mirror of
https://github.com/RPCS3/llvm-mirror.git
synced 2024-11-25 12:12:47 +01:00
6705a324ed
The hardware has created a real mess in the naming for add/sub, which have been renamed basically every generation. Switch the carry out pseudos to have the gfx9/gfx10 names. We were using the original SI/CI v_add_i32/v_sub_i32 names. Later targets reintroduced these names as carryless instructions with a saturating clamp bit, which we do not define. Do this rename so we can unambiguously add these missing instructions. The carry-in versions should also be renamed, but at least those had a consistent _u32 name to begin with. The 16-bit instructions were also renamed, but aren't ambiguous. This does regress assembler error message quality in some cases. In mismatched wave32/wave64 situations, this will switch from "unsupported instruction" to "invalid operand", with the error pointing at the wrong position. I couldn't quite follow how the assembler selects these, but the previous behavior seemed accidental to me. It looked like there was a partial attempt to handle this which was never completed (i.e. there is an AMDGPUOperand::isBoolReg but it isn't used for anything).
231 lines
8.6 KiB
YAML
231 lines
8.6 KiB
YAML
# NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py
|
|
# RUN: llc -mtriple=amdgcn-amd-amdhsa -verify-machineinstrs -run-pass si-fold-operands,dead-mi-elimination %s -o - | FileCheck -check-prefix=GCN %s
|
|
|
|
---
|
|
|
|
# First operand is FI is in a VGPR, other operand is a VGPR
|
|
name: shrink_vgpr_fi_vgpr_v_add_i32_e64_no_carry_out_use
|
|
tracksRegLiveness: true
|
|
stack:
|
|
- { id: 0, type: default, offset: 0, size: 64, alignment: 16 }
|
|
body: |
|
|
bb.0:
|
|
liveins: $vgpr0
|
|
|
|
; GCN-LABEL: name: shrink_vgpr_fi_vgpr_v_add_i32_e64_no_carry_out_use
|
|
; GCN: liveins: $vgpr0
|
|
; GCN: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 %stack.0, implicit $exec
|
|
; GCN: [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr0
|
|
; GCN: [[V_ADD_CO_U32_e32_:%[0-9]+]]:vgpr_32 = V_ADD_CO_U32_e32 [[V_MOV_B32_e32_]], [[COPY]], implicit-def $vcc, implicit $exec
|
|
; GCN: S_ENDPGM 0, implicit [[V_ADD_CO_U32_e32_]]
|
|
%0:vgpr_32 = V_MOV_B32_e32 %stack.0, implicit $exec
|
|
%1:vgpr_32 = COPY $vgpr0
|
|
%2:vgpr_32, %3:sreg_64 = V_ADD_CO_U32_e64 %0, %1, 0, implicit $exec
|
|
S_ENDPGM 0, implicit %2
|
|
|
|
...
|
|
|
|
---
|
|
|
|
# First operand is a VGPR, other operand FI is in a VGPR
|
|
name: shrink_vgpr_vgpr_fi_v_add_i32_e64_no_carry_out_use
|
|
tracksRegLiveness: true
|
|
stack:
|
|
- { id: 0, type: default, offset: 0, size: 64, alignment: 16 }
|
|
body: |
|
|
bb.0:
|
|
liveins: $vgpr0
|
|
|
|
; GCN-LABEL: name: shrink_vgpr_vgpr_fi_v_add_i32_e64_no_carry_out_use
|
|
; GCN: liveins: $vgpr0
|
|
; GCN: [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr0
|
|
; GCN: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 %stack.0, implicit $exec
|
|
; GCN: [[V_ADD_CO_U32_e32_:%[0-9]+]]:vgpr_32 = V_ADD_CO_U32_e32 [[COPY]], [[V_MOV_B32_e32_]], implicit-def $vcc, implicit $exec
|
|
; GCN: S_ENDPGM 0, implicit [[V_ADD_CO_U32_e32_]]
|
|
%0:vgpr_32 = COPY $vgpr0
|
|
%1:vgpr_32 = V_MOV_B32_e32 %stack.0, implicit $exec
|
|
%2:vgpr_32, %3:sreg_64 = V_ADD_CO_U32_e64 %0, %1, 0, implicit $exec
|
|
S_ENDPGM 0, implicit %2
|
|
|
|
...
|
|
|
|
---
|
|
|
|
# First operand is FI is in an SGPR, other operand is a VGPR
|
|
name: shrink_vgpr_fi_sgpr_v_add_i32_e64_no_carry_out_use
|
|
tracksRegLiveness: true
|
|
stack:
|
|
- { id: 0, type: default, offset: 0, size: 64, alignment: 16 }
|
|
body: |
|
|
bb.0:
|
|
liveins: $sgpr0
|
|
|
|
; GCN-LABEL: name: shrink_vgpr_fi_sgpr_v_add_i32_e64_no_carry_out_use
|
|
; GCN: liveins: $sgpr0
|
|
; GCN: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 %stack.0, implicit $exec
|
|
; GCN: [[COPY:%[0-9]+]]:sreg_32_xm0 = COPY $sgpr0
|
|
; GCN: [[V_ADD_CO_U32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_CO_U32_e64_1:%[0-9]+]]:sreg_64 = V_ADD_CO_U32_e64 [[COPY]], [[V_MOV_B32_e32_]], 0, implicit $exec
|
|
; GCN: S_ENDPGM 0, implicit [[V_ADD_CO_U32_e64_]]
|
|
%0:vgpr_32 = V_MOV_B32_e32 %stack.0, implicit $exec
|
|
%1:sreg_32_xm0 = COPY $sgpr0
|
|
%2:vgpr_32, %3:sreg_64 = V_ADD_CO_U32_e64 %0, %1, 0, implicit $exec
|
|
S_ENDPGM 0, implicit %2
|
|
|
|
...
|
|
|
|
---
|
|
|
|
# First operand is an SGPR, other operand FI is in a VGPR
|
|
name: shrink_sgpr_vgpr_fi_v_add_i32_e64_no_carry_out_use
|
|
tracksRegLiveness: true
|
|
stack:
|
|
- { id: 0, type: default, offset: 0, size: 64, alignment: 16 }
|
|
body: |
|
|
bb.0:
|
|
liveins: $sgpr0
|
|
|
|
; GCN-LABEL: name: shrink_sgpr_vgpr_fi_v_add_i32_e64_no_carry_out_use
|
|
; GCN: liveins: $sgpr0
|
|
; GCN: [[COPY:%[0-9]+]]:sreg_32_xm0 = COPY $sgpr0
|
|
; GCN: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 %stack.0, implicit $exec
|
|
; GCN: [[V_ADD_CO_U32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_CO_U32_e64_1:%[0-9]+]]:sreg_64 = V_ADD_CO_U32_e64 [[V_MOV_B32_e32_]], [[COPY]], 0, implicit $exec
|
|
; GCN: S_ENDPGM 0, implicit [[V_ADD_CO_U32_e64_]]
|
|
%0:sreg_32_xm0 = COPY $sgpr0
|
|
%1:vgpr_32 = V_MOV_B32_e32 %stack.0, implicit $exec
|
|
%2:vgpr_32, %3:sreg_64 = V_ADD_CO_U32_e64 %0, %1, 0, implicit $exec
|
|
S_ENDPGM 0, implicit %2
|
|
|
|
...
|
|
|
|
---
|
|
|
|
# First operand is FI is in an SGPR, other operand is a VGPR
|
|
name: shrink_sgpr_fi_vgpr_v_add_i32_e64_no_carry_out_use
|
|
tracksRegLiveness: true
|
|
stack:
|
|
- { id: 0, type: default, offset: 0, size: 64, alignment: 16 }
|
|
body: |
|
|
bb.0:
|
|
liveins: $vgpr0
|
|
|
|
; GCN-LABEL: name: shrink_sgpr_fi_vgpr_v_add_i32_e64_no_carry_out_use
|
|
; GCN: liveins: $vgpr0
|
|
; GCN: [[S_MOV_B32_:%[0-9]+]]:sreg_32_xm0 = S_MOV_B32 %stack.0
|
|
; GCN: [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr0
|
|
; GCN: [[V_ADD_CO_U32_e32_:%[0-9]+]]:vgpr_32 = V_ADD_CO_U32_e32 [[S_MOV_B32_]], [[COPY]], implicit-def $vcc, implicit $exec
|
|
; GCN: S_ENDPGM 0, implicit [[V_ADD_CO_U32_e32_]]
|
|
%0:sreg_32_xm0 = S_MOV_B32 %stack.0
|
|
%1:vgpr_32 = COPY $vgpr0
|
|
%2:vgpr_32, %3:sreg_64 = V_ADD_CO_U32_e64 %0, %1, 0, implicit $exec
|
|
S_ENDPGM 0, implicit %2
|
|
|
|
...
|
|
|
|
---
|
|
|
|
# First operand is a VGPR, other operand FI is in an SGPR
|
|
name: shrink_vgpr_sgpr_fi_v_add_i32_e64_no_carry_out_use
|
|
tracksRegLiveness: true
|
|
stack:
|
|
- { id: 0, type: default, offset: 0, size: 64, alignment: 16}
|
|
body: |
|
|
bb.0:
|
|
liveins: $vgpr0
|
|
|
|
; GCN-LABEL: name: shrink_vgpr_sgpr_fi_v_add_i32_e64_no_carry_out_use
|
|
; GCN: liveins: $vgpr0
|
|
; GCN: [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr0
|
|
; GCN: [[S_MOV_B32_:%[0-9]+]]:sreg_32_xm0 = S_MOV_B32 %stack.0
|
|
; GCN: [[V_ADD_CO_U32_e32_:%[0-9]+]]:vgpr_32 = V_ADD_CO_U32_e32 [[S_MOV_B32_]], [[COPY]], implicit-def $vcc, implicit $exec
|
|
; GCN: S_ENDPGM 0, implicit [[V_ADD_CO_U32_e32_]]
|
|
%0:vgpr_32 = COPY $vgpr0
|
|
%1:sreg_32_xm0 = S_MOV_B32 %stack.0
|
|
%2:vgpr_32, %3:sreg_64 = V_ADD_CO_U32_e64 %0, %1, 0, implicit $exec
|
|
S_ENDPGM 0, implicit %2
|
|
|
|
...
|
|
|
|
---
|
|
|
|
# First operand is FI is in a VGPR, other operand is an inline imm in a VGPR
|
|
name: shrink_vgpr_imm_fi_vgpr_v_add_i32_e64_no_carry_out_use
|
|
tracksRegLiveness: true
|
|
stack:
|
|
- { id: 0, type: default, offset: 0, size: 64, alignment: 16 }
|
|
body: |
|
|
bb.0:
|
|
|
|
; GCN-LABEL: name: shrink_vgpr_imm_fi_vgpr_v_add_i32_e64_no_carry_out_use
|
|
; GCN: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 %stack.0, implicit $exec
|
|
; GCN: [[V_ADD_CO_U32_e32_:%[0-9]+]]:vgpr_32 = V_ADD_CO_U32_e32 16, [[V_MOV_B32_e32_]], implicit-def $vcc, implicit $exec
|
|
; GCN: S_ENDPGM 0, implicit [[V_ADD_CO_U32_e32_]]
|
|
%0:vgpr_32 = V_MOV_B32_e32 %stack.0, implicit $exec
|
|
%1:vgpr_32 = V_MOV_B32_e32 16, implicit $exec
|
|
%2:vgpr_32, %3:sreg_64 = V_ADD_CO_U32_e64 %0, %1, 0, implicit $exec
|
|
S_ENDPGM 0, implicit %2
|
|
|
|
...
|
|
|
|
---
|
|
|
|
# First operand is an inline imm in a VGPR, other operand FI is in a VGPR
|
|
name: shrink_vgpr_imm_vgpr_fi_v_add_i32_e64_no_carry_out_use
|
|
tracksRegLiveness: true
|
|
stack:
|
|
- { id: 0, type: default, offset: 0, size: 64, alignment: 16 }
|
|
body: |
|
|
bb.0:
|
|
|
|
; GCN-LABEL: name: shrink_vgpr_imm_vgpr_fi_v_add_i32_e64_no_carry_out_use
|
|
; GCN: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 %stack.0, implicit $exec
|
|
; GCN: [[V_ADD_CO_U32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_CO_U32_e64_1:%[0-9]+]]:sreg_64 = V_ADD_CO_U32_e64 16, [[V_MOV_B32_e32_]], 0, implicit $exec
|
|
; GCN: S_ENDPGM 0, implicit [[V_ADD_CO_U32_e64_]]
|
|
%0:vgpr_32 = V_MOV_B32_e32 16, implicit $exec
|
|
%1:vgpr_32 = V_MOV_B32_e32 %stack.0, implicit $exec
|
|
%2:vgpr_32, %3:sreg_64 = V_ADD_CO_U32_e64 %0, %1, 0, implicit $exec
|
|
S_ENDPGM 0, implicit %2
|
|
|
|
...
|
|
|
|
---
|
|
|
|
# First operand is FI is in a VGPR, other operand is an literal constant in a VGPR
|
|
name: shrink_vgpr_k_fi_vgpr_v_add_i32_e64_no_carry_out_use
|
|
tracksRegLiveness: true
|
|
stack:
|
|
- { id: 0, type: default, offset: 0, size: 64, alignment: 16 }
|
|
body: |
|
|
bb.0:
|
|
|
|
; GCN-LABEL: name: shrink_vgpr_k_fi_vgpr_v_add_i32_e64_no_carry_out_use
|
|
; GCN: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 %stack.0, implicit $exec
|
|
; GCN: [[V_ADD_CO_U32_e32_:%[0-9]+]]:vgpr_32 = V_ADD_CO_U32_e32 1234, [[V_MOV_B32_e32_]], implicit-def $vcc, implicit $exec
|
|
; GCN: S_ENDPGM 0, implicit [[V_ADD_CO_U32_e32_]]
|
|
%0:vgpr_32 = V_MOV_B32_e32 %stack.0, implicit $exec
|
|
%1:vgpr_32 = V_MOV_B32_e32 1234, implicit $exec
|
|
%2:vgpr_32, %3:sreg_64 = V_ADD_CO_U32_e64 %0, %1, 0, implicit $exec
|
|
S_ENDPGM 0, implicit %2
|
|
|
|
...
|
|
|
|
---
|
|
|
|
# First operand is a literal constant in a VGPR, other operand FI is in a VGPR
|
|
name: shrink_vgpr_k_vgpr_fi_v_add_i32_e64_no_carry_out_use
|
|
tracksRegLiveness: true
|
|
stack:
|
|
- { id: 0, type: default, offset: 0, size: 64, alignment: 16 }
|
|
body: |
|
|
bb.0:
|
|
|
|
; GCN-LABEL: name: shrink_vgpr_k_vgpr_fi_v_add_i32_e64_no_carry_out_use
|
|
; GCN: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 1234, implicit $exec
|
|
; GCN: [[V_ADD_CO_U32_e32_:%[0-9]+]]:vgpr_32 = V_ADD_CO_U32_e32 %stack.0, [[V_MOV_B32_e32_]], implicit-def $vcc, implicit $exec
|
|
; GCN: S_ENDPGM 0, implicit [[V_ADD_CO_U32_e32_]]
|
|
%0:vgpr_32 = V_MOV_B32_e32 1234, implicit $exec
|
|
%1:vgpr_32 = V_MOV_B32_e32 %stack.0, implicit $exec
|
|
%2:vgpr_32, %3:sreg_64 = V_ADD_CO_U32_e64 %0, %1, 0, implicit $exec
|
|
S_ENDPGM 0, implicit %2
|
|
|
|
...
|