1
0
mirror of https://github.com/RPCS3/llvm-mirror.git synced 2024-11-26 04:32:44 +01:00
llvm-mirror/test/CodeGen/AMDGPU/vmem-to-salu-hazard.mir
Piotr Sobczak b9148f5d85 [AMDGPU] Extend buffer intrinsics with swizzling
Summary:
Extend cachepolicy operand in the new VMEM buffer intrinsics
to supply information whether the buffer data is swizzled.
Also, propagate this information to MIR.

Intrinsics updated:
int_amdgcn_raw_buffer_load
int_amdgcn_raw_buffer_load_format
int_amdgcn_raw_buffer_store
int_amdgcn_raw_buffer_store_format
int_amdgcn_raw_tbuffer_load
int_amdgcn_raw_tbuffer_store
int_amdgcn_struct_buffer_load
int_amdgcn_struct_buffer_load_format
int_amdgcn_struct_buffer_store
int_amdgcn_struct_buffer_store_format
int_amdgcn_struct_tbuffer_load
int_amdgcn_struct_tbuffer_store

Furthermore, disable merging of VMEM buffer instructions
in SI Load/Store optimizer, if the "swizzled" bit on the instruction
is on.

The default value of the bit is 0, meaning that data in buffer
is linear and buffer instructions can be merged.

There is no difference in the generated code with this commit.
However, in the future it will be expected that front-ends
use buffer intrinsics with correct "swizzled" bit set.

Reviewers: arsenm, nhaehnle, tpr

Reviewed By: nhaehnle

Subscribers: arsenm, kzhuravl, jvesely, wdng, nhaehnle, yaxunl, dstuttard, tpr, t-tye, arphaman, jfb, Petar.Avramovic, llvm-commits

Tags: #llvm

Differential Revision: https://reviews.llvm.org/D68200

llvm-svn: 373491
2019-10-02 17:22:36 +00:00

225 lines
6.2 KiB
YAML

# RUN: llc -march=amdgcn -mcpu=gfx1010 -verify-machineinstrs -run-pass post-RA-hazard-rec -o - %s | FileCheck -check-prefix=GCN %s
# GCN-LABEL: name: vmem_write_sgpr
# GCN: BUFFER_LOAD_DWORD_OFFEN
# GCN-NEXT: V_NOP
# GCN-NEXT: S_MOV_B32
---
name: vmem_write_sgpr
body: |
bb.0:
$sgpr0_sgpr1_sgpr2_sgpr3 = IMPLICIT_DEF
$sgpr4 = IMPLICIT_DEF
$vgpr0 = IMPLICIT_DEF
$vgpr1 = BUFFER_LOAD_DWORD_OFFEN $vgpr0, $sgpr0_sgpr1_sgpr2_sgpr3, $sgpr4, 0, 0, 0, 0, 0, 0, implicit $exec
$sgpr0 = S_MOV_B32 0
...
# GCN-LABEL: name: vmem_smem_write_sgpr
# GCN: BUFFER_LOAD_DWORD_OFFEN
# GCN-NEXT: V_NOP
# GCN-NEXT: S_LOAD_DWORD_IMM
---
name: vmem_smem_write_sgpr
body: |
bb.0:
$sgpr0_sgpr1_sgpr2_sgpr3 = IMPLICIT_DEF
$sgpr4 = IMPLICIT_DEF
$vgpr0 = IMPLICIT_DEF
$vgpr1 = BUFFER_LOAD_DWORD_OFFEN $vgpr0, $sgpr0_sgpr1_sgpr2_sgpr3, $sgpr4, 0, 0, 0, 0, 0, 0, implicit $exec
$sgpr0 = S_LOAD_DWORD_IMM $sgpr0_sgpr1, 0, 0, 0
...
# GCN-LABEL: name: vmem_snop_write_sgpr
# GCN: BUFFER_LOAD_DWORD_OFFEN
# GCN-NEXT: S_NOP
# GCN-NEXT: V_NOP
# GCN-NEXT: S_MOV_B32
---
name: vmem_snop_write_sgpr
body: |
bb.0:
$sgpr0_sgpr1_sgpr2_sgpr3 = IMPLICIT_DEF
$sgpr4 = IMPLICIT_DEF
$vgpr0 = IMPLICIT_DEF
$vgpr1 = BUFFER_LOAD_DWORD_OFFEN $vgpr0, $sgpr0_sgpr1_sgpr2_sgpr3, $sgpr4, 0, 0, 0, 0, 0, 0, implicit $exec
S_NOP 0
$sgpr0 = S_MOV_B32 0
...
# GCN-LABEL: name: vmem_valu_write_sgpr
# GCN: BUFFER_LOAD_DWORD_OFFEN
# GCN-NEXT: V_ADD_F32
# GCN-NEXT: S_MOV_B32
---
name: vmem_valu_write_sgpr
body: |
bb.0:
$sgpr0_sgpr1_sgpr2_sgpr3 = IMPLICIT_DEF
$sgpr4 = IMPLICIT_DEF
$vgpr0 = IMPLICIT_DEF
$vgpr1 = BUFFER_LOAD_DWORD_OFFEN $vgpr0, $sgpr0_sgpr1_sgpr2_sgpr3, $sgpr4, 0, 0, 0, 0, 0, 0, implicit $exec
$vgpr2 = V_ADD_F32_e32 $vgpr1, $vgpr1, implicit $exec
$sgpr0 = S_MOV_B32 0
...
# GCN-LABEL: name: vmem_swait0_write_sgpr
# GCN: BUFFER_LOAD_DWORD_OFFEN
# GCN-NEXT: S_WAITCNT
# GCN-NEXT: S_MOV_B32
---
name: vmem_swait0_write_sgpr
body: |
bb.0:
$sgpr0_sgpr1_sgpr2_sgpr3 = IMPLICIT_DEF
$sgpr4 = IMPLICIT_DEF
$vgpr0 = IMPLICIT_DEF
$vgpr1 = BUFFER_LOAD_DWORD_OFFEN $vgpr0, $sgpr0_sgpr1_sgpr2_sgpr3, $sgpr4, 0, 0, 0, 0, 0, 0, implicit $exec
S_WAITCNT 0
$sgpr0 = S_MOV_B32 0
...
# GCN-LABEL: name: vmem_swait_any_write_sgpr
# GCN: BUFFER_LOAD_DWORD_OFFEN
# GCN-NEXT: S_WAITCNT
# GCN-NEXT: V_NOP
# GCN-NEXT: S_MOV_B32
---
name: vmem_swait_any_write_sgpr
body: |
bb.0:
$sgpr0_sgpr1_sgpr2_sgpr3 = IMPLICIT_DEF
$sgpr4 = IMPLICIT_DEF
$vgpr0 = IMPLICIT_DEF
$vgpr1 = BUFFER_LOAD_DWORD_OFFEN $vgpr0, $sgpr0_sgpr1_sgpr2_sgpr3, $sgpr4, 0, 0, 0, 0, 0, 0, implicit $exec
S_WAITCNT 1
$sgpr0 = S_MOV_B32 0
...
# GCN-LABEL: name: vmem_write_exec_impread
# GCN: BUFFER_LOAD_DWORD_OFFEN
# GCN: V_NOP
# GCN-NEXT: S_MOV_B64
---
name: vmem_write_exec_impread
body: |
bb.0:
$sgpr0_sgpr1_sgpr2_sgpr3 = IMPLICIT_DEF
$sgpr4 = IMPLICIT_DEF
$vgpr0 = IMPLICIT_DEF
$vgpr1 = BUFFER_LOAD_DWORD_OFFEN $vgpr0, $sgpr0_sgpr1_sgpr2_sgpr3, $sgpr4, 0, 0, 0, 0, 0, 0, implicit $exec
$exec = S_MOV_B64 7
...
# GCN-LABEL: name: vmem_write_exec_expread
# GCN: BUFFER_LOAD_DWORD_OFFEN
# GCN-NEXT: V_NOP
# GCN-NEXT: S_MOV_B64
---
name: vmem_write_exec_expread
body: |
bb.0:
$sgpr0_sgpr1_sgpr2_sgpr3 = IMPLICIT_DEF
$vgpr0 = IMPLICIT_DEF
$vgpr1 = BUFFER_LOAD_DWORD_OFFEN $vgpr0, $sgpr0_sgpr1_sgpr2_sgpr3, $exec_lo, 0, 0, 0, 0, 0, 0, implicit $exec
$exec = S_MOV_B64 7
...
# GCN-LABEL: name: ds_write_m0
# GCN: DS_READ_B32
# GCN-NEXT: V_NOP
# GCN-NEXT: S_MOV_B32
---
name: ds_write_m0
body: |
bb.0:
$sgpr0_sgpr1_sgpr2_sgpr3 = IMPLICIT_DEF
$sgpr4 = IMPLICIT_DEF
$vgpr0 = IMPLICIT_DEF
$vgpr1 = DS_READ_B32 $vgpr0, 0, 0, implicit $m0, implicit $exec
$m0 = S_MOV_B32 7
...
# GCN-LABEL: name: vmem_write_sgpr_fall_through
# GCN: BUFFER_LOAD_DWORD_OFFEN
# GCN: V_NOP
# GCN-NEXT: S_MOV_B32
---
name: vmem_write_sgpr_fall_through
body: |
bb.0:
successors: %bb.1
$sgpr0_sgpr1_sgpr2_sgpr3 = IMPLICIT_DEF
$sgpr4 = IMPLICIT_DEF
$vgpr0 = IMPLICIT_DEF
$vgpr1 = BUFFER_LOAD_DWORD_OFFEN $vgpr0, $sgpr0_sgpr1_sgpr2_sgpr3, $sgpr4, 0, 0, 0, 0, 0, 0, implicit $exec
bb.1:
$sgpr0 = S_MOV_B32 0
...
# GCN-LABEL: name: vmem_write_sgpr_branch
# GCN: BUFFER_LOAD_DWORD_OFFEN
# GCN-NEXT: S_BRANCH
# GCN: V_NOP
# GCN-NEXT: S_MOV_B32
---
name: vmem_write_sgpr_branch
body: |
bb.0:
successors: %bb.1
$sgpr0_sgpr1_sgpr2_sgpr3 = IMPLICIT_DEF
$sgpr4 = IMPLICIT_DEF
$vgpr0 = IMPLICIT_DEF
$vgpr1 = BUFFER_LOAD_DWORD_OFFEN $vgpr0, $sgpr0_sgpr1_sgpr2_sgpr3, $sgpr4, 0, 0, 0, 0, 0, 0, implicit $exec
S_BRANCH %bb.1
bb.1:
$sgpr0 = S_MOV_B32 0
...
# GCN-LABEL: name: vmem_write_sgpr_branch_around
# GCN: BUFFER_LOAD_DWORD_OFFEN
# GCN-NEXT: S_BRANCH
# GCN: bb.2:
# GCN-NEXT: V_NOP
# GCN-NEXT: S_MOV_B32
---
name: vmem_write_sgpr_branch_around
body: |
bb.0:
successors: %bb.2
$sgpr0_sgpr1_sgpr2_sgpr3 = IMPLICIT_DEF
$sgpr4 = IMPLICIT_DEF
$vgpr0 = IMPLICIT_DEF
$vgpr1 = BUFFER_LOAD_DWORD_OFFEN $vgpr0, $sgpr0_sgpr1_sgpr2_sgpr3, $sgpr4, 0, 0, 0, 0, 0, 0, implicit $exec
S_BRANCH %bb.2
bb.1:
successors: %bb.2
S_WAITCNT 0
bb.2:
$sgpr0 = S_MOV_B32 0
...
# GCN-LABEL: name: vmem_write_sgpr_branch_backedge
# GCN: $vgpr0 = IMPLICIT_DEF
# GCN-NEXT: V_NOP
# GCN-NEXT: S_MOV_B32
---
name: vmem_write_sgpr_branch_backedge
body: |
bb.0:
successors: %bb.1
$sgpr0_sgpr1_sgpr2_sgpr3 = IMPLICIT_DEF
$sgpr4 = IMPLICIT_DEF
$vgpr0 = IMPLICIT_DEF
$sgpr0 = S_MOV_B32 0
bb.1:
$vgpr1 = BUFFER_LOAD_DWORD_OFFEN $vgpr0, $sgpr0_sgpr1_sgpr2_sgpr3, $sgpr4, 0, 0, 0, 0, 0, 0, implicit $exec
S_BRANCH %bb.0
...
# GCN-LABEL: name: ds_write_exec
# GCN: DS_WRITE_B32_gfx9
# GCN-NEXT: V_NOP
# GCN-NEXT: S_MOV_B32
---
name: ds_write_exec
body: |
bb.0:
$vgpr0 = IMPLICIT_DEF
$vgpr1 = IMPLICIT_DEF
DS_WRITE_B32_gfx9 $vgpr0, $vgpr1, 0, 0, implicit $exec
$exec_lo = S_MOV_B32 -1
...