diff options
Diffstat (limited to 'llvm/test/CodeGen/AMDGPU/merge-tbuffer-gfx10.mir')
| -rw-r--r-- | llvm/test/CodeGen/AMDGPU/merge-tbuffer-gfx10.mir | 1526 |
1 files changed, 1526 insertions, 0 deletions
diff --git a/llvm/test/CodeGen/AMDGPU/merge-tbuffer-gfx10.mir b/llvm/test/CodeGen/AMDGPU/merge-tbuffer-gfx10.mir new file mode 100644 index 000000000000..402c00298c8d --- /dev/null +++ b/llvm/test/CodeGen/AMDGPU/merge-tbuffer-gfx10.mir @@ -0,0 +1,1526 @@ +# NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py UTC_ARGS: --version 5 +# RUN: llc -mtriple=amdgcn -mcpu=gfx1010 -verify-machineinstrs -run-pass si-load-store-opt -o - %s | FileCheck -check-prefix=GFX10 %s + +--- +name: gfx10_tbuffer_load_x_xyz +body: | + bb.0.entry: + ; GFX10-LABEL: name: gfx10_tbuffer_load_x_xyz + ; GFX10: [[COPY:%[0-9]+]]:sgpr_32 = COPY $sgpr0 + ; GFX10-NEXT: [[COPY1:%[0-9]+]]:sgpr_32 = COPY $sgpr1 + ; GFX10-NEXT: [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr2 + ; GFX10-NEXT: [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr3 + ; GFX10-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3 + ; GFX10-NEXT: [[TBUFFER_LOAD_FORMAT_XYZW_OFFSET:%[0-9]+]]:vreg_128 = TBUFFER_LOAD_FORMAT_XYZW_OFFSET [[REG_SEQUENCE]], 0, 4, 77, 0, 0, implicit $exec :: (dereferenceable load (s128), align 1, addrspace 4) + ; GFX10-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[TBUFFER_LOAD_FORMAT_XYZW_OFFSET]].sub0 + ; GFX10-NEXT: [[COPY5:%[0-9]+]]:vreg_96 = COPY killed [[TBUFFER_LOAD_FORMAT_XYZW_OFFSET]].sub1_sub2_sub3 + %0:sgpr_32 = COPY $sgpr0 + %1:sgpr_32 = COPY $sgpr1 + %2:sgpr_32 = COPY $sgpr2 + %3:sgpr_32 = COPY $sgpr3 + %5:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3 + %7:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %5:sgpr_128, 0, 4, 22, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) + %8:vreg_96 = TBUFFER_LOAD_FORMAT_XYZ_OFFSET %5:sgpr_128, 0, 8, 74, 0, 0, implicit $exec :: (dereferenceable load (s96), align 1, addrspace 4) +... +--- + +name: gfx10_tbuffer_load_xyz_x +body: | + bb.0.entry: + ; GFX10-LABEL: name: gfx10_tbuffer_load_xyz_x + ; GFX10: [[COPY:%[0-9]+]]:sgpr_32 = COPY $sgpr0 + ; GFX10-NEXT: [[COPY1:%[0-9]+]]:sgpr_32 = COPY $sgpr1 + ; GFX10-NEXT: [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr2 + ; GFX10-NEXT: [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr3 + ; GFX10-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3 + ; GFX10-NEXT: [[TBUFFER_LOAD_FORMAT_XYZW_OFFSET:%[0-9]+]]:vreg_128 = TBUFFER_LOAD_FORMAT_XYZW_OFFSET [[REG_SEQUENCE]], 0, 4, 77, 0, 0, implicit $exec :: (dereferenceable load (s128), align 1, addrspace 4) + ; GFX10-NEXT: [[COPY4:%[0-9]+]]:vreg_96 = COPY [[TBUFFER_LOAD_FORMAT_XYZW_OFFSET]].sub0_sub1_sub2 + ; GFX10-NEXT: [[COPY5:%[0-9]+]]:vgpr_32 = COPY killed [[TBUFFER_LOAD_FORMAT_XYZW_OFFSET]].sub3 + %0:sgpr_32 = COPY $sgpr0 + %1:sgpr_32 = COPY $sgpr1 + %2:sgpr_32 = COPY $sgpr2 + %3:sgpr_32 = COPY $sgpr3 + %5:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3 + %7:vreg_96 = TBUFFER_LOAD_FORMAT_XYZ_OFFSET %5:sgpr_128, 0, 4, 74, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) + %8:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %5:sgpr_128, 0, 16, 22, 0, 0, implicit $exec :: (dereferenceable load (s96), align 1, addrspace 4) +... +--- + +name: gfx10_tbuffer_load_xy_xy +body: | + bb.0.entry: + ; GFX10-LABEL: name: gfx10_tbuffer_load_xy_xy + ; GFX10: [[COPY:%[0-9]+]]:sgpr_32 = COPY $sgpr0 + ; GFX10-NEXT: [[COPY1:%[0-9]+]]:sgpr_32 = COPY $sgpr1 + ; GFX10-NEXT: [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr2 + ; GFX10-NEXT: [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr3 + ; GFX10-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3 + ; GFX10-NEXT: [[TBUFFER_LOAD_FORMAT_XYZW_OFFSET:%[0-9]+]]:vreg_128 = TBUFFER_LOAD_FORMAT_XYZW_OFFSET [[REG_SEQUENCE]], 0, 4, 77, 0, 0, implicit $exec :: (dereferenceable load (s128), align 1, addrspace 4) + ; GFX10-NEXT: [[COPY4:%[0-9]+]]:vreg_64 = COPY [[TBUFFER_LOAD_FORMAT_XYZW_OFFSET]].sub0_sub1 + ; GFX10-NEXT: [[COPY5:%[0-9]+]]:vreg_64 = COPY killed [[TBUFFER_LOAD_FORMAT_XYZW_OFFSET]].sub2_sub3 + %0:sgpr_32 = COPY $sgpr0 + %1:sgpr_32 = COPY $sgpr1 + %2:sgpr_32 = COPY $sgpr2 + %3:sgpr_32 = COPY $sgpr3 + %5:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3 + %7:vreg_64 = TBUFFER_LOAD_FORMAT_XY_OFFSET %5:sgpr_128, 0, 4, 64, 0, 0, implicit $exec :: (dereferenceable load (s64), align 1, addrspace 4) + %8:vreg_64 = TBUFFER_LOAD_FORMAT_XY_OFFSET %5:sgpr_128, 0, 12, 64, 0, 0, implicit $exec :: (dereferenceable load (s64), align 1, addrspace 4) +... +--- + +name: gfx10_tbuffer_load_x_xy +body: | + bb.0.entry: + ; GFX10-LABEL: name: gfx10_tbuffer_load_x_xy + ; GFX10: [[COPY:%[0-9]+]]:sgpr_32 = COPY $sgpr0 + ; GFX10-NEXT: [[COPY1:%[0-9]+]]:sgpr_32 = COPY $sgpr1 + ; GFX10-NEXT: [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr2 + ; GFX10-NEXT: [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr3 + ; GFX10-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3 + ; GFX10-NEXT: [[TBUFFER_LOAD_FORMAT_XYZ_OFFSET:%[0-9]+]]:vreg_96 = TBUFFER_LOAD_FORMAT_XYZ_OFFSET [[REG_SEQUENCE]], 0, 4, 74, 0, 0, implicit $exec :: (dereferenceable load (s96), align 1, addrspace 4) + ; GFX10-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[TBUFFER_LOAD_FORMAT_XYZ_OFFSET]].sub0 + ; GFX10-NEXT: [[COPY5:%[0-9]+]]:vreg_64 = COPY killed [[TBUFFER_LOAD_FORMAT_XYZ_OFFSET]].sub1_sub2 + %0:sgpr_32 = COPY $sgpr0 + %1:sgpr_32 = COPY $sgpr1 + %2:sgpr_32 = COPY $sgpr2 + %3:sgpr_32 = COPY $sgpr3 + %5:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3 + %7:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %5:sgpr_128, 0, 4, 22, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) + %8:vreg_64 = TBUFFER_LOAD_FORMAT_XY_OFFSET %5:sgpr_128, 0, 8, 64, 0, 0, implicit $exec :: (dereferenceable load (s64), align 1, addrspace 4) +... +--- + +name: gfx10_tbuffer_load_xy_x +body: | + bb.0.entry: + ; GFX10-LABEL: name: gfx10_tbuffer_load_xy_x + ; GFX10: [[COPY:%[0-9]+]]:sgpr_32 = COPY $sgpr0 + ; GFX10-NEXT: [[COPY1:%[0-9]+]]:sgpr_32 = COPY $sgpr1 + ; GFX10-NEXT: [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr2 + ; GFX10-NEXT: [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr3 + ; GFX10-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3 + ; GFX10-NEXT: [[TBUFFER_LOAD_FORMAT_XYZ_OFFSET:%[0-9]+]]:vreg_96 = TBUFFER_LOAD_FORMAT_XYZ_OFFSET [[REG_SEQUENCE]], 0, 4, 74, 0, 0, implicit $exec :: (dereferenceable load (s96), align 1, addrspace 4) + ; GFX10-NEXT: [[COPY4:%[0-9]+]]:vreg_64 = COPY [[TBUFFER_LOAD_FORMAT_XYZ_OFFSET]].sub0_sub1 + ; GFX10-NEXT: [[COPY5:%[0-9]+]]:vgpr_32 = COPY killed [[TBUFFER_LOAD_FORMAT_XYZ_OFFSET]].sub2 + %0:sgpr_32 = COPY $sgpr0 + %1:sgpr_32 = COPY $sgpr1 + %2:sgpr_32 = COPY $sgpr2 + %3:sgpr_32 = COPY $sgpr3 + %5:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3 + %7:vreg_64 = TBUFFER_LOAD_FORMAT_XY_OFFSET %5:sgpr_128, 0, 4, 64, 0, 0, implicit $exec :: (dereferenceable load (s64), align 1, addrspace 4) + %8:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %5:sgpr_128, 0, 12, 22, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) +... +--- + +name: gfx10_tbuffer_load_x_x +body: | + bb.0.entry: + ; GFX10-LABEL: name: gfx10_tbuffer_load_x_x + ; GFX10: [[COPY:%[0-9]+]]:sgpr_32 = COPY $sgpr0 + ; GFX10-NEXT: [[COPY1:%[0-9]+]]:sgpr_32 = COPY $sgpr1 + ; GFX10-NEXT: [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr2 + ; GFX10-NEXT: [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr3 + ; GFX10-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3 + ; GFX10-NEXT: [[TBUFFER_LOAD_FORMAT_XY_OFFSET:%[0-9]+]]:vreg_64 = TBUFFER_LOAD_FORMAT_XY_OFFSET [[REG_SEQUENCE]], 0, 4, 64, 0, 0, implicit $exec :: (dereferenceable load (s64), align 1, addrspace 4) + ; GFX10-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[TBUFFER_LOAD_FORMAT_XY_OFFSET]].sub0 + ; GFX10-NEXT: [[COPY5:%[0-9]+]]:vgpr_32 = COPY killed [[TBUFFER_LOAD_FORMAT_XY_OFFSET]].sub1 + %0:sgpr_32 = COPY $sgpr0 + %1:sgpr_32 = COPY $sgpr1 + %2:sgpr_32 = COPY $sgpr2 + %3:sgpr_32 = COPY $sgpr3 + %5:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3 + %7:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %5:sgpr_128, 0, 4, 22, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) + %8:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %5:sgpr_128, 0, 8, 22, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) +... +--- + +name: gfx10_tbuffer_load_x_x_format_32_32_32_32 +body: | + bb.0.entry: + ; GFX10-LABEL: name: gfx10_tbuffer_load_x_x_format_32_32_32_32 + ; GFX10: [[COPY:%[0-9]+]]:sgpr_32 = COPY $sgpr0 + ; GFX10-NEXT: [[COPY1:%[0-9]+]]:sgpr_32 = COPY $sgpr1 + ; GFX10-NEXT: [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr2 + ; GFX10-NEXT: [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr3 + ; GFX10-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3 + ; GFX10-NEXT: [[TBUFFER_LOAD_FORMAT_XY_OFFSET:%[0-9]+]]:vreg_64 = TBUFFER_LOAD_FORMAT_XY_OFFSET [[REG_SEQUENCE]], 0, 4, 64, 0, 0, implicit $exec :: (dereferenceable load (s64), align 1, addrspace 4) + ; GFX10-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[TBUFFER_LOAD_FORMAT_XY_OFFSET]].sub0 + ; GFX10-NEXT: [[COPY5:%[0-9]+]]:vgpr_32 = COPY killed [[TBUFFER_LOAD_FORMAT_XY_OFFSET]].sub1 + %0:sgpr_32 = COPY $sgpr0 + %1:sgpr_32 = COPY $sgpr1 + %2:sgpr_32 = COPY $sgpr2 + %3:sgpr_32 = COPY $sgpr3 + %5:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3 + %7:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %5:sgpr_128, 0, 4, 77, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) + %8:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %5:sgpr_128, 0, 8, 77, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) +... +--- + +name: gfx10_tbuffer_load_float_32 +body: | + bb.0.entry: + ; GFX10-LABEL: name: gfx10_tbuffer_load_float_32 + ; GFX10: [[COPY:%[0-9]+]]:sgpr_32 = COPY $sgpr0 + ; GFX10-NEXT: [[COPY1:%[0-9]+]]:sgpr_32 = COPY $sgpr1 + ; GFX10-NEXT: [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr2 + ; GFX10-NEXT: [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr3 + ; GFX10-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3 + ; GFX10-NEXT: [[TBUFFER_LOAD_FORMAT_XY_OFFSET:%[0-9]+]]:vreg_64 = TBUFFER_LOAD_FORMAT_XY_OFFSET [[REG_SEQUENCE]], 0, 4, 64, 0, 0, implicit $exec :: (dereferenceable load (s64), align 1, addrspace 4) + ; GFX10-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[TBUFFER_LOAD_FORMAT_XY_OFFSET]].sub0 + ; GFX10-NEXT: [[COPY5:%[0-9]+]]:vgpr_32 = COPY killed [[TBUFFER_LOAD_FORMAT_XY_OFFSET]].sub1 + ; GFX10-NEXT: [[TBUFFER_LOAD_FORMAT_XYZW_OFFSET:%[0-9]+]]:vreg_128 = TBUFFER_LOAD_FORMAT_XYZW_OFFSET [[REG_SEQUENCE]], 0, 16, 77, 0, 0, implicit $exec :: (dereferenceable load (s128), align 1, addrspace 4) + ; GFX10-NEXT: [[COPY6:%[0-9]+]]:vreg_96 = COPY [[TBUFFER_LOAD_FORMAT_XYZW_OFFSET]].sub0_sub1_sub2 + ; GFX10-NEXT: [[COPY7:%[0-9]+]]:vgpr_32 = COPY killed [[TBUFFER_LOAD_FORMAT_XYZW_OFFSET]].sub3 + ; GFX10-NEXT: [[COPY8:%[0-9]+]]:vreg_64 = COPY [[COPY6]].sub0_sub1 + ; GFX10-NEXT: [[COPY9:%[0-9]+]]:vgpr_32 = COPY killed [[COPY6]].sub2 + ; GFX10-NEXT: [[COPY10:%[0-9]+]]:vgpr_32 = COPY [[COPY8]].sub0 + ; GFX10-NEXT: [[COPY11:%[0-9]+]]:vgpr_32 = COPY killed [[COPY8]].sub1 + ; GFX10-NEXT: [[TBUFFER_LOAD_FORMAT_XYZ_OFFSET:%[0-9]+]]:vreg_96 = TBUFFER_LOAD_FORMAT_XYZ_OFFSET [[REG_SEQUENCE]], 0, 36, 74, 0, 0, implicit $exec :: (dereferenceable load (s96), align 1, addrspace 4) + ; GFX10-NEXT: [[COPY12:%[0-9]+]]:vreg_64 = COPY [[TBUFFER_LOAD_FORMAT_XYZ_OFFSET]].sub0_sub1 + ; GFX10-NEXT: [[COPY13:%[0-9]+]]:vgpr_32 = COPY killed [[TBUFFER_LOAD_FORMAT_XYZ_OFFSET]].sub2 + ; GFX10-NEXT: [[COPY14:%[0-9]+]]:vgpr_32 = COPY [[COPY12]].sub0 + ; GFX10-NEXT: [[COPY15:%[0-9]+]]:vgpr_32 = COPY killed [[COPY12]].sub1 + %0:sgpr_32 = COPY $sgpr0 + %1:sgpr_32 = COPY $sgpr1 + %2:sgpr_32 = COPY $sgpr2 + %3:sgpr_32 = COPY $sgpr3 + %5:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3 + %7:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %5:sgpr_128, 0, 4, 22, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) + %8:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %5:sgpr_128, 0, 8, 22, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) + %9:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %5:sgpr_128, 0, 16, 22, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) + %10:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %5:sgpr_128, 0, 20, 22, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) + %11:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %5:sgpr_128, 0, 24, 22, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) + %12:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %5:sgpr_128, 0, 28, 22, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) + %13:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %5:sgpr_128, 0, 36, 22, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) + %14:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %5:sgpr_128, 0, 40, 22, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) + %15:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %5:sgpr_128, 0, 44, 22, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) +... +--- + +name: gfx10_tbuffer_load_sint_32 +body: | + bb.0.entry: + ; GFX10-LABEL: name: gfx10_tbuffer_load_sint_32 + ; GFX10: [[COPY:%[0-9]+]]:sgpr_32 = COPY $sgpr0 + ; GFX10-NEXT: [[COPY1:%[0-9]+]]:sgpr_32 = COPY $sgpr1 + ; GFX10-NEXT: [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr2 + ; GFX10-NEXT: [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr3 + ; GFX10-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3 + ; GFX10-NEXT: [[TBUFFER_LOAD_FORMAT_XY_OFFSET:%[0-9]+]]:vreg_64 = TBUFFER_LOAD_FORMAT_XY_OFFSET [[REG_SEQUENCE]], 0, 4, 63, 0, 0, implicit $exec :: (dereferenceable load (s64), align 1, addrspace 4) + ; GFX10-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[TBUFFER_LOAD_FORMAT_XY_OFFSET]].sub0 + ; GFX10-NEXT: [[COPY5:%[0-9]+]]:vgpr_32 = COPY killed [[TBUFFER_LOAD_FORMAT_XY_OFFSET]].sub1 + ; GFX10-NEXT: [[TBUFFER_LOAD_FORMAT_XYZW_OFFSET:%[0-9]+]]:vreg_128 = TBUFFER_LOAD_FORMAT_XYZW_OFFSET [[REG_SEQUENCE]], 0, 16, 76, 0, 0, implicit $exec :: (dereferenceable load (s128), align 1, addrspace 4) + ; GFX10-NEXT: [[COPY6:%[0-9]+]]:vreg_96 = COPY [[TBUFFER_LOAD_FORMAT_XYZW_OFFSET]].sub0_sub1_sub2 + ; GFX10-NEXT: [[COPY7:%[0-9]+]]:vgpr_32 = COPY killed [[TBUFFER_LOAD_FORMAT_XYZW_OFFSET]].sub3 + ; GFX10-NEXT: [[COPY8:%[0-9]+]]:vreg_64 = COPY [[COPY6]].sub0_sub1 + ; GFX10-NEXT: [[COPY9:%[0-9]+]]:vgpr_32 = COPY killed [[COPY6]].sub2 + ; GFX10-NEXT: [[COPY10:%[0-9]+]]:vgpr_32 = COPY [[COPY8]].sub0 + ; GFX10-NEXT: [[COPY11:%[0-9]+]]:vgpr_32 = COPY killed [[COPY8]].sub1 + ; GFX10-NEXT: [[TBUFFER_LOAD_FORMAT_XYZ_OFFSET:%[0-9]+]]:vreg_96 = TBUFFER_LOAD_FORMAT_XYZ_OFFSET [[REG_SEQUENCE]], 0, 36, 73, 0, 0, implicit $exec :: (dereferenceable load (s96), align 1, addrspace 4) + ; GFX10-NEXT: [[COPY12:%[0-9]+]]:vreg_64 = COPY [[TBUFFER_LOAD_FORMAT_XYZ_OFFSET]].sub0_sub1 + ; GFX10-NEXT: [[COPY13:%[0-9]+]]:vgpr_32 = COPY killed [[TBUFFER_LOAD_FORMAT_XYZ_OFFSET]].sub2 + ; GFX10-NEXT: [[COPY14:%[0-9]+]]:vgpr_32 = COPY [[COPY12]].sub0 + ; GFX10-NEXT: [[COPY15:%[0-9]+]]:vgpr_32 = COPY killed [[COPY12]].sub1 + %0:sgpr_32 = COPY $sgpr0 + %1:sgpr_32 = COPY $sgpr1 + %2:sgpr_32 = COPY $sgpr2 + %3:sgpr_32 = COPY $sgpr3 + %5:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3 + %7:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %5:sgpr_128, 0, 4, 21, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) + %8:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %5:sgpr_128, 0, 8, 21, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) + %9:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %5:sgpr_128, 0, 16, 21, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) + %10:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %5:sgpr_128, 0, 20, 21, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) + %11:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %5:sgpr_128, 0, 24, 21, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) + %12:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %5:sgpr_128, 0, 28, 21, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) + %13:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %5:sgpr_128, 0, 36, 21, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) + %14:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %5:sgpr_128, 0, 40, 21, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) + %15:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %5:sgpr_128, 0, 44, 21, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) +... +--- + +name: gfx10_tbuffer_load_uint_32 +body: | + bb.0.entry: + ; GFX10-LABEL: name: gfx10_tbuffer_load_uint_32 + ; GFX10: [[COPY:%[0-9]+]]:sgpr_32 = COPY $sgpr0 + ; GFX10-NEXT: [[COPY1:%[0-9]+]]:sgpr_32 = COPY $sgpr1 + ; GFX10-NEXT: [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr2 + ; GFX10-NEXT: [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr3 + ; GFX10-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3 + ; GFX10-NEXT: [[TBUFFER_LOAD_FORMAT_XY_OFFSET:%[0-9]+]]:vreg_64 = TBUFFER_LOAD_FORMAT_XY_OFFSET [[REG_SEQUENCE]], 0, 4, 62, 0, 0, implicit $exec :: (dereferenceable load (s64), align 1, addrspace 4) + ; GFX10-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[TBUFFER_LOAD_FORMAT_XY_OFFSET]].sub0 + ; GFX10-NEXT: [[COPY5:%[0-9]+]]:vgpr_32 = COPY killed [[TBUFFER_LOAD_FORMAT_XY_OFFSET]].sub1 + ; GFX10-NEXT: [[TBUFFER_LOAD_FORMAT_XYZW_OFFSET:%[0-9]+]]:vreg_128 = TBUFFER_LOAD_FORMAT_XYZW_OFFSET [[REG_SEQUENCE]], 0, 16, 75, 0, 0, implicit $exec :: (dereferenceable load (s128), align 1, addrspace 4) + ; GFX10-NEXT: [[COPY6:%[0-9]+]]:vreg_96 = COPY [[TBUFFER_LOAD_FORMAT_XYZW_OFFSET]].sub0_sub1_sub2 + ; GFX10-NEXT: [[COPY7:%[0-9]+]]:vgpr_32 = COPY killed [[TBUFFER_LOAD_FORMAT_XYZW_OFFSET]].sub3 + ; GFX10-NEXT: [[COPY8:%[0-9]+]]:vreg_64 = COPY [[COPY6]].sub0_sub1 + ; GFX10-NEXT: [[COPY9:%[0-9]+]]:vgpr_32 = COPY killed [[COPY6]].sub2 + ; GFX10-NEXT: [[COPY10:%[0-9]+]]:vgpr_32 = COPY [[COPY8]].sub0 + ; GFX10-NEXT: [[COPY11:%[0-9]+]]:vgpr_32 = COPY killed [[COPY8]].sub1 + ; GFX10-NEXT: [[TBUFFER_LOAD_FORMAT_XYZ_OFFSET:%[0-9]+]]:vreg_96 = TBUFFER_LOAD_FORMAT_XYZ_OFFSET [[REG_SEQUENCE]], 0, 36, 72, 0, 0, implicit $exec :: (dereferenceable load (s96), align 1, addrspace 4) + ; GFX10-NEXT: [[COPY12:%[0-9]+]]:vreg_64 = COPY [[TBUFFER_LOAD_FORMAT_XYZ_OFFSET]].sub0_sub1 + ; GFX10-NEXT: [[COPY13:%[0-9]+]]:vgpr_32 = COPY killed [[TBUFFER_LOAD_FORMAT_XYZ_OFFSET]].sub2 + ; GFX10-NEXT: [[COPY14:%[0-9]+]]:vgpr_32 = COPY [[COPY12]].sub0 + ; GFX10-NEXT: [[COPY15:%[0-9]+]]:vgpr_32 = COPY killed [[COPY12]].sub1 + %0:sgpr_32 = COPY $sgpr0 + %1:sgpr_32 = COPY $sgpr1 + %2:sgpr_32 = COPY $sgpr2 + %3:sgpr_32 = COPY $sgpr3 + %5:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3 + %7:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %5:sgpr_128, 0, 4, 20, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) + %8:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %5:sgpr_128, 0, 8, 20, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) + %9:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %5:sgpr_128, 0, 16, 20, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) + %10:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %5:sgpr_128, 0, 20, 20, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) + %11:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %5:sgpr_128, 0, 24, 20, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) + %12:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %5:sgpr_128, 0, 28, 20, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) + %13:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %5:sgpr_128, 0, 36, 20, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) + %14:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %5:sgpr_128, 0, 40, 20, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) + %15:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %5:sgpr_128, 0, 44, 20, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) +... +--- + +name: gfx10_tbuffer_load_not_merged_data_format_mismatch +body: | + bb.0.entry: + ; GFX10-LABEL: name: gfx10_tbuffer_load_not_merged_data_format_mismatch + ; GFX10: [[COPY:%[0-9]+]]:sgpr_32 = COPY $sgpr0 + ; GFX10-NEXT: [[COPY1:%[0-9]+]]:sgpr_32 = COPY $sgpr1 + ; GFX10-NEXT: [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr2 + ; GFX10-NEXT: [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr3 + ; GFX10-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3 + ; GFX10-NEXT: [[TBUFFER_LOAD_FORMAT_X_OFFSET:%[0-9]+]]:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET [[REG_SEQUENCE]], 0, 4, 22, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) + ; GFX10-NEXT: [[TBUFFER_LOAD_FORMAT_X_OFFSET1:%[0-9]+]]:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET [[REG_SEQUENCE]], 0, 8, 13, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) + ; GFX10-NEXT: [[TBUFFER_LOAD_FORMAT_X_OFFSET2:%[0-9]+]]:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET [[REG_SEQUENCE]], 0, 16, 22, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) + ; GFX10-NEXT: [[TBUFFER_LOAD_FORMAT_X_OFFSET3:%[0-9]+]]:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET [[REG_SEQUENCE]], 0, 20, 13, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) + ; GFX10-NEXT: [[TBUFFER_LOAD_FORMAT_X_OFFSET4:%[0-9]+]]:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET [[REG_SEQUENCE]], 0, 24, 22, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) + ; GFX10-NEXT: [[TBUFFER_LOAD_FORMAT_X_OFFSET5:%[0-9]+]]:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET [[REG_SEQUENCE]], 0, 28, 13, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) + ; GFX10-NEXT: [[TBUFFER_LOAD_FORMAT_X_OFFSET6:%[0-9]+]]:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET [[REG_SEQUENCE]], 0, 36, 22, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) + ; GFX10-NEXT: [[TBUFFER_LOAD_FORMAT_X_OFFSET7:%[0-9]+]]:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET [[REG_SEQUENCE]], 0, 40, 13, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) + ; GFX10-NEXT: [[TBUFFER_LOAD_FORMAT_X_OFFSET8:%[0-9]+]]:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET [[REG_SEQUENCE]], 0, 44, 22, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) + %0:sgpr_32 = COPY $sgpr0 + %1:sgpr_32 = COPY $sgpr1 + %2:sgpr_32 = COPY $sgpr2 + %3:sgpr_32 = COPY $sgpr3 + %5:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3 + %7:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %5:sgpr_128, 0, 4, 22, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) + %8:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %5:sgpr_128, 0, 8, 13, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) + %9:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %5:sgpr_128, 0, 16, 22, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) + %10:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %5:sgpr_128, 0, 20, 13, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) + %11:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %5:sgpr_128, 0, 24, 22, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) + %12:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %5:sgpr_128, 0, 28, 13, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) + %13:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %5:sgpr_128, 0, 36, 22, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) + %14:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %5:sgpr_128, 0, 40, 13, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) + %15:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %5:sgpr_128, 0, 44, 22, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) +... +--- + +name: gfx10_tbuffer_load_not_merged_num_format_mismatch +body: | + bb.0.entry: + ; GFX10-LABEL: name: gfx10_tbuffer_load_not_merged_num_format_mismatch + ; GFX10: [[COPY:%[0-9]+]]:sgpr_32 = COPY $sgpr0 + ; GFX10-NEXT: [[COPY1:%[0-9]+]]:sgpr_32 = COPY $sgpr1 + ; GFX10-NEXT: [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr2 + ; GFX10-NEXT: [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr3 + ; GFX10-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3 + ; GFX10-NEXT: [[TBUFFER_LOAD_FORMAT_X_OFFSET:%[0-9]+]]:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET [[REG_SEQUENCE]], 0, 4, 22, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) + ; GFX10-NEXT: [[TBUFFER_LOAD_FORMAT_X_OFFSET1:%[0-9]+]]:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET [[REG_SEQUENCE]], 0, 8, 21, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) + ; GFX10-NEXT: [[TBUFFER_LOAD_FORMAT_X_OFFSET2:%[0-9]+]]:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET [[REG_SEQUENCE]], 0, 16, 22, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) + ; GFX10-NEXT: [[TBUFFER_LOAD_FORMAT_X_OFFSET3:%[0-9]+]]:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET [[REG_SEQUENCE]], 0, 20, 21, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) + ; GFX10-NEXT: [[TBUFFER_LOAD_FORMAT_X_OFFSET4:%[0-9]+]]:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET [[REG_SEQUENCE]], 0, 24, 22, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) + ; GFX10-NEXT: [[TBUFFER_LOAD_FORMAT_X_OFFSET5:%[0-9]+]]:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET [[REG_SEQUENCE]], 0, 28, 21, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) + ; GFX10-NEXT: [[TBUFFER_LOAD_FORMAT_X_OFFSET6:%[0-9]+]]:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET [[REG_SEQUENCE]], 0, 36, 22, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) + ; GFX10-NEXT: [[TBUFFER_LOAD_FORMAT_X_OFFSET7:%[0-9]+]]:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET [[REG_SEQUENCE]], 0, 40, 21, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) + ; GFX10-NEXT: [[TBUFFER_LOAD_FORMAT_X_OFFSET8:%[0-9]+]]:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET [[REG_SEQUENCE]], 0, 44, 22, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) + %0:sgpr_32 = COPY $sgpr0 + %1:sgpr_32 = COPY $sgpr1 + %2:sgpr_32 = COPY $sgpr2 + %3:sgpr_32 = COPY $sgpr3 + %5:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3 + %7:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %5:sgpr_128, 0, 4, 22, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) + %8:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %5:sgpr_128, 0, 8, 21, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) + %9:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %5:sgpr_128, 0, 16, 22, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) + %10:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %5:sgpr_128, 0, 20, 21, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) + %11:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %5:sgpr_128, 0, 24, 22, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) + %12:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %5:sgpr_128, 0, 28, 21, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) + %13:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %5:sgpr_128, 0, 36, 22, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) + %14:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %5:sgpr_128, 0, 40, 21, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) + %15:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %5:sgpr_128, 0, 44, 22, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) +... +--- + +name: gfx10_tbuffer_store_x_xyz +body: | + bb.0.entry: + liveins: $sgpr0, $sgpr1, $sgpr2, $sgpr3, $vgpr0, $vgpr1, $vgpr2, $vgpr3 + ; GFX10-LABEL: name: gfx10_tbuffer_store_x_xyz + ; GFX10: liveins: $sgpr0, $sgpr1, $sgpr2, $sgpr3, $vgpr0, $vgpr1, $vgpr2, $vgpr3 + ; GFX10-NEXT: {{ $}} + ; GFX10-NEXT: [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr3 + ; GFX10-NEXT: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr2 + ; GFX10-NEXT: [[COPY2:%[0-9]+]]:vgpr_32 = COPY $vgpr1 + ; GFX10-NEXT: [[COPY3:%[0-9]+]]:vgpr_32 = COPY $vgpr0 + ; GFX10-NEXT: [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr3 + ; GFX10-NEXT: [[COPY5:%[0-9]+]]:sgpr_32 = COPY $sgpr2 + ; GFX10-NEXT: [[COPY6:%[0-9]+]]:sgpr_32 = COPY $sgpr1 + ; GFX10-NEXT: [[COPY7:%[0-9]+]]:sgpr_32 = COPY $sgpr0 + ; GFX10-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY7]], %subreg.sub0, [[COPY6]], %subreg.sub1, [[COPY5]], %subreg.sub2, [[COPY4]], %subreg.sub3 + ; GFX10-NEXT: [[REG_SEQUENCE1:%[0-9]+]]:vreg_96 = REG_SEQUENCE [[COPY3]], %subreg.sub0, [[COPY2]], %subreg.sub1, [[COPY1]], %subreg.sub2 + ; GFX10-NEXT: [[REG_SEQUENCE2:%[0-9]+]]:vreg_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[REG_SEQUENCE1]], %subreg.sub1_sub2_sub3 + ; GFX10-NEXT: TBUFFER_STORE_FORMAT_XYZW_OFFSET_exact killed [[REG_SEQUENCE2]], [[REG_SEQUENCE]], 0, 4, 77, 0, 0, implicit $exec :: (dereferenceable store (s128), align 1, addrspace 4) + %7:vgpr_32 = COPY $vgpr3 + %6:vgpr_32 = COPY $vgpr2 + %5:vgpr_32 = COPY $vgpr1 + %4:vgpr_32 = COPY $vgpr0 + %3:sgpr_32 = COPY $sgpr3 + %2:sgpr_32 = COPY $sgpr2 + %1:sgpr_32 = COPY $sgpr1 + %0:sgpr_32 = COPY $sgpr0 + %13:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3 + %14:vreg_96 = REG_SEQUENCE %4:vgpr_32, %subreg.sub0, %5:vgpr_32, %subreg.sub1, %6:vgpr_32, %subreg.sub2 + TBUFFER_STORE_FORMAT_X_OFFSET_exact %7:vgpr_32, %13:sgpr_128, 0, 4, 22, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4) + TBUFFER_STORE_FORMAT_XYZ_OFFSET_exact %14:vreg_96, %13:sgpr_128, 0, 8, 74, 0, 0, implicit $exec :: (dereferenceable store (s96), align 1, addrspace 4) +... +--- + +name: gfx10_tbuffer_store_xyz_x +body: | + bb.0.entry: + liveins: $sgpr0, $sgpr1, $sgpr2, $sgpr3, $vgpr0, $vgpr1, $vgpr2, $vgpr3 + ; GFX10-LABEL: name: gfx10_tbuffer_store_xyz_x + ; GFX10: liveins: $sgpr0, $sgpr1, $sgpr2, $sgpr3, $vgpr0, $vgpr1, $vgpr2, $vgpr3 + ; GFX10-NEXT: {{ $}} + ; GFX10-NEXT: [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr3 + ; GFX10-NEXT: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr2 + ; GFX10-NEXT: [[COPY2:%[0-9]+]]:vgpr_32 = COPY $vgpr1 + ; GFX10-NEXT: [[COPY3:%[0-9]+]]:vgpr_32 = COPY $vgpr0 + ; GFX10-NEXT: [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr3 + ; GFX10-NEXT: [[COPY5:%[0-9]+]]:sgpr_32 = COPY $sgpr2 + ; GFX10-NEXT: [[COPY6:%[0-9]+]]:sgpr_32 = COPY $sgpr1 + ; GFX10-NEXT: [[COPY7:%[0-9]+]]:sgpr_32 = COPY $sgpr0 + ; GFX10-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY7]], %subreg.sub0, [[COPY6]], %subreg.sub1, [[COPY5]], %subreg.sub2, [[COPY4]], %subreg.sub3 + ; GFX10-NEXT: [[REG_SEQUENCE1:%[0-9]+]]:vreg_96 = REG_SEQUENCE [[COPY3]], %subreg.sub0, [[COPY2]], %subreg.sub1, [[COPY1]], %subreg.sub2 + ; GFX10-NEXT: [[REG_SEQUENCE2:%[0-9]+]]:vreg_128 = REG_SEQUENCE [[REG_SEQUENCE1]], %subreg.sub0_sub1_sub2, [[COPY]], %subreg.sub3 + ; GFX10-NEXT: TBUFFER_STORE_FORMAT_XYZW_OFFSET_exact killed [[REG_SEQUENCE2]], [[REG_SEQUENCE]], 0, 4, 77, 0, 0, implicit $exec :: (dereferenceable store (s128), align 1, addrspace 4) + %7:vgpr_32 = COPY $vgpr3 + %6:vgpr_32 = COPY $vgpr2 + %5:vgpr_32 = COPY $vgpr1 + %4:vgpr_32 = COPY $vgpr0 + %3:sgpr_32 = COPY $sgpr3 + %2:sgpr_32 = COPY $sgpr2 + %1:sgpr_32 = COPY $sgpr1 + %0:sgpr_32 = COPY $sgpr0 + %13:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3 + %14:vreg_96 = REG_SEQUENCE %4:vgpr_32, %subreg.sub0, %5:vgpr_32, %subreg.sub1, %6:vgpr_32, %subreg.sub2 + TBUFFER_STORE_FORMAT_XYZ_OFFSET_exact %14:vreg_96, %13:sgpr_128, 0, 4, 74, 0, 0, implicit $exec :: (dereferenceable store (s96), align 1, addrspace 4) + TBUFFER_STORE_FORMAT_X_OFFSET_exact %7:vgpr_32, %13:sgpr_128, 0, 16, 22, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4) +... +--- + +name: gfx10_tbuffer_store_xy_xy +body: | + bb.0.entry: + liveins: $sgpr0, $sgpr1, $sgpr2, $sgpr3, $vgpr0, $vgpr1, $vgpr2, $vgpr3 + ; GFX10-LABEL: name: gfx10_tbuffer_store_xy_xy + ; GFX10: liveins: $sgpr0, $sgpr1, $sgpr2, $sgpr3, $vgpr0, $vgpr1, $vgpr2, $vgpr3 + ; GFX10-NEXT: {{ $}} + ; GFX10-NEXT: [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr3 + ; GFX10-NEXT: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr2 + ; GFX10-NEXT: [[COPY2:%[0-9]+]]:vgpr_32 = COPY $vgpr1 + ; GFX10-NEXT: [[COPY3:%[0-9]+]]:vgpr_32 = COPY $vgpr0 + ; GFX10-NEXT: [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr3 + ; GFX10-NEXT: [[COPY5:%[0-9]+]]:sgpr_32 = COPY $sgpr2 + ; GFX10-NEXT: [[COPY6:%[0-9]+]]:sgpr_32 = COPY $sgpr1 + ; GFX10-NEXT: [[COPY7:%[0-9]+]]:sgpr_32 = COPY $sgpr0 + ; GFX10-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY7]], %subreg.sub0, [[COPY6]], %subreg.sub1, [[COPY5]], %subreg.sub2, [[COPY4]], %subreg.sub3 + ; GFX10-NEXT: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[COPY3]], %subreg.sub0, [[COPY2]], %subreg.sub1 + ; GFX10-NEXT: [[REG_SEQUENCE2:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[COPY1]], %subreg.sub0, [[COPY]], %subreg.sub1 + ; GFX10-NEXT: [[REG_SEQUENCE3:%[0-9]+]]:vreg_128 = REG_SEQUENCE [[REG_SEQUENCE1]], %subreg.sub0_sub1, [[REG_SEQUENCE2]], %subreg.sub2_sub3 + ; GFX10-NEXT: TBUFFER_STORE_FORMAT_XYZW_OFFSET_exact killed [[REG_SEQUENCE3]], [[REG_SEQUENCE]], 0, 4, 77, 0, 0, implicit $exec :: (dereferenceable store (s128), align 1, addrspace 4) + %7:vgpr_32 = COPY $vgpr3 + %6:vgpr_32 = COPY $vgpr2 + %5:vgpr_32 = COPY $vgpr1 + %4:vgpr_32 = COPY $vgpr0 + %3:sgpr_32 = COPY $sgpr3 + %2:sgpr_32 = COPY $sgpr2 + %1:sgpr_32 = COPY $sgpr1 + %0:sgpr_32 = COPY $sgpr0 + %13:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3 + %14:vreg_64 = REG_SEQUENCE %4:vgpr_32, %subreg.sub0, %5:vgpr_32, %subreg.sub1 + %15:vreg_64 = REG_SEQUENCE %6:vgpr_32, %subreg.sub0, %7:vgpr_32, %subreg.sub1 + TBUFFER_STORE_FORMAT_XY_OFFSET_exact %14:vreg_64, %13:sgpr_128, 0, 4, 64, 0, 0, implicit $exec :: (dereferenceable store (s64), align 1, addrspace 4) + TBUFFER_STORE_FORMAT_XY_OFFSET_exact %15:vreg_64, %13:sgpr_128, 0, 12, 64, 0, 0, implicit $exec :: (dereferenceable store (s64), align 1, addrspace 4) +... +--- + +name: gfx10_tbuffer_store_x_xy +body: | + bb.0.entry: + liveins: $sgpr0, $sgpr1, $sgpr2, $sgpr3, $vgpr0, $vgpr1, $vgpr2, $vgpr3 + ; GFX10-LABEL: name: gfx10_tbuffer_store_x_xy + ; GFX10: liveins: $sgpr0, $sgpr1, $sgpr2, $sgpr3, $vgpr0, $vgpr1, $vgpr2, $vgpr3 + ; GFX10-NEXT: {{ $}} + ; GFX10-NEXT: [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr3 + ; GFX10-NEXT: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr2 + ; GFX10-NEXT: [[COPY2:%[0-9]+]]:vgpr_32 = COPY $vgpr1 + ; GFX10-NEXT: [[COPY3:%[0-9]+]]:vgpr_32 = COPY $vgpr0 + ; GFX10-NEXT: [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr3 + ; GFX10-NEXT: [[COPY5:%[0-9]+]]:sgpr_32 = COPY $sgpr2 + ; GFX10-NEXT: [[COPY6:%[0-9]+]]:sgpr_32 = COPY $sgpr1 + ; GFX10-NEXT: [[COPY7:%[0-9]+]]:sgpr_32 = COPY $sgpr0 + ; GFX10-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY7]], %subreg.sub0, [[COPY6]], %subreg.sub1, [[COPY5]], %subreg.sub2, [[COPY4]], %subreg.sub3 + ; GFX10-NEXT: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[COPY3]], %subreg.sub0, [[COPY2]], %subreg.sub1 + ; GFX10-NEXT: [[REG_SEQUENCE2:%[0-9]+]]:vreg_96 = REG_SEQUENCE [[COPY]], %subreg.sub0, %10:vreg_64, %subreg.sub1_sub2 + ; GFX10-NEXT: TBUFFER_STORE_FORMAT_XYZ_OFFSET_exact killed [[REG_SEQUENCE2]], [[REG_SEQUENCE]], 0, 4, 74, 0, 0, implicit $exec :: (dereferenceable store (s96), align 1, addrspace 4) + %7:vgpr_32 = COPY $vgpr3 + %6:vgpr_32 = COPY $vgpr2 + %5:vgpr_32 = COPY $vgpr1 + %4:vgpr_32 = COPY $vgpr0 + %3:sgpr_32 = COPY $sgpr3 + %2:sgpr_32 = COPY $sgpr2 + %1:sgpr_32 = COPY $sgpr1 + %0:sgpr_32 = COPY $sgpr0 + %13:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3 + %14:vreg_64 = REG_SEQUENCE %4:vgpr_32, %subreg.sub0, %5:vgpr_32, %subreg.sub1 + TBUFFER_STORE_FORMAT_X_OFFSET_exact %7:vgpr_32, %13:sgpr_128, 0, 4, 22, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4) + TBUFFER_STORE_FORMAT_XY_OFFSET_exact %15:vreg_64, %13:sgpr_128, 0, 8, 64, 0, 0, implicit $exec :: (dereferenceable store (s64), align 1, addrspace 4) +... +--- + +name: gfx10_tbuffer_store_xy_x +body: | + bb.0.entry: + liveins: $sgpr0, $sgpr1, $sgpr2, $sgpr3, $vgpr0, $vgpr1, $vgpr2, $vgpr3 + liveins: $sgpr0, $sgpr1, $sgpr2, $sgpr3, $vgpr0, $vgpr1, $vgpr2, $vgpr3 + ; GFX10-LABEL: name: gfx10_tbuffer_store_xy_x + ; GFX10: liveins: $sgpr0, $sgpr1, $sgpr2, $sgpr3, $vgpr0, $vgpr1, $vgpr2, $vgpr3, $sgpr0, $sgpr1, $sgpr2, $sgpr3, $vgpr0, $vgpr1, $vgpr2, $vgpr3 + ; GFX10-NEXT: {{ $}} + ; GFX10-NEXT: [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr3 + ; GFX10-NEXT: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr2 + ; GFX10-NEXT: [[COPY2:%[0-9]+]]:vgpr_32 = COPY $vgpr1 + ; GFX10-NEXT: [[COPY3:%[0-9]+]]:vgpr_32 = COPY $vgpr0 + ; GFX10-NEXT: [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr3 + ; GFX10-NEXT: [[COPY5:%[0-9]+]]:sgpr_32 = COPY $sgpr2 + ; GFX10-NEXT: [[COPY6:%[0-9]+]]:sgpr_32 = COPY $sgpr1 + ; GFX10-NEXT: [[COPY7:%[0-9]+]]:sgpr_32 = COPY $sgpr0 + ; GFX10-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY7]], %subreg.sub0, [[COPY6]], %subreg.sub1, [[COPY5]], %subreg.sub2, [[COPY4]], %subreg.sub3 + ; GFX10-NEXT: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[COPY3]], %subreg.sub0, [[COPY2]], %subreg.sub1 + ; GFX10-NEXT: [[REG_SEQUENCE2:%[0-9]+]]:vreg_96 = REG_SEQUENCE [[REG_SEQUENCE1]], %subreg.sub0_sub1, [[COPY]], %subreg.sub2 + ; GFX10-NEXT: TBUFFER_STORE_FORMAT_XYZ_OFFSET_exact killed [[REG_SEQUENCE2]], [[REG_SEQUENCE]], 0, 4, 74, 0, 0, implicit $exec :: (dereferenceable store (s96), align 1, addrspace 4) + %7:vgpr_32 = COPY $vgpr3 + %6:vgpr_32 = COPY $vgpr2 + %5:vgpr_32 = COPY $vgpr1 + %4:vgpr_32 = COPY $vgpr0 + %3:sgpr_32 = COPY $sgpr3 + %2:sgpr_32 = COPY $sgpr2 + %1:sgpr_32 = COPY $sgpr1 + %0:sgpr_32 = COPY $sgpr0 + %13:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3 + %14:vreg_64 = REG_SEQUENCE %4:vgpr_32, %subreg.sub0, %5:vgpr_32, %subreg.sub1 + TBUFFER_STORE_FORMAT_XY_OFFSET_exact %14:vreg_64, %13:sgpr_128, 0, 4, 64, 0, 0, implicit $exec :: (dereferenceable store (s64), align 1, addrspace 4) + TBUFFER_STORE_FORMAT_X_OFFSET_exact %7:vgpr_32, %13:sgpr_128, 0, 12, 22, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4) +... +--- + +name: gfx10_tbuffer_store_x_x +body: | + bb.0.entry: + liveins: $sgpr0, $sgpr1, $sgpr2, $sgpr3, $vgpr0, $vgpr1, $vgpr2, $vgpr3 + ; GFX10-LABEL: name: gfx10_tbuffer_store_x_x + ; GFX10: liveins: $sgpr0, $sgpr1, $sgpr2, $sgpr3, $vgpr0, $vgpr1, $vgpr2, $vgpr3 + ; GFX10-NEXT: {{ $}} + ; GFX10-NEXT: [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr3 + ; GFX10-NEXT: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr2 + ; GFX10-NEXT: [[COPY2:%[0-9]+]]:vgpr_32 = COPY $vgpr1 + ; GFX10-NEXT: [[COPY3:%[0-9]+]]:vgpr_32 = COPY $vgpr0 + ; GFX10-NEXT: [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr3 + ; GFX10-NEXT: [[COPY5:%[0-9]+]]:sgpr_32 = COPY $sgpr2 + ; GFX10-NEXT: [[COPY6:%[0-9]+]]:sgpr_32 = COPY $sgpr1 + ; GFX10-NEXT: [[COPY7:%[0-9]+]]:sgpr_32 = COPY $sgpr0 + ; GFX10-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY7]], %subreg.sub0, [[COPY6]], %subreg.sub1, [[COPY5]], %subreg.sub2, [[COPY4]], %subreg.sub3 + ; GFX10-NEXT: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[COPY1]], %subreg.sub0, [[COPY]], %subreg.sub1 + ; GFX10-NEXT: TBUFFER_STORE_FORMAT_XY_OFFSET_exact killed [[REG_SEQUENCE1]], [[REG_SEQUENCE]], 0, 4, 64, 0, 0, implicit $exec :: (dereferenceable store (s64), align 1, addrspace 4) + %7:vgpr_32 = COPY $vgpr3 + %6:vgpr_32 = COPY $vgpr2 + %5:vgpr_32 = COPY $vgpr1 + %4:vgpr_32 = COPY $vgpr0 + %3:sgpr_32 = COPY $sgpr3 + %2:sgpr_32 = COPY $sgpr2 + %1:sgpr_32 = COPY $sgpr1 + %0:sgpr_32 = COPY $sgpr0 + %13:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3 + TBUFFER_STORE_FORMAT_X_OFFSET_exact %6:vgpr_32, %13:sgpr_128, 0, 4, 22, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4) + TBUFFER_STORE_FORMAT_X_OFFSET_exact %7:vgpr_32, %13:sgpr_128, 0, 8, 22, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4) +... +--- + +name: gfx10_tbuffer_store_x_x_format_32_32_32_32 +body: | + bb.0.entry: + liveins: $sgpr0, $sgpr1, $sgpr2, $sgpr3, $vgpr0, $vgpr1, $vgpr2, $vgpr3 + ; GFX10-LABEL: name: gfx10_tbuffer_store_x_x_format_32_32_32_32 + ; GFX10: liveins: $sgpr0, $sgpr1, $sgpr2, $sgpr3, $vgpr0, $vgpr1, $vgpr2, $vgpr3 + ; GFX10-NEXT: {{ $}} + ; GFX10-NEXT: [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr3 + ; GFX10-NEXT: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr2 + ; GFX10-NEXT: [[COPY2:%[0-9]+]]:vgpr_32 = COPY $vgpr1 + ; GFX10-NEXT: [[COPY3:%[0-9]+]]:vgpr_32 = COPY $vgpr0 + ; GFX10-NEXT: [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr3 + ; GFX10-NEXT: [[COPY5:%[0-9]+]]:sgpr_32 = COPY $sgpr2 + ; GFX10-NEXT: [[COPY6:%[0-9]+]]:sgpr_32 = COPY $sgpr1 + ; GFX10-NEXT: [[COPY7:%[0-9]+]]:sgpr_32 = COPY $sgpr0 + ; GFX10-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY7]], %subreg.sub0, [[COPY6]], %subreg.sub1, [[COPY5]], %subreg.sub2, [[COPY4]], %subreg.sub3 + ; GFX10-NEXT: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[COPY1]], %subreg.sub0, [[COPY]], %subreg.sub1 + ; GFX10-NEXT: TBUFFER_STORE_FORMAT_XY_OFFSET_exact killed [[REG_SEQUENCE1]], [[REG_SEQUENCE]], 0, 4, 64, 0, 0, implicit $exec :: (dereferenceable store (s64), align 1, addrspace 4) + %7:vgpr_32 = COPY $vgpr3 + %6:vgpr_32 = COPY $vgpr2 + %5:vgpr_32 = COPY $vgpr1 + %4:vgpr_32 = COPY $vgpr0 + %3:sgpr_32 = COPY $sgpr3 + %2:sgpr_32 = COPY $sgpr2 + %1:sgpr_32 = COPY $sgpr1 + %0:sgpr_32 = COPY $sgpr0 + %13:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3 + TBUFFER_STORE_FORMAT_X_OFFSET_exact %6:vgpr_32, %13:sgpr_128, 0, 4, 77, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4) + TBUFFER_STORE_FORMAT_X_OFFSET_exact %7:vgpr_32, %13:sgpr_128, 0, 8, 77, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4) +... +--- + +name: gfx10_tbuffer_store_float32 +body: | + bb.0.entry: + liveins: $sgpr0, $sgpr1, $sgpr2, $sgpr3, $vgpr0, $vgpr1, $vgpr2, $vgpr3, $vgpr4, $vgpr5, $vgpr6, $vgpr7, $vgpr8 + ; GFX10-LABEL: name: gfx10_tbuffer_store_float32 + ; GFX10: liveins: $sgpr0, $sgpr1, $sgpr2, $sgpr3, $vgpr0, $vgpr1, $vgpr2, $vgpr3, $vgpr4, $vgpr5, $vgpr6, $vgpr7, $vgpr8 + ; GFX10-NEXT: {{ $}} + ; GFX10-NEXT: [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr8 + ; GFX10-NEXT: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr7 + ; GFX10-NEXT: [[COPY2:%[0-9]+]]:vgpr_32 = COPY $vgpr6 + ; GFX10-NEXT: [[COPY3:%[0-9]+]]:vgpr_32 = COPY $vgpr5 + ; GFX10-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY $vgpr4 + ; GFX10-NEXT: [[COPY5:%[0-9]+]]:vgpr_32 = COPY $vgpr3 + ; GFX10-NEXT: [[COPY6:%[0-9]+]]:vgpr_32 = COPY $vgpr2 + ; GFX10-NEXT: [[COPY7:%[0-9]+]]:vgpr_32 = COPY $vgpr1 + ; GFX10-NEXT: [[COPY8:%[0-9]+]]:vgpr_32 = COPY $vgpr0 + ; GFX10-NEXT: [[COPY9:%[0-9]+]]:sgpr_32 = COPY $sgpr3 + ; GFX10-NEXT: [[COPY10:%[0-9]+]]:sgpr_32 = COPY $sgpr2 + ; GFX10-NEXT: [[COPY11:%[0-9]+]]:sgpr_32 = COPY $sgpr1 + ; GFX10-NEXT: [[COPY12:%[0-9]+]]:sgpr_32 = COPY $sgpr0 + ; GFX10-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY12]], %subreg.sub0, [[COPY11]], %subreg.sub1, [[COPY10]], %subreg.sub2, [[COPY9]], %subreg.sub3 + ; GFX10-NEXT: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[COPY8]], %subreg.sub0, [[COPY7]], %subreg.sub1 + ; GFX10-NEXT: TBUFFER_STORE_FORMAT_XY_OFFSET_exact killed [[REG_SEQUENCE1]], [[REG_SEQUENCE]], 0, 4, 64, 0, 0, implicit $exec :: (dereferenceable store (s64), align 1, addrspace 4) + ; GFX10-NEXT: [[REG_SEQUENCE2:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[COPY6]], %subreg.sub0, [[COPY5]], %subreg.sub1 + ; GFX10-NEXT: [[REG_SEQUENCE3:%[0-9]+]]:vreg_96 = REG_SEQUENCE killed [[REG_SEQUENCE2]], %subreg.sub0_sub1, [[COPY4]], %subreg.sub2 + ; GFX10-NEXT: [[REG_SEQUENCE4:%[0-9]+]]:vreg_128 = REG_SEQUENCE killed [[REG_SEQUENCE3]], %subreg.sub0_sub1_sub2, [[COPY3]], %subreg.sub3 + ; GFX10-NEXT: TBUFFER_STORE_FORMAT_XYZW_OFFSET_exact killed [[REG_SEQUENCE4]], [[REG_SEQUENCE]], 0, 16, 77, 0, 0, implicit $exec :: (dereferenceable store (s128), align 1, addrspace 4) + ; GFX10-NEXT: [[REG_SEQUENCE5:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[COPY2]], %subreg.sub0, [[COPY1]], %subreg.sub1 + ; GFX10-NEXT: [[REG_SEQUENCE6:%[0-9]+]]:vreg_96 = REG_SEQUENCE killed [[REG_SEQUENCE5]], %subreg.sub0_sub1, [[COPY]], %subreg.sub2 + ; GFX10-NEXT: TBUFFER_STORE_FORMAT_XYZ_OFFSET_exact killed [[REG_SEQUENCE6]], [[REG_SEQUENCE]], 0, 36, 74, 0, 0, implicit $exec :: (dereferenceable store (s96), align 1, addrspace 4) + %12:vgpr_32 = COPY $vgpr8 + %11:vgpr_32 = COPY $vgpr7 + %10:vgpr_32 = COPY $vgpr6 + %9:vgpr_32 = COPY $vgpr5 + %8:vgpr_32 = COPY $vgpr4 + %7:vgpr_32 = COPY $vgpr3 + %6:vgpr_32 = COPY $vgpr2 + %5:vgpr_32 = COPY $vgpr1 + %4:vgpr_32 = COPY $vgpr0 + %3:sgpr_32 = COPY $sgpr3 + %2:sgpr_32 = COPY $sgpr2 + %1:sgpr_32 = COPY $sgpr1 + %0:sgpr_32 = COPY $sgpr0 + %13:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3 + TBUFFER_STORE_FORMAT_X_OFFSET_exact %4:vgpr_32, %13:sgpr_128, 0, 4, 22, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4) + TBUFFER_STORE_FORMAT_X_OFFSET_exact %5:vgpr_32, %13:sgpr_128, 0, 8, 22, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4) + TBUFFER_STORE_FORMAT_X_OFFSET_exact %6:vgpr_32, %13:sgpr_128, 0, 16, 22, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4) + TBUFFER_STORE_FORMAT_X_OFFSET_exact %7:vgpr_32, %13:sgpr_128, 0, 20, 22, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4) + TBUFFER_STORE_FORMAT_X_OFFSET_exact %8:vgpr_32, %13:sgpr_128, 0, 24, 22, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4) + TBUFFER_STORE_FORMAT_X_OFFSET_exact %9:vgpr_32, %13:sgpr_128, 0, 28, 22, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4) + TBUFFER_STORE_FORMAT_X_OFFSET_exact %10:vgpr_32, %13:sgpr_128, 0, 36, 22, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4) + TBUFFER_STORE_FORMAT_X_OFFSET_exact %11:vgpr_32, %13:sgpr_128, 0, 40, 22, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4) + TBUFFER_STORE_FORMAT_X_OFFSET_exact %12:vgpr_32, %13:sgpr_128, 0, 44, 22, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4) +... +--- + +name: gfx10_tbuffer_store_sint32 +body: | + bb.0.entry: + liveins: $sgpr0, $sgpr1, $sgpr2, $sgpr3, $vgpr0, $vgpr1, $vgpr2, $vgpr3, $vgpr4, $vgpr5, $vgpr6, $vgpr7, $vgpr8 + ; GFX10-LABEL: name: gfx10_tbuffer_store_sint32 + ; GFX10: liveins: $sgpr0, $sgpr1, $sgpr2, $sgpr3, $vgpr0, $vgpr1, $vgpr2, $vgpr3, $vgpr4, $vgpr5, $vgpr6, $vgpr7, $vgpr8 + ; GFX10-NEXT: {{ $}} + ; GFX10-NEXT: [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr8 + ; GFX10-NEXT: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr7 + ; GFX10-NEXT: [[COPY2:%[0-9]+]]:vgpr_32 = COPY $vgpr6 + ; GFX10-NEXT: [[COPY3:%[0-9]+]]:vgpr_32 = COPY $vgpr5 + ; GFX10-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY $vgpr4 + ; GFX10-NEXT: [[COPY5:%[0-9]+]]:vgpr_32 = COPY $vgpr3 + ; GFX10-NEXT: [[COPY6:%[0-9]+]]:vgpr_32 = COPY $vgpr2 + ; GFX10-NEXT: [[COPY7:%[0-9]+]]:vgpr_32 = COPY $vgpr1 + ; GFX10-NEXT: [[COPY8:%[0-9]+]]:vgpr_32 = COPY $vgpr0 + ; GFX10-NEXT: [[COPY9:%[0-9]+]]:sgpr_32 = COPY $sgpr3 + ; GFX10-NEXT: [[COPY10:%[0-9]+]]:sgpr_32 = COPY $sgpr2 + ; GFX10-NEXT: [[COPY11:%[0-9]+]]:sgpr_32 = COPY $sgpr1 + ; GFX10-NEXT: [[COPY12:%[0-9]+]]:sgpr_32 = COPY $sgpr0 + ; GFX10-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY12]], %subreg.sub0, [[COPY11]], %subreg.sub1, [[COPY10]], %subreg.sub2, [[COPY9]], %subreg.sub3 + ; GFX10-NEXT: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[COPY8]], %subreg.sub0, [[COPY7]], %subreg.sub1 + ; GFX10-NEXT: TBUFFER_STORE_FORMAT_XY_OFFSET_exact killed [[REG_SEQUENCE1]], [[REG_SEQUENCE]], 0, 4, 63, 0, 0, implicit $exec :: (dereferenceable store (s64), align 1, addrspace 4) + ; GFX10-NEXT: [[REG_SEQUENCE2:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[COPY6]], %subreg.sub0, [[COPY5]], %subreg.sub1 + ; GFX10-NEXT: [[REG_SEQUENCE3:%[0-9]+]]:vreg_96 = REG_SEQUENCE killed [[REG_SEQUENCE2]], %subreg.sub0_sub1, [[COPY4]], %subreg.sub2 + ; GFX10-NEXT: [[REG_SEQUENCE4:%[0-9]+]]:vreg_128 = REG_SEQUENCE killed [[REG_SEQUENCE3]], %subreg.sub0_sub1_sub2, [[COPY3]], %subreg.sub3 + ; GFX10-NEXT: TBUFFER_STORE_FORMAT_XYZW_OFFSET_exact killed [[REG_SEQUENCE4]], [[REG_SEQUENCE]], 0, 16, 76, 0, 0, implicit $exec :: (dereferenceable store (s128), align 1, addrspace 4) + ; GFX10-NEXT: [[REG_SEQUENCE5:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[COPY2]], %subreg.sub0, [[COPY1]], %subreg.sub1 + ; GFX10-NEXT: [[REG_SEQUENCE6:%[0-9]+]]:vreg_96 = REG_SEQUENCE killed [[REG_SEQUENCE5]], %subreg.sub0_sub1, [[COPY]], %subreg.sub2 + ; GFX10-NEXT: TBUFFER_STORE_FORMAT_XYZ_OFFSET_exact killed [[REG_SEQUENCE6]], [[REG_SEQUENCE]], 0, 36, 73, 0, 0, implicit $exec :: (dereferenceable store (s96), align 1, addrspace 4) + %12:vgpr_32 = COPY $vgpr8 + %11:vgpr_32 = COPY $vgpr7 + %10:vgpr_32 = COPY $vgpr6 + %9:vgpr_32 = COPY $vgpr5 + %8:vgpr_32 = COPY $vgpr4 + %7:vgpr_32 = COPY $vgpr3 + %6:vgpr_32 = COPY $vgpr2 + %5:vgpr_32 = COPY $vgpr1 + %4:vgpr_32 = COPY $vgpr0 + %3:sgpr_32 = COPY $sgpr3 + %2:sgpr_32 = COPY $sgpr2 + %1:sgpr_32 = COPY $sgpr1 + %0:sgpr_32 = COPY $sgpr0 + %13:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3 + TBUFFER_STORE_FORMAT_X_OFFSET_exact %4:vgpr_32, %13:sgpr_128, 0, 4, 21, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4) + TBUFFER_STORE_FORMAT_X_OFFSET_exact %5:vgpr_32, %13:sgpr_128, 0, 8, 21, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4) + TBUFFER_STORE_FORMAT_X_OFFSET_exact %6:vgpr_32, %13:sgpr_128, 0, 16, 21, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4) + TBUFFER_STORE_FORMAT_X_OFFSET_exact %7:vgpr_32, %13:sgpr_128, 0, 20, 21, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4) + TBUFFER_STORE_FORMAT_X_OFFSET_exact %8:vgpr_32, %13:sgpr_128, 0, 24, 21, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4) + TBUFFER_STORE_FORMAT_X_OFFSET_exact %9:vgpr_32, %13:sgpr_128, 0, 28, 21, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4) + TBUFFER_STORE_FORMAT_X_OFFSET_exact %10:vgpr_32, %13:sgpr_128, 0, 36, 21, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4) + TBUFFER_STORE_FORMAT_X_OFFSET_exact %11:vgpr_32, %13:sgpr_128, 0, 40, 21, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4) + TBUFFER_STORE_FORMAT_X_OFFSET_exact %12:vgpr_32, %13:sgpr_128, 0, 44, 21, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4) +... +--- + +name: gfx10_tbuffer_store_uint32 +body: | + bb.0.entry: + liveins: $sgpr0, $sgpr1, $sgpr2, $sgpr3, $vgpr0, $vgpr1, $vgpr2, $vgpr3, $vgpr4, $vgpr5, $vgpr6, $vgpr7, $vgpr8 + ; GFX10-LABEL: name: gfx10_tbuffer_store_uint32 + ; GFX10: liveins: $sgpr0, $sgpr1, $sgpr2, $sgpr3, $vgpr0, $vgpr1, $vgpr2, $vgpr3, $vgpr4, $vgpr5, $vgpr6, $vgpr7, $vgpr8 + ; GFX10-NEXT: {{ $}} + ; GFX10-NEXT: [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr8 + ; GFX10-NEXT: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr7 + ; GFX10-NEXT: [[COPY2:%[0-9]+]]:vgpr_32 = COPY $vgpr6 + ; GFX10-NEXT: [[COPY3:%[0-9]+]]:vgpr_32 = COPY $vgpr5 + ; GFX10-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY $vgpr4 + ; GFX10-NEXT: [[COPY5:%[0-9]+]]:vgpr_32 = COPY $vgpr3 + ; GFX10-NEXT: [[COPY6:%[0-9]+]]:vgpr_32 = COPY $vgpr2 + ; GFX10-NEXT: [[COPY7:%[0-9]+]]:vgpr_32 = COPY $vgpr1 + ; GFX10-NEXT: [[COPY8:%[0-9]+]]:vgpr_32 = COPY $vgpr0 + ; GFX10-NEXT: [[COPY9:%[0-9]+]]:sgpr_32 = COPY $sgpr3 + ; GFX10-NEXT: [[COPY10:%[0-9]+]]:sgpr_32 = COPY $sgpr2 + ; GFX10-NEXT: [[COPY11:%[0-9]+]]:sgpr_32 = COPY $sgpr1 + ; GFX10-NEXT: [[COPY12:%[0-9]+]]:sgpr_32 = COPY $sgpr0 + ; GFX10-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY12]], %subreg.sub0, [[COPY11]], %subreg.sub1, [[COPY10]], %subreg.sub2, [[COPY9]], %subreg.sub3 + ; GFX10-NEXT: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[COPY8]], %subreg.sub0, [[COPY7]], %subreg.sub1 + ; GFX10-NEXT: TBUFFER_STORE_FORMAT_XY_OFFSET_exact killed [[REG_SEQUENCE1]], [[REG_SEQUENCE]], 0, 4, 62, 0, 0, implicit $exec :: (dereferenceable store (s64), align 1, addrspace 4) + ; GFX10-NEXT: [[REG_SEQUENCE2:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[COPY6]], %subreg.sub0, [[COPY5]], %subreg.sub1 + ; GFX10-NEXT: [[REG_SEQUENCE3:%[0-9]+]]:vreg_96 = REG_SEQUENCE killed [[REG_SEQUENCE2]], %subreg.sub0_sub1, [[COPY4]], %subreg.sub2 + ; GFX10-NEXT: [[REG_SEQUENCE4:%[0-9]+]]:vreg_128 = REG_SEQUENCE killed [[REG_SEQUENCE3]], %subreg.sub0_sub1_sub2, [[COPY3]], %subreg.sub3 + ; GFX10-NEXT: TBUFFER_STORE_FORMAT_XYZW_OFFSET_exact killed [[REG_SEQUENCE4]], [[REG_SEQUENCE]], 0, 16, 75, 0, 0, implicit $exec :: (dereferenceable store (s128), align 1, addrspace 4) + ; GFX10-NEXT: [[REG_SEQUENCE5:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[COPY2]], %subreg.sub0, [[COPY1]], %subreg.sub1 + ; GFX10-NEXT: [[REG_SEQUENCE6:%[0-9]+]]:vreg_96 = REG_SEQUENCE killed [[REG_SEQUENCE5]], %subreg.sub0_sub1, [[COPY]], %subreg.sub2 + ; GFX10-NEXT: TBUFFER_STORE_FORMAT_XYZ_OFFSET_exact killed [[REG_SEQUENCE6]], [[REG_SEQUENCE]], 0, 36, 72, 0, 0, implicit $exec :: (dereferenceable store (s96), align 1, addrspace 4) + %12:vgpr_32 = COPY $vgpr8 + %11:vgpr_32 = COPY $vgpr7 + %10:vgpr_32 = COPY $vgpr6 + %9:vgpr_32 = COPY $vgpr5 + %8:vgpr_32 = COPY $vgpr4 + %7:vgpr_32 = COPY $vgpr3 + %6:vgpr_32 = COPY $vgpr2 + %5:vgpr_32 = COPY $vgpr1 + %4:vgpr_32 = COPY $vgpr0 + %3:sgpr_32 = COPY $sgpr3 + %2:sgpr_32 = COPY $sgpr2 + %1:sgpr_32 = COPY $sgpr1 + %0:sgpr_32 = COPY $sgpr0 + %13:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3 + TBUFFER_STORE_FORMAT_X_OFFSET_exact %4:vgpr_32, %13:sgpr_128, 0, 4, 20, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4) + TBUFFER_STORE_FORMAT_X_OFFSET_exact %5:vgpr_32, %13:sgpr_128, 0, 8, 20, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4) + TBUFFER_STORE_FORMAT_X_OFFSET_exact %6:vgpr_32, %13:sgpr_128, 0, 16, 20, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4) + TBUFFER_STORE_FORMAT_X_OFFSET_exact %7:vgpr_32, %13:sgpr_128, 0, 20, 20, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4) + TBUFFER_STORE_FORMAT_X_OFFSET_exact %8:vgpr_32, %13:sgpr_128, 0, 24, 20, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4) + TBUFFER_STORE_FORMAT_X_OFFSET_exact %9:vgpr_32, %13:sgpr_128, 0, 28, 20, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4) + TBUFFER_STORE_FORMAT_X_OFFSET_exact %10:vgpr_32, %13:sgpr_128, 0, 36, 20, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4) + TBUFFER_STORE_FORMAT_X_OFFSET_exact %11:vgpr_32, %13:sgpr_128, 0, 40, 20, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4) + TBUFFER_STORE_FORMAT_X_OFFSET_exact %12:vgpr_32, %13:sgpr_128, 0, 44, 20, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4) +... +--- + +name: gfx10_tbuffer_store_not_merged_data_format_mismatch +body: | + bb.0.entry: + liveins: $sgpr0, $sgpr1, $sgpr2, $sgpr3, $vgpr0, $vgpr1, $vgpr2, $vgpr3, $vgpr4, $vgpr5, $vgpr6, $vgpr7, $vgpr8 + ; GFX10-LABEL: name: gfx10_tbuffer_store_not_merged_data_format_mismatch + ; GFX10: liveins: $sgpr0, $sgpr1, $sgpr2, $sgpr3, $vgpr0, $vgpr1, $vgpr2, $vgpr3, $vgpr4, $vgpr5, $vgpr6, $vgpr7, $vgpr8 + ; GFX10-NEXT: {{ $}} + ; GFX10-NEXT: [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr8 + ; GFX10-NEXT: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr7 + ; GFX10-NEXT: [[COPY2:%[0-9]+]]:vgpr_32 = COPY $vgpr6 + ; GFX10-NEXT: [[COPY3:%[0-9]+]]:vgpr_32 = COPY $vgpr5 + ; GFX10-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY $vgpr4 + ; GFX10-NEXT: [[COPY5:%[0-9]+]]:vgpr_32 = COPY $vgpr3 + ; GFX10-NEXT: [[COPY6:%[0-9]+]]:vgpr_32 = COPY $vgpr2 + ; GFX10-NEXT: [[COPY7:%[0-9]+]]:vgpr_32 = COPY $vgpr1 + ; GFX10-NEXT: [[COPY8:%[0-9]+]]:vgpr_32 = COPY $vgpr0 + ; GFX10-NEXT: [[COPY9:%[0-9]+]]:sgpr_32 = COPY $sgpr3 + ; GFX10-NEXT: [[COPY10:%[0-9]+]]:sgpr_32 = COPY $sgpr2 + ; GFX10-NEXT: [[COPY11:%[0-9]+]]:sgpr_32 = COPY $sgpr1 + ; GFX10-NEXT: [[COPY12:%[0-9]+]]:sgpr_32 = COPY $sgpr0 + ; GFX10-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY12]], %subreg.sub0, [[COPY11]], %subreg.sub1, [[COPY10]], %subreg.sub2, [[COPY9]], %subreg.sub3 + ; GFX10-NEXT: TBUFFER_STORE_FORMAT_X_OFFSET_exact [[COPY8]], [[REG_SEQUENCE]], 0, 4, 22, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4) + ; GFX10-NEXT: TBUFFER_STORE_FORMAT_X_OFFSET_exact [[COPY7]], [[REG_SEQUENCE]], 0, 8, 21, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4) + ; GFX10-NEXT: TBUFFER_STORE_FORMAT_X_OFFSET_exact [[COPY6]], [[REG_SEQUENCE]], 0, 16, 22, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4) + ; GFX10-NEXT: TBUFFER_STORE_FORMAT_X_OFFSET_exact [[COPY5]], [[REG_SEQUENCE]], 0, 20, 21, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4) + ; GFX10-NEXT: TBUFFER_STORE_FORMAT_X_OFFSET_exact [[COPY4]], [[REG_SEQUENCE]], 0, 24, 22, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4) + ; GFX10-NEXT: TBUFFER_STORE_FORMAT_X_OFFSET_exact [[COPY3]], [[REG_SEQUENCE]], 0, 28, 21, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4) + ; GFX10-NEXT: TBUFFER_STORE_FORMAT_X_OFFSET_exact [[COPY2]], [[REG_SEQUENCE]], 0, 36, 22, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4) + ; GFX10-NEXT: TBUFFER_STORE_FORMAT_X_OFFSET_exact [[COPY1]], [[REG_SEQUENCE]], 0, 40, 21, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4) + ; GFX10-NEXT: TBUFFER_STORE_FORMAT_X_OFFSET_exact [[COPY]], [[REG_SEQUENCE]], 0, 44, 22, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4) + %12:vgpr_32 = COPY $vgpr8 + %11:vgpr_32 = COPY $vgpr7 + %10:vgpr_32 = COPY $vgpr6 + %9:vgpr_32 = COPY $vgpr5 + %8:vgpr_32 = COPY $vgpr4 + %7:vgpr_32 = COPY $vgpr3 + %6:vgpr_32 = COPY $vgpr2 + %5:vgpr_32 = COPY $vgpr1 + %4:vgpr_32 = COPY $vgpr0 + %3:sgpr_32 = COPY $sgpr3 + %2:sgpr_32 = COPY $sgpr2 + %1:sgpr_32 = COPY $sgpr1 + %0:sgpr_32 = COPY $sgpr0 + %13:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3 + TBUFFER_STORE_FORMAT_X_OFFSET_exact %4:vgpr_32, %13:sgpr_128, 0, 4, 22, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4) + TBUFFER_STORE_FORMAT_X_OFFSET_exact %5:vgpr_32, %13:sgpr_128, 0, 8, 21, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4) + TBUFFER_STORE_FORMAT_X_OFFSET_exact %6:vgpr_32, %13:sgpr_128, 0, 16, 22, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4) + TBUFFER_STORE_FORMAT_X_OFFSET_exact %7:vgpr_32, %13:sgpr_128, 0, 20, 21, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4) + TBUFFER_STORE_FORMAT_X_OFFSET_exact %8:vgpr_32, %13:sgpr_128, 0, 24, 22, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4) + TBUFFER_STORE_FORMAT_X_OFFSET_exact %9:vgpr_32, %13:sgpr_128, 0, 28, 21, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4) + TBUFFER_STORE_FORMAT_X_OFFSET_exact %10:vgpr_32, %13:sgpr_128, 0, 36, 22, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4) + TBUFFER_STORE_FORMAT_X_OFFSET_exact %11:vgpr_32, %13:sgpr_128, 0, 40, 21, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4) + TBUFFER_STORE_FORMAT_X_OFFSET_exact %12:vgpr_32, %13:sgpr_128, 0, 44, 22, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4) +... +--- + +name: gfx10_tbuffer_store_not_merged_num_format_mismatch +body: | + bb.0.entry: + liveins: $sgpr0, $sgpr1, $sgpr2, $sgpr3, $vgpr0, $vgpr1, $vgpr2, $vgpr3, $vgpr4, $vgpr5, $vgpr6, $vgpr7, $vgpr8 + ; GFX10-LABEL: name: gfx10_tbuffer_store_not_merged_num_format_mismatch + ; GFX10: liveins: $sgpr0, $sgpr1, $sgpr2, $sgpr3, $vgpr0, $vgpr1, $vgpr2, $vgpr3, $vgpr4, $vgpr5, $vgpr6, $vgpr7, $vgpr8 + ; GFX10-NEXT: {{ $}} + ; GFX10-NEXT: [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr8 + ; GFX10-NEXT: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr7 + ; GFX10-NEXT: [[COPY2:%[0-9]+]]:vgpr_32 = COPY $vgpr6 + ; GFX10-NEXT: [[COPY3:%[0-9]+]]:vgpr_32 = COPY $vgpr5 + ; GFX10-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY $vgpr4 + ; GFX10-NEXT: [[COPY5:%[0-9]+]]:vgpr_32 = COPY $vgpr3 + ; GFX10-NEXT: [[COPY6:%[0-9]+]]:vgpr_32 = COPY $vgpr2 + ; GFX10-NEXT: [[COPY7:%[0-9]+]]:vgpr_32 = COPY $vgpr1 + ; GFX10-NEXT: [[COPY8:%[0-9]+]]:vgpr_32 = COPY $vgpr0 + ; GFX10-NEXT: [[COPY9:%[0-9]+]]:sgpr_32 = COPY $sgpr3 + ; GFX10-NEXT: [[COPY10:%[0-9]+]]:sgpr_32 = COPY $sgpr2 + ; GFX10-NEXT: [[COPY11:%[0-9]+]]:sgpr_32 = COPY $sgpr1 + ; GFX10-NEXT: [[COPY12:%[0-9]+]]:sgpr_32 = COPY $sgpr0 + ; GFX10-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY12]], %subreg.sub0, [[COPY11]], %subreg.sub1, [[COPY10]], %subreg.sub2, [[COPY9]], %subreg.sub3 + ; GFX10-NEXT: TBUFFER_STORE_FORMAT_X_OFFSET_exact [[COPY8]], [[REG_SEQUENCE]], 0, 4, 22, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4) + ; GFX10-NEXT: TBUFFER_STORE_FORMAT_X_OFFSET_exact [[COPY7]], [[REG_SEQUENCE]], 0, 8, 13, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4) + ; GFX10-NEXT: TBUFFER_STORE_FORMAT_X_OFFSET_exact [[COPY6]], [[REG_SEQUENCE]], 0, 16, 22, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4) + ; GFX10-NEXT: TBUFFER_STORE_FORMAT_X_OFFSET_exact [[COPY5]], [[REG_SEQUENCE]], 0, 20, 13, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4) + ; GFX10-NEXT: TBUFFER_STORE_FORMAT_X_OFFSET_exact [[COPY4]], [[REG_SEQUENCE]], 0, 24, 22, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4) + ; GFX10-NEXT: TBUFFER_STORE_FORMAT_X_OFFSET_exact [[COPY3]], [[REG_SEQUENCE]], 0, 28, 13, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4) + ; GFX10-NEXT: TBUFFER_STORE_FORMAT_X_OFFSET_exact [[COPY2]], [[REG_SEQUENCE]], 0, 36, 22, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4) + ; GFX10-NEXT: TBUFFER_STORE_FORMAT_X_OFFSET_exact [[COPY1]], [[REG_SEQUENCE]], 0, 40, 13, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4) + ; GFX10-NEXT: TBUFFER_STORE_FORMAT_X_OFFSET_exact [[COPY]], [[REG_SEQUENCE]], 0, 44, 22, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4) + %12:vgpr_32 = COPY $vgpr8 + %11:vgpr_32 = COPY $vgpr7 + %10:vgpr_32 = COPY $vgpr6 + %9:vgpr_32 = COPY $vgpr5 + %8:vgpr_32 = COPY $vgpr4 + %7:vgpr_32 = COPY $vgpr3 + %6:vgpr_32 = COPY $vgpr2 + %5:vgpr_32 = COPY $vgpr1 + %4:vgpr_32 = COPY $vgpr0 + %3:sgpr_32 = COPY $sgpr3 + %2:sgpr_32 = COPY $sgpr2 + %1:sgpr_32 = COPY $sgpr1 + %0:sgpr_32 = COPY $sgpr0 + %13:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3 + TBUFFER_STORE_FORMAT_X_OFFSET_exact %4:vgpr_32, %13:sgpr_128, 0, 4, 22, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4) + TBUFFER_STORE_FORMAT_X_OFFSET_exact %5:vgpr_32, %13:sgpr_128, 0, 8, 13, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4) + TBUFFER_STORE_FORMAT_X_OFFSET_exact %6:vgpr_32, %13:sgpr_128, 0, 16, 22, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4) + TBUFFER_STORE_FORMAT_X_OFFSET_exact %7:vgpr_32, %13:sgpr_128, 0, 20, 13, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4) + TBUFFER_STORE_FORMAT_X_OFFSET_exact %8:vgpr_32, %13:sgpr_128, 0, 24, 22, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4) + TBUFFER_STORE_FORMAT_X_OFFSET_exact %9:vgpr_32, %13:sgpr_128, 0, 28, 13, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4) + TBUFFER_STORE_FORMAT_X_OFFSET_exact %10:vgpr_32, %13:sgpr_128, 0, 36, 22, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4) + TBUFFER_STORE_FORMAT_X_OFFSET_exact %11:vgpr_32, %13:sgpr_128, 0, 40, 13, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4) + TBUFFER_STORE_FORMAT_X_OFFSET_exact %12:vgpr_32, %13:sgpr_128, 0, 44, 22, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4) +... +--- + +name: gfx10_tbuffer_load_not_merged_swizzled_0 +body: | + bb.0.entry: + ; GFX10-LABEL: name: gfx10_tbuffer_load_not_merged_swizzled_0 + ; GFX10: [[COPY:%[0-9]+]]:sgpr_32 = COPY $sgpr0 + ; GFX10-NEXT: [[COPY1:%[0-9]+]]:sgpr_32 = COPY $sgpr1 + ; GFX10-NEXT: [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr2 + ; GFX10-NEXT: [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr3 + ; GFX10-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3 + ; GFX10-NEXT: [[TBUFFER_LOAD_FORMAT_X_OFFSET:%[0-9]+]]:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET [[REG_SEQUENCE]], 0, 4, 22, 0, 1, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) + ; GFX10-NEXT: [[TBUFFER_LOAD_FORMAT_X_OFFSET1:%[0-9]+]]:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET [[REG_SEQUENCE]], 0, 8, 22, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) + %0:sgpr_32 = COPY $sgpr0 + %1:sgpr_32 = COPY $sgpr1 + %2:sgpr_32 = COPY $sgpr2 + %3:sgpr_32 = COPY $sgpr3 + %5:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3 + %7:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %5:sgpr_128, 0, 4, 22, 0, 1, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) + %8:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %5:sgpr_128, 0, 8, 22, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) +... +--- + +name: gfx10_tbuffer_load_not_merged_swizzled_1 +body: | + bb.0.entry: + ; GFX10-LABEL: name: gfx10_tbuffer_load_not_merged_swizzled_1 + ; GFX10: [[COPY:%[0-9]+]]:sgpr_32 = COPY $sgpr0 + ; GFX10-NEXT: [[COPY1:%[0-9]+]]:sgpr_32 = COPY $sgpr1 + ; GFX10-NEXT: [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr2 + ; GFX10-NEXT: [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr3 + ; GFX10-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3 + ; GFX10-NEXT: [[TBUFFER_LOAD_FORMAT_X_OFFSET:%[0-9]+]]:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET [[REG_SEQUENCE]], 0, 4, 22, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) + ; GFX10-NEXT: [[TBUFFER_LOAD_FORMAT_X_OFFSET1:%[0-9]+]]:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET [[REG_SEQUENCE]], 0, 8, 22, 0, 1, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) + %0:sgpr_32 = COPY $sgpr0 + %1:sgpr_32 = COPY $sgpr1 + %2:sgpr_32 = COPY $sgpr2 + %3:sgpr_32 = COPY $sgpr3 + %5:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3 + %7:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %5:sgpr_128, 0, 4, 22, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) + %8:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %5:sgpr_128, 0, 8, 22, 0, 1, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) +... +--- + +name: gfx10_tbuffer_load_merge_across_swizzle +body: | + bb.0.entry: + ; GFX10-LABEL: name: gfx10_tbuffer_load_merge_across_swizzle + ; GFX10: [[COPY:%[0-9]+]]:sgpr_32 = COPY $sgpr0 + ; GFX10-NEXT: [[COPY1:%[0-9]+]]:sgpr_32 = COPY $sgpr1 + ; GFX10-NEXT: [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr2 + ; GFX10-NEXT: [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr3 + ; GFX10-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3 + ; GFX10-NEXT: [[TBUFFER_LOAD_FORMAT_XY_OFFSET:%[0-9]+]]:vreg_64 = TBUFFER_LOAD_FORMAT_XY_OFFSET [[REG_SEQUENCE]], 0, 4, 64, 0, 0, implicit $exec :: (dereferenceable load (s64), align 1, addrspace 4) + ; GFX10-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[TBUFFER_LOAD_FORMAT_XY_OFFSET]].sub0 + ; GFX10-NEXT: [[COPY5:%[0-9]+]]:vgpr_32 = COPY killed [[TBUFFER_LOAD_FORMAT_XY_OFFSET]].sub1 + ; GFX10-NEXT: [[TBUFFER_LOAD_FORMAT_X_OFFSET:%[0-9]+]]:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET [[REG_SEQUENCE]], 0, 12, 22, 0, 1, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) + %0:sgpr_32 = COPY $sgpr0 + %1:sgpr_32 = COPY $sgpr1 + %2:sgpr_32 = COPY $sgpr2 + %3:sgpr_32 = COPY $sgpr3 + %4:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3 + %5:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %4:sgpr_128, 0, 4, 22, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) + %6:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %4:sgpr_128, 0, 12, 22, 0, 1, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) + %7:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %4:sgpr_128, 0, 8, 22, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) +... +--- + +name: gfx10_tbuffer_load_x_x_idxen +body: | + bb.0.entry: + ; GFX10-LABEL: name: gfx10_tbuffer_load_x_x_idxen + ; GFX10: [[COPY:%[0-9]+]]:sgpr_32 = COPY $sgpr0 + ; GFX10-NEXT: [[COPY1:%[0-9]+]]:sgpr_32 = COPY $sgpr1 + ; GFX10-NEXT: [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr2 + ; GFX10-NEXT: [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr3 + ; GFX10-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY $vgpr0 + ; GFX10-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3 + ; GFX10-NEXT: [[TBUFFER_LOAD_FORMAT_XY_IDXEN:%[0-9]+]]:vreg_64 = TBUFFER_LOAD_FORMAT_XY_IDXEN [[COPY4]], [[REG_SEQUENCE]], 0, 0, 64, 0, 0, implicit $exec :: (dereferenceable load (s64), align 1, addrspace 4) + ; GFX10-NEXT: [[COPY5:%[0-9]+]]:vgpr_32 = COPY [[TBUFFER_LOAD_FORMAT_XY_IDXEN]].sub0 + ; GFX10-NEXT: [[COPY6:%[0-9]+]]:vgpr_32 = COPY killed [[TBUFFER_LOAD_FORMAT_XY_IDXEN]].sub1 + %0:sgpr_32 = COPY $sgpr0 + %1:sgpr_32 = COPY $sgpr1 + %2:sgpr_32 = COPY $sgpr2 + %3:sgpr_32 = COPY $sgpr3 + %4:vgpr_32 = COPY $vgpr0 + %5:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3 + %7:vgpr_32 = TBUFFER_LOAD_FORMAT_X_IDXEN %4, %5:sgpr_128, 0, 0, 22, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) + %8:vgpr_32 = TBUFFER_LOAD_FORMAT_X_IDXEN %4, %5:sgpr_128, 0, 4, 22, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) +... +--- + +name: gfx10_tbuffer_load_x_xy_idxen +body: | + bb.0.entry: + ; GFX10-LABEL: name: gfx10_tbuffer_load_x_xy_idxen + ; GFX10: [[COPY:%[0-9]+]]:sgpr_32 = COPY $sgpr0 + ; GFX10-NEXT: [[COPY1:%[0-9]+]]:sgpr_32 = COPY $sgpr1 + ; GFX10-NEXT: [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr2 + ; GFX10-NEXT: [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr3 + ; GFX10-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY $vgpr0 + ; GFX10-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3 + ; GFX10-NEXT: [[TBUFFER_LOAD_FORMAT_XYZ_IDXEN:%[0-9]+]]:vreg_96 = TBUFFER_LOAD_FORMAT_XYZ_IDXEN [[COPY4]], [[REG_SEQUENCE]], 0, 0, 74, 0, 0, implicit $exec :: (dereferenceable load (s64), align 1, addrspace 4) + ; GFX10-NEXT: [[COPY5:%[0-9]+]]:vgpr_32 = COPY [[TBUFFER_LOAD_FORMAT_XYZ_IDXEN]].sub0 + ; GFX10-NEXT: [[COPY6:%[0-9]+]]:vreg_64 = COPY killed [[TBUFFER_LOAD_FORMAT_XYZ_IDXEN]].sub1_sub2 + %0:sgpr_32 = COPY $sgpr0 + %1:sgpr_32 = COPY $sgpr1 + %2:sgpr_32 = COPY $sgpr2 + %3:sgpr_32 = COPY $sgpr3 + %4:vgpr_32 = COPY $vgpr0 + %5:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3 + %7:vgpr_32 = TBUFFER_LOAD_FORMAT_X_IDXEN %4, %5:sgpr_128, 0, 0, 22, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) + %8:vreg_64 = TBUFFER_LOAD_FORMAT_XY_IDXEN %4, %5:sgpr_128, 0, 4, 64, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) +... +--- + +name: gfx10_tbuffer_load_xy_xy_idxen +body: | + bb.0.entry: + ; GFX10-LABEL: name: gfx10_tbuffer_load_xy_xy_idxen + ; GFX10: [[COPY:%[0-9]+]]:sgpr_32 = COPY $sgpr0 + ; GFX10-NEXT: [[COPY1:%[0-9]+]]:sgpr_32 = COPY $sgpr1 + ; GFX10-NEXT: [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr2 + ; GFX10-NEXT: [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr3 + ; GFX10-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY $vgpr0 + ; GFX10-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3 + ; GFX10-NEXT: [[TBUFFER_LOAD_FORMAT_XYZW_IDXEN:%[0-9]+]]:vreg_128 = TBUFFER_LOAD_FORMAT_XYZW_IDXEN [[COPY4]], [[REG_SEQUENCE]], 0, 4, 77, 0, 0, implicit $exec :: (dereferenceable load (s64), align 1, addrspace 4) + ; GFX10-NEXT: [[COPY5:%[0-9]+]]:vreg_64 = COPY [[TBUFFER_LOAD_FORMAT_XYZW_IDXEN]].sub0_sub1 + ; GFX10-NEXT: [[COPY6:%[0-9]+]]:vreg_64 = COPY killed [[TBUFFER_LOAD_FORMAT_XYZW_IDXEN]].sub2_sub3 + %0:sgpr_32 = COPY $sgpr0 + %1:sgpr_32 = COPY $sgpr1 + %2:sgpr_32 = COPY $sgpr2 + %3:sgpr_32 = COPY $sgpr3 + %4:vgpr_32 = COPY $vgpr0 + %5:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3 + %7:vreg_64 = TBUFFER_LOAD_FORMAT_XY_IDXEN %4, %5:sgpr_128, 0, 4, 64, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) + %8:vreg_64 = TBUFFER_LOAD_FORMAT_XY_IDXEN %4, %5:sgpr_128, 0, 12, 64, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) +... +--- + +name: gfx10_tbuffer_load_x_xyz_idxen +body: | + bb.0.entry: + ; GFX10-LABEL: name: gfx10_tbuffer_load_x_xyz_idxen + ; GFX10: [[COPY:%[0-9]+]]:sgpr_32 = COPY $sgpr0 + ; GFX10-NEXT: [[COPY1:%[0-9]+]]:sgpr_32 = COPY $sgpr1 + ; GFX10-NEXT: [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr2 + ; GFX10-NEXT: [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr3 + ; GFX10-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY $vgpr0 + ; GFX10-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3 + ; GFX10-NEXT: [[TBUFFER_LOAD_FORMAT_XYZW_IDXEN:%[0-9]+]]:vreg_128 = TBUFFER_LOAD_FORMAT_XYZW_IDXEN [[COPY4]], [[REG_SEQUENCE]], 0, 4, 77, 0, 0, implicit $exec :: (dereferenceable load (s64), align 1, addrspace 4) + ; GFX10-NEXT: [[COPY5:%[0-9]+]]:vgpr_32 = COPY [[TBUFFER_LOAD_FORMAT_XYZW_IDXEN]].sub0 + ; GFX10-NEXT: [[COPY6:%[0-9]+]]:vreg_96 = COPY killed [[TBUFFER_LOAD_FORMAT_XYZW_IDXEN]].sub1_sub2_sub3 + %0:sgpr_32 = COPY $sgpr0 + %1:sgpr_32 = COPY $sgpr1 + %2:sgpr_32 = COPY $sgpr2 + %3:sgpr_32 = COPY $sgpr3 + %4:vgpr_32 = COPY $vgpr0 + %5:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3 + %7:vgpr_32 = TBUFFER_LOAD_FORMAT_X_IDXEN %4, %5:sgpr_128, 0, 4, 22, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) + %8:vreg_96 = TBUFFER_LOAD_FORMAT_XYZ_IDXEN %4, %5:sgpr_128, 0, 8, 74, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) +... +--- + +name: gfx10_tbuffer_load_x_x_x_idxen_exact +body: | + bb.0.entry: + ; GFX10-LABEL: name: gfx10_tbuffer_load_x_x_x_idxen_exact + ; GFX10: [[COPY:%[0-9]+]]:sgpr_32 = COPY $sgpr0 + ; GFX10-NEXT: [[COPY1:%[0-9]+]]:sgpr_32 = COPY $sgpr1 + ; GFX10-NEXT: [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr2 + ; GFX10-NEXT: [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr3 + ; GFX10-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY $vgpr0 + ; GFX10-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3 + ; GFX10-NEXT: [[TBUFFER_LOAD_FORMAT_XYZ_IDXEN_exact:%[0-9]+]]:vreg_96 = TBUFFER_LOAD_FORMAT_XYZ_IDXEN_exact [[COPY4]], [[REG_SEQUENCE]], 0, 4, 74, 0, 0, implicit $exec :: (dereferenceable load (s96), align 1, addrspace 4) + ; GFX10-NEXT: [[COPY5:%[0-9]+]]:vreg_64 = COPY [[TBUFFER_LOAD_FORMAT_XYZ_IDXEN_exact]].sub0_sub1 + ; GFX10-NEXT: [[COPY6:%[0-9]+]]:vgpr_32 = COPY killed [[TBUFFER_LOAD_FORMAT_XYZ_IDXEN_exact]].sub2 + ; GFX10-NEXT: [[COPY7:%[0-9]+]]:vgpr_32 = COPY [[COPY5]].sub0 + ; GFX10-NEXT: [[COPY8:%[0-9]+]]:vgpr_32 = COPY killed [[COPY5]].sub1 + %0:sgpr_32 = COPY $sgpr0 + %1:sgpr_32 = COPY $sgpr1 + %2:sgpr_32 = COPY $sgpr2 + %3:sgpr_32 = COPY $sgpr3 + %4:vgpr_32 = COPY $vgpr0 + %5:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3 + %7:vgpr_32 = TBUFFER_LOAD_FORMAT_X_IDXEN_exact %4, %5:sgpr_128, 0, 4, 22, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) + %8:vgpr_32 = TBUFFER_LOAD_FORMAT_X_IDXEN_exact %4, %5:sgpr_128, 0, 8, 22, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) + %9:vgpr_32 = TBUFFER_LOAD_FORMAT_X_IDXEN_exact %4, %5:sgpr_128, 0, 12, 22, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) +... +--- + +name: gfx10_tbuffer_load_x_x_x_idxen_exact_swizzled_0 +body: | + bb.0.entry: + ; GFX10-LABEL: name: gfx10_tbuffer_load_x_x_x_idxen_exact_swizzled_0 + ; GFX10: [[COPY:%[0-9]+]]:sgpr_32 = COPY $sgpr0 + ; GFX10-NEXT: [[COPY1:%[0-9]+]]:sgpr_32 = COPY $sgpr1 + ; GFX10-NEXT: [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr2 + ; GFX10-NEXT: [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr3 + ; GFX10-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY $vgpr0 + ; GFX10-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3 + ; GFX10-NEXT: [[TBUFFER_LOAD_FORMAT_X_IDXEN_exact:%[0-9]+]]:vgpr_32 = TBUFFER_LOAD_FORMAT_X_IDXEN_exact [[COPY4]], [[REG_SEQUENCE]], 0, 4, 22, 0, 1, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) + ; GFX10-NEXT: [[TBUFFER_LOAD_FORMAT_XY_IDXEN_exact:%[0-9]+]]:vreg_64 = TBUFFER_LOAD_FORMAT_XY_IDXEN_exact [[COPY4]], [[REG_SEQUENCE]], 0, 8, 64, 0, 0, implicit $exec :: (dereferenceable load (s64), align 1, addrspace 4) + ; GFX10-NEXT: [[COPY5:%[0-9]+]]:vgpr_32 = COPY [[TBUFFER_LOAD_FORMAT_XY_IDXEN_exact]].sub0 + ; GFX10-NEXT: [[COPY6:%[0-9]+]]:vgpr_32 = COPY killed [[TBUFFER_LOAD_FORMAT_XY_IDXEN_exact]].sub1 + %0:sgpr_32 = COPY $sgpr0 + %1:sgpr_32 = COPY $sgpr1 + %2:sgpr_32 = COPY $sgpr2 + %3:sgpr_32 = COPY $sgpr3 + %4:vgpr_32 = COPY $vgpr0 + %5:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3 + %7:vgpr_32 = TBUFFER_LOAD_FORMAT_X_IDXEN_exact %4, %5:sgpr_128, 0, 4, 22, 0, 1, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) + %8:vgpr_32 = TBUFFER_LOAD_FORMAT_X_IDXEN_exact %4, %5:sgpr_128, 0, 8, 22, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) + %9:vgpr_32 = TBUFFER_LOAD_FORMAT_X_IDXEN_exact %4, %5:sgpr_128, 0, 12, 22, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) +... +--- + +name: gfx10_tbuffer_load_x_x_bothen +body: | + bb.0.entry: + ; GFX10-LABEL: name: gfx10_tbuffer_load_x_x_bothen + ; GFX10: [[COPY:%[0-9]+]]:sgpr_32 = COPY $sgpr0 + ; GFX10-NEXT: [[COPY1:%[0-9]+]]:sgpr_32 = COPY $sgpr1 + ; GFX10-NEXT: [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr2 + ; GFX10-NEXT: [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr3 + ; GFX10-NEXT: [[COPY4:%[0-9]+]]:vreg_64 = COPY $vgpr0 + ; GFX10-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3 + ; GFX10-NEXT: [[TBUFFER_LOAD_FORMAT_XY_BOTHEN:%[0-9]+]]:vreg_64 = TBUFFER_LOAD_FORMAT_XY_BOTHEN [[COPY4]], [[REG_SEQUENCE]], 0, 0, 64, 0, 0, implicit $exec :: (dereferenceable load (s64), align 1, addrspace 4) + ; GFX10-NEXT: [[COPY5:%[0-9]+]]:vgpr_32 = COPY [[TBUFFER_LOAD_FORMAT_XY_BOTHEN]].sub0 + ; GFX10-NEXT: [[COPY6:%[0-9]+]]:vgpr_32 = COPY killed [[TBUFFER_LOAD_FORMAT_XY_BOTHEN]].sub1 + %0:sgpr_32 = COPY $sgpr0 + %1:sgpr_32 = COPY $sgpr1 + %2:sgpr_32 = COPY $sgpr2 + %3:sgpr_32 = COPY $sgpr3 + %4:vreg_64 = COPY $vgpr0 + %5:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3 + %7:vgpr_32 = TBUFFER_LOAD_FORMAT_X_BOTHEN %4, %5:sgpr_128, 0, 0, 22, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) + %8:vgpr_32 = TBUFFER_LOAD_FORMAT_X_BOTHEN %4, %5:sgpr_128, 0, 4, 22, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) +... +--- + +name: gfx10_tbuffer_load_x_xy_bothen +body: | + bb.0.entry: + ; GFX10-LABEL: name: gfx10_tbuffer_load_x_xy_bothen + ; GFX10: [[COPY:%[0-9]+]]:sgpr_32 = COPY $sgpr0 + ; GFX10-NEXT: [[COPY1:%[0-9]+]]:sgpr_32 = COPY $sgpr1 + ; GFX10-NEXT: [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr2 + ; GFX10-NEXT: [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr3 + ; GFX10-NEXT: [[COPY4:%[0-9]+]]:vreg_64 = COPY $vgpr0 + ; GFX10-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3 + ; GFX10-NEXT: [[TBUFFER_LOAD_FORMAT_XYZ_BOTHEN:%[0-9]+]]:vreg_96 = TBUFFER_LOAD_FORMAT_XYZ_BOTHEN [[COPY4]], [[REG_SEQUENCE]], 0, 0, 74, 0, 0, implicit $exec :: (dereferenceable load (s64), align 1, addrspace 4) + ; GFX10-NEXT: [[COPY5:%[0-9]+]]:vgpr_32 = COPY [[TBUFFER_LOAD_FORMAT_XYZ_BOTHEN]].sub0 + ; GFX10-NEXT: [[COPY6:%[0-9]+]]:vreg_64 = COPY killed [[TBUFFER_LOAD_FORMAT_XYZ_BOTHEN]].sub1_sub2 + %0:sgpr_32 = COPY $sgpr0 + %1:sgpr_32 = COPY $sgpr1 + %2:sgpr_32 = COPY $sgpr2 + %3:sgpr_32 = COPY $sgpr3 + %4:vreg_64 = COPY $vgpr0 + %5:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3 + %7:vgpr_32 = TBUFFER_LOAD_FORMAT_X_BOTHEN %4, %5:sgpr_128, 0, 0, 22, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) + %8:vreg_64 = TBUFFER_LOAD_FORMAT_XY_BOTHEN %4, %5:sgpr_128, 0, 4, 64, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) +... +--- + +name: gfx10_tbuffer_load_xy_xy_bothen +body: | + bb.0.entry: + ; GFX10-LABEL: name: gfx10_tbuffer_load_xy_xy_bothen + ; GFX10: [[COPY:%[0-9]+]]:sgpr_32 = COPY $sgpr0 + ; GFX10-NEXT: [[COPY1:%[0-9]+]]:sgpr_32 = COPY $sgpr1 + ; GFX10-NEXT: [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr2 + ; GFX10-NEXT: [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr3 + ; GFX10-NEXT: [[COPY4:%[0-9]+]]:vreg_64 = COPY $vgpr0 + ; GFX10-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3 + ; GFX10-NEXT: [[TBUFFER_LOAD_FORMAT_XYZW_BOTHEN:%[0-9]+]]:vreg_128 = TBUFFER_LOAD_FORMAT_XYZW_BOTHEN [[COPY4]], [[REG_SEQUENCE]], 0, 0, 77, 0, 0, implicit $exec :: (dereferenceable load (s64), align 1, addrspace 4) + ; GFX10-NEXT: [[COPY5:%[0-9]+]]:vreg_64 = COPY [[TBUFFER_LOAD_FORMAT_XYZW_BOTHEN]].sub0_sub1 + ; GFX10-NEXT: [[COPY6:%[0-9]+]]:vreg_64 = COPY killed [[TBUFFER_LOAD_FORMAT_XYZW_BOTHEN]].sub2_sub3 + %0:sgpr_32 = COPY $sgpr0 + %1:sgpr_32 = COPY $sgpr1 + %2:sgpr_32 = COPY $sgpr2 + %3:sgpr_32 = COPY $sgpr3 + %4:vreg_64 = COPY $vgpr0 + %5:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3 + %7:vreg_64 = TBUFFER_LOAD_FORMAT_XY_BOTHEN %4, %5:sgpr_128, 0, 0, 64, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) + %8:vreg_64 = TBUFFER_LOAD_FORMAT_XY_BOTHEN %4, %5:sgpr_128, 0, 8, 64, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) +... +--- + +name: gfx10_tbuffer_load_x_xyz_bothen +body: | + bb.0.entry: + ; GFX10-LABEL: name: gfx10_tbuffer_load_x_xyz_bothen + ; GFX10: [[COPY:%[0-9]+]]:sgpr_32 = COPY $sgpr0 + ; GFX10-NEXT: [[COPY1:%[0-9]+]]:sgpr_32 = COPY $sgpr1 + ; GFX10-NEXT: [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr2 + ; GFX10-NEXT: [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr3 + ; GFX10-NEXT: [[COPY4:%[0-9]+]]:vreg_64 = COPY $vgpr0 + ; GFX10-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3 + ; GFX10-NEXT: [[TBUFFER_LOAD_FORMAT_XYZW_BOTHEN:%[0-9]+]]:vreg_128 = TBUFFER_LOAD_FORMAT_XYZW_BOTHEN [[COPY4]], [[REG_SEQUENCE]], 0, 0, 77, 0, 0, implicit $exec :: (dereferenceable load (s64), align 1, addrspace 4) + ; GFX10-NEXT: [[COPY5:%[0-9]+]]:vgpr_32 = COPY [[TBUFFER_LOAD_FORMAT_XYZW_BOTHEN]].sub0 + ; GFX10-NEXT: [[COPY6:%[0-9]+]]:vreg_96 = COPY killed [[TBUFFER_LOAD_FORMAT_XYZW_BOTHEN]].sub1_sub2_sub3 + %0:sgpr_32 = COPY $sgpr0 + %1:sgpr_32 = COPY $sgpr1 + %2:sgpr_32 = COPY $sgpr2 + %3:sgpr_32 = COPY $sgpr3 + %4:vreg_64 = COPY $vgpr0 + %5:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3 + %7:vgpr_32 = TBUFFER_LOAD_FORMAT_X_BOTHEN %4, %5:sgpr_128, 0, 0, 22, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) + %8:vreg_96 = TBUFFER_LOAD_FORMAT_XYZ_BOTHEN %4, %5:sgpr_128, 0, 4, 74, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) +... +--- + +name: gfx10_tbuffer_load_x_x_idxen_exact +body: | + bb.0.entry: + ; GFX10-LABEL: name: gfx10_tbuffer_load_x_x_idxen_exact + ; GFX10: [[COPY:%[0-9]+]]:sgpr_32 = COPY $sgpr0 + ; GFX10-NEXT: [[COPY1:%[0-9]+]]:sgpr_32 = COPY $sgpr1 + ; GFX10-NEXT: [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr2 + ; GFX10-NEXT: [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr3 + ; GFX10-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY $vgpr0 + ; GFX10-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3 + ; GFX10-NEXT: [[TBUFFER_LOAD_FORMAT_XY_IDXEN_exact:%[0-9]+]]:vreg_64 = TBUFFER_LOAD_FORMAT_XY_IDXEN_exact [[COPY4]], [[REG_SEQUENCE]], 0, 0, 64, 0, 0, implicit $exec :: (dereferenceable load (s64), align 1, addrspace 4) + ; GFX10-NEXT: [[COPY5:%[0-9]+]]:vgpr_32 = COPY [[TBUFFER_LOAD_FORMAT_XY_IDXEN_exact]].sub0 + ; GFX10-NEXT: [[COPY6:%[0-9]+]]:vgpr_32 = COPY killed [[TBUFFER_LOAD_FORMAT_XY_IDXEN_exact]].sub1 + %0:sgpr_32 = COPY $sgpr0 + %1:sgpr_32 = COPY $sgpr1 + %2:sgpr_32 = COPY $sgpr2 + %3:sgpr_32 = COPY $sgpr3 + %4:vgpr_32 = COPY $vgpr0 + %5:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3 + %7:vgpr_32 = TBUFFER_LOAD_FORMAT_X_IDXEN_exact %4, %5:sgpr_128, 0, 0, 22, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) + %8:vgpr_32 = TBUFFER_LOAD_FORMAT_X_IDXEN_exact %4, %5:sgpr_128, 0, 4, 22, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) +... +--- + +name: gfx10_tbuffer_load_x_xy_idxen_exact +body: | + bb.0.entry: + ; GFX10-LABEL: name: gfx10_tbuffer_load_x_xy_idxen_exact + ; GFX10: [[COPY:%[0-9]+]]:sgpr_32 = COPY $sgpr0 + ; GFX10-NEXT: [[COPY1:%[0-9]+]]:sgpr_32 = COPY $sgpr1 + ; GFX10-NEXT: [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr2 + ; GFX10-NEXT: [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr3 + ; GFX10-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY $vgpr0 + ; GFX10-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3 + ; GFX10-NEXT: [[TBUFFER_LOAD_FORMAT_XYZ_IDXEN_exact:%[0-9]+]]:vreg_96 = TBUFFER_LOAD_FORMAT_XYZ_IDXEN_exact [[COPY4]], [[REG_SEQUENCE]], 0, 0, 74, 0, 0, implicit $exec :: (dereferenceable load (s64), align 1, addrspace 4) + ; GFX10-NEXT: [[COPY5:%[0-9]+]]:vgpr_32 = COPY [[TBUFFER_LOAD_FORMAT_XYZ_IDXEN_exact]].sub0 + ; GFX10-NEXT: [[COPY6:%[0-9]+]]:vreg_64 = COPY killed [[TBUFFER_LOAD_FORMAT_XYZ_IDXEN_exact]].sub1_sub2 + %0:sgpr_32 = COPY $sgpr0 + %1:sgpr_32 = COPY $sgpr1 + %2:sgpr_32 = COPY $sgpr2 + %3:sgpr_32 = COPY $sgpr3 + %4:vgpr_32 = COPY $vgpr0 + %5:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3 + %7:vgpr_32 = TBUFFER_LOAD_FORMAT_X_IDXEN_exact %4, %5:sgpr_128, 0, 0, 22, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) + %8:vreg_64 = TBUFFER_LOAD_FORMAT_XY_IDXEN_exact %4, %5:sgpr_128, 0, 4, 64, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) +... +--- + +name: gfx10_tbuffer_load_xy_xy_idxen_exact +body: | + bb.0.entry: + ; GFX10-LABEL: name: gfx10_tbuffer_load_xy_xy_idxen_exact + ; GFX10: [[COPY:%[0-9]+]]:sgpr_32 = COPY $sgpr0 + ; GFX10-NEXT: [[COPY1:%[0-9]+]]:sgpr_32 = COPY $sgpr1 + ; GFX10-NEXT: [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr2 + ; GFX10-NEXT: [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr3 + ; GFX10-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY $vgpr0 + ; GFX10-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3 + ; GFX10-NEXT: [[TBUFFER_LOAD_FORMAT_XYZW_IDXEN_exact:%[0-9]+]]:vreg_128 = TBUFFER_LOAD_FORMAT_XYZW_IDXEN_exact [[COPY4]], [[REG_SEQUENCE]], 0, 4, 77, 0, 0, implicit $exec :: (dereferenceable load (s64), align 1, addrspace 4) + ; GFX10-NEXT: [[COPY5:%[0-9]+]]:vreg_64 = COPY [[TBUFFER_LOAD_FORMAT_XYZW_IDXEN_exact]].sub0_sub1 + ; GFX10-NEXT: [[COPY6:%[0-9]+]]:vreg_64 = COPY killed [[TBUFFER_LOAD_FORMAT_XYZW_IDXEN_exact]].sub2_sub3 + %0:sgpr_32 = COPY $sgpr0 + %1:sgpr_32 = COPY $sgpr1 + %2:sgpr_32 = COPY $sgpr2 + %3:sgpr_32 = COPY $sgpr3 + %4:vgpr_32 = COPY $vgpr0 + %5:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3 + %7:vreg_64 = TBUFFER_LOAD_FORMAT_XY_IDXEN_exact %4, %5:sgpr_128, 0, 4, 64, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) + %8:vreg_64 = TBUFFER_LOAD_FORMAT_XY_IDXEN_exact %4, %5:sgpr_128, 0, 12, 64, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) +... +--- + +name: gfx10_tbuffer_load_x_xyz_idxen_exact +body: | + bb.0.entry: + ; GFX10-LABEL: name: gfx10_tbuffer_load_x_xyz_idxen_exact + ; GFX10: [[COPY:%[0-9]+]]:sgpr_32 = COPY $sgpr0 + ; GFX10-NEXT: [[COPY1:%[0-9]+]]:sgpr_32 = COPY $sgpr1 + ; GFX10-NEXT: [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr2 + ; GFX10-NEXT: [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr3 + ; GFX10-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY $vgpr0 + ; GFX10-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3 + ; GFX10-NEXT: [[TBUFFER_LOAD_FORMAT_XYZW_IDXEN_exact:%[0-9]+]]:vreg_128 = TBUFFER_LOAD_FORMAT_XYZW_IDXEN_exact [[COPY4]], [[REG_SEQUENCE]], 0, 4, 77, 0, 0, implicit $exec :: (dereferenceable load (s64), align 1, addrspace 4) + ; GFX10-NEXT: [[COPY5:%[0-9]+]]:vgpr_32 = COPY [[TBUFFER_LOAD_FORMAT_XYZW_IDXEN_exact]].sub0 + ; GFX10-NEXT: [[COPY6:%[0-9]+]]:vreg_96 = COPY killed [[TBUFFER_LOAD_FORMAT_XYZW_IDXEN_exact]].sub1_sub2_sub3 + %0:sgpr_32 = COPY $sgpr0 + %1:sgpr_32 = COPY $sgpr1 + %2:sgpr_32 = COPY $sgpr2 + %3:sgpr_32 = COPY $sgpr3 + %4:vgpr_32 = COPY $vgpr0 + %5:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3 + %7:vgpr_32 = TBUFFER_LOAD_FORMAT_X_IDXEN_exact %4, %5:sgpr_128, 0, 4, 22, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) + %8:vreg_96 = TBUFFER_LOAD_FORMAT_XYZ_IDXEN_exact %4, %5:sgpr_128, 0, 8, 74, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) +... +--- + +name: gfx10_tbuffer_load_x_x_bothen_exact +body: | + bb.0.entry: + ; GFX10-LABEL: name: gfx10_tbuffer_load_x_x_bothen_exact + ; GFX10: [[COPY:%[0-9]+]]:sgpr_32 = COPY $sgpr0 + ; GFX10-NEXT: [[COPY1:%[0-9]+]]:sgpr_32 = COPY $sgpr1 + ; GFX10-NEXT: [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr2 + ; GFX10-NEXT: [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr3 + ; GFX10-NEXT: [[COPY4:%[0-9]+]]:vreg_64 = COPY $vgpr0 + ; GFX10-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3 + ; GFX10-NEXT: [[TBUFFER_LOAD_FORMAT_XY_BOTHEN_exact:%[0-9]+]]:vreg_64 = TBUFFER_LOAD_FORMAT_XY_BOTHEN_exact [[COPY4]], [[REG_SEQUENCE]], 0, 0, 64, 0, 0, implicit $exec :: (dereferenceable load (s64), align 1, addrspace 4) + ; GFX10-NEXT: [[COPY5:%[0-9]+]]:vgpr_32 = COPY [[TBUFFER_LOAD_FORMAT_XY_BOTHEN_exact]].sub0 + ; GFX10-NEXT: [[COPY6:%[0-9]+]]:vgpr_32 = COPY killed [[TBUFFER_LOAD_FORMAT_XY_BOTHEN_exact]].sub1 + %0:sgpr_32 = COPY $sgpr0 + %1:sgpr_32 = COPY $sgpr1 + %2:sgpr_32 = COPY $sgpr2 + %3:sgpr_32 = COPY $sgpr3 + %4:vreg_64 = COPY $vgpr0 + %5:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3 + %7:vgpr_32 = TBUFFER_LOAD_FORMAT_X_BOTHEN_exact %4, %5:sgpr_128, 0, 0, 22, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) + %8:vgpr_32 = TBUFFER_LOAD_FORMAT_X_BOTHEN_exact %4, %5:sgpr_128, 0, 4, 22, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) +... +--- + +name: gfx10_tbuffer_load_x_xy_bothen_exact +body: | + bb.0.entry: + ; GFX10-LABEL: name: gfx10_tbuffer_load_x_xy_bothen_exact + ; GFX10: [[COPY:%[0-9]+]]:sgpr_32 = COPY $sgpr0 + ; GFX10-NEXT: [[COPY1:%[0-9]+]]:sgpr_32 = COPY $sgpr1 + ; GFX10-NEXT: [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr2 + ; GFX10-NEXT: [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr3 + ; GFX10-NEXT: [[COPY4:%[0-9]+]]:vreg_64 = COPY $vgpr0 + ; GFX10-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3 + ; GFX10-NEXT: [[TBUFFER_LOAD_FORMAT_XYZ_BOTHEN_exact:%[0-9]+]]:vreg_96 = TBUFFER_LOAD_FORMAT_XYZ_BOTHEN_exact [[COPY4]], [[REG_SEQUENCE]], 0, 0, 74, 0, 0, implicit $exec :: (dereferenceable load (s64), align 1, addrspace 4) + ; GFX10-NEXT: [[COPY5:%[0-9]+]]:vgpr_32 = COPY [[TBUFFER_LOAD_FORMAT_XYZ_BOTHEN_exact]].sub0 + ; GFX10-NEXT: [[COPY6:%[0-9]+]]:vreg_64 = COPY killed [[TBUFFER_LOAD_FORMAT_XYZ_BOTHEN_exact]].sub1_sub2 + %0:sgpr_32 = COPY $sgpr0 + %1:sgpr_32 = COPY $sgpr1 + %2:sgpr_32 = COPY $sgpr2 + %3:sgpr_32 = COPY $sgpr3 + %4:vreg_64 = COPY $vgpr0 + %5:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3 + %7:vgpr_32 = TBUFFER_LOAD_FORMAT_X_BOTHEN_exact %4, %5:sgpr_128, 0, 0, 22, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) + %8:vreg_64 = TBUFFER_LOAD_FORMAT_XY_BOTHEN_exact %4, %5:sgpr_128, 0, 4, 64, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) +... +--- + +name: gfx10_tbuffer_load_xy_xy_bothen_exact +body: | + bb.0.entry: + ; GFX10-LABEL: name: gfx10_tbuffer_load_xy_xy_bothen_exact + ; GFX10: [[COPY:%[0-9]+]]:sgpr_32 = COPY $sgpr0 + ; GFX10-NEXT: [[COPY1:%[0-9]+]]:sgpr_32 = COPY $sgpr1 + ; GFX10-NEXT: [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr2 + ; GFX10-NEXT: [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr3 + ; GFX10-NEXT: [[COPY4:%[0-9]+]]:vreg_64 = COPY $vgpr0 + ; GFX10-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3 + ; GFX10-NEXT: [[TBUFFER_LOAD_FORMAT_XYZW_BOTHEN_exact:%[0-9]+]]:vreg_128 = TBUFFER_LOAD_FORMAT_XYZW_BOTHEN_exact [[COPY4]], [[REG_SEQUENCE]], 0, 0, 77, 0, 0, implicit $exec :: (dereferenceable load (s64), align 1, addrspace 4) + ; GFX10-NEXT: [[COPY5:%[0-9]+]]:vreg_64 = COPY [[TBUFFER_LOAD_FORMAT_XYZW_BOTHEN_exact]].sub0_sub1 + ; GFX10-NEXT: [[COPY6:%[0-9]+]]:vreg_64 = COPY killed [[TBUFFER_LOAD_FORMAT_XYZW_BOTHEN_exact]].sub2_sub3 + %0:sgpr_32 = COPY $sgpr0 + %1:sgpr_32 = COPY $sgpr1 + %2:sgpr_32 = COPY $sgpr2 + %3:sgpr_32 = COPY $sgpr3 + %4:vreg_64 = COPY $vgpr0 + %5:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3 + %7:vreg_64 = TBUFFER_LOAD_FORMAT_XY_BOTHEN_exact %4, %5:sgpr_128, 0, 0, 64, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) + %8:vreg_64 = TBUFFER_LOAD_FORMAT_XY_BOTHEN_exact %4, %5:sgpr_128, 0, 8, 64, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) +... +--- + +name: gfx10_tbuffer_load_x_xyz_bothen_exact +body: | + bb.0.entry: + ; GFX10-LABEL: name: gfx10_tbuffer_load_x_xyz_bothen_exact + ; GFX10: [[COPY:%[0-9]+]]:sgpr_32 = COPY $sgpr0 + ; GFX10-NEXT: [[COPY1:%[0-9]+]]:sgpr_32 = COPY $sgpr1 + ; GFX10-NEXT: [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr2 + ; GFX10-NEXT: [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr3 + ; GFX10-NEXT: [[COPY4:%[0-9]+]]:vreg_64 = COPY $vgpr0 + ; GFX10-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3 + ; GFX10-NEXT: [[TBUFFER_LOAD_FORMAT_XYZW_BOTHEN_exact:%[0-9]+]]:vreg_128 = TBUFFER_LOAD_FORMAT_XYZW_BOTHEN_exact [[COPY4]], [[REG_SEQUENCE]], 0, 0, 77, 0, 0, implicit $exec :: (dereferenceable load (s64), align 1, addrspace 4) + ; GFX10-NEXT: [[COPY5:%[0-9]+]]:vgpr_32 = COPY [[TBUFFER_LOAD_FORMAT_XYZW_BOTHEN_exact]].sub0 + ; GFX10-NEXT: [[COPY6:%[0-9]+]]:vreg_96 = COPY killed [[TBUFFER_LOAD_FORMAT_XYZW_BOTHEN_exact]].sub1_sub2_sub3 + %0:sgpr_32 = COPY $sgpr0 + %1:sgpr_32 = COPY $sgpr1 + %2:sgpr_32 = COPY $sgpr2 + %3:sgpr_32 = COPY $sgpr3 + %4:vreg_64 = COPY $vgpr0 + %5:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3 + %7:vgpr_32 = TBUFFER_LOAD_FORMAT_X_BOTHEN_exact %4, %5:sgpr_128, 0, 0, 22, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) + %8:vreg_96 = TBUFFER_LOAD_FORMAT_XYZ_BOTHEN_exact %4, %5:sgpr_128, 0, 4, 74, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) +... +--- + +name: gfx10_tbuffer_load_x_x_x_bothen_exact +body: | + bb.0.entry: + ; GFX10-LABEL: name: gfx10_tbuffer_load_x_x_x_bothen_exact + ; GFX10: [[COPY:%[0-9]+]]:sgpr_32 = COPY $sgpr0 + ; GFX10-NEXT: [[COPY1:%[0-9]+]]:sgpr_32 = COPY $sgpr1 + ; GFX10-NEXT: [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr2 + ; GFX10-NEXT: [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr3 + ; GFX10-NEXT: [[COPY4:%[0-9]+]]:vreg_64 = COPY $vgpr0 + ; GFX10-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3 + ; GFX10-NEXT: [[TBUFFER_LOAD_FORMAT_XYZ_BOTHEN_exact:%[0-9]+]]:vreg_96 = TBUFFER_LOAD_FORMAT_XYZ_BOTHEN_exact [[COPY4]], [[REG_SEQUENCE]], 0, 4, 74, 0, 0, implicit $exec :: (dereferenceable load (s96), align 1, addrspace 4) + ; GFX10-NEXT: [[COPY5:%[0-9]+]]:vreg_64 = COPY [[TBUFFER_LOAD_FORMAT_XYZ_BOTHEN_exact]].sub0_sub1 + ; GFX10-NEXT: [[COPY6:%[0-9]+]]:vgpr_32 = COPY killed [[TBUFFER_LOAD_FORMAT_XYZ_BOTHEN_exact]].sub2 + ; GFX10-NEXT: [[COPY7:%[0-9]+]]:vgpr_32 = COPY [[COPY5]].sub0 + ; GFX10-NEXT: [[COPY8:%[0-9]+]]:vgpr_32 = COPY killed [[COPY5]].sub1 + %0:sgpr_32 = COPY $sgpr0 + %1:sgpr_32 = COPY $sgpr1 + %2:sgpr_32 = COPY $sgpr2 + %3:sgpr_32 = COPY $sgpr3 + %4:vreg_64 = COPY $vgpr0 + %5:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3 + %7:vgpr_32 = TBUFFER_LOAD_FORMAT_X_BOTHEN_exact %4, %5:sgpr_128, 0, 4, 22, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) + %8:vgpr_32 = TBUFFER_LOAD_FORMAT_X_BOTHEN_exact %4, %5:sgpr_128, 0, 8, 22, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) + %9:vgpr_32 = TBUFFER_LOAD_FORMAT_X_BOTHEN_exact %4, %5:sgpr_128, 0, 12, 22, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) +... +--- + +name: gfx10_tbuffer_load_x_x_x_bothen_exact_swizzled_0 +body: | + bb.0.entry: + ; GFX10-LABEL: name: gfx10_tbuffer_load_x_x_x_bothen_exact_swizzled_0 + ; GFX10: [[COPY:%[0-9]+]]:sgpr_32 = COPY $sgpr0 + ; GFX10-NEXT: [[COPY1:%[0-9]+]]:sgpr_32 = COPY $sgpr1 + ; GFX10-NEXT: [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr2 + ; GFX10-NEXT: [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr3 + ; GFX10-NEXT: [[COPY4:%[0-9]+]]:vreg_64 = COPY $vgpr0 + ; GFX10-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3 + ; GFX10-NEXT: [[TBUFFER_LOAD_FORMAT_X_BOTHEN_exact:%[0-9]+]]:vgpr_32 = TBUFFER_LOAD_FORMAT_X_BOTHEN_exact [[COPY4]], [[REG_SEQUENCE]], 0, 4, 22, 0, 1, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) + ; GFX10-NEXT: [[TBUFFER_LOAD_FORMAT_XY_BOTHEN_exact:%[0-9]+]]:vreg_64 = TBUFFER_LOAD_FORMAT_XY_BOTHEN_exact [[COPY4]], [[REG_SEQUENCE]], 0, 8, 64, 0, 0, implicit $exec :: (dereferenceable load (s64), align 1, addrspace 4) + ; GFX10-NEXT: [[COPY5:%[0-9]+]]:vgpr_32 = COPY [[TBUFFER_LOAD_FORMAT_XY_BOTHEN_exact]].sub0 + ; GFX10-NEXT: [[COPY6:%[0-9]+]]:vgpr_32 = COPY killed [[TBUFFER_LOAD_FORMAT_XY_BOTHEN_exact]].sub1 + %0:sgpr_32 = COPY $sgpr0 + %1:sgpr_32 = COPY $sgpr1 + %2:sgpr_32 = COPY $sgpr2 + %3:sgpr_32 = COPY $sgpr3 + %4:vreg_64 = COPY $vgpr0 + %5:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3 + %7:vgpr_32 = TBUFFER_LOAD_FORMAT_X_BOTHEN_exact %4, %5:sgpr_128, 0, 4, 22, 0, 1, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) + %8:vgpr_32 = TBUFFER_LOAD_FORMAT_X_BOTHEN_exact %4, %5:sgpr_128, 0, 8, 22, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) + %9:vgpr_32 = TBUFFER_LOAD_FORMAT_X_BOTHEN_exact %4, %5:sgpr_128, 0, 12, 22, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) +... +--- + +name: gfx10_tbuffer_load_xy_xy_bothen_exact_diff_vaddr +body: | + bb.0.entry: + ; GFX10-LABEL: name: gfx10_tbuffer_load_xy_xy_bothen_exact_diff_vaddr + ; GFX10: [[COPY:%[0-9]+]]:sgpr_32 = COPY $sgpr0 + ; GFX10-NEXT: [[COPY1:%[0-9]+]]:sgpr_32 = COPY $sgpr1 + ; GFX10-NEXT: [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr2 + ; GFX10-NEXT: [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr3 + ; GFX10-NEXT: [[COPY4:%[0-9]+]]:vreg_64 = COPY $vgpr0 + ; GFX10-NEXT: [[COPY5:%[0-9]+]]:vreg_64 = COPY $vgpr1 + ; GFX10-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3 + ; GFX10-NEXT: [[TBUFFER_LOAD_FORMAT_XY_BOTHEN_exact:%[0-9]+]]:vreg_64 = TBUFFER_LOAD_FORMAT_XY_BOTHEN_exact [[COPY4]], [[REG_SEQUENCE]], 0, 4, 64, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) + ; GFX10-NEXT: [[TBUFFER_LOAD_FORMAT_XY_BOTHEN_exact1:%[0-9]+]]:vreg_64 = TBUFFER_LOAD_FORMAT_XY_BOTHEN_exact [[COPY5]], [[REG_SEQUENCE]], 0, 12, 64, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) + %0:sgpr_32 = COPY $sgpr0 + %1:sgpr_32 = COPY $sgpr1 + %2:sgpr_32 = COPY $sgpr2 + %3:sgpr_32 = COPY $sgpr3 + %4:vreg_64 = COPY $vgpr0 + %5:vreg_64 = COPY $vgpr1 + %6:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3 + %8:vreg_64 = TBUFFER_LOAD_FORMAT_XY_BOTHEN_exact %4, %6:sgpr_128, 0, 4, 64, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) + %9:vreg_64 = TBUFFER_LOAD_FORMAT_XY_BOTHEN_exact %5, %6:sgpr_128, 0, 12, 64, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) +... +--- + +name: gfx10_tbuffer_load_xy_xy_bothen_exact_diff_srsrc +body: | + bb.0.entry: + ; GFX10-LABEL: name: gfx10_tbuffer_load_xy_xy_bothen_exact_diff_srsrc + ; GFX10: [[COPY:%[0-9]+]]:sgpr_32 = COPY $sgpr0 + ; GFX10-NEXT: [[COPY1:%[0-9]+]]:sgpr_32 = COPY $sgpr1 + ; GFX10-NEXT: [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr2 + ; GFX10-NEXT: [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr3 + ; GFX10-NEXT: [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr4 + ; GFX10-NEXT: [[COPY5:%[0-9]+]]:vreg_64 = COPY $vgpr0 + ; GFX10-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3 + ; GFX10-NEXT: [[REG_SEQUENCE1:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY1]], %subreg.sub0, [[COPY2]], %subreg.sub1, [[COPY3]], %subreg.sub2, [[COPY4]], %subreg.sub3 + ; GFX10-NEXT: [[TBUFFER_LOAD_FORMAT_XY_BOTHEN_exact:%[0-9]+]]:vreg_64 = TBUFFER_LOAD_FORMAT_XY_BOTHEN_exact [[COPY5]], [[REG_SEQUENCE]], 0, 4, 64, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) + ; GFX10-NEXT: [[TBUFFER_LOAD_FORMAT_XY_BOTHEN_exact1:%[0-9]+]]:vreg_64 = TBUFFER_LOAD_FORMAT_XY_BOTHEN_exact [[COPY5]], [[REG_SEQUENCE1]], 0, 12, 64, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) + %0:sgpr_32 = COPY $sgpr0 + %1:sgpr_32 = COPY $sgpr1 + %2:sgpr_32 = COPY $sgpr2 + %3:sgpr_32 = COPY $sgpr3 + %4:sgpr_32 = COPY $sgpr4 + %5:vreg_64 = COPY $vgpr0 + %6:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3 + %7:sgpr_128 = REG_SEQUENCE %1:sgpr_32, %subreg.sub0, %2:sgpr_32, %subreg.sub1, %3:sgpr_32, %subreg.sub2, %4:sgpr_32, %subreg.sub3 + %8:vreg_64 = TBUFFER_LOAD_FORMAT_XY_BOTHEN_exact %5, %6:sgpr_128, 0, 4, 64, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) + %9:vreg_64 = TBUFFER_LOAD_FORMAT_XY_BOTHEN_exact %5, %7:sgpr_128, 0, 12, 64, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) +... +--- + +name: gfx10_tbuffer_load_xy_xy_idxen_exact_diff_vaddr +body: | + bb.0.entry: + ; GFX10-LABEL: name: gfx10_tbuffer_load_xy_xy_idxen_exact_diff_vaddr + ; GFX10: [[COPY:%[0-9]+]]:sgpr_32 = COPY $sgpr0 + ; GFX10-NEXT: [[COPY1:%[0-9]+]]:sgpr_32 = COPY $sgpr1 + ; GFX10-NEXT: [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr2 + ; GFX10-NEXT: [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr3 + ; GFX10-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY $vgpr0 + ; GFX10-NEXT: [[COPY5:%[0-9]+]]:vgpr_32 = COPY $vgpr1 + ; GFX10-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3 + ; GFX10-NEXT: [[TBUFFER_LOAD_FORMAT_XY_IDXEN_exact:%[0-9]+]]:vreg_64 = TBUFFER_LOAD_FORMAT_XY_IDXEN_exact [[COPY4]], [[REG_SEQUENCE]], 0, 4, 64, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) + ; GFX10-NEXT: [[TBUFFER_LOAD_FORMAT_XY_IDXEN_exact1:%[0-9]+]]:vreg_64 = TBUFFER_LOAD_FORMAT_XY_IDXEN_exact [[COPY5]], [[REG_SEQUENCE]], 0, 12, 64, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) + %0:sgpr_32 = COPY $sgpr0 + %1:sgpr_32 = COPY $sgpr1 + %2:sgpr_32 = COPY $sgpr2 + %3:sgpr_32 = COPY $sgpr3 + %4:vgpr_32 = COPY $vgpr0 + %5:vgpr_32 = COPY $vgpr1 + %6:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3 + %8:vreg_64 = TBUFFER_LOAD_FORMAT_XY_IDXEN_exact %4, %6:sgpr_128, 0, 4, 64, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) + %9:vreg_64 = TBUFFER_LOAD_FORMAT_XY_IDXEN_exact %5, %6:sgpr_128, 0, 12, 64, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) +... +--- + +name: gfx10_tbuffer_load_xy_xy_idxen_exact_diff_srsrc +body: | + bb.0.entry: + ; GFX10-LABEL: name: gfx10_tbuffer_load_xy_xy_idxen_exact_diff_srsrc + ; GFX10: [[COPY:%[0-9]+]]:sgpr_32 = COPY $sgpr0 + ; GFX10-NEXT: [[COPY1:%[0-9]+]]:sgpr_32 = COPY $sgpr1 + ; GFX10-NEXT: [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr2 + ; GFX10-NEXT: [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr3 + ; GFX10-NEXT: [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr4 + ; GFX10-NEXT: [[COPY5:%[0-9]+]]:vgpr_32 = COPY $vgpr0 + ; GFX10-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3 + ; GFX10-NEXT: [[REG_SEQUENCE1:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY1]], %subreg.sub0, [[COPY2]], %subreg.sub1, [[COPY3]], %subreg.sub2, [[COPY4]], %subreg.sub3 + ; GFX10-NEXT: [[TBUFFER_LOAD_FORMAT_XY_IDXEN_exact:%[0-9]+]]:vreg_64 = TBUFFER_LOAD_FORMAT_XY_IDXEN_exact [[COPY5]], [[REG_SEQUENCE]], 0, 4, 64, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) + ; GFX10-NEXT: [[TBUFFER_LOAD_FORMAT_XY_IDXEN_exact1:%[0-9]+]]:vreg_64 = TBUFFER_LOAD_FORMAT_XY_IDXEN_exact [[COPY5]], [[REG_SEQUENCE1]], 0, 12, 64, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) + %0:sgpr_32 = COPY $sgpr0 + %1:sgpr_32 = COPY $sgpr1 + %2:sgpr_32 = COPY $sgpr2 + %3:sgpr_32 = COPY $sgpr3 + %4:sgpr_32 = COPY $sgpr4 + %5:vgpr_32 = COPY $vgpr0 + %6:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3 + %7:sgpr_128 = REG_SEQUENCE %1:sgpr_32, %subreg.sub0, %2:sgpr_32, %subreg.sub1, %3:sgpr_32, %subreg.sub2, %4:sgpr_32, %subreg.sub3 + %8:vreg_64 = TBUFFER_LOAD_FORMAT_XY_IDXEN_exact %5, %6:sgpr_128, 0, 4, 64, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) + %9:vreg_64 = TBUFFER_LOAD_FORMAT_XY_IDXEN_exact %5, %7:sgpr_128, 0, 12, 64, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4) +... |