implicit $exec |
| ; GFX9-NEXT: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 -1, implicit $exec |
| ; GFX9-NEXT: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1 |
| ; GFX9-NEXT: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0 |
| ; GFX9-NEXT: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0 |
| ; GFX9-NEXT: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1 |
| ; GFX9-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1 |
| ; GFX9-NEXT: [[V_ADD_CO_U32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_CO_U32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADD_CO_U32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec |
| ; GFX9-NEXT: [[V_ADDC_U32_e64_:%[0-9]+]]:vgpr_32, dead [[V_ADDC_U32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_CO_U32_e64_1]], 0, implicit $exec |
| ; GFX9-NEXT: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_CO_U32_e64_]], %subreg.sub0, [[V_ADDC_U32_e64_]], %subreg.sub1 |
| ; GFX9-NEXT: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, implicit $exec, implicit $flat_scr :: (load (s8)) |
| ; GFX9-NEXT: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]] |
| ; GFX10-LABEL: name: load_flat_s32_from_1_gep_m8191 |
| ; GFX10: liveins: $vgpr0_vgpr1 |
| ; GFX10-NEXT: {{ $}} |
| ; GFX10-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1 |
| ; GFX10-NEXT: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 4294959105, implicit $exec |
| ; GFX10-NEXT: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 -1, implicit $exec |
| ; GFX10-NEXT: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1 |
| ; GFX10-NEXT: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0 |
| ; GFX10-NEXT: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0 |
| ; GFX10-NEXT: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1 |
| ; GFX10-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1 |
| ; GFX10-NEXT: [[V_ADD_CO_U32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_CO_U32_e64_1:%[0-9]+]]:sreg_32_xm0_xexec = V_ADD_CO_U32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec |
| ; GFX10-NEXT: [[V_ADDC_U32_e64_:%[0-9]+]]:vgpr_32, dead [[V_ADDC_U32_e64_1:%[0-9]+]]:sreg_32_xm0_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_CO_U32_e64_1]], 0, implicit $exec |
| ; GFX10-NEXT: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_CO_U32_e64_]], %subreg.sub0, [[V_ADDC_U32_e64_]], %subreg.sub1 |
| ; GFX10-NEXT: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, implicit $exec, implicit $flat_scr :: (load (s8)) |
| ; GFX10-NEXT: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]] |
| ; GFX11-LABEL: name: load_flat_s32_from_1_gep_m8191 |
| ; GFX11: liveins: $vgpr0_vgpr1 |
| ; GFX11-NEXT: {{ $}} |
| ; GFX11-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1 |
| ; GFX11-NEXT: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 4294959105, implicit $exec |
| ; GFX11-NEXT: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 -1, implicit $exec |
| ; GFX11-NEXT: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1 |
| ; GFX11-NEXT: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0 |
| ; GFX11-NEXT: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0 |
| ; GFX11-NEXT: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1 |
| ; GFX11-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1 |
| ; GFX11-NEXT: [[V_ADD_CO_U32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_CO_U32_e64_1:%[0-9]+]]:sreg_32_xm0_xexec = V_ADD_CO_U32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec |
| ; GFX11-NEXT: [[V_ADDC_U32_e64_:%[0-9]+]]:vgpr_32, dead [[V_ADDC_U32_e64_1:%[0-9]+]]:sreg_32_xm0_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_CO_U32_e64_1]], 0, implicit $exec |
| ; GFX11-NEXT: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_CO_U32_e64_]], %subreg.sub0, [[V_ADDC_U32_e64_]], %subreg.sub1 |
| ; GFX11-NEXT: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, implicit $exec, implicit $flat_scr :: (load (s8)) |
| ; GFX11-NEXT: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]] |
| %0:vgpr(p1) = COPY $vgpr0_vgpr1 |
| %1:vgpr(s64) = G_CONSTANT i64 -8191 |
| %2:vgpr(p1) = G_PTR_ADD %0, %1 |
| %3:vgpr(s32) = G_LOAD %2 :: (load (s8), align 1, addrspace 0) |
| $vgpr0 = COPY %3 |
| |
| ... |
| |
| --- |
| |
| name: load_flat_s32_from_1_gep_m8192 |
| legalized: true |
| regBankSelected: true |
| tracksRegLiveness: true |
| |
| body: | |
| bb.0: |
| liveins: $vgpr0_vgpr1 |
| |
| ; GFX7-LABEL: name: load_flat_s32_from_1_gep_m8192 |
| ; GFX7: liveins: $vgpr0_vgpr1 |
| ; GFX7-NEXT: {{ $}} |
| ; GFX7-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1 |
| ; GFX7-NEXT: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 4294959104, implicit $exec |
| ; GFX7-NEXT: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 -1, implicit $exec |
| ; GFX7-NEXT: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1 |
| ; GFX7-NEXT: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0 |
| ; GFX7-NEXT: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0 |
| ; GFX7-NEXT: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1 |
| ; GFX7-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1 |
| ; GFX7-NEXT: [[V_ADD_CO_U32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_CO_U32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADD_CO_U32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec |
| ; GFX7-NEXT: [[V_ADDC_U32_e64_:%[0-9]+]]:vgpr_32, dead [[V_ADDC_U32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_CO_U32_e64_1]], 0, implicit $exec |
| ; GFX7-NEXT: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_CO_U32_e64_]], %subreg.sub0, [[V_ADDC_U32_e64_]], %subreg.sub1 |
| ; GFX7-NEXT: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, implicit $exec, implicit $flat_scr :: (load (s8)) |
| ; GFX7-NEXT: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]] |
| ; GFX8-LABEL: name: load_flat_s32_from_1_gep_m8192 |
| ; GFX8: liveins: $vgpr0_vgpr1 |
| ; GFX8-NEXT: {{ $}} |
| ; GFX8-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1 |
| ; GFX8-NEXT: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 4294959104, implicit $exec |
| ; GFX8-NEXT: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 -1, implicit $exec |
| ; GFX8-NEXT: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1 |
| ; GFX8-NEXT: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0 |
| ; GFX8-NEXT: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0 |
| ; GFX8-NEXT: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1 |
| ; GFX8-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1 |
| ; GFX8-NEXT: [[V_ADD_CO_U32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_CO_U32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADD_CO_U32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec |
| ; GFX8-NEXT: [[V_ADDC_U32_e64_:%[0-9]+]]:vgpr_32, dead [[V_ADDC_U32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_CO_U32_e64_1]], 0, implicit $exec |
| ; GFX8-NEXT: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_CO_U32_e64_]], %subreg.sub0, [[V_ADDC_U32_e64_]], %subreg.sub1 |
| ; GFX8-NEXT: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, implicit $exec, implicit $flat_scr :: (load (s8)) |
| ; GFX8-NEXT: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]] |
| ; GFX9-LABEL: name: load_flat_s32_from_1_gep_m8192 |
| ; GFX9: liveins: $vgpr0_vgpr1 |
| ; GFX9-NEXT: {{ $}} |
| ; GFX9-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1 |
| ; GFX9-NEXT: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 4294959104, implicit $exec |
| ; GFX9-NEXT: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 -1, implicit $exec |
| ; GFX9-NEXT: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1 |
| ; GFX9-NEXT: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0 |
| ; GFX9-NEXT: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0 |
| ; GFX9-NEXT: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1 |
| ; GFX9-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1 |
| ; GFX9-NEXT: [[V_ADD_CO_U32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_CO_U32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADD_CO_U32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec |
| ; GFX9-NEXT: [[V_ADDC_U32_e64_:%[0-9]+]]:vgpr_32, dead [[V_ADDC_U32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_CO_U32_e64_1]], 0, implicit $exec |
| ; GFX9-NEXT: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_CO_U32_e64_]], %subreg.sub0, [[V_ADDC_U32_e64_]], %subreg.sub1 |
| ; GFX9-NEXT: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, implicit $exec, implicit $flat_scr :: (load (s8)) |
| ; GFX9-NEXT: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]] |
| ; GFX10-LABEL: name: load_flat_s32_from_1_gep_m8192 |
| ; GFX10: liveins: $vgpr0_vgpr1 |
| ; GFX10-NEXT: {{ $}} |
| ; GFX10-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1 |
| ; GFX10-NEXT: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 4294959104, implicit $exec |
| ; GFX10-NEXT: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 -1, implicit $exec |
| ; GFX10-NEXT: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1 |
| ; GFX10-NEXT: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0 |
| ; GFX10-NEXT: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0 |
| ; GFX10-NEXT: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1 |
| ; GFX10-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1 |
| ; GFX10-NEXT: [[V_ADD_CO_U32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_CO_U32_e64_1:%[0-9]+]]:sreg_32_xm0_xexec = V_ADD_CO_U32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec |
| ; GFX10-NEXT: [[V_ADDC_U32_e64_:%[0-9]+]]:vgpr_32, dead [[V_ADDC_U32_e64_1:%[0-9]+]]:sreg_32_xm0_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_CO_U32_e64_1]], 0, implicit $exec |
| ; GFX10-NEXT: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_CO_U32_e64_]], %subreg.sub0, [[V_ADDC_U32_e64_]], %subreg.sub1 |
| ; GFX10-NEXT: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, implicit $exec, implicit $flat_scr :: (load (s8)) |
| ; GFX10-NEXT: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]] |
| ; GFX11-LABEL: name: load_flat_s32_from_1_gep_m8192 |
| ; GFX11: liveins: $vgpr0_vgpr1 |
| ; GFX11-NEXT: {{ $}} |
| ; GFX11-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1 |
| ; GFX11-NEXT: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 4294959104, implicit $exec |
| ; GFX11-NEXT: [[V_MOV_B32_e32_1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 -1, implicit $exec |
| ; GFX11-NEXT: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_MOV_B32_e32_]], %subreg.sub0, [[V_MOV_B32_e32_1]], %subreg.sub1 |
| ; GFX11-NEXT: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0 |
| ; GFX11-NEXT: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub0 |
| ; GFX11-NEXT: [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1 |
| ; GFX11-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE]].sub1 |
| ; GFX11-NEXT: [[V_ADD_CO_U32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_CO_U32_e64_1:%[0-9]+]]:sreg_32_xm0_xexec = V_ADD_CO_U32_e64 [[COPY1]], [[COPY2]], 0, implicit $exec |
| ; GFX11-NEXT: [[V_ADDC_U32_e64_:%[0-9]+]]:vgpr_32, dead [[V_ADDC_U32_e64_1:%[0-9]+]]:sreg_32_xm0_xexec = V_ADDC_U32_e64 [[COPY3]], [[COPY4]], killed [[V_ADD_CO_U32_e64_1]], 0, implicit $exec |
| ; GFX11-NEXT: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_CO_U32_e64_]], %subreg.sub0, [[V_ADDC_U32_e64_]], %subreg.sub1 |
| ; GFX11-NEXT: [[FLAT_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = FLAT_LOAD_UBYTE [[REG_SEQUENCE1]], 0, 0, implicit $exec, implicit $flat_scr :: (load (s8)) |
| ; GFX11-NEXT: $vgpr0 = COPY [[FLAT_LOAD_UBYTE]] |
| %0:vgpr(p1) = COPY $vgpr0_vgpr1 |
| %1:vgpr(s64) = G_CONSTANT i64 -8192 |
| %2:vgpr(p1) = G_PTR_ADD %0, %1 |
| %3:vgpr(s32) = G_LOAD %2 :: (load (s8), align 1, addrspace 0) |
| $vgpr0 = COPY %3 |
| |
| ... |
