llvm/test/CodeGen/AMDGPU/break-vmem-soft-clauses.mir

# RUN: llc -march=amdgcn -mcpu=carrizo -verify-machineinstrs -run-pass post-RA-hazard-rec %s -o - | FileCheck -check-prefixes=GCN,XNACK %s
# RUN: llc -march=amdgcn -mcpu=fiji -verify-machineinstrs -run-pass post-RA-hazard-rec %s -o - | FileCheck -check-prefixes=GCN,NOXNACK %s
---
# Trivial clause at beginning of program
name: trivial_clause_load_flat4_x1

body: |
  bb.0:
    ; GCN-LABEL: name: trivial_clause_load_flat4_x1
    ; GCN: $vgpr0 = FLAT_LOAD_DWORD $vgpr2_vgpr3, 0, 0, 0, implicit $exec, implicit $flat_scr
    ; GCN-NEXT: S_ENDPGM

    $vgpr0 = FLAT_LOAD_DWORD $vgpr2_vgpr3, 0, 0, 0, implicit $exec, implicit $flat_scr
    S_ENDPGM
...
---
# Trivial clause at beginning of program
name: trivial_clause_load_flat4_x2

body: |
  bb.0:
    ; GCN-LABEL: name: trivial_clause_load_flat4_x2
    ; GCN: $vgpr0 = FLAT_LOAD_DWORD $vgpr2_vgpr3, 0, 0, 0, implicit $exec, implicit $flat_scr
    ; GCN-NEXT: $vgpr1 = FLAT_LOAD_DWORD $vgpr4_vgpr5, 0, 0, 0, implicit $exec, implicit $flat_scr
    ; GCN-NEXT: S_ENDPGM

    $vgpr0 = FLAT_LOAD_DWORD $vgpr2_vgpr3, 0, 0, 0, implicit $exec, implicit $flat_scr
    $vgpr1 = FLAT_LOAD_DWORD $vgpr4_vgpr5, 0, 0, 0, implicit $exec, implicit $flat_scr
    S_ENDPGM
...
---
# Trivial clause at beginning of program
name: trivial_clause_load_flat4_x3

body: |
  bb.0:
    ; GCN-LABEL: name: trivial_clause_load_flat4_x3
    ; GCN: $vgpr0 = FLAT_LOAD_DWORD $vgpr3_vgpr4, 0, 0, 0, implicit $exec, implicit $flat_scr
    ; GCN-NEXT: $vgpr1 = FLAT_LOAD_DWORD $vgpr5_vgpr6, 0, 0, 0, implicit $exec, implicit $flat_scr
    ; GCN-NEXT: $vgpr2 = FLAT_LOAD_DWORD $vgpr7_vgpr8, 0, 0, 0, implicit $exec, implicit $flat_scr
    ; GCN-NEXT: S_ENDPGM

    $vgpr0 = FLAT_LOAD_DWORD $vgpr3_vgpr4, 0, 0, 0, implicit $exec, implicit $flat_scr
    $vgpr1 = FLAT_LOAD_DWORD $vgpr5_vgpr6, 0, 0, 0, implicit $exec, implicit $flat_scr
    $vgpr2 = FLAT_LOAD_DWORD $vgpr7_vgpr8, 0, 0, 0, implicit $exec, implicit $flat_scr
    S_ENDPGM
...
---
# Trivial clause at beginning of program
name: trivial_clause_load_flat4_x4

body: |
  bb.0:
    ; GCN-LABEL: name: trivial_clause_load_flat4_x4
    ; GCN: $vgpr0 = FLAT_LOAD_DWORD $vgpr4_vgpr5, 0, 0, 0, implicit $exec, implicit $flat_scr
    ; GCN-NEXT: $vgpr1 = FLAT_LOAD_DWORD $vgpr6_vgpr7, 0, 0, 0, implicit $exec, implicit $flat_scr
    ; GCN-NEXT: $vgpr2 = FLAT_LOAD_DWORD $vgpr8_vgpr9, 0, 0, 0, implicit $exec, implicit $flat_scr
    ; GCN-NEXT: $vgpr3 = FLAT_LOAD_DWORD $vgpr10_vgpr11, 0, 0, 0, implicit $exec, implicit $flat_scr
    ; GCN-NEXT: S_ENDPGM

    $vgpr0 = FLAT_LOAD_DWORD $vgpr4_vgpr5, 0, 0, 0, implicit $exec, implicit $flat_scr
    $vgpr1 = FLAT_LOAD_DWORD $vgpr6_vgpr7, 0, 0, 0, implicit $exec, implicit $flat_scr
    $vgpr2 = FLAT_LOAD_DWORD $vgpr8_vgpr9, 0, 0, 0, implicit $exec, implicit $flat_scr
    $vgpr3 = FLAT_LOAD_DWORD $vgpr10_vgpr11, 0, 0, 0, implicit $exec, implicit $flat_scr
    S_ENDPGM
...
---
# Reuse of same input pointer is OK

name: trivial_clause_load_flat4_x2_sameptr
body: |
  bb.0:
    ; GCN-LABEL: name: trivial_clause_load_flat4_x2_sameptr
    ; GCN: $vgpr0 = FLAT_LOAD_DWORD $vgpr2_vgpr3, 0, 0, 0, implicit $exec, implicit $flat_scr
    ; GCN-NEXT: $vgpr1 = FLAT_LOAD_DWORD $vgpr2_vgpr3, 0, 0, 0, implicit $exec, implicit $flat_scr
    ; GCN-NEXT: S_ENDPGM

    $vgpr0 = FLAT_LOAD_DWORD $vgpr2_vgpr3, 0, 0, 0, implicit $exec, implicit $flat_scr
    $vgpr1 = FLAT_LOAD_DWORD $vgpr2_vgpr3, 0, 0, 0, implicit $exec, implicit $flat_scr
    S_ENDPGM
...
---
# 32-bit load partially clobbers its own ptr reg
name: flat_load4_overwrite_ptr_lo

body: |
  bb.0:
    ; GCN-LABEL: name: flat_load4_overwrite_ptr_lo
    ; GCN: $vgpr0 = FLAT_LOAD_DWORD $vgpr0_vgpr1, 0, 0, 0, implicit $exec, implicit $flat_scr
    ; GCN-NEXT: S_ENDPGM

    $vgpr0 = FLAT_LOAD_DWORD $vgpr0_vgpr1, 0, 0, 0, implicit $exec, implicit $flat_scr
    S_ENDPGM
...
---
# 32-bit load partially clobbers its own ptr reg
name: flat_load4_overwrite_ptr_hi

body: |
  bb.0:
    ; GCN-LABEL: name: flat_load4_overwrite_ptr_hi
    ; GCN: $vgpr1 = FLAT_LOAD_DWORD $vgpr0_vgpr1, 0, 0, 0, implicit $exec, implicit $flat_scr
    ; GCN-NEXT: S_ENDPGM

    $vgpr1 = FLAT_LOAD_DWORD $vgpr0_vgpr1, 0, 0, 0, implicit $exec, implicit $flat_scr
    S_ENDPGM
...
---
# 64-bit load clobbers its own ptr reg
name: flat_load8_overwrite_ptr

body: |
  bb.0:
    ; GCN-LABEL: name: flat_load8_overwrite_ptr
    ; GCN: $vgpr2_vgpr3 = FLAT_LOAD_DWORDX2 $vgpr2_vgpr3, 0, 0, 0, implicit $exec, implicit $flat_scr
    ; GCN-NEXT: S_ENDPGM

    $vgpr2_vgpr3 = FLAT_LOAD_DWORDX2 $vgpr2_vgpr3, 0, 0, 0, implicit $exec, implicit $flat_scr
    S_ENDPGM
...
---
# vmcnt has 4 bits, so maximum 16 outstanding loads. The waitcnt
# breaks the clause.


name: break_clause_at_max_clause_size_flat_load4

body: |
  bb.0:
    ; GCN-LABEL: name: break_clause_at_max_clause_size_flat_load4
    ; GCN: $vgpr2 = FLAT_LOAD_DWORD $vgpr0_vgpr1, 0, 0, 0, implicit $exec, implicit $flat_scr
    ; GCN-NEXT: $vgpr3 = FLAT_LOAD_DWORD $vgpr0_vgpr1, 0, 0, 0, implicit $exec, implicit $flat_scr
    ; GCN-NEXT: $vgpr4 = FLAT_LOAD_DWORD $vgpr0_vgpr1, 0, 0, 0, implicit $exec, implicit $flat_scr
    ; GCN-NEXT: $vgpr5 = FLAT_LOAD_DWORD $vgpr0_vgpr1, 0, 0, 0, implicit $exec, implicit $flat_scr
    ; GCN-NEXT: $vgpr6 = FLAT_LOAD_DWORD $vgpr0_vgpr1, 0, 0, 0, implicit $exec, implicit $flat_scr
    ; GCN-NEXT: $vgpr7 = FLAT_LOAD_DWORD $vgpr0_vgpr1, 0, 0, 0, implicit $exec, implicit $flat_scr
    ; GCN-NEXT: $vgpr8 = FLAT_LOAD_DWORD $vgpr0_vgpr1, 0, 0, 0, implicit $exec, implicit $flat_scr
    ; GCN-NEXT: $vgpr9 = FLAT_LOAD_DWORD $vgpr0_vgpr1, 0, 0, 0, implicit $exec, implicit $flat_scr
    ; GCN-NEXT: $vgpr10 = FLAT_LOAD_DWORD $vgpr0_vgpr1, 0, 0, 0, implicit $exec, implicit $flat_scr
    ; GCN-NEXT: $vgpr11 = FLAT_LOAD_DWORD $vgpr0_vgpr1, 0, 0, 0, implicit $exec, implicit $flat_scr
    ; GCN-NEXT: $vgpr12 = FLAT_LOAD_DWORD $vgpr0_vgpr1, 0, 0, 0, implicit $exec, implicit $flat_scr
    ; GCN-NEXT: $vgpr13 = FLAT_LOAD_DWORD $vgpr0_vgpr1, 0, 0, 0, implicit $exec, implicit $flat_scr
    ; GCN-NEXT: $vgpr14 = FLAT_LOAD_DWORD $vgpr0_vgpr1, 0, 0, 0, implicit $exec, implicit $flat_scr
    ; GCN-NEXT: $vgpr15 = FLAT_LOAD_DWORD $vgpr0_vgpr1, 0, 0, 0, implicit $exec, implicit $flat_scr
    ; GCN-NEXT: $vgpr16 = FLAT_LOAD_DWORD $vgpr0_vgpr1, 0, 0, 0, implicit $exec, implicit $flat_scr
    ; GCN-NEXT: $vgpr17 = FLAT_LOAD_DWORD $vgpr0_vgpr1, 0, 0, 0, implicit $exec, implicit $flat_scr
    ; XNACK-NEXT: S_NOP 0
    ; GCN-NEXT: $vgpr0 = FLAT_LOAD_DWORD $vgpr2_vgpr3, 0, 0, 0, implicit $exec, implicit $flat_scr
    ; GCN-NEXT: $sgpr0 = S_MOV_B32 $sgpr0, implicit $vgpr2, implicit $vgpr3, implicit $vgpr4, implicit $vgpr5, implicit $vgpr6, implicit $vgpr7, implicit $vgpr8, implicit $vgpr9, implicit $vgpr10, implicit $vgpr11, implicit $vgpr12, implicit $vgpr13, implicit $vgpr14, implicit $vgpr15, implicit $vgpr16, implicit $vgpr17, implicit $vgpr18
    ; GCN-NEXT: S_ENDPGM

    $vgpr2 = FLAT_LOAD_DWORD $vgpr0_vgpr1, 0, 0, 0, implicit $exec, implicit $flat_scr
    $vgpr3 = FLAT_LOAD_DWORD $vgpr0_vgpr1, 0, 0, 0, implicit $exec, implicit $flat_scr
    $vgpr4 = FLAT_LOAD_DWORD $vgpr0_vgpr1, 0, 0, 0, implicit $exec, implicit $flat_scr
    $vgpr5 = FLAT_LOAD_DWORD $vgpr0_vgpr1, 0, 0, 0, implicit $exec, implicit $flat_scr

    $vgpr6 = FLAT_LOAD_DWORD $vgpr0_vgpr1, 0, 0, 0, implicit $exec, implicit $flat_scr
    $vgpr7 = FLAT_LOAD_DWORD $vgpr0_vgpr1, 0, 0, 0, implicit $exec, implicit $flat_scr
    $vgpr8 = FLAT_LOAD_DWORD $vgpr0_vgpr1, 0, 0, 0, implicit $exec, implicit $flat_scr
    $vgpr9 = FLAT_LOAD_DWORD $vgpr0_vgpr1, 0, 0, 0, implicit $exec, implicit $flat_scr

    $vgpr10 = FLAT_LOAD_DWORD $vgpr0_vgpr1, 0, 0, 0, implicit $exec, implicit $flat_scr
    $vgpr11 = FLAT_LOAD_DWORD $vgpr0_vgpr1, 0, 0, 0, implicit $exec, implicit $flat_scr
    $vgpr12 = FLAT_LOAD_DWORD $vgpr0_vgpr1, 0, 0, 0, implicit $exec, implicit $flat_scr
    $vgpr13 = FLAT_LOAD_DWORD $vgpr0_vgpr1, 0, 0, 0, implicit $exec, implicit $flat_scr

    $vgpr14 = FLAT_LOAD_DWORD $vgpr0_vgpr1, 0, 0, 0, implicit $exec, implicit $flat_scr
    $vgpr15 = FLAT_LOAD_DWORD $vgpr0_vgpr1, 0, 0, 0, implicit $exec, implicit $flat_scr
    $vgpr16 = FLAT_LOAD_DWORD $vgpr0_vgpr1, 0, 0, 0, implicit $exec, implicit $flat_scr
    $vgpr17 = FLAT_LOAD_DWORD $vgpr0_vgpr1, 0, 0, 0, implicit $exec, implicit $flat_scr

    $vgpr0 = FLAT_LOAD_DWORD $vgpr2_vgpr3, 0, 0, 0, implicit $exec, implicit $flat_scr
    $sgpr0 = S_MOV_B32 $sgpr0, implicit $vgpr2, implicit $vgpr3, implicit $vgpr4, implicit $vgpr5, implicit $vgpr6, implicit $vgpr7, implicit $vgpr8, implicit $vgpr9, implicit $vgpr10, implicit $vgpr11, implicit $vgpr12, implicit $vgpr13, implicit $vgpr14, implicit $vgpr15, implicit $vgpr16, implicit $vgpr17, implicit $vgpr18
    S_ENDPGM
...
---

name: break_clause_simple_load_flat4_lo_ptr

body: |
  bb.0:
    ; GCN-LABEL: name: break_clause_simple_load_flat4_lo_ptr
    ; GCN: $vgpr0 = FLAT_LOAD_DWORD $vgpr2_vgpr3, 0, 0, 0, implicit $exec, implicit $flat_scr
    ; XNACK-NEXT: S_NOP 0
    ; GCN-NEXT: $vgpr2 = FLAT_LOAD_DWORD $vgpr4_vgpr5, 0, 0, 0, implicit $exec, implicit $flat_scr
    ; GCN-NEXT: S_ENDPGM

    $vgpr0 = FLAT_LOAD_DWORD $vgpr2_vgpr3, 0, 0, 0, implicit $exec, implicit $flat_scr
    $vgpr2 = FLAT_LOAD_DWORD $vgpr4_vgpr5, 0, 0, 0, implicit $exec, implicit $flat_scr
    S_ENDPGM
...
---

name: break_clause_simple_load_flat4_hi_ptr

body: |
  bb.0:
    ; GCN-LABEL: name: break_clause_simple_load_flat4_hi_ptr
    ; GCN: $vgpr0 = FLAT_LOAD_DWORD $vgpr2_vgpr3, 0, 0, 0, implicit $exec, implicit $flat_scr
    ; XNACK-NEXT: S_NOP 0
    ; GCN-NEXT: $vgpr3 = FLAT_LOAD_DWORD $vgpr4_vgpr5, 0, 0, 0, implicit $exec, implicit $flat_scr
    ; GCN-NEXT: S_ENDPGM

    $vgpr0 = FLAT_LOAD_DWORD $vgpr2_vgpr3, 0, 0, 0, implicit $exec, implicit $flat_scr
    $vgpr3 = FLAT_LOAD_DWORD $vgpr4_vgpr5, 0, 0, 0, implicit $exec, implicit $flat_scr
    S_ENDPGM
...
---

name: break_clause_simple_load_flat8_ptr

body: |
  bb.0:
    ; GCN-LABEL: name: break_clause_simple_load_flat8_ptr
    ; GCN: $vgpr0_vgpr1 = FLAT_LOAD_DWORDX2 $vgpr2_vgpr3, 0, 0, 0, implicit $exec, implicit $flat_scr
    ; XNACK-NEXT: S_NOP 0
    ; GCN-NEXT: $vgpr2_vgpr3 = FLAT_LOAD_DWORDX2 $vgpr4_vgpr5, 0, 0, 0, implicit $exec, implicit $flat_scr
    ; GCN-NEXT: S_ENDPGM

    $vgpr0_vgpr1 = FLAT_LOAD_DWORDX2 $vgpr2_vgpr3, 0, 0, 0, implicit $exec, implicit $flat_scr
    $vgpr2_vgpr3 = FLAT_LOAD_DWORDX2 $vgpr4_vgpr5, 0, 0, 0, implicit $exec, implicit $flat_scr
    S_ENDPGM
...
---


name: break_clause_simple_load_flat16_ptr

body: |
  bb.0:
    ; GCN-LABEL: name: break_clause_simple_load_flat16_ptr
    ; GCN: $vgpr0_vgpr1 = FLAT_LOAD_DWORDX2 $vgpr2_vgpr3, 0, 0, 0, implicit $exec, implicit $flat_scr
    ; XNACK-NEXT: S_NOP 0
    ; GCN-NEXT: $vgpr2_vgpr3_vgpr4_vgpr5 = FLAT_LOAD_DWORDX4 $vgpr6_vgpr7, 0, 0, 0, implicit $exec, implicit $flat_scr
    ; GCN-NEXT: S_ENDPGM
    $vgpr0_vgpr1 = FLAT_LOAD_DWORDX2 $vgpr2_vgpr3, 0, 0, 0, implicit $exec, implicit $flat_scr
    $vgpr2_vgpr3_vgpr4_vgpr5 = FLAT_LOAD_DWORDX4 $vgpr6_vgpr7, 0, 0, 0, implicit $exec, implicit $flat_scr
    S_ENDPGM
...
---

# The clause is broken by the waitcnt inserted at the end of the
# block, so no nop is needed.


name: break_clause_block_boundary_load_flat8_ptr

body: |
  ; GCN-LABEL: name: break_clause_block_boundary_load_flat8_ptr
  ; GCN: bb.0:
  ; GCN-NEXT:   successors: %bb.1(0x80000000)
  ; GCN:   $vgpr0_vgpr1 = FLAT_LOAD_DWORDX2 $vgpr2_vgpr3, 0, 0, 0, implicit $exec, implicit $flat_scr
  ; GCN: bb.1:
  ; XNACK-NEXT:  S_NOP 0
  ; GCN-NEXT:   $vgpr2_vgpr3 = FLAT_LOAD_DWORDX2 $vgpr4_vgpr5, 0, 0, 0, implicit $exec, implicit $flat_scr
  ; GCN-NEXT:   S_ENDPGM

  bb.0:
    $vgpr0_vgpr1 = FLAT_LOAD_DWORDX2 $vgpr2_vgpr3, 0, 0, 0, implicit $exec, implicit $flat_scr

  bb.1:
    $vgpr2_vgpr3 = FLAT_LOAD_DWORDX2 $vgpr4_vgpr5, 0, 0, 0, implicit $exec, implicit $flat_scr
    S_ENDPGM
...
---
# The load clobbers the pointer of the store, so it needs to break.

name: break_clause_store_load_into_ptr_flat4

body: |
  bb.0:
    ; GCN-LABEL: name: break_clause_store_load_into_ptr_flat4
    ; GCN: FLAT_STORE_DWORD $vgpr2_vgpr3, $vgpr0, 0, 0, 0, implicit $exec, implicit $flat_scr
    ; GCN-NEXT: $vgpr2 = FLAT_LOAD_DWORD $vgpr4_vgpr5, 0, 0, 0, implicit $exec, implicit $flat_scr
    ; GCN-NEXT: S_ENDPGM

    FLAT_STORE_DWORD $vgpr2_vgpr3, $vgpr0, 0, 0, 0, implicit $exec, implicit $flat_scr
    $vgpr2 = FLAT_LOAD_DWORD $vgpr4_vgpr5, 0, 0, 0, implicit $exec, implicit $flat_scr
    S_ENDPGM
...
---
# The load clobbers the data of the store, so it needs to break.
# FIXME: Would it be better to s_nop and wait later?

name: break_clause_store_load_into_data_flat4

body: |
  bb.0:
    ; GCN-LABEL: name: break_clause_store_load_into_data_flat4
    ; GCN: FLAT_STORE_DWORD $vgpr2_vgpr3, $vgpr0, 0, 0, 0, implicit $exec, implicit $flat_scr
    ; GCN-NEXT: $vgpr0 = FLAT_LOAD_DWORD $vgpr4_vgpr5, 0, 0, 0, implicit $exec, implicit $flat_scr
    ; GCN-NEXT: S_ENDPGM

    FLAT_STORE_DWORD $vgpr2_vgpr3, $vgpr0, 0, 0, 0, implicit $exec, implicit $flat_scr
    $vgpr0 = FLAT_LOAD_DWORD $vgpr4_vgpr5, 0, 0, 0, implicit $exec, implicit $flat_scr
    S_ENDPGM
...
---
# Regular VALU instruction breaks clause, no nop needed

name: valu_inst_breaks_clause

body: |
  bb.0:
    ; GCN-LABEL: name: valu_inst_breaks_clause
    ; GCN: $vgpr0 = FLAT_LOAD_DWORD $vgpr2_vgpr3, 0, 0, 0, implicit $exec, implicit $flat_scr
    ; GCN-NEXT: $vgpr8 = V_MOV_B32_e32 0, implicit $exec
    ; XNACK-NEXT: S_NOP 0
    ; GCN-NEXT: $vgpr2 = FLAT_LOAD_DWORD $vgpr4_vgpr5, 0, 0, 0, implicit $exec, implicit $flat_scr
    ; GCN-NEXT: S_ENDPGM

    $vgpr0 = FLAT_LOAD_DWORD $vgpr2_vgpr3, 0, 0, 0, implicit $exec, implicit $flat_scr
    $vgpr8 = V_MOV_B32_e32 0, implicit $exec
    $vgpr2 = FLAT_LOAD_DWORD $vgpr4_vgpr5, 0, 0, 0, implicit $exec, implicit $flat_scr
    S_ENDPGM
...
---
# Regular SALU instruction breaks clause, no nop needed

name: salu_inst_breaks_clause

body: |
  bb.0:
    ; GCN-LABEL: name: salu_inst_breaks_clause
    ; GCN: $vgpr0 = FLAT_LOAD_DWORD $vgpr2_vgpr3, 0, 0, 0, implicit $exec, implicit $flat_scr
    ; GCN-NEXT: $sgpr8 = S_MOV_B32 0
    ; XNACK-NEXT: S_NOP 0
    ; GCN-NEXT: $vgpr2 = FLAT_LOAD_DWORD $vgpr4_vgpr5, 0, 0, 0, implicit $exec, implicit $flat_scr
    ; GCN-NEXT: S_ENDPGM

    $vgpr0 = FLAT_LOAD_DWORD $vgpr2_vgpr3, 0, 0, 0, implicit $exec, implicit $flat_scr
    $sgpr8 = S_MOV_B32 0
    $vgpr2 = FLAT_LOAD_DWORD $vgpr4_vgpr5, 0, 0, 0, implicit $exec, implicit $flat_scr
    S_ENDPGM
...
---

name: ds_inst_breaks_clause

body: |
  bb.0:
    ; GCN-LABEL: name: ds_inst_breaks_clause
    ; GCN: $vgpr0 = FLAT_LOAD_DWORD $vgpr2_vgpr3, 0, 0, 0, implicit $exec, implicit $flat_scr
    ; GCN-NEXT: $vgpr8 = DS_READ_B32 $vgpr9, 0, 0, implicit $m0, implicit $exec
    ; XNACK-NEXT: S_NOP 0
    ; GCN-NEXT: $vgpr2 = FLAT_LOAD_DWORD $vgpr4_vgpr5, 0, 0, 0, implicit $exec, implicit $flat_scr
    ; GCN-NEXT: S_ENDPGM

    $vgpr0 = FLAT_LOAD_DWORD $vgpr2_vgpr3, 0, 0, 0, implicit $exec, implicit $flat_scr
    $vgpr8 = DS_READ_B32 $vgpr9, 0, 0, implicit $m0, implicit $exec
    $vgpr2 = FLAT_LOAD_DWORD $vgpr4_vgpr5, 0, 0, 0, implicit $exec, implicit $flat_scr
    S_ENDPGM
...
---

name: smrd_inst_breaks_clause

body: |
  bb.0:
    ; GCN-LABEL: name: smrd_inst_breaks_clause
    ; GCN: $vgpr0 = FLAT_LOAD_DWORD $vgpr2_vgpr3, 0, 0, 0, implicit $exec, implicit $flat_scr
    ; GCN-NEXT: $sgpr8 = S_LOAD_DWORD_IMM $sgpr0_sgpr1, 0, 0
    ; GCN-NEXT: $vgpr2 = FLAT_LOAD_DWORD $vgpr4_vgpr5, 0, 0, 0, implicit $exec, implicit $flat_scr
    ; GCN-NEXT: S_ENDPGM

    $vgpr0 = FLAT_LOAD_DWORD $vgpr2_vgpr3, 0, 0, 0, implicit $exec, implicit $flat_scr
    $sgpr8 = S_LOAD_DWORD_IMM $sgpr0_sgpr1, 0, 0
    $vgpr2 = FLAT_LOAD_DWORD $vgpr4_vgpr5, 0, 0, 0, implicit $exec, implicit $flat_scr
    S_ENDPGM
...
---
# FIXME: Should this be handled?
name: implicit_use_breaks_clause

body: |
  bb.0:
    ; GCN-LABEL: name: implicit_use_breaks_clause
    ; GCN: $vgpr0_vgpr1 = FLAT_LOAD_DWORDX2 $vgpr2_vgpr3, 0, 0, 0, implicit $exec, implicit $flat_scr, implicit $vgpr4_vgpr5
    ; XNACK-NEXT: S_NOP 0
    ; GCN-NEXT: $vgpr4_vgpr5 = FLAT_LOAD_DWORDX2 $vgpr6_vgpr7, 0, 0, 0, implicit $exec, implicit $flat_scr
    ; GCN-NEXT: S_ENDPGM

    $vgpr0_vgpr1 = FLAT_LOAD_DWORDX2 $vgpr2_vgpr3, 0, 0, 0, implicit $exec, implicit $flat_scr, implicit $vgpr4_vgpr5
    $vgpr4_vgpr5 = FLAT_LOAD_DWORDX2 $vgpr6_vgpr7, 0, 0, 0, implicit $exec, implicit $flat_scr
    S_ENDPGM
...
---
name: trivial_clause_load_mubuf4_x2

body: |
  bb.0:
    ; GCN-LABEL: name: trivial_clause_load_mubuf4_x2
    ; GCN: $vgpr1 = BUFFER_LOAD_DWORD_OFFEN $vgpr2, $sgpr0_sgpr1_sgpr2_sgpr3, $sgpr4, 0, 0, 0, 0, implicit $exec
    ; GCN-NEXT: $vgpr3 = BUFFER_LOAD_DWORD_OFFEN $vgpr4, $sgpr0_sgpr1_sgpr2_sgpr3, $sgpr4, 0, 0, 0, 0, implicit $exec
    ; GCN-NEXT: S_ENDPGM

    $vgpr1 = BUFFER_LOAD_DWORD_OFFEN $vgpr2, $sgpr0_sgpr1_sgpr2_sgpr3, $sgpr4, 0, 0, 0, 0, implicit $exec
    $vgpr3 = BUFFER_LOAD_DWORD_OFFEN $vgpr4, $sgpr0_sgpr1_sgpr2_sgpr3, $sgpr4, 0, 0, 0, 0, implicit $exec
    S_ENDPGM
...
---
name: break_clause_simple_load_mubuf_offen_ptr

body: |
  bb.0:
    ; GCN-LABEL: name: break_clause_simple_load_mubuf_offen_ptr
    ; GCN: $vgpr1 = BUFFER_LOAD_DWORD_OFFEN $vgpr2, $sgpr0_sgpr1_sgpr2_sgpr3, $sgpr4, 0, 0, 0, 0, implicit $exec
    ; XNACK-NEXT: S_NOP 0
    ; GCN-NEXT: $vgpr2 = BUFFER_LOAD_DWORD_OFFEN $vgpr3, $sgpr0_sgpr1_sgpr2_sgpr3, $sgpr4, 0, 0, 0, 0, implicit $exec
    ; GCN-NEXT: S_ENDPGM

    $vgpr1 = BUFFER_LOAD_DWORD_OFFEN $vgpr2, $sgpr0_sgpr1_sgpr2_sgpr3, $sgpr4, 0, 0, 0, 0, implicit $exec
    $vgpr2 = BUFFER_LOAD_DWORD_OFFEN $vgpr3, $sgpr0_sgpr1_sgpr2_sgpr3, $sgpr4, 0, 0, 0, 0, implicit $exec
    S_ENDPGM
...
---
# BUFFER instructions overwriting their own inputs is supposedly OK.

name: mubuf_load4_overwrite_ptr

body: |
  bb.0:
    ; GCN-LABEL: name: mubuf_load4_overwrite_ptr
    ; GCN: $vgpr0 = BUFFER_LOAD_DWORD_OFFEN $vgpr0, $sgpr0_sgpr1_sgpr2_sgpr3, $sgpr4, 0, 0, 0, 0, implicit $exec
    ; GCN-NEXT: $vgpr1 = V_MOV_B32_e32 0, implicit $exec
    ; GCN-NEXT: $vgpr2 = V_MOV_B32_e32 $vgpr0, implicit $exec
    ; GCN-NEXT: S_ENDPGM
    $vgpr0 = BUFFER_LOAD_DWORD_OFFEN $vgpr0, $sgpr0_sgpr1_sgpr2_sgpr3, $sgpr4, 0, 0, 0, 0, implicit $exec
    $vgpr1 = V_MOV_B32_e32 0, implicit $exec
    $vgpr2 = V_MOV_B32_e32 $vgpr0, implicit $exec
    S_ENDPGM
...
---
# Break a clause from interference between mubuf and flat instructions

name: break_clause_flat_load_mubuf_load

body: |
  bb.0:
    ; GCN-LABEL: name: break_clause_flat_load_mubuf_load
    ; GCN: $vgpr0 = FLAT_LOAD_DWORD $vgpr2_vgpr3, 0, 0, 0, implicit $exec, implicit $flat_scr
    ; XNACK-NEXT: S_NOP 0
    ; GCN-NEXT: $vgpr2 = BUFFER_LOAD_DWORD_OFFEN $vgpr1, $sgpr0_sgpr1_sgpr2_sgpr3, $sgpr4, 0, 0, 0, 0, implicit $exec
    ; GCN-NEXT: S_ENDPGM

    $vgpr0 = FLAT_LOAD_DWORD $vgpr2_vgpr3, 0, 0, 0, implicit $exec, implicit $flat_scr
    $vgpr2 = BUFFER_LOAD_DWORD_OFFEN $vgpr1, $sgpr0_sgpr1_sgpr2_sgpr3, $sgpr4, 0, 0, 0, 0, implicit $exec
    S_ENDPGM
...
# Break a clause from interference between mubuf and flat instructions

# GCN-LABEL: name: break_clause_mubuf_load_flat_load
# GCN: bb.0:
# GCN-NEXT: $vgpr0 = BUFFER_LOAD_DWORD_OFFEN $vgpr1, $sgpr0_sgpr1_sgpr2_sgpr3, $sgpr4
# XNACK-NEXT: S_NOP 0
# GCN-NEXT: $vgpr1 = FLAT_LOAD_DWORD $vgpr2_vgpr3
# GCN-NEXT: S_ENDPGM
name: break_clause_mubuf_load_flat_load

body: |
  bb.0:
    $vgpr0 = BUFFER_LOAD_DWORD_OFFEN $vgpr1, $sgpr0_sgpr1_sgpr2_sgpr3, $sgpr4, 0, 0, 0, 0, implicit $exec
    $vgpr1 = FLAT_LOAD_DWORD $vgpr2_vgpr3, 0, 0, 0, implicit $exec, implicit $flat_scr

    S_ENDPGM
...
---

name: break_clause_atomic_rtn_into_ptr_flat4

body: |
  bb.0:
    ; GCN-LABEL: name: break_clause_atomic_rtn_into_ptr_flat4
    ; GCN: $vgpr2 = FLAT_LOAD_DWORD $vgpr4_vgpr5, 0, 0, 0, implicit $exec, implicit $flat_scr
    ; XNACK-NEXT: S_NOP 0
    ; GCN-NEXT: $vgpr4 = FLAT_ATOMIC_ADD_RTN $vgpr5_vgpr6, $vgpr7, 0, 0, implicit $exec, implicit $flat_scr
    ; GCN-NEXT: S_ENDPGM

    $vgpr2 = FLAT_LOAD_DWORD $vgpr4_vgpr5, 0, 0, 0, implicit $exec, implicit $flat_scr
    $vgpr4 = FLAT_ATOMIC_ADD_RTN $vgpr5_vgpr6, $vgpr7, 0, 0, implicit $exec, implicit $flat_scr
    S_ENDPGM
...
---
name: break_clause_atomic_nortn_ptr_load_flat4

body: |
  bb.0:
    ; GCN-LABEL: name: break_clause_atomic_nortn_ptr_load_flat4
    ; GCN: FLAT_ATOMIC_ADD $vgpr0_vgpr1, $vgpr2, 0, 0, implicit $exec, implicit $flat_scr
    ; GCN-NEXT: $vgpr2 = FLAT_LOAD_DWORD $vgpr3_vgpr4, 0, 0, 0, implicit $exec, implicit $flat_scr
    ; GCN-NEXT: S_ENDPGM

    FLAT_ATOMIC_ADD $vgpr0_vgpr1, $vgpr2, 0, 0, implicit $exec, implicit $flat_scr
    $vgpr2 = FLAT_LOAD_DWORD $vgpr3_vgpr4, 0, 0, 0, implicit $exec, implicit $flat_scr
    S_ENDPGM
...
---

name: break_clause_atomic_rtn_into_ptr_mubuf4

body: |
  bb.0:
    ; GCN-LABEL: name: break_clause_atomic_rtn_into_ptr_mubuf4
    ; GCN: $vgpr1 = BUFFER_LOAD_DWORD_OFFEN $vgpr2, $sgpr0_sgpr1_sgpr2_sgpr3, $sgpr4, 0, 0, 0, 0, implicit $exec
    ; XNACK-NEXT: S_NOP 0
    ; GCN-NEXT: $vgpr2 = BUFFER_ATOMIC_ADD_OFFEN_RTN $vgpr2, $vgpr5, $sgpr0_sgpr1_sgpr2_sgpr3, $sgpr4, 0, 0, implicit $exec
    ; GCN-NEXT: S_ENDPGM

    $vgpr1 = BUFFER_LOAD_DWORD_OFFEN $vgpr2, $sgpr0_sgpr1_sgpr2_sgpr3, $sgpr4, 0, 0, 0, 0, implicit $exec
    $vgpr2 = BUFFER_ATOMIC_ADD_OFFEN_RTN $vgpr2, $vgpr5, $sgpr0_sgpr1_sgpr2_sgpr3, $sgpr4, 0, 0, implicit $exec
    S_ENDPGM
...
---

name: break_clause_atomic_nortn_ptr_load_mubuf4

body: |
  bb.0:
    ; GCN-LABEL: name: break_clause_atomic_nortn_ptr_load_mubuf4
    ; GCN: BUFFER_ATOMIC_ADD_OFFEN $vgpr0, $vgpr1, $sgpr0_sgpr1_sgpr2_sgpr3, $sgpr4, 0, 0, implicit $exec
    ; GCN-NEXT: $vgpr1 = BUFFER_LOAD_DWORD_OFFEN $vgpr2, $sgpr0_sgpr1_sgpr2_sgpr3, $sgpr4, 0, 0, 0, 0, implicit $exec
    ; GCN-NEXT: S_ENDPGM

    BUFFER_ATOMIC_ADD_OFFEN $vgpr0, $vgpr1, $sgpr0_sgpr1_sgpr2_sgpr3, $sgpr4, 0, 0, implicit $exec
    $vgpr1 = BUFFER_LOAD_DWORD_OFFEN $vgpr2, $sgpr0_sgpr1_sgpr2_sgpr3, $sgpr4, 0, 0, 0, 0, implicit $exec
    S_ENDPGM
...
---
# Make sure there is no assert on mubuf instructions which do not have
# vaddr, and don't add register to track.
name: no_break_clause_mubuf_load_novaddr

body: |
  bb.0:
    ; GCN-LABEL: name: no_break_clause_mubuf_load_novaddr
    ; GCN: $vgpr1 = BUFFER_LOAD_DWORD_OFFSET $sgpr0_sgpr1_sgpr2_sgpr3, $sgpr4, 0, 0, 0, 0, implicit $exec
    ; GCN-NEXT: $vgpr3 = BUFFER_LOAD_DWORD_OFFSET $sgpr0_sgpr1_sgpr2_sgpr3, $sgpr4, 0, 0, 0, 0, implicit $exec
    ; GCN-NEXT: S_ENDPGM
    $vgpr1 = BUFFER_LOAD_DWORD_OFFSET $sgpr0_sgpr1_sgpr2_sgpr3, $sgpr4, 0, 0, 0, 0, implicit $exec
    $vgpr3 = BUFFER_LOAD_DWORD_OFFSET $sgpr0_sgpr1_sgpr2_sgpr3, $sgpr4, 0, 0, 0, 0, implicit $exec
    S_ENDPGM
...
---
# Loads and stores using different addresses theoretically does not
# need a nop
name: mix_load_store_clause
body: |
  bb.0:
    ; GCN-LABEL: name: mix_load_store_clause
    ; GCN: FLAT_STORE_DWORD $vgpr0_vgpr1, $vgpr5, 0, 0, 0, implicit $exec, implicit $flat_scr
    ; GCN-NEXT: $vgpr10 = FLAT_LOAD_DWORD $vgpr2_vgpr3, 0, 0, 0, implicit $exec, implicit $flat_scr
    ; XNACK-NEXT: S_NOP 0
    ; GCN-NEXT: FLAT_STORE_DWORD $vgpr2_vgpr3, $vgpr6, 0, 0, 0, implicit $exec, implicit $flat_scr
    ; GCN-NEXT: $vgpr11 = FLAT_LOAD_DWORD $vgpr4_vgpr5, 0, 0, 0, implicit $exec, implicit $flat_scr

    FLAT_STORE_DWORD $vgpr0_vgpr1, $vgpr5, 0, 0, 0, implicit $exec, implicit $flat_scr
    $vgpr10 = FLAT_LOAD_DWORD $vgpr2_vgpr3, 0, 0, 0, implicit $exec, implicit $flat_scr
    FLAT_STORE_DWORD $vgpr2_vgpr3, $vgpr6, 0, 0, 0, implicit $exec, implicit $flat_scr
    $vgpr11 = FLAT_LOAD_DWORD $vgpr4_vgpr5, 0, 0, 0, implicit $exec, implicit $flat_scr
    S_ENDPGM
...
---
# Loads and stores using the same address needs a nop.

name: mix_load_store_clause_same_address
body: |
  bb.0:
    ; GCN-LABEL: name: mix_load_store_clause_same_address
    ; GCN: FLAT_STORE_DWORD $vgpr0_vgpr1, $vgpr5, 0, 0, 0, implicit $exec, implicit $flat_scr
    ; GCN-NEXT: $vgpr10 = FLAT_LOAD_DWORD $vgpr2_vgpr3, 0, 0, 0, implicit $exec, implicit $flat_scr
    ; XNACK-NEXT: S_NOP 0
    ; GCN-NEXT: FLAT_STORE_DWORD $vgpr0_vgpr1, $vgpr6, 0, 0, 0, implicit $exec, implicit $flat_scr
    ; GCN-NEXT: $vgpr11 = FLAT_LOAD_DWORD $vgpr4_vgpr5, 0, 0, 0, implicit $exec, implicit $flat_scr

    FLAT_STORE_DWORD $vgpr0_vgpr1, $vgpr5, 0, 0, 0, implicit $exec, implicit $flat_scr
    $vgpr10 = FLAT_LOAD_DWORD $vgpr2_vgpr3, 0, 0, 0, implicit $exec, implicit $flat_scr
    FLAT_STORE_DWORD $vgpr0_vgpr1, $vgpr6, 0, 0, 0, implicit $exec, implicit $flat_scr
    $vgpr11 = FLAT_LOAD_DWORD $vgpr4_vgpr5, 0, 0, 0, implicit $exec, implicit $flat_scr
    S_ENDPGM
...