359a792f9b
Previously when combining two loads this pass would sink the first one down to the second one, putting the combined load where the second one was. It would also sink any intervening instructions which depended on the first load down to just after the combined load. For example, if we started with this sequence of instructions (code flowing from left to right): X A B C D E F Y After combining loads X and Y into XY we might end up with: A B C D E F XY But if B D and F depended on X, we would get: A C E XY B D F Now if the original code had some short disjoint live ranges from A to B, C to D and E to F, in the transformed code these live ranges will be long and overlapping. In this way a single merge of two loads could cause an unbounded increase in register pressure. To fix this, change the way the way that loads are moved in order to merge them so that: - The second load is moved up to the first one. (But when merging stores, we still move the first store down to the second one.) - Intervening instructions are never moved. - Instead, if we find an intervening instruction that would need to be moved, give up on the merge. But this case should now be pretty rare because normal stores have no outputs, and normal loads only have address register inputs, but these will be identical for any pair of loads that we try to merge. As well as fixing the unbounded register pressure increase problem, moving loads up and stores down seems like it should usually be a win for memory latency reasons. Differential Revision: https://reviews.llvm.org/D119006
1615 lines
106 KiB
YAML
1615 lines
106 KiB
YAML
# RUN: llc -march=amdgcn -mcpu=gfx900 -verify-machineinstrs -run-pass si-load-store-opt -o - %s | FileCheck -check-prefix=GFX9 %s
|
|
# RUN: llc -march=amdgcn -mcpu=gfx1010 -verify-machineinstrs -run-pass si-load-store-opt -o - %s | FileCheck -check-prefix=GFX10 %s
|
|
|
|
#
|
|
# GFX9 tests
|
|
#
|
|
|
|
# GFX9-LABEL: name: gfx9_tbuffer_load_x_xyz
|
|
# GFX9: %{{[0-9]+}}:vreg_128 = TBUFFER_LOAD_FORMAT_XYZW_OFFSET %4, 0, 4, 126, 0, 0, 0, implicit $exec :: (dereferenceable load (s128), align 1, addrspace 4)
|
|
# GFX9: %{{[0-9]+}}:vgpr_32 = COPY %7.sub0
|
|
# GFX9: %{{[0-9]+}}:vreg_96 = COPY killed %7.sub1_sub2_sub3
|
|
name: gfx9_tbuffer_load_x_xyz
|
|
body: |
|
|
bb.0.entry:
|
|
%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, 116, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
|
|
%8:vreg_96 = TBUFFER_LOAD_FORMAT_XYZ_OFFSET %5:sgpr_128, 0, 8, 125, 0, 0, 0, implicit $exec :: (dereferenceable load (s96), align 1, addrspace 4)
|
|
...
|
|
---
|
|
|
|
# GFX9-LABEL: name: gfx9_tbuffer_load_xyz_x
|
|
# GFX9: %{{[0-9]+}}:vreg_128 = TBUFFER_LOAD_FORMAT_XYZW_OFFSET %4, 0, 4, 126, 0, 0, 0, implicit $exec :: (dereferenceable load (s128), align 1, addrspace 4)
|
|
# GFX9: %{{[0-9]+}}:vreg_96 = COPY %7.sub0_sub1_sub2
|
|
# GFX9: %{{[0-9]+}}:vgpr_32 = COPY killed %7.sub3
|
|
name: gfx9_tbuffer_load_xyz_x
|
|
body: |
|
|
bb.0.entry:
|
|
%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, 125, 0, 0, 0, implicit $exec :: (dereferenceable load (s96), align 1, addrspace 4)
|
|
%8:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %5:sgpr_128, 0, 16, 116, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
|
|
...
|
|
---
|
|
|
|
# GFX9-LABEL: name: gfx9_tbuffer_load_xy_xy
|
|
# GFX9: %{{[0-9]+}}:vreg_128 = TBUFFER_LOAD_FORMAT_XYZW_OFFSET %4, 0, 4, 126, 0, 0, 0, implicit $exec :: (dereferenceable load (s128), align 1, addrspace 4)
|
|
# GFX9: %{{[0-9]+}}:vreg_64 = COPY %7.sub0_sub1
|
|
# GFX9: %{{[0-9]+}}:vreg_64 = COPY killed %7.sub2_sub3
|
|
name: gfx9_tbuffer_load_xy_xy
|
|
body: |
|
|
bb.0.entry:
|
|
%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, 123, 0, 0, 0, implicit $exec :: (dereferenceable load (s64), align 1, addrspace 4)
|
|
%8:vreg_64 = TBUFFER_LOAD_FORMAT_XY_OFFSET %5:sgpr_128, 0, 12, 123, 0, 0, 0, implicit $exec :: (dereferenceable load (s64), align 1, addrspace 4)
|
|
...
|
|
---
|
|
|
|
# GFX9-LABEL: name: gfx9_tbuffer_load_x_xy
|
|
# GFX9: %{{[0-9]+}}:vreg_96 = TBUFFER_LOAD_FORMAT_XYZ_OFFSET %4, 0, 4, 125, 0, 0, 0, implicit $exec :: (dereferenceable load (s96), align 1, addrspace 4)
|
|
# GFX9: %{{[0-9]+}}:vgpr_32 = COPY %7.sub0
|
|
# GFX9: %{{[0-9]+}}:vreg_64 = COPY killed %7.sub1_sub2
|
|
name: gfx9_tbuffer_load_x_xy
|
|
body: |
|
|
bb.0.entry:
|
|
%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, 116, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
|
|
%8:vreg_64 = TBUFFER_LOAD_FORMAT_XY_OFFSET %5:sgpr_128, 0, 8, 123, 0, 0, 0, implicit $exec :: (dereferenceable load (s64), align 1, addrspace 4)
|
|
...
|
|
---
|
|
|
|
# GFX9-LABEL: name: gfx9_tbuffer_load_xy_x
|
|
# GFX9: %{{[0-9]+}}:vreg_96 = TBUFFER_LOAD_FORMAT_XYZ_OFFSET %4, 0, 4, 125, 0, 0, 0, implicit $exec :: (dereferenceable load (s96), align 1, addrspace 4)
|
|
# GFX9: %{{[0-9]+}}:vreg_64 = COPY %7.sub0_sub1
|
|
# GFX9: %{{[0-9]+}}:vgpr_32 = COPY killed %7.sub2
|
|
name: gfx9_tbuffer_load_xy_x
|
|
body: |
|
|
bb.0.entry:
|
|
%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, 123, 0, 0, 0, implicit $exec :: (dereferenceable load (s64), align 1, addrspace 4)
|
|
%8:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %5:sgpr_128, 0, 12, 116, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
|
|
...
|
|
---
|
|
|
|
|
|
# GFX9-LABEL: name: gfx9_tbuffer_load_x_x
|
|
# GFX9: %{{[0-9]+}}:vreg_64 = TBUFFER_LOAD_FORMAT_XY_OFFSET %4, 0, 4, 123, 0, 0, 0, implicit $exec :: (dereferenceable load (s64), align 1, addrspace 4)
|
|
# GFX9: %{{[0-9]+}}:vgpr_32 = COPY %7.sub0
|
|
# GFX9: %{{[0-9]+}}:vgpr_32 = COPY killed %7.sub1
|
|
|
|
name: gfx9_tbuffer_load_x_x
|
|
body: |
|
|
bb.0.entry:
|
|
%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, 116, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
|
|
%8:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %5:sgpr_128, 0, 8, 116, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
|
|
...
|
|
---
|
|
|
|
# GFX9-LABEL: name: gfx9_tbuffer_load_x_x_format_32_32_32_32
|
|
# GFX9: %{{[0-9]+}}:vreg_64 = TBUFFER_LOAD_FORMAT_XY_OFFSET %4, 0, 4, 123, 0, 0, 0, implicit $exec :: (dereferenceable load (s64), align 1, addrspace 4)
|
|
# GFX9: %{{[0-9]+}}:vgpr_32 = COPY %7.sub0
|
|
# GFX9: %{{[0-9]+}}:vgpr_32 = COPY killed %7.sub1
|
|
|
|
name: gfx9_tbuffer_load_x_x_format_32_32_32_32
|
|
body: |
|
|
bb.0.entry:
|
|
%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, 126, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
|
|
%8:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %5:sgpr_128, 0, 8, 126, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
|
|
...
|
|
---
|
|
|
|
|
|
# GFX9-LABEL: name: gfx9_tbuffer_load_float_32
|
|
# GFX9: %{{[0-9]+}}:vreg_64 = TBUFFER_LOAD_FORMAT_XY_OFFSET %4, 0, 4, 123, 0, 0, 0, implicit $exec :: (dereferenceable load (s64), align 1, addrspace 4)
|
|
# GFX9: %{{[0-9]+}}:vgpr_32 = COPY %14.sub0
|
|
# GFX9: %{{[0-9]+}}:vgpr_32 = COPY killed %14.sub1
|
|
# GFX9: %{{[0-9]+}}:vreg_128 = TBUFFER_LOAD_FORMAT_XYZW_OFFSET %4, 0, 16, 126, 0, 0, 0, implicit $exec :: (dereferenceable load (s128), align 1, addrspace 4)
|
|
# GFX9: %{{[0-9]+}}:vreg_96 = COPY %17.sub0_sub1_sub2
|
|
# GFX9: %{{[0-9]+}}:vgpr_32 = COPY killed %17.sub3
|
|
# GFX9: %{{[0-9]+}}:vreg_64 = COPY %16.sub0_sub1
|
|
# GFX9: %{{[0-9]+}}:vgpr_32 = COPY killed %16.sub2
|
|
# GFX9: %{{[0-9]+}}:vgpr_32 = COPY %15.sub0
|
|
# GFX9: %{{[0-9]+}}:vgpr_32 = COPY killed %15.sub1
|
|
# GFX9: %{{[0-9]+}}:vreg_96 = TBUFFER_LOAD_FORMAT_XYZ_OFFSET %4, 0, 36, 125, 0, 0, 0, implicit $exec :: (dereferenceable load (s96), align 1, addrspace 4)
|
|
# GFX9: %{{[0-9]+}}:vreg_64 = COPY %19.sub0_sub1
|
|
# GFX9: %{{[0-9]+}}:vgpr_32 = COPY killed %19.sub2
|
|
# GFX9: %{{[0-9]+}}:vgpr_32 = COPY %18.sub0
|
|
# GFX9: %{{[0-9]+}}:vgpr_32 = COPY killed %18.sub1
|
|
|
|
name: gfx9_tbuffer_load_float_32
|
|
body: |
|
|
bb.0.entry:
|
|
%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, 116, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
|
|
%8:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %5:sgpr_128, 0, 8, 116, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
|
|
%9:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %5:sgpr_128, 0, 16, 116, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
|
|
%10:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %5:sgpr_128, 0, 20, 116, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
|
|
%11:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %5:sgpr_128, 0, 24, 116, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
|
|
%12:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %5:sgpr_128, 0, 28, 116, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
|
|
%13:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %5:sgpr_128, 0, 36, 116, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
|
|
%14:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %5:sgpr_128, 0, 40, 116, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
|
|
%15:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %5:sgpr_128, 0, 44, 116, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
|
|
...
|
|
---
|
|
|
|
# GFX9-LABEL: name: gfx9_tbuffer_load_sint_32
|
|
# GFX9: %{{[0-9]+}}:vreg_64 = TBUFFER_LOAD_FORMAT_XY_OFFSET %4, 0, 4, 91, 0, 0, 0, implicit $exec :: (dereferenceable load (s64), align 1, addrspace 4)
|
|
# GFX9: %{{[0-9]+}}:vgpr_32 = COPY %14.sub0
|
|
# GFX9: %{{[0-9]+}}:vgpr_32 = COPY killed %14.sub1
|
|
# GFX9: %{{[0-9]+}}:vreg_128 = TBUFFER_LOAD_FORMAT_XYZW_OFFSET %4, 0, 16, 94, 0, 0, 0, implicit $exec :: (dereferenceable load (s128), align 1, addrspace 4)
|
|
# GFX9: %{{[0-9]+}}:vreg_96 = COPY %17.sub0_sub1_sub2
|
|
# GFX9: %{{[0-9]+}}:vgpr_32 = COPY killed %17.sub3
|
|
# GFX9: %{{[0-9]+}}:vreg_64 = COPY %16.sub0_sub1
|
|
# GFX9: %{{[0-9]+}}:vgpr_32 = COPY killed %16.sub2
|
|
# GFX9: %{{[0-9]+}}:vgpr_32 = COPY %15.sub0
|
|
# GFX9: %{{[0-9]+}}:vgpr_32 = COPY killed %15.sub1
|
|
# GFX9: %{{[0-9]+}}:vreg_96 = TBUFFER_LOAD_FORMAT_XYZ_OFFSET %4, 0, 36, 93, 0, 0, 0, implicit $exec :: (dereferenceable load (s96), align 1, addrspace 4)
|
|
# GFX9: %{{[0-9]+}}:vreg_64 = COPY %19.sub0_sub1
|
|
# GFX9: %{{[0-9]+}}:vgpr_32 = COPY killed %19.sub2
|
|
# GFX9: %{{[0-9]+}}:vgpr_32 = COPY %18.sub0
|
|
# GFX9: %{{[0-9]+}}:vgpr_32 = COPY killed %18.sub1
|
|
|
|
name: gfx9_tbuffer_load_sint_32
|
|
body: |
|
|
bb.0.entry:
|
|
%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, 84, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
|
|
%8:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %5:sgpr_128, 0, 8, 84, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
|
|
%9:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %5:sgpr_128, 0, 16, 84, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
|
|
%10:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %5:sgpr_128, 0, 20, 84, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
|
|
%11:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %5:sgpr_128, 0, 24, 84, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
|
|
%12:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %5:sgpr_128, 0, 28, 84, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
|
|
%13:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %5:sgpr_128, 0, 36, 84, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
|
|
%14:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %5:sgpr_128, 0, 40, 84, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
|
|
%15:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %5:sgpr_128, 0, 44, 84, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
|
|
...
|
|
---
|
|
|
|
# GFX9-LABEL: name: gfx9_tbuffer_load_uint_32
|
|
# GFX9: %{{[0-9]+}}:vreg_64 = TBUFFER_LOAD_FORMAT_XY_OFFSET %4, 0, 4, 75, 0, 0, 0, implicit $exec :: (dereferenceable load (s64), align 1, addrspace 4)
|
|
# GFX9: %{{[0-9]+}}:vgpr_32 = COPY %14.sub0
|
|
# GFX9: %{{[0-9]+}}:vgpr_32 = COPY killed %14.sub1
|
|
# GFX9: %{{[0-9]+}}:vreg_128 = TBUFFER_LOAD_FORMAT_XYZW_OFFSET %4, 0, 16, 78, 0, 0, 0, implicit $exec :: (dereferenceable load (s128), align 1, addrspace 4)
|
|
# GFX9: %{{[0-9]+}}:vreg_96 = COPY %17.sub0_sub1_sub2
|
|
# GFX9: %{{[0-9]+}}:vgpr_32 = COPY killed %17.sub3
|
|
# GFX9: %{{[0-9]+}}:vreg_64 = COPY %16.sub0_sub1
|
|
# GFX9: %{{[0-9]+}}:vgpr_32 = COPY killed %16.sub2
|
|
# GFX9: %{{[0-9]+}}:vgpr_32 = COPY %15.sub0
|
|
# GFX9: %{{[0-9]+}}:vgpr_32 = COPY killed %15.sub1
|
|
# GFX9: %{{[0-9]+}}:vreg_96 = TBUFFER_LOAD_FORMAT_XYZ_OFFSET %4, 0, 36, 77, 0, 0, 0, implicit $exec :: (dereferenceable load (s96), align 1, addrspace 4)
|
|
# GFX9: %{{[0-9]+}}:vreg_64 = COPY %19.sub0_sub1
|
|
# GFX9: %{{[0-9]+}}:vgpr_32 = COPY killed %19.sub2
|
|
# GFX9: %{{[0-9]+}}:vgpr_32 = COPY %18.sub0
|
|
# GFX9: %{{[0-9]+}}:vgpr_32 = COPY killed %18.sub1
|
|
|
|
name: gfx9_tbuffer_load_uint_32
|
|
body: |
|
|
bb.0.entry:
|
|
%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, 68, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
|
|
%8:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %5:sgpr_128, 0, 8, 68, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
|
|
%9:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %5:sgpr_128, 0, 16, 68, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
|
|
%10:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %5:sgpr_128, 0, 20, 68, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
|
|
%11:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %5:sgpr_128, 0, 24, 68, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
|
|
%12:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %5:sgpr_128, 0, 28, 68, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
|
|
%13:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %5:sgpr_128, 0, 36, 68, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
|
|
%14:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %5:sgpr_128, 0, 40, 68, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
|
|
%15:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %5:sgpr_128, 0, 44, 68, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
|
|
...
|
|
---
|
|
|
|
# GFX9-LABEL: name: gfx9_tbuffer_load_not_merged_data_format_mismatch
|
|
# GFX9: %{{[0-9]+}}:sgpr_32 = COPY $sgpr0
|
|
# GFX9: %{{[0-9]+}}:sgpr_32 = COPY $sgpr1
|
|
# GFX9: %{{[0-9]+}}:sgpr_32 = COPY $sgpr2
|
|
# GFX9: %{{[0-9]+}}:sgpr_32 = COPY $sgpr3
|
|
# GFX9: %{{[0-9]+}}:sgpr_128 = REG_SEQUENCE %0, %subreg.sub0, %1, %subreg.sub1, %2, %subreg.sub2, %3, %subreg.sub3
|
|
# GFX9: %{{[0-9]+}}:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %4, 0, 4, 116, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
|
|
# GFX9: %{{[0-9]+}}:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %4, 0, 8, 114, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
|
|
# GFX9: %{{[0-9]+}}:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %4, 0, 16, 116, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
|
|
# GFX9: %{{[0-9]+}}:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %4, 0, 20, 114, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
|
|
# GFX9: %{{[0-9]+}}:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %4, 0, 24, 116, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
|
|
# GFX9: %{{[0-9]+}}:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %4, 0, 28, 114, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
|
|
# GFX9: %{{[0-9]+}}:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %4, 0, 36, 116, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
|
|
# GFX9: %{{[0-9]+}}:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %4, 0, 40, 114, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
|
|
# GFX9: %{{[0-9]+}}:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %4, 0, 44, 116, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
|
|
|
|
|
|
name: gfx9_tbuffer_load_not_merged_data_format_mismatch
|
|
body: |
|
|
bb.0.entry:
|
|
%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, 116, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
|
|
%8:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %5:sgpr_128, 0, 8, 114, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
|
|
%9:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %5:sgpr_128, 0, 16, 116, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
|
|
%10:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %5:sgpr_128, 0, 20, 114, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
|
|
%11:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %5:sgpr_128, 0, 24, 116, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
|
|
%12:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %5:sgpr_128, 0, 28, 114, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
|
|
%13:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %5:sgpr_128, 0, 36, 116, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
|
|
%14:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %5:sgpr_128, 0, 40, 114, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
|
|
%15:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %5:sgpr_128, 0, 44, 116, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
|
|
...
|
|
---
|
|
|
|
# GFX9-LABEL: name: gfx9_tbuffer_load_not_merged_num_format_mismatch
|
|
# GFX9: %{{[0-9]+}}:sgpr_32 = COPY $sgpr0
|
|
# GFX9: %{{[0-9]+}}:sgpr_32 = COPY $sgpr1
|
|
# GFX9: %{{[0-9]+}}:sgpr_32 = COPY $sgpr2
|
|
# GFX9: %{{[0-9]+}}:sgpr_32 = COPY $sgpr3
|
|
# GFX9: %{{[0-9]+}}:sgpr_128 = REG_SEQUENCE %0, %subreg.sub0, %1, %subreg.sub1, %2, %subreg.sub2, %3, %subreg.sub3
|
|
# GFX9: %{{[0-9]+}}:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %4, 0, 4, 116, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
|
|
# GFX9: %{{[0-9]+}}:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %4, 0, 8, 84, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
|
|
# GFX9: %{{[0-9]+}}:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %4, 0, 16, 116, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
|
|
# GFX9: %{{[0-9]+}}:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %4, 0, 20, 84, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
|
|
# GFX9: %{{[0-9]+}}:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %4, 0, 24, 116, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
|
|
# GFX9: %{{[0-9]+}}:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %4, 0, 28, 84, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
|
|
# GFX9: %{{[0-9]+}}:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %4, 0, 36, 116, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
|
|
# GFX9: %{{[0-9]+}}:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %4, 0, 40, 84, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
|
|
# GFX9: %{{[0-9]+}}:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %4, 0, 44, 116, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
|
|
name: gfx9_tbuffer_load_not_merged_num_format_mismatch
|
|
body: |
|
|
bb.0.entry:
|
|
%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, 116, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
|
|
%8:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %5:sgpr_128, 0, 8, 84, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
|
|
%9:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %5:sgpr_128, 0, 16, 116, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
|
|
%10:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %5:sgpr_128, 0, 20, 84, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
|
|
%11:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %5:sgpr_128, 0, 24, 116, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
|
|
%12:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %5:sgpr_128, 0, 28, 84, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
|
|
%13:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %5:sgpr_128, 0, 36, 116, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
|
|
%14:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %5:sgpr_128, 0, 40, 84, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
|
|
%15:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %5:sgpr_128, 0, 44, 116, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
|
|
...
|
|
---
|
|
|
|
# GFX9-LABEL: name: gfx9_tbuffer_store_x_xyz
|
|
# GFX9: %{{[0-9]+}}:vreg_96 = REG_SEQUENCE %3, %subreg.sub0, %2, %subreg.sub1, %1, %subreg.sub2
|
|
# GFX9: %{{[0-9]+}}:vreg_128 = REG_SEQUENCE %0, %subreg.sub0, %9, %subreg.sub1_sub2_sub3
|
|
# GFX9: TBUFFER_STORE_FORMAT_XYZW_OFFSET_exact killed %10, %8, 0, 4, 126, 0, 0, 0, implicit $exec :: (dereferenceable store (s128), align 1, addrspace 4)
|
|
name: gfx9_tbuffer_store_x_xyz
|
|
body: |
|
|
bb.0.entry:
|
|
liveins: $sgpr0, $sgpr1, $sgpr2, $sgpr3, $vgpr0, $vgpr1, $vgpr2, $vgpr3
|
|
%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, 116, 0, 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, 125, 0, 0, 0, implicit $exec :: (dereferenceable store (s96), align 1, addrspace 4)
|
|
...
|
|
---
|
|
|
|
|
|
|
|
# GFX9-LABEL: name: gfx9_tbuffer_store_xyz_x
|
|
# GFX9: %{{[0-9]+}}:vreg_96 = REG_SEQUENCE %3, %subreg.sub0, %2, %subreg.sub1, %1, %subreg.sub2
|
|
# GFX9: %{{[0-9]+}}:vreg_128 = REG_SEQUENCE %9, %subreg.sub0_sub1_sub2, %0, %subreg.sub3
|
|
# GFX9: TBUFFER_STORE_FORMAT_XYZW_OFFSET_exact killed %10, %8, 0, 4, 126, 0, 0, 0, implicit $exec :: (dereferenceable store (s128), align 1, addrspace 4)
|
|
name: gfx9_tbuffer_store_xyz_x
|
|
body: |
|
|
bb.0.entry:
|
|
liveins: $sgpr0, $sgpr1, $sgpr2, $sgpr3, $vgpr0, $vgpr1, $vgpr2, $vgpr3
|
|
%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, 125, 0, 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, 116, 0, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4)
|
|
...
|
|
---
|
|
|
|
# GFX9-LABEL: name: gfx9_tbuffer_store_xy_xy
|
|
# GFX9: %{{[0-9]+}}:vreg_64 = REG_SEQUENCE %3, %subreg.sub0, %2, %subreg.sub1
|
|
# GFX9: %{{[0-9]+}}:vreg_64 = REG_SEQUENCE %1, %subreg.sub0, %0, %subreg.sub1
|
|
# GFX9: %{{[0-9]+}}:vreg_128 = REG_SEQUENCE %9, %subreg.sub0_sub1, %10, %subreg.sub2_sub3
|
|
# GFX9: TBUFFER_STORE_FORMAT_XYZW_OFFSET_exact killed %11, %8, 0, 4, 126, 0, 0, 0, implicit $exec :: (dereferenceable store (s128), align 1, addrspace 4)
|
|
name: gfx9_tbuffer_store_xy_xy
|
|
body: |
|
|
bb.0.entry:
|
|
liveins: $sgpr0, $sgpr1, $sgpr2, $sgpr3, $vgpr0, $vgpr1, $vgpr2, $vgpr3
|
|
%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, 123, 0, 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, 123, 0, 0, 0, implicit $exec :: (dereferenceable store (s64), align 1, addrspace 4)
|
|
...
|
|
---
|
|
|
|
# GFX9-LABEL: name: gfx9_tbuffer_store_x_xy
|
|
# GFX9: %{{[0-9]+}}:vreg_64 = REG_SEQUENCE %3, %subreg.sub0, %2, %subreg.sub1
|
|
# GFX9: %{{[0-9]+}}:vreg_64, %subreg.sub1_sub2
|
|
# GFX9: TBUFFER_STORE_FORMAT_XYZ_OFFSET_exact killed %11, %8, 0, 4, 125, 0, 0, 0, implicit $exec :: (dereferenceable store (s96), align 1, addrspace 4)
|
|
name: gfx9_tbuffer_store_x_xy
|
|
body: |
|
|
bb.0.entry:
|
|
liveins: $sgpr0, $sgpr1, $sgpr2, $sgpr3, $vgpr0, $vgpr1, $vgpr2, $vgpr3
|
|
%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, 116, 0, 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, 123, 0, 0, 0, implicit $exec :: (dereferenceable store (s64), align 1, addrspace 4)
|
|
...
|
|
---
|
|
|
|
# GFX9-LABEL: name: gfx9_tbuffer_store_xy_x
|
|
# GFX9: %{{[0-9]+}}:vreg_64 = REG_SEQUENCE %3, %subreg.sub0, %2, %subreg.sub1
|
|
# GFX9: %{{[0-9]+}}:vreg_96 = REG_SEQUENCE %9, %subreg.sub0_sub1, %0, %subreg.sub2
|
|
# GFX9: TBUFFER_STORE_FORMAT_XYZ_OFFSET_exact killed %10, %8, 0, 4, 125, 0, 0, 0, implicit $exec :: (dereferenceable store (s96), align 1, addrspace 4)
|
|
name: gfx9_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
|
|
%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, 123, 0, 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, 116, 0, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4)
|
|
...
|
|
---
|
|
|
|
|
|
# GFX9-LABEL: name: gfx9_tbuffer_store_x_x
|
|
# GFX9: %{{[0-9]+}}:vreg_64 = REG_SEQUENCE %1, %subreg.sub0, %0, %subreg.sub1
|
|
# GFX9: TBUFFER_STORE_FORMAT_XY_OFFSET_exact killed %9, %8, 0, 4, 123, 0, 0, 0, implicit $exec :: (dereferenceable store (s64), align 1, addrspace 4)
|
|
name: gfx9_tbuffer_store_x_x
|
|
body: |
|
|
bb.0.entry:
|
|
liveins: $sgpr0, $sgpr1, $sgpr2, $sgpr3, $vgpr0, $vgpr1, $vgpr2, $vgpr3
|
|
%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, 116, 0, 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, 116, 0, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4)
|
|
...
|
|
---
|
|
|
|
# GFX9-LABEL: name: gfx9_tbuffer_store_x_x_format_32_32_32_32
|
|
# GFX9: %{{[0-9]+}}:vreg_64 = REG_SEQUENCE %1, %subreg.sub0, %0, %subreg.sub1
|
|
# GFX9: TBUFFER_STORE_FORMAT_XY_OFFSET_exact killed %9, %8, 0, 4, 123, 0, 0, 0, implicit $exec :: (dereferenceable store (s64), align 1, addrspace 4)
|
|
name: gfx9_tbuffer_store_x_x_format_32_32_32_32
|
|
body: |
|
|
bb.0.entry:
|
|
liveins: $sgpr0, $sgpr1, $sgpr2, $sgpr3, $vgpr0, $vgpr1, $vgpr2, $vgpr3
|
|
%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, 126, 0, 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, 126, 0, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4)
|
|
...
|
|
---
|
|
|
|
# GFX9-LABEL: name: gfx9_tbuffer_store_float32
|
|
# GFX9: %{{[0-9]+}}:vgpr_32 = COPY $vgpr8
|
|
# GFX9: %{{[0-9]+}}:vgpr_32 = COPY $vgpr7
|
|
# GFX9: %{{[0-9]+}}:vgpr_32 = COPY $vgpr6
|
|
# GFX9: %{{[0-9]+}}:vgpr_32 = COPY $vgpr5
|
|
# GFX9: %{{[0-9]+}}:vgpr_32 = COPY $vgpr4
|
|
# GFX9: %{{[0-9]+}}:vgpr_32 = COPY $vgpr3
|
|
# GFX9: %{{[0-9]+}}:vgpr_32 = COPY $vgpr2
|
|
# GFX9: %{{[0-9]+}}:vgpr_32 = COPY $vgpr1
|
|
# GFX9: %{{[0-9]+}}:vgpr_32 = COPY $vgpr0
|
|
# GFX9: %{{[0-9]+}}:sgpr_32 = COPY $sgpr3
|
|
# GFX9: %{{[0-9]+}}:sgpr_32 = COPY $sgpr2
|
|
# GFX9: %{{[0-9]+}}:sgpr_32 = COPY $sgpr1
|
|
# GFX9: %{{[0-9]+}}:sgpr_32 = COPY $sgpr0
|
|
# GFX9: %{{[0-9]+}}:sgpr_128 = REG_SEQUENCE %12, %subreg.sub0, %11, %subreg.sub1, %10, %subreg.sub2, %9, %subreg.sub3
|
|
# GFX9: %{{[0-9]+}}:vreg_64 = REG_SEQUENCE %8, %subreg.sub0, %7, %subreg.sub1
|
|
# GFX9: TBUFFER_STORE_FORMAT_XY_OFFSET_exact killed %14, %13, 0, 4, 123, 0, 0, 0, implicit $exec :: (dereferenceable store (s64), align 1, addrspace 4)
|
|
# GFX9: %{{[0-9]+}}:vreg_64 = REG_SEQUENCE %6, %subreg.sub0, %5, %subreg.sub1
|
|
# GFX9: %{{[0-9]+}}:vreg_96 = REG_SEQUENCE killed %15, %subreg.sub0_sub1, %4, %subreg.sub2
|
|
# GFX9: %{{[0-9]+}}:vreg_128 = REG_SEQUENCE killed %16, %subreg.sub0_sub1_sub2, %3, %subreg.sub3
|
|
# GFX9: TBUFFER_STORE_FORMAT_XYZW_OFFSET_exact killed %17, %13, 0, 16, 126, 0, 0, 0, implicit $exec :: (dereferenceable store (s128), align 1, addrspace 4)
|
|
# GFX9: %{{[0-9]+}}:vreg_64 = REG_SEQUENCE %2, %subreg.sub0, %1, %subreg.sub1
|
|
# GFX9: %{{[0-9]+}}:vreg_96 = REG_SEQUENCE killed %18, %subreg.sub0_sub1, %0, %subreg.sub2
|
|
# GFX9: TBUFFER_STORE_FORMAT_XYZ_OFFSET_exact killed %19, %13, 0, 36, 125, 0, 0, 0, implicit $exec :: (dereferenceable store (s96), align 1, addrspace 4)
|
|
name: gfx9_tbuffer_store_float32
|
|
body: |
|
|
bb.0.entry:
|
|
liveins: $sgpr0, $sgpr1, $sgpr2, $sgpr3, $vgpr0, $vgpr1, $vgpr2, $vgpr3, $vgpr4, $vgpr5, $vgpr6, $vgpr7, $vgpr8
|
|
%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, 116, 0, 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, 116, 0, 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, 116, 0, 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, 116, 0, 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, 116, 0, 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, 116, 0, 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, 116, 0, 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, 116, 0, 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, 116, 0, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4)
|
|
...
|
|
---
|
|
|
|
# GFX9-LABEL: name: gfx9_tbuffer_store_sint32
|
|
# GFX9: %{{[0-9]+}}:vgpr_32 = COPY $vgpr8
|
|
# GFX9: %{{[0-9]+}}:vgpr_32 = COPY $vgpr7
|
|
# GFX9: %{{[0-9]+}}:vgpr_32 = COPY $vgpr6
|
|
# GFX9: %{{[0-9]+}}:vgpr_32 = COPY $vgpr5
|
|
# GFX9: %{{[0-9]+}}:vgpr_32 = COPY $vgpr4
|
|
# GFX9: %{{[0-9]+}}:vgpr_32 = COPY $vgpr3
|
|
# GFX9: %{{[0-9]+}}:vgpr_32 = COPY $vgpr2
|
|
# GFX9: %{{[0-9]+}}:vgpr_32 = COPY $vgpr1
|
|
# GFX9: %{{[0-9]+}}:vgpr_32 = COPY $vgpr0
|
|
# GFX9: %{{[0-9]+}}:sgpr_32 = COPY $sgpr3
|
|
# GFX9: %{{[0-9]+}}:sgpr_32 = COPY $sgpr2
|
|
# GFX9: %{{[0-9]+}}:sgpr_32 = COPY $sgpr1
|
|
# GFX9: %{{[0-9]+}}:sgpr_32 = COPY $sgpr0
|
|
# GFX9: %{{[0-9]+}}:sgpr_128 = REG_SEQUENCE %12, %subreg.sub0, %11, %subreg.sub1, %10, %subreg.sub2, %9, %subreg.sub3
|
|
# GFX9: %{{[0-9]+}}:vreg_64 = REG_SEQUENCE %8, %subreg.sub0, %7, %subreg.sub1
|
|
# GFX9: TBUFFER_STORE_FORMAT_XY_OFFSET_exact killed %14, %13, 0, 4, 91, 0, 0, 0, implicit $exec :: (dereferenceable store (s64), align 1, addrspace 4)
|
|
# GFX9: %{{[0-9]+}}:vreg_64 = REG_SEQUENCE %6, %subreg.sub0, %5, %subreg.sub1
|
|
# GFX9: %{{[0-9]+}}:vreg_96 = REG_SEQUENCE killed %15, %subreg.sub0_sub1, %4, %subreg.sub2
|
|
# GFX9: %{{[0-9]+}}:vreg_128 = REG_SEQUENCE killed %16, %subreg.sub0_sub1_sub2, %3, %subreg.sub3
|
|
# GFX9: TBUFFER_STORE_FORMAT_XYZW_OFFSET_exact killed %17, %13, 0, 16, 94, 0, 0, 0, implicit $exec :: (dereferenceable store (s128), align 1, addrspace 4)
|
|
# GFX9: %{{[0-9]+}}:vreg_64 = REG_SEQUENCE %2, %subreg.sub0, %1, %subreg.sub1
|
|
# GFX9: %{{[0-9]+}}:vreg_96 = REG_SEQUENCE killed %18, %subreg.sub0_sub1, %0, %subreg.sub2
|
|
# GFX9: TBUFFER_STORE_FORMAT_XYZ_OFFSET_exact killed %19, %13, 0, 36, 93, 0, 0, 0, implicit $exec :: (dereferenceable store (s96), align 1, addrspace 4)
|
|
name: gfx9_tbuffer_store_sint32
|
|
body: |
|
|
bb.0.entry:
|
|
liveins: $sgpr0, $sgpr1, $sgpr2, $sgpr3, $vgpr0, $vgpr1, $vgpr2, $vgpr3, $vgpr4, $vgpr5, $vgpr6, $vgpr7, $vgpr8
|
|
%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, 84, 0, 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, 84, 0, 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, 84, 0, 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, 84, 0, 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, 84, 0, 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, 84, 0, 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, 84, 0, 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, 84, 0, 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, 84, 0, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4)
|
|
...
|
|
---
|
|
|
|
# GFX9-LABEL: name: gfx9_tbuffer_store_uint32
|
|
# GFX9: %{{[0-9]+}}:vgpr_32 = COPY $vgpr8
|
|
# GFX9: %{{[0-9]+}}:vgpr_32 = COPY $vgpr7
|
|
# GFX9: %{{[0-9]+}}:vgpr_32 = COPY $vgpr6
|
|
# GFX9: %{{[0-9]+}}:vgpr_32 = COPY $vgpr5
|
|
# GFX9: %{{[0-9]+}}:vgpr_32 = COPY $vgpr4
|
|
# GFX9: %{{[0-9]+}}:vgpr_32 = COPY $vgpr3
|
|
# GFX9: %{{[0-9]+}}:vgpr_32 = COPY $vgpr2
|
|
# GFX9: %{{[0-9]+}}:vgpr_32 = COPY $vgpr1
|
|
# GFX9: %{{[0-9]+}}:vgpr_32 = COPY $vgpr0
|
|
# GFX9: %{{[0-9]+}}:sgpr_32 = COPY $sgpr3
|
|
# GFX9: %{{[0-9]+}}:sgpr_32 = COPY $sgpr2
|
|
# GFX9: %{{[0-9]+}}:sgpr_32 = COPY $sgpr1
|
|
# GFX9: %{{[0-9]+}}:sgpr_32 = COPY $sgpr0
|
|
# GFX9: %{{[0-9]+}}:sgpr_128 = REG_SEQUENCE %12, %subreg.sub0, %11, %subreg.sub1, %10, %subreg.sub2, %9, %subreg.sub3
|
|
# GFX9: %{{[0-9]+}}:vreg_64 = REG_SEQUENCE %8, %subreg.sub0, %7, %subreg.sub1
|
|
# GFX9: TBUFFER_STORE_FORMAT_XY_OFFSET_exact killed %14, %13, 0, 4, 75, 0, 0, 0, implicit $exec :: (dereferenceable store (s64), align 1, addrspace 4)
|
|
# GFX9: %{{[0-9]+}}:vreg_64 = REG_SEQUENCE %6, %subreg.sub0, %5, %subreg.sub1
|
|
# GFX9: %{{[0-9]+}}:vreg_96 = REG_SEQUENCE killed %15, %subreg.sub0_sub1, %4, %subreg.sub2
|
|
# GFX9: %{{[0-9]+}}:vreg_128 = REG_SEQUENCE killed %16, %subreg.sub0_sub1_sub2, %3, %subreg.sub3
|
|
# GFX9: TBUFFER_STORE_FORMAT_XYZW_OFFSET_exact killed %17, %13, 0, 16, 78, 0, 0, 0, implicit $exec :: (dereferenceable store (s128), align 1, addrspace 4)
|
|
# GFX9: %{{[0-9]+}}:vreg_64 = REG_SEQUENCE %2, %subreg.sub0, %1, %subreg.sub1
|
|
# GFX9: %{{[0-9]+}}:vreg_96 = REG_SEQUENCE killed %18, %subreg.sub0_sub1, %0, %subreg.sub2
|
|
# GFX9: TBUFFER_STORE_FORMAT_XYZ_OFFSET_exact killed %19, %13, 0, 36, 77, 0, 0, 0, implicit $exec :: (dereferenceable store (s96), align 1, addrspace 4)
|
|
name: gfx9_tbuffer_store_uint32
|
|
body: |
|
|
bb.0.entry:
|
|
liveins: $sgpr0, $sgpr1, $sgpr2, $sgpr3, $vgpr0, $vgpr1, $vgpr2, $vgpr3, $vgpr4, $vgpr5, $vgpr6, $vgpr7, $vgpr8
|
|
%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, 68, 0, 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, 68, 0, 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, 68, 0, 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, 68, 0, 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, 68, 0, 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, 68, 0, 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, 68, 0, 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, 68, 0, 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, 68, 0, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4)
|
|
...
|
|
---
|
|
|
|
# GFX9-LABEL: name: gfx9_tbuffer_store_not_merged_data_format_mismatch
|
|
# GFX9: %{{[0-9]+}}:vgpr_32 = COPY $vgpr8
|
|
# GFX9: %{{[0-9]+}}:vgpr_32 = COPY $vgpr7
|
|
# GFX9: %{{[0-9]+}}:vgpr_32 = COPY $vgpr6
|
|
# GFX9: %{{[0-9]+}}:vgpr_32 = COPY $vgpr5
|
|
# GFX9: %{{[0-9]+}}:vgpr_32 = COPY $vgpr4
|
|
# GFX9: %{{[0-9]+}}:vgpr_32 = COPY $vgpr3
|
|
# GFX9: %{{[0-9]+}}:vgpr_32 = COPY $vgpr2
|
|
# GFX9: %{{[0-9]+}}:vgpr_32 = COPY $vgpr1
|
|
# GFX9: %{{[0-9]+}}:vgpr_32 = COPY $vgpr0
|
|
# GFX9: %{{[0-9]+}}:sgpr_32 = COPY $sgpr3
|
|
# GFX9: %{{[0-9]+}}:sgpr_32 = COPY $sgpr2
|
|
# GFX9: %{{[0-9]+}}:sgpr_32 = COPY $sgpr1
|
|
# GFX9: %{{[0-9]+}}:sgpr_32 = COPY $sgpr0
|
|
# GFX9: %{{[0-9]+}}:sgpr_128 = REG_SEQUENCE %12, %subreg.sub0, %11, %subreg.sub1, %10, %subreg.sub2, %9, %subreg.sub3
|
|
# GFX9: TBUFFER_STORE_FORMAT_X_OFFSET_exact %8, %13, 0, 4, 116, 0, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4)
|
|
# GFX9: TBUFFER_STORE_FORMAT_X_OFFSET_exact %7, %13, 0, 8, 84, 0, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4)
|
|
# GFX9: TBUFFER_STORE_FORMAT_X_OFFSET_exact %6, %13, 0, 16, 116, 0, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4)
|
|
# GFX9: TBUFFER_STORE_FORMAT_X_OFFSET_exact %5, %13, 0, 20, 84, 0, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4)
|
|
# GFX9: TBUFFER_STORE_FORMAT_X_OFFSET_exact %4, %13, 0, 24, 116, 0, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4)
|
|
# GFX9: TBUFFER_STORE_FORMAT_X_OFFSET_exact %3, %13, 0, 28, 84, 0, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4)
|
|
# GFX9: TBUFFER_STORE_FORMAT_X_OFFSET_exact %2, %13, 0, 36, 116, 0, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4)
|
|
# GFX9: TBUFFER_STORE_FORMAT_X_OFFSET_exact %1, %13, 0, 40, 84, 0, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4)
|
|
# GFX9: TBUFFER_STORE_FORMAT_X_OFFSET_exact %0, %13, 0, 44, 116, 0, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4)
|
|
name: gfx9_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
|
|
%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, 116, 0, 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, 84, 0, 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, 116, 0, 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, 84, 0, 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, 116, 0, 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, 84, 0, 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, 116, 0, 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, 84, 0, 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, 116, 0, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4)
|
|
...
|
|
---
|
|
|
|
# GFX9-LABEL: name: gfx9_tbuffer_store_not_merged_num_format_mismatch
|
|
# GFX9: %{{[0-9]+}}:vgpr_32 = COPY $vgpr8
|
|
# GFX9: %{{[0-9]+}}:vgpr_32 = COPY $vgpr7
|
|
# GFX9: %{{[0-9]+}}:vgpr_32 = COPY $vgpr6
|
|
# GFX9: %{{[0-9]+}}:vgpr_32 = COPY $vgpr5
|
|
# GFX9: %{{[0-9]+}}:vgpr_32 = COPY $vgpr4
|
|
# GFX9: %{{[0-9]+}}:vgpr_32 = COPY $vgpr3
|
|
# GFX9: %{{[0-9]+}}:vgpr_32 = COPY $vgpr2
|
|
# GFX9: %{{[0-9]+}}:vgpr_32 = COPY $vgpr1
|
|
# GFX9: %{{[0-9]+}}:vgpr_32 = COPY $vgpr0
|
|
# GFX9: %{{[0-9]+}}:sgpr_32 = COPY $sgpr3
|
|
# GFX9: %{{[0-9]+}}:sgpr_32 = COPY $sgpr2
|
|
# GFX9: %{{[0-9]+}}:sgpr_32 = COPY $sgpr1
|
|
# GFX9: %{{[0-9]+}}:sgpr_32 = COPY $sgpr0
|
|
# GFX9: %{{[0-9]+}}:sgpr_128 = REG_SEQUENCE %12, %subreg.sub0, %11, %subreg.sub1, %10, %subreg.sub2, %9, %subreg.sub3
|
|
# GFX9: TBUFFER_STORE_FORMAT_X_OFFSET_exact %8, %13, 0, 4, 116, 0, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4)
|
|
# GFX9: TBUFFER_STORE_FORMAT_X_OFFSET_exact %7, %13, 0, 8, 114, 0, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4)
|
|
# GFX9: TBUFFER_STORE_FORMAT_X_OFFSET_exact %6, %13, 0, 16, 116, 0, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4)
|
|
# GFX9: TBUFFER_STORE_FORMAT_X_OFFSET_exact %5, %13, 0, 20, 114, 0, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4)
|
|
# GFX9: TBUFFER_STORE_FORMAT_X_OFFSET_exact %4, %13, 0, 24, 116, 0, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4)
|
|
# GFX9: TBUFFER_STORE_FORMAT_X_OFFSET_exact %3, %13, 0, 28, 114, 0, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4)
|
|
# GFX9: TBUFFER_STORE_FORMAT_X_OFFSET_exact %2, %13, 0, 36, 116, 0, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4)
|
|
# GFX9: TBUFFER_STORE_FORMAT_X_OFFSET_exact %1, %13, 0, 40, 114, 0, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4)
|
|
# GFX9: TBUFFER_STORE_FORMAT_X_OFFSET_exact %0, %13, 0, 44, 116, 0, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4)
|
|
name: gfx9_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
|
|
%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, 116, 0, 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, 114, 0, 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, 116, 0, 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, 114, 0, 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, 116, 0, 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, 114, 0, 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, 116, 0, 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, 114, 0, 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, 116, 0, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4)
|
|
...
|
|
---
|
|
|
|
|
|
# GFX9-LABEL: name: gfx9_tbuffer_load_not_merged_swizzled_0
|
|
# GFX9: %{{[0-9]+}}:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %4, 0, 4, 116, 0, 0, 1, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
|
|
# GFX9: %{{[0-9]+}}:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %4, 0, 8, 116, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
|
|
name: gfx9_tbuffer_load_not_merged_swizzled_0
|
|
body: |
|
|
bb.0.entry:
|
|
%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, 116, 0, 0, 1, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
|
|
%8:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %5:sgpr_128, 0, 8, 116, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
|
|
...
|
|
---
|
|
|
|
|
|
# GFX9-LABEL: name: gfx9_tbuffer_load_not_merged_swizzled_1
|
|
# GFX9: %{{[0-9]+}}:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %4, 0, 4, 116, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
|
|
# GFX9: %{{[0-9]+}}:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %4, 0, 8, 116, 0, 0, 1, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
|
|
name: gfx9_tbuffer_load_not_merged_swizzled_1
|
|
body: |
|
|
bb.0.entry:
|
|
%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, 116, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
|
|
%8:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %5:sgpr_128, 0, 8, 116, 0, 0, 1, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
|
|
...
|
|
---
|
|
|
|
|
|
# GFX9-LABEL: name: gfx9_tbuffer_load_merge_across_swizzle
|
|
# GFX9-DAG: %{{[0-9]+}}:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %4, 0, 12, 116, 0, 0, 1, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
|
|
# GFX9-DAG: %{{[0-9]+}}:vreg_64 = TBUFFER_LOAD_FORMAT_XY_OFFSET %4, 0, 4, 123, 0, 0, 0, implicit $exec :: (dereferenceable load (s64), align 1, addrspace 4)
|
|
name: gfx9_tbuffer_load_merge_across_swizzle
|
|
body: |
|
|
bb.0.entry:
|
|
%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, 116, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
|
|
%6:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %4:sgpr_128, 0, 12, 116, 0, 0, 1, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
|
|
%7:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %4:sgpr_128, 0, 8, 116, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
|
|
...
|
|
---
|
|
|
|
|
|
# GFX9-LABEL: name: gfx9_tbuffer_load_merge_across_swizzled_store
|
|
# GFX9: %{{[0-9]+}}:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %5, 0, 4, 116, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
|
|
# GFX9: TBUFFER_STORE_FORMAT_X_OFFSET_exact %4, %5, 0, 6, 116, 0, 0, 1, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4)
|
|
# GFX9: %{{[0-9]+}}:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %5, 0, 8, 116, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
|
|
name: gfx9_tbuffer_load_merge_across_swizzled_store
|
|
body: |
|
|
bb.0.entry:
|
|
%0:sgpr_32 = COPY $sgpr0
|
|
%1:sgpr_32 = COPY $sgpr1
|
|
%2:sgpr_32 = COPY $sgpr2
|
|
%3:sgpr_32 = COPY $sgpr3
|
|
%4:vgpr_32 = COPY $vgpr0
|
|
%5:sgpr_128 = REG_SEQUENCE %0:sgpr_32, %subreg.sub0, %1:sgpr_32, %subreg.sub1, %2:sgpr_32, %subreg.sub2, %3:sgpr_32, %subreg.sub3
|
|
%6:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %5:sgpr_128, 0, 4, 116, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
|
|
TBUFFER_STORE_FORMAT_X_OFFSET_exact %4:vgpr_32, %5:sgpr_128, 0, 6, 116, 0, 0, 1, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4)
|
|
%7:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %5:sgpr_128, 0, 8, 116, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
|
|
...
|
|
---
|
|
|
|
|
|
#
|
|
# GFX10 tests
|
|
#
|
|
|
|
# GFX10-LABEL: name: gfx10_tbuffer_load_x_xyz
|
|
# GFX10: %{{[0-9]+}}:vreg_128 = TBUFFER_LOAD_FORMAT_XYZW_OFFSET %4, 0, 4, 77, 0, 0, 0, implicit $exec :: (dereferenceable load (s128), align 1, addrspace 4)
|
|
# GFX10: %{{[0-9]+}}:vgpr_32 = COPY %7.sub0
|
|
# GFX10: %{{[0-9]+}}:vreg_96 = COPY killed %7.sub1_sub2_sub3
|
|
name: gfx10_tbuffer_load_x_xyz
|
|
body: |
|
|
bb.0.entry:
|
|
%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, 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, 0, implicit $exec :: (dereferenceable load (s96), align 1, addrspace 4)
|
|
...
|
|
---
|
|
|
|
# GFX10-LABEL: name: gfx10_tbuffer_load_xyz_x
|
|
# GFX10: %{{[0-9]+}}:vreg_128 = TBUFFER_LOAD_FORMAT_XYZW_OFFSET %4, 0, 4, 77, 0, 0, 0, implicit $exec :: (dereferenceable load (s128), align 1, addrspace 4)
|
|
# GFX10: %{{[0-9]+}}:vreg_96 = COPY %7.sub0_sub1_sub2
|
|
# GFX10: %{{[0-9]+}}:vgpr_32 = COPY killed %7.sub3
|
|
name: gfx10_tbuffer_load_xyz_x
|
|
body: |
|
|
bb.0.entry:
|
|
%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, 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, 0, implicit $exec :: (dereferenceable load (s96), align 1, addrspace 4)
|
|
...
|
|
---
|
|
|
|
# GFX10-LABEL: name: gfx10_tbuffer_load_xy_xy
|
|
# GFX10: %{{[0-9]+}}:vreg_128 = TBUFFER_LOAD_FORMAT_XYZW_OFFSET %4, 0, 4, 77, 0, 0, 0, implicit $exec :: (dereferenceable load (s128), align 1, addrspace 4)
|
|
# GFX10: %{{[0-9]+}}:vreg_64 = COPY %7.sub0_sub1
|
|
# GFX10: %{{[0-9]+}}:vreg_64 = COPY killed %7.sub2_sub3
|
|
name: gfx10_tbuffer_load_xy_xy
|
|
body: |
|
|
bb.0.entry:
|
|
%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, 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, 0, implicit $exec :: (dereferenceable load (s64), align 1, addrspace 4)
|
|
...
|
|
---
|
|
|
|
# GFX10-LABEL: name: gfx10_tbuffer_load_x_xy
|
|
# GFX10: %{{[0-9]+}}:vreg_96 = TBUFFER_LOAD_FORMAT_XYZ_OFFSET %4, 0, 4, 74, 0, 0, 0, implicit $exec :: (dereferenceable load (s96), align 1, addrspace 4)
|
|
# GFX10: %{{[0-9]+}}:vgpr_32 = COPY %7.sub0
|
|
# GFX10: %{{[0-9]+}}:vreg_64 = COPY killed %7.sub1_sub2
|
|
name: gfx10_tbuffer_load_x_xy
|
|
body: |
|
|
bb.0.entry:
|
|
%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, 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, 0, implicit $exec :: (dereferenceable load (s64), align 1, addrspace 4)
|
|
...
|
|
---
|
|
|
|
# GFX10-LABEL: name: gfx10_tbuffer_load_xy_x
|
|
# GFX10: %{{[0-9]+}}:vreg_96 = TBUFFER_LOAD_FORMAT_XYZ_OFFSET %4, 0, 4, 74, 0, 0, 0, implicit $exec :: (dereferenceable load (s96), align 1, addrspace 4)
|
|
# GFX10: %{{[0-9]+}}:vreg_64 = COPY %7.sub0_sub1
|
|
# GFX10: %{{[0-9]+}}:vgpr_32 = COPY killed %7.sub2
|
|
name: gfx10_tbuffer_load_xy_x
|
|
body: |
|
|
bb.0.entry:
|
|
%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, 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, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
|
|
...
|
|
---
|
|
|
|
|
|
# GFX10-LABEL: name: gfx10_tbuffer_load_x_x
|
|
# GFX10: %{{[0-9]+}}:vreg_64 = TBUFFER_LOAD_FORMAT_XY_OFFSET %4, 0, 4, 64, 0, 0, 0, implicit $exec :: (dereferenceable load (s64), align 1, addrspace 4)
|
|
# GFX10: %{{[0-9]+}}:vgpr_32 = COPY %7.sub0
|
|
# GFX10: %{{[0-9]+}}:vgpr_32 = COPY killed %7.sub1
|
|
|
|
name: gfx10_tbuffer_load_x_x
|
|
body: |
|
|
bb.0.entry:
|
|
%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, 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, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
|
|
...
|
|
---
|
|
|
|
# GFX10-LABEL: name: gfx10_tbuffer_load_x_x_format_32_32_32_32
|
|
# GFX10: %{{[0-9]+}}:vreg_64 = TBUFFER_LOAD_FORMAT_XY_OFFSET %4, 0, 4, 64, 0, 0, 0, implicit $exec :: (dereferenceable load (s64), align 1, addrspace 4)
|
|
# GFX10: %{{[0-9]+}}:vgpr_32 = COPY %7.sub0
|
|
# GFX10: %{{[0-9]+}}:vgpr_32 = COPY killed %7.sub1
|
|
|
|
name: gfx10_tbuffer_load_x_x_format_32_32_32_32
|
|
body: |
|
|
bb.0.entry:
|
|
%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, 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, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
|
|
...
|
|
---
|
|
|
|
|
|
# GFX10-LABEL: name: gfx10_tbuffer_load_float_32
|
|
# GFX10: %{{[0-9]+}}:vreg_64 = TBUFFER_LOAD_FORMAT_XY_OFFSET %4, 0, 4, 64, 0, 0, 0, implicit $exec :: (dereferenceable load (s64), align 1, addrspace 4)
|
|
# GFX10: %{{[0-9]+}}:vgpr_32 = COPY %14.sub0
|
|
# GFX10: %{{[0-9]+}}:vgpr_32 = COPY killed %14.sub1
|
|
# GFX10: %{{[0-9]+}}:vreg_128 = TBUFFER_LOAD_FORMAT_XYZW_OFFSET %4, 0, 16, 77, 0, 0, 0, implicit $exec :: (dereferenceable load (s128), align 1, addrspace 4)
|
|
# GFX10: %{{[0-9]+}}:vreg_96 = COPY %17.sub0_sub1_sub2
|
|
# GFX10: %{{[0-9]+}}:vgpr_32 = COPY killed %17.sub3
|
|
# GFX10: %{{[0-9]+}}:vreg_64 = COPY %16.sub0_sub1
|
|
# GFX10: %{{[0-9]+}}:vgpr_32 = COPY killed %16.sub2
|
|
# GFX10: %{{[0-9]+}}:vgpr_32 = COPY %15.sub0
|
|
# GFX10: %{{[0-9]+}}:vgpr_32 = COPY killed %15.sub1
|
|
# GFX10: %{{[0-9]+}}:vreg_96 = TBUFFER_LOAD_FORMAT_XYZ_OFFSET %4, 0, 36, 74, 0, 0, 0, implicit $exec :: (dereferenceable load (s96), align 1, addrspace 4)
|
|
# GFX10: %{{[0-9]+}}:vreg_64 = COPY %19.sub0_sub1
|
|
# GFX10: %{{[0-9]+}}:vgpr_32 = COPY killed %19.sub2
|
|
# GFX10: %{{[0-9]+}}:vgpr_32 = COPY %18.sub0
|
|
# GFX10: %{{[0-9]+}}:vgpr_32 = COPY killed %18.sub1
|
|
|
|
name: gfx10_tbuffer_load_float_32
|
|
body: |
|
|
bb.0.entry:
|
|
%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, 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, 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, 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, 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, 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, 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, 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, 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, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
|
|
...
|
|
---
|
|
|
|
# GFX10-LABEL: name: gfx10_tbuffer_load_sint_32
|
|
# GFX10: %{{[0-9]+}}:vreg_64 = TBUFFER_LOAD_FORMAT_XY_OFFSET %4, 0, 4, 63, 0, 0, 0, implicit $exec :: (dereferenceable load (s64), align 1, addrspace 4)
|
|
# GFX10: %{{[0-9]+}}:vgpr_32 = COPY %14.sub0
|
|
# GFX10: %{{[0-9]+}}:vgpr_32 = COPY killed %14.sub1
|
|
# GFX10: %{{[0-9]+}}:vreg_128 = TBUFFER_LOAD_FORMAT_XYZW_OFFSET %4, 0, 16, 76, 0, 0, 0, implicit $exec :: (dereferenceable load (s128), align 1, addrspace 4)
|
|
# GFX10: %{{[0-9]+}}:vreg_96 = COPY %17.sub0_sub1_sub2
|
|
# GFX10: %{{[0-9]+}}:vgpr_32 = COPY killed %17.sub3
|
|
# GFX10: %{{[0-9]+}}:vreg_64 = COPY %16.sub0_sub1
|
|
# GFX10: %{{[0-9]+}}:vgpr_32 = COPY killed %16.sub2
|
|
# GFX10: %{{[0-9]+}}:vgpr_32 = COPY %15.sub0
|
|
# GFX10: %{{[0-9]+}}:vgpr_32 = COPY killed %15.sub1
|
|
# GFX10: %{{[0-9]+}}:vreg_96 = TBUFFER_LOAD_FORMAT_XYZ_OFFSET %4, 0, 36, 73, 0, 0, 0, implicit $exec :: (dereferenceable load (s96), align 1, addrspace 4)
|
|
# GFX10: %{{[0-9]+}}:vreg_64 = COPY %19.sub0_sub1
|
|
# GFX10: %{{[0-9]+}}:vgpr_32 = COPY killed %19.sub2
|
|
# GFX10: %{{[0-9]+}}:vgpr_32 = COPY %18.sub0
|
|
# GFX10: %{{[0-9]+}}:vgpr_32 = COPY killed %18.sub1
|
|
|
|
name: gfx10_tbuffer_load_sint_32
|
|
body: |
|
|
bb.0.entry:
|
|
%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, 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, 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, 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, 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, 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, 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, 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, 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, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
|
|
...
|
|
---
|
|
|
|
# GFX10-LABEL: name: gfx10_tbuffer_load_uint_32
|
|
# GFX10: %{{[0-9]+}}:vreg_64 = TBUFFER_LOAD_FORMAT_XY_OFFSET %4, 0, 4, 62, 0, 0, 0, implicit $exec :: (dereferenceable load (s64), align 1, addrspace 4)
|
|
# GFX10: %{{[0-9]+}}:vgpr_32 = COPY %14.sub0
|
|
# GFX10: %{{[0-9]+}}:vgpr_32 = COPY killed %14.sub1
|
|
# GFX10: %{{[0-9]+}}:vreg_128 = TBUFFER_LOAD_FORMAT_XYZW_OFFSET %4, 0, 16, 75, 0, 0, 0, implicit $exec :: (dereferenceable load (s128), align 1, addrspace 4)
|
|
# GFX10: %{{[0-9]+}}:vreg_96 = COPY %17.sub0_sub1_sub2
|
|
# GFX10: %{{[0-9]+}}:vgpr_32 = COPY killed %17.sub3
|
|
# GFX10: %{{[0-9]+}}:vreg_64 = COPY %16.sub0_sub1
|
|
# GFX10: %{{[0-9]+}}:vgpr_32 = COPY killed %16.sub2
|
|
# GFX10: %{{[0-9]+}}:vgpr_32 = COPY %15.sub0
|
|
# GFX10: %{{[0-9]+}}:vgpr_32 = COPY killed %15.sub1
|
|
# GFX10: %{{[0-9]+}}:vreg_96 = TBUFFER_LOAD_FORMAT_XYZ_OFFSET %4, 0, 36, 72, 0, 0, 0, implicit $exec :: (dereferenceable load (s96), align 1, addrspace 4)
|
|
# GFX10: %{{[0-9]+}}:vreg_64 = COPY %19.sub0_sub1
|
|
# GFX10: %{{[0-9]+}}:vgpr_32 = COPY killed %19.sub2
|
|
# GFX10: %{{[0-9]+}}:vgpr_32 = COPY %18.sub0
|
|
# GFX10: %{{[0-9]+}}:vgpr_32 = COPY killed %18.sub1
|
|
|
|
name: gfx10_tbuffer_load_uint_32
|
|
body: |
|
|
bb.0.entry:
|
|
%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, 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, 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, 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, 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, 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, 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, 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, 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, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
|
|
...
|
|
---
|
|
|
|
# GFX10-LABEL: name: gfx10_tbuffer_load_not_merged_data_format_mismatch
|
|
# GFX10: %{{[0-9]+}}:sgpr_32 = COPY $sgpr0
|
|
# GFX10: %{{[0-9]+}}:sgpr_32 = COPY $sgpr1
|
|
# GFX10: %{{[0-9]+}}:sgpr_32 = COPY $sgpr2
|
|
# GFX10: %{{[0-9]+}}:sgpr_32 = COPY $sgpr3
|
|
# GFX10: %{{[0-9]+}}:sgpr_128 = REG_SEQUENCE %0, %subreg.sub0, %1, %subreg.sub1, %2, %subreg.sub2, %3, %subreg.sub3
|
|
# GFX10: %{{[0-9]+}}:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %4, 0, 4, 22, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
|
|
# GFX10: %{{[0-9]+}}:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %4, 0, 8, 13, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
|
|
# GFX10: %{{[0-9]+}}:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %4, 0, 16, 22, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
|
|
# GFX10: %{{[0-9]+}}:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %4, 0, 20, 13, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
|
|
# GFX10: %{{[0-9]+}}:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %4, 0, 24, 22, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
|
|
# GFX10: %{{[0-9]+}}:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %4, 0, 28, 13, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
|
|
# GFX10: %{{[0-9]+}}:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %4, 0, 36, 22, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
|
|
# GFX10: %{{[0-9]+}}:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %4, 0, 40, 13, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
|
|
# GFX10: %{{[0-9]+}}:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %4, 0, 44, 22, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
|
|
|
|
|
|
name: gfx10_tbuffer_load_not_merged_data_format_mismatch
|
|
body: |
|
|
bb.0.entry:
|
|
%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, 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, 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, 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, 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, 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, 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, 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, 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, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
|
|
...
|
|
---
|
|
|
|
# GFX10-LABEL: name: gfx10_tbuffer_load_not_merged_num_format_mismatch
|
|
# GFX10: %{{[0-9]+}}:sgpr_32 = COPY $sgpr0
|
|
# GFX10: %{{[0-9]+}}:sgpr_32 = COPY $sgpr1
|
|
# GFX10: %{{[0-9]+}}:sgpr_32 = COPY $sgpr2
|
|
# GFX10: %{{[0-9]+}}:sgpr_32 = COPY $sgpr3
|
|
# GFX10: %{{[0-9]+}}:sgpr_128 = REG_SEQUENCE %0, %subreg.sub0, %1, %subreg.sub1, %2, %subreg.sub2, %3, %subreg.sub3
|
|
# GFX10: %{{[0-9]+}}:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %4, 0, 4, 22, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
|
|
# GFX10: %{{[0-9]+}}:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %4, 0, 8, 21, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
|
|
# GFX10: %{{[0-9]+}}:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %4, 0, 16, 22, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
|
|
# GFX10: %{{[0-9]+}}:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %4, 0, 20, 21, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
|
|
# GFX10: %{{[0-9]+}}:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %4, 0, 24, 22, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
|
|
# GFX10: %{{[0-9]+}}:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %4, 0, 28, 21, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
|
|
# GFX10: %{{[0-9]+}}:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %4, 0, 36, 22, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
|
|
# GFX10: %{{[0-9]+}}:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %4, 0, 40, 21, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
|
|
# GFX10: %{{[0-9]+}}:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %4, 0, 44, 22, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
|
|
name: gfx10_tbuffer_load_not_merged_num_format_mismatch
|
|
body: |
|
|
bb.0.entry:
|
|
%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, 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, 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, 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, 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, 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, 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, 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, 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, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
|
|
...
|
|
---
|
|
|
|
|
|
|
|
# GFX10-LABEL: name: gfx10_tbuffer_store_x_xyz
|
|
# GFX10: %{{[0-9]+}}:vreg_96 = REG_SEQUENCE %3, %subreg.sub0, %2, %subreg.sub1, %1, %subreg.sub2
|
|
# GFX10: %{{[0-9]+}}:vreg_128 = REG_SEQUENCE %0, %subreg.sub0, %9, %subreg.sub1_sub2_sub3
|
|
# GFX10: TBUFFER_STORE_FORMAT_XYZW_OFFSET_exact killed %10, %8, 0, 4, 77, 0, 0, 0, implicit $exec :: (dereferenceable store (s128), align 1, addrspace 4)
|
|
name: gfx10_tbuffer_store_x_xyz
|
|
body: |
|
|
bb.0.entry:
|
|
liveins: $sgpr0, $sgpr1, $sgpr2, $sgpr3, $vgpr0, $vgpr1, $vgpr2, $vgpr3
|
|
%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, 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, 0, implicit $exec :: (dereferenceable store (s96), align 1, addrspace 4)
|
|
...
|
|
---
|
|
|
|
|
|
# GFX10-LABEL: name: gfx10_tbuffer_store_xyz_x
|
|
# GFX10: %{{[0-9]+}}:vreg_96 = REG_SEQUENCE %3, %subreg.sub0, %2, %subreg.sub1, %1, %subreg.sub2
|
|
# GFX10: %{{[0-9]+}}:vreg_128 = REG_SEQUENCE %9, %subreg.sub0_sub1_sub2, %0, %subreg.sub3
|
|
# GFX10: TBUFFER_STORE_FORMAT_XYZW_OFFSET_exact killed %10, %8, 0, 4, 77, 0, 0, 0, implicit $exec :: (dereferenceable store (s128), align 1, addrspace 4)
|
|
name: gfx10_tbuffer_store_xyz_x
|
|
body: |
|
|
bb.0.entry:
|
|
liveins: $sgpr0, $sgpr1, $sgpr2, $sgpr3, $vgpr0, $vgpr1, $vgpr2, $vgpr3
|
|
%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, 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, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4)
|
|
...
|
|
---
|
|
|
|
# GFX10-LABEL: name: gfx10_tbuffer_store_xy_xy
|
|
# GFX10: %{{[0-9]+}}:vreg_64 = REG_SEQUENCE %3, %subreg.sub0, %2, %subreg.sub1
|
|
# GFX10: %{{[0-9]+}}:vreg_64 = REG_SEQUENCE %1, %subreg.sub0, %0, %subreg.sub1
|
|
# GFX10: %{{[0-9]+}}:vreg_128 = REG_SEQUENCE %9, %subreg.sub0_sub1, %10, %subreg.sub2_sub3
|
|
# GFX10: TBUFFER_STORE_FORMAT_XYZW_OFFSET_exact killed %11, %8, 0, 4, 77, 0, 0, 0, implicit $exec :: (dereferenceable store (s128), align 1, addrspace 4)
|
|
name: gfx10_tbuffer_store_xy_xy
|
|
body: |
|
|
bb.0.entry:
|
|
liveins: $sgpr0, $sgpr1, $sgpr2, $sgpr3, $vgpr0, $vgpr1, $vgpr2, $vgpr3
|
|
%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, 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, 0, implicit $exec :: (dereferenceable store (s64), align 1, addrspace 4)
|
|
...
|
|
---
|
|
|
|
# GFX10-LABEL: name: gfx10_tbuffer_store_x_xy
|
|
# GFX10: %{{[0-9]+}}:vreg_64 = REG_SEQUENCE %3, %subreg.sub0, %2, %subreg.sub1
|
|
# GFX10: %{{[0-9]+}}:vreg_64, %subreg.sub1_sub2
|
|
# GFX10: TBUFFER_STORE_FORMAT_XYZ_OFFSET_exact killed %11, %8, 0, 4, 74, 0, 0, 0, implicit $exec :: (dereferenceable store (s96), align 1, addrspace 4)
|
|
name: gfx10_tbuffer_store_x_xy
|
|
body: |
|
|
bb.0.entry:
|
|
liveins: $sgpr0, $sgpr1, $sgpr2, $sgpr3, $vgpr0, $vgpr1, $vgpr2, $vgpr3
|
|
%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, 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, 0, implicit $exec :: (dereferenceable store (s64), align 1, addrspace 4)
|
|
...
|
|
---
|
|
|
|
# GFX10-LABEL: name: gfx10_tbuffer_store_xy_x
|
|
# GFX10: %{{[0-9]+}}:vreg_64 = REG_SEQUENCE %3, %subreg.sub0, %2, %subreg.sub1
|
|
# GFX10: %{{[0-9]+}}:vreg_96 = REG_SEQUENCE %9, %subreg.sub0_sub1, %0, %subreg.sub2
|
|
# GFX10: TBUFFER_STORE_FORMAT_XYZ_OFFSET_exact killed %10, %8, 0, 4, 74, 0, 0, 0, implicit $exec :: (dereferenceable store (s96), 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
|
|
%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, 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, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4)
|
|
...
|
|
---
|
|
|
|
|
|
# GFX10-LABEL: name: gfx10_tbuffer_store_x_x
|
|
# GFX10: %{{[0-9]+}}:vreg_64 = REG_SEQUENCE %1, %subreg.sub0, %0, %subreg.sub1
|
|
# GFX10: TBUFFER_STORE_FORMAT_XY_OFFSET_exact killed %9, %8, 0, 4, 64, 0, 0, 0, implicit $exec :: (dereferenceable store (s64), align 1, addrspace 4)
|
|
name: gfx10_tbuffer_store_x_x
|
|
body: |
|
|
bb.0.entry:
|
|
liveins: $sgpr0, $sgpr1, $sgpr2, $sgpr3, $vgpr0, $vgpr1, $vgpr2, $vgpr3
|
|
%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, 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, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4)
|
|
...
|
|
---
|
|
|
|
# GFX10-LABEL: name: gfx10_tbuffer_store_x_x_format_32_32_32_32
|
|
# GFX10: %{{[0-9]+}}:vreg_64 = REG_SEQUENCE %1, %subreg.sub0, %0, %subreg.sub1
|
|
# GFX10: TBUFFER_STORE_FORMAT_XY_OFFSET_exact killed %9, %8, 0, 4, 64, 0, 0, 0, implicit $exec :: (dereferenceable store (s64), 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
|
|
%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, 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, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4)
|
|
...
|
|
---
|
|
|
|
# GFX10-LABEL: name: gfx10_tbuffer_store_float32
|
|
# GFX10: %{{[0-9]+}}:vgpr_32 = COPY $vgpr8
|
|
# GFX10: %{{[0-9]+}}:vgpr_32 = COPY $vgpr7
|
|
# GFX10: %{{[0-9]+}}:vgpr_32 = COPY $vgpr6
|
|
# GFX10: %{{[0-9]+}}:vgpr_32 = COPY $vgpr5
|
|
# GFX10: %{{[0-9]+}}:vgpr_32 = COPY $vgpr4
|
|
# GFX10: %{{[0-9]+}}:vgpr_32 = COPY $vgpr3
|
|
# GFX10: %{{[0-9]+}}:vgpr_32 = COPY $vgpr2
|
|
# GFX10: %{{[0-9]+}}:vgpr_32 = COPY $vgpr1
|
|
# GFX10: %{{[0-9]+}}:vgpr_32 = COPY $vgpr0
|
|
# GFX10: %{{[0-9]+}}:sgpr_32 = COPY $sgpr3
|
|
# GFX10: %{{[0-9]+}}:sgpr_32 = COPY $sgpr2
|
|
# GFX10: %{{[0-9]+}}:sgpr_32 = COPY $sgpr1
|
|
# GFX10: %{{[0-9]+}}:sgpr_32 = COPY $sgpr0
|
|
# GFX10: %{{[0-9]+}}:sgpr_128 = REG_SEQUENCE %12, %subreg.sub0, %11, %subreg.sub1, %10, %subreg.sub2, %9, %subreg.sub3
|
|
# GFX10: %{{[0-9]+}}:vreg_64 = REG_SEQUENCE %8, %subreg.sub0, %7, %subreg.sub1
|
|
# GFX10: TBUFFER_STORE_FORMAT_XY_OFFSET_exact killed %14, %13, 0, 4, 64, 0, 0, 0, implicit $exec :: (dereferenceable store (s64), align 1, addrspace 4)
|
|
# GFX10: %{{[0-9]+}}:vreg_64 = REG_SEQUENCE %6, %subreg.sub0, %5, %subreg.sub1
|
|
# GFX10: %{{[0-9]+}}:vreg_96 = REG_SEQUENCE killed %15, %subreg.sub0_sub1, %4, %subreg.sub2
|
|
# GFX10: %{{[0-9]+}}:vreg_128 = REG_SEQUENCE killed %16, %subreg.sub0_sub1_sub2, %3, %subreg.sub3
|
|
# GFX10: TBUFFER_STORE_FORMAT_XYZW_OFFSET_exact killed %17, %13, 0, 16, 77, 0, 0, 0, implicit $exec :: (dereferenceable store (s128), align 1, addrspace 4)
|
|
# GFX10: %{{[0-9]+}}:vreg_64 = REG_SEQUENCE %2, %subreg.sub0, %1, %subreg.sub1
|
|
# GFX10: %{{[0-9]+}}:vreg_96 = REG_SEQUENCE killed %18, %subreg.sub0_sub1, %0, %subreg.sub2
|
|
# GFX10: TBUFFER_STORE_FORMAT_XYZ_OFFSET_exact killed %19, %13, 0, 36, 74, 0, 0, 0, implicit $exec :: (dereferenceable store (s96), 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
|
|
%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, 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, 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, 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, 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, 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, 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, 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, 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, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4)
|
|
...
|
|
---
|
|
|
|
# GFX10-LABEL: name: gfx10_tbuffer_store_sint32
|
|
# GFX10: %{{[0-9]+}}:vgpr_32 = COPY $vgpr8
|
|
# GFX10: %{{[0-9]+}}:vgpr_32 = COPY $vgpr7
|
|
# GFX10: %{{[0-9]+}}:vgpr_32 = COPY $vgpr6
|
|
# GFX10: %{{[0-9]+}}:vgpr_32 = COPY $vgpr5
|
|
# GFX10: %{{[0-9]+}}:vgpr_32 = COPY $vgpr4
|
|
# GFX10: %{{[0-9]+}}:vgpr_32 = COPY $vgpr3
|
|
# GFX10: %{{[0-9]+}}:vgpr_32 = COPY $vgpr2
|
|
# GFX10: %{{[0-9]+}}:vgpr_32 = COPY $vgpr1
|
|
# GFX10: %{{[0-9]+}}:vgpr_32 = COPY $vgpr0
|
|
# GFX10: %{{[0-9]+}}:sgpr_32 = COPY $sgpr3
|
|
# GFX10: %{{[0-9]+}}:sgpr_32 = COPY $sgpr2
|
|
# GFX10: %{{[0-9]+}}:sgpr_32 = COPY $sgpr1
|
|
# GFX10: %{{[0-9]+}}:sgpr_32 = COPY $sgpr0
|
|
# GFX10: %{{[0-9]+}}:sgpr_128 = REG_SEQUENCE %12, %subreg.sub0, %11, %subreg.sub1, %10, %subreg.sub2, %9, %subreg.sub3
|
|
# GFX10: %{{[0-9]+}}:vreg_64 = REG_SEQUENCE %8, %subreg.sub0, %7, %subreg.sub1
|
|
# GFX10: TBUFFER_STORE_FORMAT_XY_OFFSET_exact killed %14, %13, 0, 4, 63, 0, 0, 0, implicit $exec :: (dereferenceable store (s64), align 1, addrspace 4)
|
|
# GFX10: %{{[0-9]+}}:vreg_64 = REG_SEQUENCE %6, %subreg.sub0, %5, %subreg.sub1
|
|
# GFX10: %{{[0-9]+}}:vreg_96 = REG_SEQUENCE killed %15, %subreg.sub0_sub1, %4, %subreg.sub2
|
|
# GFX10: %{{[0-9]+}}:vreg_128 = REG_SEQUENCE killed %16, %subreg.sub0_sub1_sub2, %3, %subreg.sub3
|
|
# GFX10: TBUFFER_STORE_FORMAT_XYZW_OFFSET_exact killed %17, %13, 0, 16, 76, 0, 0, 0, implicit $exec :: (dereferenceable store (s128), align 1, addrspace 4)
|
|
# GFX10: %{{[0-9]+}}:vreg_64 = REG_SEQUENCE %2, %subreg.sub0, %1, %subreg.sub1
|
|
# GFX10: %{{[0-9]+}}:vreg_96 = REG_SEQUENCE killed %18, %subreg.sub0_sub1, %0, %subreg.sub2
|
|
# GFX10: TBUFFER_STORE_FORMAT_XYZ_OFFSET_exact killed %19, %13, 0, 36, 73, 0, 0, 0, implicit $exec :: (dereferenceable store (s96), 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
|
|
%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, 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, 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, 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, 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, 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, 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, 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, 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, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4)
|
|
...
|
|
---
|
|
|
|
# GFX10-LABEL: name: gfx10_tbuffer_store_uint32
|
|
# GFX10: %{{[0-9]+}}:vgpr_32 = COPY $vgpr8
|
|
# GFX10: %{{[0-9]+}}:vgpr_32 = COPY $vgpr7
|
|
# GFX10: %{{[0-9]+}}:vgpr_32 = COPY $vgpr6
|
|
# GFX10: %{{[0-9]+}}:vgpr_32 = COPY $vgpr5
|
|
# GFX10: %{{[0-9]+}}:vgpr_32 = COPY $vgpr4
|
|
# GFX10: %{{[0-9]+}}:vgpr_32 = COPY $vgpr3
|
|
# GFX10: %{{[0-9]+}}:vgpr_32 = COPY $vgpr2
|
|
# GFX10: %{{[0-9]+}}:vgpr_32 = COPY $vgpr1
|
|
# GFX10: %{{[0-9]+}}:vgpr_32 = COPY $vgpr0
|
|
# GFX10: %{{[0-9]+}}:sgpr_32 = COPY $sgpr3
|
|
# GFX10: %{{[0-9]+}}:sgpr_32 = COPY $sgpr2
|
|
# GFX10: %{{[0-9]+}}:sgpr_32 = COPY $sgpr1
|
|
# GFX10: %{{[0-9]+}}:sgpr_32 = COPY $sgpr0
|
|
# GFX10: %{{[0-9]+}}:sgpr_128 = REG_SEQUENCE %12, %subreg.sub0, %11, %subreg.sub1, %10, %subreg.sub2, %9, %subreg.sub3
|
|
# GFX10: %{{[0-9]+}}:vreg_64 = REG_SEQUENCE %8, %subreg.sub0, %7, %subreg.sub1
|
|
# GFX10: TBUFFER_STORE_FORMAT_XY_OFFSET_exact killed %14, %13, 0, 4, 62, 0, 0, 0, implicit $exec :: (dereferenceable store (s64), align 1, addrspace 4)
|
|
# GFX10: %{{[0-9]+}}:vreg_64 = REG_SEQUENCE %6, %subreg.sub0, %5, %subreg.sub1
|
|
# GFX10: %{{[0-9]+}}:vreg_96 = REG_SEQUENCE killed %15, %subreg.sub0_sub1, %4, %subreg.sub2
|
|
# GFX10: %{{[0-9]+}}:vreg_128 = REG_SEQUENCE killed %16, %subreg.sub0_sub1_sub2, %3, %subreg.sub3
|
|
# GFX10: TBUFFER_STORE_FORMAT_XYZW_OFFSET_exact killed %17, %13, 0, 16, 75, 0, 0, 0, implicit $exec :: (dereferenceable store (s128), align 1, addrspace 4)
|
|
# GFX10: %{{[0-9]+}}:vreg_64 = REG_SEQUENCE %2, %subreg.sub0, %1, %subreg.sub1
|
|
# GFX10: %{{[0-9]+}}:vreg_96 = REG_SEQUENCE killed %18, %subreg.sub0_sub1, %0, %subreg.sub2
|
|
# GFX10: TBUFFER_STORE_FORMAT_XYZ_OFFSET_exact killed %19, %13, 0, 36, 72, 0, 0, 0, implicit $exec :: (dereferenceable store (s96), 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
|
|
%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, 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, 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, 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, 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, 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, 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, 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, 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, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4)
|
|
...
|
|
---
|
|
|
|
# GFX10-LABEL: name: gfx10_tbuffer_store_not_merged_data_format_mismatch
|
|
# GFX10: %{{[0-9]+}}:vgpr_32 = COPY $vgpr8
|
|
# GFX10: %{{[0-9]+}}:vgpr_32 = COPY $vgpr7
|
|
# GFX10: %{{[0-9]+}}:vgpr_32 = COPY $vgpr6
|
|
# GFX10: %{{[0-9]+}}:vgpr_32 = COPY $vgpr5
|
|
# GFX10: %{{[0-9]+}}:vgpr_32 = COPY $vgpr4
|
|
# GFX10: %{{[0-9]+}}:vgpr_32 = COPY $vgpr3
|
|
# GFX10: %{{[0-9]+}}:vgpr_32 = COPY $vgpr2
|
|
# GFX10: %{{[0-9]+}}:vgpr_32 = COPY $vgpr1
|
|
# GFX10: %{{[0-9]+}}:vgpr_32 = COPY $vgpr0
|
|
# GFX10: %{{[0-9]+}}:sgpr_32 = COPY $sgpr3
|
|
# GFX10: %{{[0-9]+}}:sgpr_32 = COPY $sgpr2
|
|
# GFX10: %{{[0-9]+}}:sgpr_32 = COPY $sgpr1
|
|
# GFX10: %{{[0-9]+}}:sgpr_32 = COPY $sgpr0
|
|
# GFX10: %{{[0-9]+}}:sgpr_128 = REG_SEQUENCE %12, %subreg.sub0, %11, %subreg.sub1, %10, %subreg.sub2, %9, %subreg.sub3
|
|
# GFX10: TBUFFER_STORE_FORMAT_X_OFFSET_exact %8, %13, 0, 4, 22, 0, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4)
|
|
# GFX10: TBUFFER_STORE_FORMAT_X_OFFSET_exact %7, %13, 0, 8, 21, 0, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4)
|
|
# GFX10: TBUFFER_STORE_FORMAT_X_OFFSET_exact %6, %13, 0, 16, 22, 0, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4)
|
|
# GFX10: TBUFFER_STORE_FORMAT_X_OFFSET_exact %5, %13, 0, 20, 21, 0, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4)
|
|
# GFX10: TBUFFER_STORE_FORMAT_X_OFFSET_exact %4, %13, 0, 24, 22, 0, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4)
|
|
# GFX10: TBUFFER_STORE_FORMAT_X_OFFSET_exact %3, %13, 0, 28, 21, 0, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4)
|
|
# GFX10: TBUFFER_STORE_FORMAT_X_OFFSET_exact %2, %13, 0, 36, 22, 0, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4)
|
|
# GFX10: TBUFFER_STORE_FORMAT_X_OFFSET_exact %1, %13, 0, 40, 21, 0, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4)
|
|
# GFX10: TBUFFER_STORE_FORMAT_X_OFFSET_exact %0, %13, 0, 44, 22, 0, 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
|
|
%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, 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, 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, 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, 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, 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, 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, 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, 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, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4)
|
|
...
|
|
---
|
|
|
|
# GFX10-LABEL: name: gfx10_tbuffer_store_not_merged_num_format_mismatch
|
|
# GFX10: %{{[0-9]+}}:vgpr_32 = COPY $vgpr8
|
|
# GFX10: %{{[0-9]+}}:vgpr_32 = COPY $vgpr7
|
|
# GFX10: %{{[0-9]+}}:vgpr_32 = COPY $vgpr6
|
|
# GFX10: %{{[0-9]+}}:vgpr_32 = COPY $vgpr5
|
|
# GFX10: %{{[0-9]+}}:vgpr_32 = COPY $vgpr4
|
|
# GFX10: %{{[0-9]+}}:vgpr_32 = COPY $vgpr3
|
|
# GFX10: %{{[0-9]+}}:vgpr_32 = COPY $vgpr2
|
|
# GFX10: %{{[0-9]+}}:vgpr_32 = COPY $vgpr1
|
|
# GFX10: %{{[0-9]+}}:vgpr_32 = COPY $vgpr0
|
|
# GFX10: %{{[0-9]+}}:sgpr_32 = COPY $sgpr3
|
|
# GFX10: %{{[0-9]+}}:sgpr_32 = COPY $sgpr2
|
|
# GFX10: %{{[0-9]+}}:sgpr_32 = COPY $sgpr1
|
|
# GFX10: %{{[0-9]+}}:sgpr_32 = COPY $sgpr0
|
|
# GFX10: %{{[0-9]+}}:sgpr_128 = REG_SEQUENCE %12, %subreg.sub0, %11, %subreg.sub1, %10, %subreg.sub2, %9, %subreg.sub3
|
|
# GFX10: TBUFFER_STORE_FORMAT_X_OFFSET_exact %8, %13, 0, 4, 22, 0, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4)
|
|
# GFX10: TBUFFER_STORE_FORMAT_X_OFFSET_exact %7, %13, 0, 8, 13, 0, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4)
|
|
# GFX10: TBUFFER_STORE_FORMAT_X_OFFSET_exact %6, %13, 0, 16, 22, 0, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4)
|
|
# GFX10: TBUFFER_STORE_FORMAT_X_OFFSET_exact %5, %13, 0, 20, 13, 0, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4)
|
|
# GFX10: TBUFFER_STORE_FORMAT_X_OFFSET_exact %4, %13, 0, 24, 22, 0, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4)
|
|
# GFX10: TBUFFER_STORE_FORMAT_X_OFFSET_exact %3, %13, 0, 28, 13, 0, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4)
|
|
# GFX10: TBUFFER_STORE_FORMAT_X_OFFSET_exact %2, %13, 0, 36, 22, 0, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4)
|
|
# GFX10: TBUFFER_STORE_FORMAT_X_OFFSET_exact %1, %13, 0, 40, 13, 0, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4)
|
|
# GFX10: TBUFFER_STORE_FORMAT_X_OFFSET_exact %0, %13, 0, 44, 22, 0, 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
|
|
%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, 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, 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, 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, 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, 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, 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, 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, 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, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4)
|
|
...
|
|
---
|
|
|
|
|
|
# GFX10-LABEL: name: gfx10_tbuffer_load_not_merged_swizzled_0
|
|
# GFX10: %{{[0-9]+}}:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %4, 0, 4, 22, 0, 0, 1, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
|
|
# GFX10: %{{[0-9]+}}:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %4, 0, 8, 22, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
|
|
name: gfx10_tbuffer_load_not_merged_swizzled_0
|
|
body: |
|
|
bb.0.entry:
|
|
%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, 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, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
|
|
...
|
|
---
|
|
|
|
|
|
# GFX10-LABEL: name: gfx10_tbuffer_load_not_merged_swizzled_1
|
|
# GFX10: %{{[0-9]+}}:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %4, 0, 4, 22, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
|
|
# GFX10: %{{[0-9]+}}:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %4, 0, 8, 22, 0, 0, 1, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
|
|
name: gfx10_tbuffer_load_not_merged_swizzled_1
|
|
body: |
|
|
bb.0.entry:
|
|
%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, 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, 1, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
|
|
...
|
|
---
|
|
|
|
|
|
# GFX10-LABEL: name: gfx10_tbuffer_load_merge_across_swizzle
|
|
# GFX10-DAG: %{{[0-9]+}}:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFSET %4, 0, 12, 22, 0, 0, 1, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
|
|
# GFX10-DAG: %{{[0-9]+}}:vreg_64 = TBUFFER_LOAD_FORMAT_XY_OFFSET %4, 0, 4, 64, 0, 0, 0, implicit $exec :: (dereferenceable load (s64), align 1, addrspace 4)
|
|
name: gfx10_tbuffer_load_merge_across_swizzle
|
|
body: |
|
|
bb.0.entry:
|
|
%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, 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, 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, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
|
|
...
|
|
---
|