4b4496312e
This is the groundwork required to implement strictfp. For now, this should be NFC for regular instructoins (many instructions just gain an extra use of a reserved register). Regalloc won't rematerialize instructions with reads of physical registers, but we were suffering from that anyway with the exec reads. Should add it for all the related FP uses (possibly with some extras). I did not add it to either the gpr index mode instructions (or every single VALU instruction) since it's a ridiculous feature already modeled as an arbitrary side effect. Also work towards marking instructions with FP exceptions. This doesn't actually set the bit yet since this would start to change codegen. It seems nofpexcept is currently not implied from the regular IR FP operations. Add it to some MIR tests where I think it might matter.
675 lines
18 KiB
YAML
675 lines
18 KiB
YAML
# RUN: llc -march=amdgcn -mcpu=gfx900 -run-pass si-shrink-instructions -verify-machineinstrs %s -o - | FileCheck -check-prefix=GCN %s
|
|
|
|
# GCN-LABEL: name: swap_phys_condensed
|
|
# GCN: bb.0:
|
|
# GCN-NEXT: $vgpr0, $vgpr1 = V_SWAP_B32 $vgpr1, $vgpr0, implicit $exec
|
|
# GCN-NEXT: S_SETPC_B64_return
|
|
---
|
|
name: swap_phys_condensed
|
|
body: |
|
|
bb.0:
|
|
liveins: $vgpr0, $vgpr1, $sgpr30_sgpr31
|
|
$vgpr2 = V_MOV_B32_e32 killed $vgpr0, implicit $exec
|
|
$vgpr0 = V_MOV_B32_e32 killed $vgpr1, implicit $exec
|
|
$vgpr1 = V_MOV_B32_e32 killed $vgpr2, implicit $exec
|
|
S_SETPC_B64_return $sgpr30_sgpr31, implicit $vgpr0, implicit $vgpr1
|
|
...
|
|
|
|
# GCN-LABEL: name: swap_phys_sparse
|
|
# GCN: bb.0:
|
|
# GCN-NEXT: $vgpr3 = V_MOV_B32_e32 killed $vgpr4, implicit $exec
|
|
# GCN-NEXT: $vgpr0, $vgpr1 = V_SWAP_B32 $vgpr1, $vgpr0, implicit $exec
|
|
# GCN-NEXT: $vgpr5 = V_MOV_B32_e32 killed $vgpr6, implicit $exec
|
|
# GCN-NEXT: S_SETPC_B64_return
|
|
---
|
|
name: swap_phys_sparse
|
|
body: |
|
|
bb.0:
|
|
liveins: $vgpr0, $vgpr1, $sgpr30_sgpr31
|
|
$vgpr2 = V_MOV_B32_e32 killed $vgpr0, implicit $exec
|
|
$vgpr3 = V_MOV_B32_e32 killed $vgpr4, implicit $exec
|
|
$vgpr0 = V_MOV_B32_e32 killed $vgpr1, implicit $exec
|
|
$vgpr5 = V_MOV_B32_e32 killed $vgpr6, implicit $exec
|
|
$vgpr1 = V_MOV_B32_e32 killed $vgpr2, implicit $exec
|
|
S_SETPC_B64_return $sgpr30_sgpr31, implicit $vgpr0, implicit $vgpr1
|
|
...
|
|
|
|
# GCN-LABEL: name: swap_phys_liveout
|
|
# GCN: bb.0:
|
|
# GCN-NEXT: $vgpr2 = V_MOV_B32_e32 $vgpr0, implicit $exec
|
|
# GCN-NEXT: $vgpr0, $vgpr1 = V_SWAP_B32 $vgpr1, $vgpr0, implicit $exec
|
|
# GCN-NEXT: S_SETPC_B64_return
|
|
---
|
|
name: swap_phys_liveout
|
|
body: |
|
|
bb.0:
|
|
liveins: $vgpr0, $vgpr1, $sgpr30_sgpr31
|
|
$vgpr2 = V_MOV_B32_e32 killed $vgpr0, implicit $exec
|
|
$vgpr0 = V_MOV_B32_e32 killed $vgpr1, implicit $exec
|
|
$vgpr1 = V_MOV_B32_e32 killed $vgpr2, implicit $exec
|
|
S_SETPC_B64_return $sgpr30_sgpr31, implicit $vgpr0, implicit $vgpr2, implicit $vgpr1
|
|
...
|
|
|
|
# GCN-LABEL: name: swap_phys_b64
|
|
# GCN: bb.0:
|
|
# GCN-NEXT: $vgpr0, $vgpr2 = V_SWAP_B32 $vgpr2, $vgpr0, implicit $exec
|
|
# GCN-NEXT: $vgpr1, $vgpr3 = V_SWAP_B32 $vgpr3, $vgpr1, implicit $exec
|
|
---
|
|
name: swap_phys_b64
|
|
body: |
|
|
bb.0:
|
|
$vgpr4_vgpr5 = COPY killed $vgpr0_vgpr1
|
|
$vgpr0_vgpr1 = COPY killed $vgpr2_vgpr3
|
|
$vgpr2_vgpr3 = COPY killed $vgpr4_vgpr5
|
|
...
|
|
|
|
# GCN-LABEL: name: swap_phys_overlap_x
|
|
# GCN: bb.0:
|
|
# GCN-NEXT: $vgpr2 = V_MOV_B32_e32 $vgpr0, implicit $exec
|
|
# GCN-NEXT: $vgpr3_vgpr4 = V_ADD_F64 0, $vgpr0_vgpr1, 0, $vgpr3_vgpr4, 0, 0, implicit $mode, implicit $exec
|
|
# GCN-NEXT: $vgpr0 = V_MOV_B32_e32 killed $vgpr1, implicit $exec
|
|
# GCN-NEXT: $vgpr1 = V_MOV_B32_e32 killed $vgpr2, implicit $exec
|
|
---
|
|
name: swap_phys_overlap_x
|
|
body: |
|
|
bb.0:
|
|
$vgpr2 = V_MOV_B32_e32 $vgpr0, implicit $exec
|
|
$vgpr3_vgpr4 = V_ADD_F64 0, $vgpr0_vgpr1, 0, $vgpr3_vgpr4, 0, 0, implicit $mode, implicit $exec
|
|
$vgpr0 = V_MOV_B32_e32 killed $vgpr1, implicit $exec
|
|
$vgpr1 = V_MOV_B32_e32 killed $vgpr2, implicit $exec
|
|
...
|
|
|
|
# GCN-LABEL: name: swap_phys_clobber_y
|
|
# GCN: bb.0:
|
|
# GCN-NEXT: $vgpr2 = V_MOV_B32_e32 $vgpr0, implicit $exec
|
|
# GCN-NEXT: $vgpr0 = V_MOV_B32_e32 killed $vgpr1, implicit $exec
|
|
# GCN-NEXT: $vgpr1 = V_MOV_B32_e32 0, implicit $exec
|
|
# GCN-NEXT: $vgpr1 = V_MOV_B32_e32 killed $vgpr2, implicit $exec
|
|
---
|
|
name: swap_phys_clobber_y
|
|
body: |
|
|
bb.0:
|
|
$vgpr2 = V_MOV_B32_e32 $vgpr0, implicit $exec
|
|
$vgpr0 = V_MOV_B32_e32 killed $vgpr1, implicit $exec
|
|
$vgpr1 = V_MOV_B32_e32 0, implicit $exec
|
|
$vgpr1 = V_MOV_B32_e32 killed $vgpr2, implicit $exec
|
|
S_ENDPGM 0
|
|
...
|
|
|
|
# GCN-LABEL: name: swap_virt_copy_condense
|
|
# GCN: %0:vgpr_32, %1:vgpr_32 = V_SWAP_B32 %1, %0, implicit $exec
|
|
---
|
|
name: swap_virt_copy_condense
|
|
registers:
|
|
- { id: 0, class: vgpr_32 }
|
|
- { id: 1, class: vgpr_32 }
|
|
- { id: 2, class: vgpr_32 }
|
|
body: |
|
|
bb.0:
|
|
%0 = IMPLICIT_DEF
|
|
%1 = IMPLICIT_DEF
|
|
%2 = COPY %0
|
|
%0 = COPY %1
|
|
%1 = COPY %2
|
|
...
|
|
|
|
# GCN-LABEL: name: swap_virt_copy_sparse
|
|
# GCN: %0:vgpr_32, %1:vgpr_32 = V_SWAP_B32 %1, %0, implicit $exec
|
|
---
|
|
name: swap_virt_copy_sparse
|
|
registers:
|
|
- { id: 0, class: vgpr_32 }
|
|
- { id: 1, class: vgpr_32 }
|
|
- { id: 2, class: vgpr_32 }
|
|
body: |
|
|
bb.0:
|
|
%0 = IMPLICIT_DEF
|
|
%1 = IMPLICIT_DEF
|
|
%2 = COPY %0
|
|
S_NOP 0
|
|
%0 = COPY %1
|
|
S_NOP 0
|
|
%1 = COPY %2
|
|
...
|
|
|
|
# GCN-LABEL: name: swap_virt_copy_subreg
|
|
# GCN: %0.sub0:vreg_64, %1.sub0:vreg_64 = V_SWAP_B32 %1.sub0, %0.sub0, implicit $exec
|
|
---
|
|
name: swap_virt_copy_subreg
|
|
registers:
|
|
- { id: 0, class: vreg_64 }
|
|
- { id: 1, class: vreg_64 }
|
|
- { id: 2, class: vreg_64 }
|
|
body: |
|
|
bb.0:
|
|
%0 = IMPLICIT_DEF
|
|
%1 = IMPLICIT_DEF
|
|
%2.sub0 = COPY %0.sub0
|
|
%2.sub1 = COPY %0.sub1
|
|
%0.sub0 = COPY %1.sub0
|
|
%0.sub1 = COPY %1.sub1
|
|
%1.sub0 = COPY %2.sub0
|
|
...
|
|
|
|
# GCN-LABEL: name: swap_virt_mov
|
|
# GCN: %0:vgpr_32, %1:vgpr_32 = V_SWAP_B32 %1, %0, implicit $exec
|
|
---
|
|
name: swap_virt_mov
|
|
registers:
|
|
- { id: 0, class: vgpr_32 }
|
|
- { id: 1, class: vgpr_32 }
|
|
- { id: 2, class: vgpr_32 }
|
|
body: |
|
|
bb.0:
|
|
%0 = IMPLICIT_DEF
|
|
%1 = IMPLICIT_DEF
|
|
%2 = V_MOV_B32_e32 %0, implicit $exec
|
|
%0 = V_MOV_B32_e32 %1, implicit $exec
|
|
%1 = V_MOV_B32_e32 %2, implicit $exec
|
|
...
|
|
|
|
# GCN-LABEL: name: swap_virt_read_x
|
|
# GCN: bb.0:
|
|
# GCN-NEXT: %0:vgpr_32 = IMPLICIT_DEF
|
|
# GCN-NEXT: %1:vgpr_32 = IMPLICIT_DEF
|
|
# GCN-NEXT: %2:vgpr_32 = COPY %0
|
|
# GCN-NEXT: %3:vgpr_32 = COPY %0
|
|
# GCN-NEXT: %0:vgpr_32 = COPY %1
|
|
# GCN-NEXT: %1:vgpr_32 = COPY %2
|
|
# GCN-NEXT: S_ENDPGM 0
|
|
|
|
---
|
|
name: swap_virt_read_x
|
|
registers:
|
|
- { id: 0, class: vgpr_32 }
|
|
- { id: 1, class: vgpr_32 }
|
|
- { id: 2, class: vgpr_32 }
|
|
- { id: 3, class: vgpr_32 }
|
|
body: |
|
|
bb.0:
|
|
%0 = IMPLICIT_DEF
|
|
%1 = IMPLICIT_DEF
|
|
%2 = COPY %0
|
|
%3 = COPY %0
|
|
%0 = COPY %1
|
|
%1 = COPY %2
|
|
S_ENDPGM 0
|
|
...
|
|
|
|
# GCN-LABEL: name: swap_virt_read_t_twice
|
|
# GCN: bb.0:
|
|
# GCN-NEXT: %0:vgpr_32 = IMPLICIT_DEF
|
|
# GCN-NEXT: %1:vgpr_32 = IMPLICIT_DEF
|
|
# GCN-NEXT: %2:vgpr_32 = COPY %0
|
|
# GCN-NEXT: %3:vgpr_32 = COPY %2
|
|
# GCN-NEXT: %0:vgpr_32, %1:vgpr_32 = V_SWAP_B32 %1, %0, implicit $exec
|
|
# GCN-NEXT: S_ENDPGM 0
|
|
|
|
---
|
|
name: swap_virt_read_t_twice
|
|
registers:
|
|
- { id: 0, class: vgpr_32 }
|
|
- { id: 1, class: vgpr_32 }
|
|
- { id: 2, class: vgpr_32 }
|
|
- { id: 3, class: vgpr_32 }
|
|
body: |
|
|
bb.0:
|
|
%0 = IMPLICIT_DEF
|
|
%1 = IMPLICIT_DEF
|
|
%2 = COPY %0
|
|
%3 = COPY %2
|
|
%0 = COPY %1
|
|
%1 = COPY %2
|
|
S_ENDPGM 0
|
|
...
|
|
|
|
# GCN-LABEL: name: swap_virt_clobber_y
|
|
# GCN: bb.0:
|
|
# GCN-NEXT: %0:vgpr_32 = IMPLICIT_DEF
|
|
# GCN-NEXT: %1:vgpr_32 = IMPLICIT_DEF
|
|
# GCN-NEXT: %2:vgpr_32 = COPY %0
|
|
# GCN-NEXT: %0:vgpr_32 = COPY %1
|
|
# GCN-NEXT: %1:vgpr_32 = IMPLICIT_DEF
|
|
# GCN-NEXT: %1:vgpr_32 = COPY %2
|
|
# GCN-NEXT: S_ENDPGM 0
|
|
|
|
---
|
|
name: swap_virt_clobber_y
|
|
registers:
|
|
- { id: 0, class: vgpr_32 }
|
|
- { id: 1, class: vgpr_32 }
|
|
- { id: 2, class: vgpr_32 }
|
|
body: |
|
|
bb.0:
|
|
%0 = IMPLICIT_DEF
|
|
%1 = IMPLICIT_DEF
|
|
%2 = COPY %0
|
|
%0 = COPY %1
|
|
%1 = IMPLICIT_DEF
|
|
%1 = COPY %2
|
|
S_ENDPGM 0
|
|
...
|
|
|
|
# GCN-LABEL: name: swap_virt_clobber_x1
|
|
# GCN: bb.0:
|
|
# GCN-NEXT: %0:vgpr_32 = IMPLICIT_DEF
|
|
# GCN-NEXT: %1:vgpr_32 = IMPLICIT_DEF
|
|
# GCN-NEXT: %2:vgpr_32 = COPY %0
|
|
# GCN-NEXT: %0:vgpr_32 = COPY %1
|
|
# GCN-NEXT: %0:vgpr_32 = IMPLICIT_DEF
|
|
# GCN-NEXT: %1:vgpr_32 = COPY %2
|
|
# GCN-NEXT: S_ENDPGM 0
|
|
|
|
---
|
|
name: swap_virt_clobber_x1
|
|
registers:
|
|
- { id: 0, class: vgpr_32 }
|
|
- { id: 1, class: vgpr_32 }
|
|
- { id: 2, class: vgpr_32 }
|
|
body: |
|
|
bb.0:
|
|
%0 = IMPLICIT_DEF
|
|
%1 = IMPLICIT_DEF
|
|
%2 = COPY %0
|
|
%0 = COPY %1
|
|
%0 = IMPLICIT_DEF
|
|
%1 = COPY %2
|
|
S_ENDPGM 0
|
|
...
|
|
|
|
# GCN-LABEL: name: swap_virt_clobber_x2
|
|
# GCN: bb.0:
|
|
# GCN-NEXT: %0:vgpr_32 = IMPLICIT_DEF
|
|
# GCN-NEXT: %1:vgpr_32 = IMPLICIT_DEF
|
|
# GCN-NEXT: %2:vgpr_32 = COPY %0
|
|
# GCN-NEXT: %0:vgpr_32 = IMPLICIT_DEF
|
|
# GCN-NEXT: %0:vgpr_32 = COPY %1
|
|
# GCN-NEXT: %1:vgpr_32 = COPY %2
|
|
# GCN-NEXT: S_ENDPGM 0
|
|
|
|
---
|
|
name: swap_virt_clobber_x2
|
|
registers:
|
|
- { id: 0, class: vgpr_32 }
|
|
- { id: 1, class: vgpr_32 }
|
|
- { id: 2, class: vgpr_32 }
|
|
body: |
|
|
bb.0:
|
|
%0 = IMPLICIT_DEF
|
|
%1 = IMPLICIT_DEF
|
|
%2 = COPY %0
|
|
%0 = IMPLICIT_DEF
|
|
%0 = COPY %1
|
|
%1 = COPY %2
|
|
S_ENDPGM 0
|
|
...
|
|
|
|
# GCN-LABEL: name: swap_virt_clobber_t
|
|
# GCN: bb.0:
|
|
# GCN-NEXT: %0:vgpr_32 = IMPLICIT_DEF
|
|
# GCN-NEXT: %1:vgpr_32 = IMPLICIT_DEF
|
|
# GCN-NEXT: %2:vgpr_32 = COPY %0
|
|
# GCN-NEXT: %0:vgpr_32 = COPY %1
|
|
# GCN-NEXT: %2:vgpr_32 = IMPLICIT_DEF
|
|
# GCN-NEXT: %1:vgpr_32 = COPY %2
|
|
# GCN-NEXT: S_ENDPGM 0
|
|
|
|
---
|
|
name: swap_virt_clobber_t
|
|
registers:
|
|
- { id: 0, class: vgpr_32 }
|
|
- { id: 1, class: vgpr_32 }
|
|
- { id: 2, class: vgpr_32 }
|
|
body: |
|
|
bb.0:
|
|
%0 = IMPLICIT_DEF
|
|
%1 = IMPLICIT_DEF
|
|
%2 = COPY %0
|
|
%0 = COPY %1
|
|
%2 = IMPLICIT_DEF
|
|
%1 = COPY %2
|
|
S_ENDPGM 0
|
|
...
|
|
|
|
# GCN-LABEL: name: swap_virt_copy_subreg_overlap_x_full
|
|
# GCN: bb.0:
|
|
# GCN-NEXT: %0:vreg_64 = IMPLICIT_DEF
|
|
# GCN-NEXT: %1:vreg_64 = IMPLICIT_DEF
|
|
# GCN-NEXT: %2.sub0:vreg_64 = COPY %0.sub0
|
|
# GCN-NEXT: %3:vreg_64 = COPY %0
|
|
# GCN-NEXT: %0.sub0:vreg_64 = COPY %1.sub0
|
|
# GCN-NEXT: %1.sub0:vreg_64 = COPY %2.sub0
|
|
---
|
|
name: swap_virt_copy_subreg_overlap_x_full
|
|
registers:
|
|
- { id: 0, class: vreg_64 }
|
|
- { id: 1, class: vreg_64 }
|
|
- { id: 2, class: vreg_64 }
|
|
- { id: 3, class: vreg_64 }
|
|
body: |
|
|
bb.0:
|
|
%0 = IMPLICIT_DEF
|
|
%1 = IMPLICIT_DEF
|
|
%2.sub0 = COPY %0.sub0
|
|
%3 = COPY %0
|
|
%0.sub0 = COPY %1.sub0
|
|
%1.sub0 = COPY %2.sub0
|
|
...
|
|
|
|
# GCN-LABEL: name: swap_virt_copy_subreg_overlap_x_part
|
|
# GCN: bb.0:
|
|
# GCN-NEXT: %0:vreg_128 = IMPLICIT_DEF
|
|
# GCN-NEXT: %1:vreg_64 = IMPLICIT_DEF
|
|
# GCN-NEXT: %2.sub0:vreg_64 = COPY %0.sub0
|
|
# GCN-NEXT: %3:vreg_64 = COPY %0.sub0_sub1
|
|
# GCN-NEXT: %0.sub0:vreg_128 = COPY %1.sub0
|
|
# GCN-NEXT: %1.sub0:vreg_64 = COPY %2.sub0
|
|
---
|
|
name: swap_virt_copy_subreg_overlap_x_part
|
|
registers:
|
|
- { id: 0, class: vreg_128 }
|
|
- { id: 1, class: vreg_64 }
|
|
- { id: 2, class: vreg_64 }
|
|
- { id: 3, class: vreg_64 }
|
|
body: |
|
|
bb.0:
|
|
%0 = IMPLICIT_DEF
|
|
%1 = IMPLICIT_DEF
|
|
%2.sub0 = COPY %0.sub0
|
|
%3 = COPY %0.sub0_sub1
|
|
%0.sub0 = COPY %1.sub0
|
|
%1.sub0 = COPY %2.sub0
|
|
...
|
|
|
|
# GCN-LABEL: name: swap_virt_copy_subreg_wide_y
|
|
# GCN: bb.0:
|
|
# GCN-NEXT: %0:vreg_64 = IMPLICIT_DEF
|
|
# GCN-NEXT: %1:vreg_64 = IMPLICIT_DEF
|
|
# GCN-NEXT: %2.sub0:vreg_64 = COPY %0.sub0
|
|
# GCN-NEXT: %0.sub0:vreg_64 = COPY %1.sub0
|
|
# GCN-NEXT: %1:vreg_64 = COPY %2
|
|
---
|
|
name: swap_virt_copy_subreg_wide_y
|
|
registers:
|
|
- { id: 0, class: vreg_64 }
|
|
- { id: 1, class: vreg_64 }
|
|
- { id: 2, class: vreg_64 }
|
|
body: |
|
|
bb.0:
|
|
%0 = IMPLICIT_DEF
|
|
%1 = IMPLICIT_DEF
|
|
%2.sub0 = COPY %0.sub0
|
|
%0.sub0 = COPY %1.sub0
|
|
%1 = COPY %2
|
|
...
|
|
|
|
# GCN-LABEL: name: swap_virt_b64
|
|
# GCN: bb.0:
|
|
# GCN-NEXT: %0:vreg_64 = IMPLICIT_DEF
|
|
# GCN-NEXT: %1:vreg_64 = IMPLICIT_DEF
|
|
# GCN-NEXT: %0.sub0:vreg_64, %1.sub0:vreg_64 = V_SWAP_B32 %1.sub0, %0.sub0, implicit $exec
|
|
# GCN-NEXT: %0.sub1:vreg_64, %1.sub1:vreg_64 = V_SWAP_B32 %1.sub1, %0.sub1, implicit $exec
|
|
---
|
|
name: swap_virt_b64
|
|
registers:
|
|
- { id: 0, class: vreg_64 }
|
|
- { id: 1, class: vreg_64 }
|
|
- { id: 2, class: vreg_64 }
|
|
body: |
|
|
bb.0:
|
|
%0 = IMPLICIT_DEF
|
|
%1 = IMPLICIT_DEF
|
|
%2 = COPY %0
|
|
%0 = COPY %1
|
|
%1 = COPY %2
|
|
...
|
|
|
|
# GCN-LABEL: name: swap_virt_b128
|
|
# GCN: bb.0:
|
|
# GCN-NEXT: %0:vreg_128 = IMPLICIT_DEF
|
|
# GCN-NEXT: %1:vreg_128 = IMPLICIT_DEF
|
|
# GCN-NEXT: %0.sub0:vreg_128, %1.sub0:vreg_128 = V_SWAP_B32 %1.sub0, %0.sub0, implicit $exec
|
|
# GCN-NEXT: %0.sub1:vreg_128, %1.sub1:vreg_128 = V_SWAP_B32 %1.sub1, %0.sub1, implicit $exec
|
|
# GCN-NEXT: %0.sub2:vreg_128, %1.sub2:vreg_128 = V_SWAP_B32 %1.sub2, %0.sub2, implicit $exec
|
|
# GCN-NEXT: %0.sub3:vreg_128, %1.sub3:vreg_128 = V_SWAP_B32 %1.sub3, %0.sub3, implicit $exec
|
|
---
|
|
name: swap_virt_b128
|
|
registers:
|
|
- { id: 0, class: vreg_128 }
|
|
- { id: 1, class: vreg_128 }
|
|
- { id: 2, class: vreg_128 }
|
|
body: |
|
|
bb.0:
|
|
%0 = IMPLICIT_DEF
|
|
%1 = IMPLICIT_DEF
|
|
%2 = COPY %0
|
|
%0 = COPY %1
|
|
%1 = COPY %2
|
|
...
|
|
|
|
# GCN-LABEL: name: swap_virt_b128_sub0_1
|
|
# GCN: bb.0:
|
|
# GCN-NEXT: %0:vreg_128 = IMPLICIT_DEF
|
|
# GCN-NEXT: %1:vreg_128 = IMPLICIT_DEF
|
|
# GCN-NEXT: %0.sub0:vreg_128, %1.sub0:vreg_128 = V_SWAP_B32 %1.sub0, %0.sub0, implicit $exec
|
|
# GCN-NEXT: %0.sub1:vreg_128, %1.sub1:vreg_128 = V_SWAP_B32 %1.sub1, %0.sub1, implicit $exec
|
|
# GCN-NEXT: S_ENDPGM 0
|
|
---
|
|
name: swap_virt_b128_sub0_1
|
|
registers:
|
|
- { id: 0, class: vreg_128 }
|
|
- { id: 1, class: vreg_128 }
|
|
- { id: 2, class: vreg_128 }
|
|
body: |
|
|
bb.0:
|
|
%0 = IMPLICIT_DEF
|
|
%1 = IMPLICIT_DEF
|
|
%2.sub0_sub1 = COPY %0.sub0_sub1
|
|
%0.sub0_sub1 = COPY %1.sub0_sub1
|
|
%1.sub0_sub1 = COPY %2.sub0_sub1
|
|
S_ENDPGM 0
|
|
...
|
|
|
|
# GCN-LABEL: name: swap_virt_b128_sub2_3
|
|
# GCN: bb.0:
|
|
# GCN-NEXT: %0:vreg_128 = IMPLICIT_DEF
|
|
# GCN-NEXT: %1:vreg_128 = IMPLICIT_DEF
|
|
# GCN-NEXT: %0.sub2:vreg_128, %1.sub2:vreg_128 = V_SWAP_B32 %1.sub2, %0.sub2, implicit $exec
|
|
# GCN-NEXT: %0.sub3:vreg_128, %1.sub3:vreg_128 = V_SWAP_B32 %1.sub3, %0.sub3, implicit $exec
|
|
# GCN-NEXT: S_ENDPGM 0
|
|
---
|
|
name: swap_virt_b128_sub2_3
|
|
registers:
|
|
- { id: 0, class: vreg_128 }
|
|
- { id: 1, class: vreg_128 }
|
|
- { id: 2, class: vreg_128 }
|
|
body: |
|
|
bb.0:
|
|
%0 = IMPLICIT_DEF
|
|
%1 = IMPLICIT_DEF
|
|
%2.sub2_sub3 = COPY %0.sub2_sub3
|
|
%0.sub2_sub3 = COPY %1.sub2_sub3
|
|
%1.sub2_sub3 = COPY %2.sub2_sub3
|
|
S_ENDPGM 0
|
|
...
|
|
|
|
|
|
# GCN-LABEL: name: swap_virt_s_to_s
|
|
# GCN: bb.0:
|
|
# GCN-NEXT: %0:sgpr_32 = IMPLICIT_DEF
|
|
# GCN-NEXT: %1:sgpr_32 = IMPLICIT_DEF
|
|
# GCN-NEXT: %2:sgpr_32 = COPY %0
|
|
# GCN-NEXT: %0:sgpr_32 = COPY %1
|
|
# GCN-NEXT: %1:sgpr_32 = COPY %2
|
|
---
|
|
name: swap_virt_s_to_s
|
|
registers:
|
|
- { id: 0, class: sgpr_32 }
|
|
- { id: 1, class: sgpr_32 }
|
|
- { id: 2, class: sgpr_32 }
|
|
body: |
|
|
bb.0:
|
|
%0 = IMPLICIT_DEF
|
|
%1 = IMPLICIT_DEF
|
|
%2 = COPY %0
|
|
%0 = COPY %1
|
|
%1 = COPY %2
|
|
...
|
|
|
|
# GCN-LABEL: name: swap_virt_copy_subreg_impdef_super
|
|
# GCN: %0.sub0:vreg_64, %1.sub0:vreg_64 = V_SWAP_B32 %1.sub0, %0.sub0, implicit $exec
|
|
---
|
|
name: swap_virt_copy_subreg_impdef_super
|
|
registers:
|
|
- { id: 0, class: vreg_64 }
|
|
- { id: 1, class: vreg_64 }
|
|
- { id: 2, class: vreg_64 }
|
|
body: |
|
|
bb.0:
|
|
%0 = IMPLICIT_DEF
|
|
%1 = IMPLICIT_DEF
|
|
%2.sub0 = COPY %0.sub0, implicit-def %2, implicit $exec
|
|
%2.sub1 = COPY %0.sub1
|
|
%0.sub0 = COPY %1.sub0
|
|
%0.sub1 = COPY %1.sub1
|
|
%1.sub0 = COPY %2.sub0
|
|
...
|
|
|
|
# GCN-LABEL: name: swap_virt_copy_subreg_impuse_x
|
|
# GCN: bb.0:
|
|
# GCN-NEXT: %0:vreg_64 = IMPLICIT_DEF
|
|
# GCN-NEXT: %1:vreg_64 = IMPLICIT_DEF
|
|
# GCN-NEXT: %2.sub0:vreg_64 = COPY %0.sub0
|
|
# GCN-NEXT: %2.sub1:vreg_64 = COPY %0.sub1
|
|
# GCN-NEXT: %0.sub0:vreg_64 = COPY %1.sub0, implicit %0
|
|
# GCN-NEXT: %0.sub1:vreg_64 = COPY %1.sub1
|
|
# GCN-NEXT: %1.sub0:vreg_64 = COPY %2.sub0
|
|
# GCN-NEXT: S_ENDPGM 0
|
|
---
|
|
name: swap_virt_copy_subreg_impuse_x
|
|
registers:
|
|
- { id: 0, class: vreg_64 }
|
|
- { id: 1, class: vreg_64 }
|
|
- { id: 2, class: vreg_64 }
|
|
body: |
|
|
bb.0:
|
|
%0 = IMPLICIT_DEF
|
|
%1 = IMPLICIT_DEF
|
|
%2.sub0 = COPY %0.sub0
|
|
%2.sub1 = COPY %0.sub1
|
|
%0.sub0 = COPY %1.sub0, implicit %0
|
|
%0.sub1 = COPY %1.sub1
|
|
%1.sub0 = COPY %2.sub0
|
|
S_ENDPGM 0
|
|
...
|
|
|
|
# GCN-LABEL: name: swap_exact_max_insns_apart
|
|
# GCN: bb.0:
|
|
# GCN-NEXT: %0:vgpr_32 = IMPLICIT_DEF
|
|
# GCN-NEXT: %1:vgpr_32 = IMPLICIT_DEF
|
|
# GCN-NEXT: %3:vgpr_32 = IMPLICIT_DEF
|
|
# GCN-NEXT: %4:vgpr_32 = COPY %3
|
|
# GCN-NEXT: %3:vgpr_32 = COPY %4
|
|
# GCN-NEXT: %4:vgpr_32 = COPY %3
|
|
# GCN-NEXT: %3:vgpr_32 = COPY %4
|
|
# GCN-NEXT: %4:vgpr_32 = COPY %3
|
|
# GCN-NEXT: %3:vgpr_32 = COPY %4
|
|
# GCN-NEXT: %4:vgpr_32 = COPY %3
|
|
# GCN-NEXT: %3:vgpr_32 = COPY %4
|
|
# GCN-NEXT: %4:vgpr_32 = COPY %3
|
|
# GCN-NEXT: %3:vgpr_32 = COPY %4
|
|
# GCN-NEXT: %4:vgpr_32 = COPY %3
|
|
# GCN-NEXT: %3:vgpr_32 = COPY %4
|
|
# GCN-NEXT: %4:vgpr_32 = COPY %3
|
|
# GCN-NEXT: %0:vgpr_32, %1:vgpr_32 = V_SWAP_B32 %1, %0, implicit $exec
|
|
# GCN-NEXT: S_ENDPGM 0
|
|
|
|
---
|
|
name: swap_exact_max_insns_apart
|
|
registers:
|
|
- { id: 0, class: vgpr_32 }
|
|
- { id: 1, class: vgpr_32 }
|
|
- { id: 2, class: vgpr_32 }
|
|
- { id: 3, class: vgpr_32 }
|
|
- { id: 4, class: vgpr_32 }
|
|
body: |
|
|
bb.0:
|
|
%0 = IMPLICIT_DEF
|
|
%1 = IMPLICIT_DEF
|
|
%2 = COPY %0
|
|
%3 = IMPLICIT_DEF
|
|
%4 = COPY %3
|
|
%3 = COPY %4
|
|
%4 = COPY %3
|
|
%3 = COPY %4
|
|
%4 = COPY %3
|
|
%3 = COPY %4
|
|
%4 = COPY %3
|
|
%3 = COPY %4
|
|
%4 = COPY %3
|
|
%3 = COPY %4
|
|
%4 = COPY %3
|
|
%3 = COPY %4
|
|
%4 = COPY %3
|
|
%0 = COPY %1
|
|
%1 = COPY %2
|
|
S_ENDPGM 0
|
|
...
|
|
|
|
# GCN-LABEL: name: swap_too_far
|
|
# GCN: bb.0:
|
|
# GCN-NEXT: %0:vgpr_32 = IMPLICIT_DEF
|
|
# GCN-NEXT: %1:vgpr_32 = IMPLICIT_DEF
|
|
# GCN-NEXT: %2:vgpr_32 = COPY %0
|
|
# GCN-NEXT: %3:vgpr_32 = IMPLICIT_DEF
|
|
# GCN-NEXT: %4:vgpr_32 = COPY %3
|
|
# GCN-NEXT: %3:vgpr_32 = COPY %4
|
|
# GCN-NEXT: %4:vgpr_32 = COPY %3
|
|
# GCN-NEXT: %3:vgpr_32 = COPY %4
|
|
# GCN-NEXT: %4:vgpr_32 = COPY %3
|
|
# GCN-NEXT: %3:vgpr_32 = COPY %4
|
|
# GCN-NEXT: %4:vgpr_32 = COPY %3
|
|
# GCN-NEXT: %3:vgpr_32 = COPY %4
|
|
# GCN-NEXT: %4:vgpr_32 = COPY %3
|
|
# GCN-NEXT: %3:vgpr_32 = COPY %4
|
|
# GCN-NEXT: %4:vgpr_32 = COPY %3
|
|
# GCN-NEXT: %3:vgpr_32 = COPY %4
|
|
# GCN-NEXT: %4:vgpr_32 = COPY %3
|
|
# GCN-NEXT: %3:vgpr_32 = COPY %4
|
|
# GCN-NEXT: %0:vgpr_32 = COPY %1
|
|
# GCN-NEXT: %1:vgpr_32 = COPY %2
|
|
# GCN-NEXT: S_ENDPGM 0
|
|
|
|
---
|
|
name: swap_too_far
|
|
registers:
|
|
- { id: 0, class: vgpr_32 }
|
|
- { id: 1, class: vgpr_32 }
|
|
- { id: 2, class: vgpr_32 }
|
|
- { id: 3, class: vgpr_32 }
|
|
- { id: 4, class: vgpr_32 }
|
|
body: |
|
|
bb.0:
|
|
%0 = IMPLICIT_DEF
|
|
%1 = IMPLICIT_DEF
|
|
%2 = COPY %0
|
|
%3 = IMPLICIT_DEF
|
|
%4 = COPY %3
|
|
%3 = COPY %4
|
|
%4 = COPY %3
|
|
%3 = COPY %4
|
|
%4 = COPY %3
|
|
%3 = COPY %4
|
|
%4 = COPY %3
|
|
%3 = COPY %4
|
|
%4 = COPY %3
|
|
%3 = COPY %4
|
|
%4 = COPY %3
|
|
%3 = COPY %4
|
|
%4 = COPY %3
|
|
%3 = COPY %4
|
|
%0 = COPY %1
|
|
%1 = COPY %2
|
|
S_ENDPGM 0
|
|
...
|