# RUN: llc -march=amdgcn -mcpu=fiji -run-pass si-insert-waits  %s -o - | FileCheck %s

--- |
  define amdgpu_kernel void @flat_zero_waitcnt(i32 addrspace(1)* %global4,
                                 <4 x i32> addrspace(1)* %global16,
                                 i32 addrspace(4)* %flat4,
                                 <4 x i32> addrspace(4)* %flat16) {
    ret void
  }

  define amdgpu_kernel void @single_fallthrough_successor_no_end_block_wait() {
    ret void
  }

  define amdgpu_kernel void @single_branch_successor_not_next_block() {
    ret void
  }

...
---

# CHECK-LABEL: name: flat_zero_waitcnt

# CHECK-LABEL: bb.0:
# CHECK: FLAT_LOAD_DWORD
# CHECK: FLAT_LOAD_DWORDX4
# Global loads will return in order so we should:
# s_waitcnt vmcnt(1) lgkmcnt(0)
# CHECK-NEXT: S_WAITCNT 113

# CHECK-LABEL: bb.1:
# CHECK: FLAT_LOAD_DWORD
# CHECK: S_WAITCNT 3952
# CHECK: FLAT_LOAD_DWORDX4
# The first load has no mem operand, so we should assume it accesses the flat
# address space.
# s_waitcnt vmcnt(0) lgkmcnt(0)
# CHECK-NEXT: S_WAITCNT 127

# CHECK-LABEL: bb.2:
# CHECK: FLAT_LOAD_DWORD
# CHECK: S_WAITCNT 3952
# CHECK: FLAT_LOAD_DWORDX4

# One outstand loads access the flat address space.
# s_waitcnt vmcnt(0) lgkmcnt(0)
# CHECK-NEXT: S_WAITCNT 127

name: flat_zero_waitcnt

body: |
  bb.0:
    successors: %bb.1
    %vgpr0 = FLAT_LOAD_DWORD %vgpr1_vgpr2, 0, 0, 0, implicit %exec, implicit %flat_scr :: (load 4 from %ir.global4)
    %vgpr3_vgpr4_vgpr5_vgpr6 = FLAT_LOAD_DWORDX4 %vgpr7_vgpr8, 0, 0, 0, implicit %exec, implicit %flat_scr :: (load 16 from %ir.global16)
    %vgpr0 = V_MOV_B32_e32 %vgpr1, implicit %exec
    S_BRANCH %bb.1

  bb.1:
    successors: %bb.2
    %vgpr0 = FLAT_LOAD_DWORD %vgpr1_vgpr2, 0, 0, 0, implicit %exec, implicit %flat_scr
    %vgpr3_vgpr4_vgpr5_vgpr6 = FLAT_LOAD_DWORDX4 %vgpr7_vgpr8, 0, 0, 0, implicit %exec, implicit %flat_scr :: (load 16 from %ir.global16)
    %vgpr0 = V_MOV_B32_e32 %vgpr1, implicit %exec
    S_BRANCH %bb.2

  bb.2:
    %vgpr0 = FLAT_LOAD_DWORD %vgpr1_vgpr2, 0, 0, 0, implicit %exec, implicit %flat_scr :: (load 4 from %ir.flat4)
    %vgpr3_vgpr4_vgpr5_vgpr6 = FLAT_LOAD_DWORDX4 %vgpr7_vgpr8, 0, 0, 0, implicit %exec, implicit %flat_scr :: (load 16 from %ir.flat16)
    %vgpr0 = V_MOV_B32_e32 %vgpr1, implicit %exec
    S_ENDPGM
...
---
# There is only a single fallthrough successor block, so there's no
# need to wait immediately.

# CHECK-LABEL: name: single_fallthrough_successor_no_end_block_wait
# CHECK:   %vgpr0 = FLAT_LOAD_DWORD %vgpr1_vgpr2
# CHECK-NOT: S_WAITCNT

# CHECK: bb.1:
# CHECK-NEXT: V_LSHLREV_B64
# CHECK-NEXT: S_WAITCNT 112
# CHECK-NEXT: FLAT_STORE_DWORD
name: single_fallthrough_successor_no_end_block_wait

body: |
  bb.0:
    successors: %bb.1
    %vgpr0 = FLAT_LOAD_DWORD %vgpr1_vgpr2, 0, 0, 0, implicit %exec, implicit %flat_scr

  bb.1:
    %vgpr3_vgpr4 = V_LSHLREV_B64 4, %vgpr7_vgpr8, implicit %exec
    FLAT_STORE_DWORD %vgpr3_vgpr4, %vgpr0, 0, 0, 0, implicit %exec, implicit %flat_scr
    S_ENDPGM
...
---
# The block has a single predecessor with a single successor, but it
# is not the next block so it's non-obvious that the wait is not needed.


# CHECK-LABEL: name: single_branch_successor_not_next_block
# CHECK:   %vgpr0 = FLAT_LOAD_DWORD %vgpr1_vgpr2
# CHECK-NEXT: S_WAITCNT 112

# CHECK: bb.1
# CHECK-NEXT: FLAT_STORE_DWORD
# CHECK-NEXT: S_ENDPGM

# CHECK: bb.2:
# CHECK-NEXT: V_LSHLREV_B64
# CHECK-NEXT: FLAT_STORE_DWORD
name: single_branch_successor_not_next_block

body: |
  bb.0:
    successors: %bb.2
    %vgpr0 = FLAT_LOAD_DWORD %vgpr1_vgpr2, 0, 0, 0, implicit %exec, implicit %flat_scr
   S_BRANCH %bb.2

  bb.1:
    FLAT_STORE_DWORD %vgpr8_vgpr9, %vgpr10, 0, 0, 0, implicit %exec, implicit %flat_scr
    S_ENDPGM

  bb.2:
     %vgpr3_vgpr4 = V_LSHLREV_B64 4, %vgpr7_vgpr8, implicit %exec
    FLAT_STORE_DWORD %vgpr3_vgpr4, %vgpr0, 0, 0, 0, implicit %exec, implicit %flat_scr
    S_ENDPGM
...