llvm/test/CodeGen/AMDGPU/lds-dma-war.ll

; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 6
; RUN: llc -mtriple=amdgpu9.42 < %s | FileCheck %s

declare void @llvm.amdgcn.global.load.lds(ptr addrspace(1) nocapture, ptr addrspace(3) nocapture, i32, i32, i32)
declare void @llvm.amdgcn.raw.buffer.load.lds(<4 x i32>, ptr addrspace(3) nocapture, i32, i32, i32, i32, i32)

define i32 @lds_then_dma.global.singlethread(ptr addrspace(1) %g, ptr addrspace(3) inreg %lds) #0 {
; CHECK-LABEL: lds_then_dma.global.singlethread:
; CHECK:       ; %bb.0:
; CHECK-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; CHECK-NEXT:    v_mov_b32_e32 v2, s0
; CHECK-NEXT:    s_mov_b32 m0, s0
; CHECK-NEXT:    ds_read_b32 v2, v2
; CHECK-NEXT:    global_load_lds_dword v[0:1], off
; CHECK-NEXT:    s_waitcnt lgkmcnt(0)
; CHECK-NEXT:    v_mov_b32_e32 v0, v2
; CHECK-NEXT:    s_waitcnt vmcnt(0)
; CHECK-NEXT:    s_setpc_b64 s[30:31]
  %v = load i32, ptr addrspace(3) %lds, align 4
  fence syncscope("singlethread") release, !mmra !0
  call void @llvm.amdgcn.global.load.lds(ptr addrspace(1) %g, ptr addrspace(3) %lds, i32 4, i32 0, i32 0)
  ret i32 %v
}

define i32 @lds_then_dma.global.wave(ptr addrspace(1) %g, ptr addrspace(3) inreg %lds) #0 {
; CHECK-LABEL: lds_then_dma.global.wave:
; CHECK:       ; %bb.0:
; CHECK-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; CHECK-NEXT:    v_mov_b32_e32 v2, s0
; CHECK-NEXT:    s_mov_b32 m0, s0
; CHECK-NEXT:    ds_read_b32 v2, v2
; CHECK-NEXT:    global_load_lds_dword v[0:1], off
; CHECK-NEXT:    s_waitcnt lgkmcnt(0)
; CHECK-NEXT:    v_mov_b32_e32 v0, v2
; CHECK-NEXT:    s_waitcnt vmcnt(0)
; CHECK-NEXT:    s_setpc_b64 s[30:31]
  %v = load i32, ptr addrspace(3) %lds, align 4
  fence syncscope("wavefront") release, !mmra !0
  call void @llvm.amdgcn.global.load.lds(ptr addrspace(1) %g, ptr addrspace(3) %lds, i32 4, i32 0, i32 0)
  ret i32 %v
}

define i32 @lds_then_dma.global.wg(ptr addrspace(1) %g, ptr addrspace(3) inreg %lds) #0 {
; CHECK-LABEL: lds_then_dma.global.wg:
; CHECK:       ; %bb.0:
; CHECK-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; CHECK-NEXT:    v_mov_b32_e32 v2, s0
; CHECK-NEXT:    s_mov_b32 m0, s0
; CHECK-NEXT:    ds_read_b32 v2, v2
; CHECK-NEXT:    s_waitcnt lgkmcnt(0)
; CHECK-NEXT:    global_load_lds_dword v[0:1], off
; CHECK-NEXT:    v_mov_b32_e32 v0, v2
; CHECK-NEXT:    s_waitcnt vmcnt(0)
; CHECK-NEXT:    s_setpc_b64 s[30:31]
  %v = load i32, ptr addrspace(3) %lds, align 4
  fence syncscope("workgroup") release, !mmra !0
  call void @llvm.amdgcn.global.load.lds(ptr addrspace(1) %g, ptr addrspace(3) %lds, i32 4, i32 0, i32 0)
  ret i32 %v
}

define i32 @lds_then_dma.global.agent(ptr addrspace(1) %g, ptr addrspace(3) inreg %lds) #0 {
; CHECK-LABEL: lds_then_dma.global.agent:
; CHECK:       ; %bb.0:
; CHECK-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; CHECK-NEXT:    v_mov_b32_e32 v2, s0
; CHECK-NEXT:    ds_read_b32 v2, v2
; CHECK-NEXT:    s_mov_b32 m0, s0
; CHECK-NEXT:    s_waitcnt lgkmcnt(0)
; CHECK-NEXT:    global_load_lds_dword v[0:1], off
; CHECK-NEXT:    v_mov_b32_e32 v0, v2
; CHECK-NEXT:    s_waitcnt vmcnt(0)
; CHECK-NEXT:    s_setpc_b64 s[30:31]
  %v = load i32, ptr addrspace(3) %lds, align 4
  fence syncscope("agent") release, !mmra !0
  call void @llvm.amdgcn.global.load.lds(ptr addrspace(1) %g, ptr addrspace(3) %lds, i32 4, i32 0, i32 0)
  ret i32 %v
}

define i32 @lds_then_dma.global.system(ptr addrspace(1) %g, ptr addrspace(3) inreg %lds) #0 {
; CHECK-LABEL: lds_then_dma.global.system:
; CHECK:       ; %bb.0:
; CHECK-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; CHECK-NEXT:    v_mov_b32_e32 v2, s0
; CHECK-NEXT:    ds_read_b32 v2, v2
; CHECK-NEXT:    s_mov_b32 m0, s0
; CHECK-NEXT:    s_waitcnt lgkmcnt(0)
; CHECK-NEXT:    global_load_lds_dword v[0:1], off
; CHECK-NEXT:    v_mov_b32_e32 v0, v2
; CHECK-NEXT:    s_waitcnt vmcnt(0)
; CHECK-NEXT:    s_setpc_b64 s[30:31]
  %v = load i32, ptr addrspace(3) %lds, align 4
  fence release, !mmra !0
  call void @llvm.amdgcn.global.load.lds(ptr addrspace(1) %g, ptr addrspace(3) %lds, i32 4, i32 0, i32 0)
  ret i32 %v
}

define i32 @lds_then_dma.global.singlethread.partial(ptr addrspace(1) %g, ptr addrspace(3) inreg %lds, ptr addrspace(3) inreg %lds2) #0 {
; CHECK-LABEL: lds_then_dma.global.singlethread.partial:
; CHECK:       ; %bb.0:
; CHECK-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; CHECK-NEXT:    s_mov_b32 m0, s0
; CHECK-NEXT:    v_mov_b32_e32 v2, s0
; CHECK-NEXT:    v_mov_b32_e32 v3, s1
; CHECK-NEXT:    ds_read_b32 v2, v2
; CHECK-NEXT:    ds_read_b32 v3, v3
; CHECK-NEXT:    global_load_lds_dword v[0:1], off
; CHECK-NEXT:    s_waitcnt lgkmcnt(0)
; CHECK-NEXT:    v_add_u32_e32 v0, v2, v3
; CHECK-NEXT:    s_waitcnt vmcnt(0)
; CHECK-NEXT:    s_setpc_b64 s[30:31]
  %v1 = load i32, ptr addrspace(3) %lds, align 4
  fence syncscope("singlethread") release, !mmra !0
  %v2 = load i32, ptr addrspace(3) %lds2, align 4
  call void @llvm.amdgcn.global.load.lds(ptr addrspace(1) %g, ptr addrspace(3) %lds, i32 4, i32 0, i32 0)
  %s = add i32 %v1, %v2
  ret i32 %s
}

define i32 @lds_then_dma.global.wave.partial(ptr addrspace(1) %g, ptr addrspace(3) inreg %lds, ptr addrspace(3) inreg %lds2) #0 {
; CHECK-LABEL: lds_then_dma.global.wave.partial:
; CHECK:       ; %bb.0:
; CHECK-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; CHECK-NEXT:    s_mov_b32 m0, s0
; CHECK-NEXT:    v_mov_b32_e32 v2, s0
; CHECK-NEXT:    v_mov_b32_e32 v3, s1
; CHECK-NEXT:    ds_read_b32 v2, v2
; CHECK-NEXT:    ds_read_b32 v3, v3
; CHECK-NEXT:    global_load_lds_dword v[0:1], off
; CHECK-NEXT:    s_waitcnt lgkmcnt(0)
; CHECK-NEXT:    v_add_u32_e32 v0, v2, v3
; CHECK-NEXT:    s_waitcnt vmcnt(0)
; CHECK-NEXT:    s_setpc_b64 s[30:31]
  %v1 = load i32, ptr addrspace(3) %lds, align 4
  fence syncscope("wavefront") release, !mmra !0
  %v2 = load i32, ptr addrspace(3) %lds2, align 4
  call void @llvm.amdgcn.global.load.lds(ptr addrspace(1) %g, ptr addrspace(3) %lds, i32 4, i32 0, i32 0)
  %s = add i32 %v1, %v2
  ret i32 %s
}

define i32 @lds_then_dma.global.wg.partial(ptr addrspace(1) %g, ptr addrspace(3) inreg %lds, ptr addrspace(3) inreg %lds2) #0 {
; CHECK-LABEL: lds_then_dma.global.wg.partial:
; CHECK:       ; %bb.0:
; CHECK-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; CHECK-NEXT:    s_mov_b32 m0, s0
; CHECK-NEXT:    v_mov_b32_e32 v2, s0
; CHECK-NEXT:    v_mov_b32_e32 v3, s1
; CHECK-NEXT:    ds_read_b32 v2, v2
; CHECK-NEXT:    s_waitcnt lgkmcnt(0)
; CHECK-NEXT:    ds_read_b32 v3, v3
; CHECK-NEXT:    global_load_lds_dword v[0:1], off
; CHECK-NEXT:    s_waitcnt lgkmcnt(0)
; CHECK-NEXT:    v_add_u32_e32 v0, v2, v3
; CHECK-NEXT:    s_waitcnt vmcnt(0)
; CHECK-NEXT:    s_setpc_b64 s[30:31]
  %v1 = load i32, ptr addrspace(3) %lds, align 4
  fence syncscope("workgroup") release, !mmra !0
  %v2 = load i32, ptr addrspace(3) %lds2, align 4
  call void @llvm.amdgcn.global.load.lds(ptr addrspace(1) %g, ptr addrspace(3) %lds, i32 4, i32 0, i32 0)
  %s = add i32 %v1, %v2
  ret i32 %s
}

define i32 @lds_then_dma.global.agent.partial(ptr addrspace(1) %g, ptr addrspace(3) inreg %lds, ptr addrspace(3) inreg %lds2) #0 {
; CHECK-LABEL: lds_then_dma.global.agent.partial:
; CHECK:       ; %bb.0:
; CHECK-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; CHECK-NEXT:    v_mov_b32_e32 v2, s0
; CHECK-NEXT:    ds_read_b32 v2, v2
; CHECK-NEXT:    s_mov_b32 m0, s0
; CHECK-NEXT:    v_mov_b32_e32 v3, s1
; CHECK-NEXT:    s_waitcnt lgkmcnt(0)
; CHECK-NEXT:    ds_read_b32 v3, v3
; CHECK-NEXT:    global_load_lds_dword v[0:1], off
; CHECK-NEXT:    s_waitcnt lgkmcnt(0)
; CHECK-NEXT:    v_add_u32_e32 v0, v2, v3
; CHECK-NEXT:    s_waitcnt vmcnt(0)
; CHECK-NEXT:    s_setpc_b64 s[30:31]
  %v1 = load i32, ptr addrspace(3) %lds, align 4
  fence syncscope("agent") release, !mmra !0
  %v2 = load i32, ptr addrspace(3) %lds2, align 4
  call void @llvm.amdgcn.global.load.lds(ptr addrspace(1) %g, ptr addrspace(3) %lds, i32 4, i32 0, i32 0)
  %s = add i32 %v1, %v2
  ret i32 %s
}

define i32 @lds_then_dma.global.system.partial(ptr addrspace(1) %g, ptr addrspace(3) inreg %lds, ptr addrspace(3) inreg %lds2) #0 {
; CHECK-LABEL: lds_then_dma.global.system.partial:
; CHECK:       ; %bb.0:
; CHECK-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; CHECK-NEXT:    v_mov_b32_e32 v2, s0
; CHECK-NEXT:    ds_read_b32 v2, v2
; CHECK-NEXT:    s_mov_b32 m0, s0
; CHECK-NEXT:    v_mov_b32_e32 v3, s1
; CHECK-NEXT:    s_waitcnt lgkmcnt(0)
; CHECK-NEXT:    ds_read_b32 v3, v3
; CHECK-NEXT:    global_load_lds_dword v[0:1], off
; CHECK-NEXT:    s_waitcnt lgkmcnt(0)
; CHECK-NEXT:    v_add_u32_e32 v0, v2, v3
; CHECK-NEXT:    s_waitcnt vmcnt(0)
; CHECK-NEXT:    s_setpc_b64 s[30:31]
  %v1 = load i32, ptr addrspace(3) %lds, align 4
  fence release, !mmra !0
  %v2 = load i32, ptr addrspace(3) %lds2, align 4
  call void @llvm.amdgcn.global.load.lds(ptr addrspace(1) %g, ptr addrspace(3) %lds, i32 4, i32 0, i32 0)
  %s = add i32 %v1, %v2
  ret i32 %s
}

define i32 @lds_then_dma.buffer.singlethread(<4 x i32> inreg %rsrc, ptr addrspace(3) inreg %lds) #0 {
; CHECK-LABEL: lds_then_dma.buffer.singlethread:
; CHECK:       ; %bb.0:
; CHECK-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; CHECK-NEXT:    v_mov_b32_e32 v0, s16
; CHECK-NEXT:    s_mov_b32 m0, s16
; CHECK-NEXT:    ds_read_b32 v0, v0
; CHECK-NEXT:    buffer_load_dword off, s[0:3], 0 lds
; CHECK-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
; CHECK-NEXT:    s_setpc_b64 s[30:31]
  %v = load i32, ptr addrspace(3) %lds, align 4
  fence syncscope("singlethread") release, !mmra !0
  call void @llvm.amdgcn.raw.buffer.load.lds(<4 x i32> %rsrc, ptr addrspace(3) %lds, i32 4, i32 0, i32 0, i32 0, i32 0)
  ret i32 %v
}

define i32 @lds_then_dma.buffer.wave(<4 x i32> inreg %rsrc, ptr addrspace(3) inreg %lds) #0 {
; CHECK-LABEL: lds_then_dma.buffer.wave:
; CHECK:       ; %bb.0:
; CHECK-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; CHECK-NEXT:    v_mov_b32_e32 v0, s16
; CHECK-NEXT:    s_mov_b32 m0, s16
; CHECK-NEXT:    ds_read_b32 v0, v0
; CHECK-NEXT:    buffer_load_dword off, s[0:3], 0 lds
; CHECK-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
; CHECK-NEXT:    s_setpc_b64 s[30:31]
  %v = load i32, ptr addrspace(3) %lds, align 4
  fence syncscope("wavefront") release, !mmra !0
  call void @llvm.amdgcn.raw.buffer.load.lds(<4 x i32> %rsrc, ptr addrspace(3) %lds, i32 4, i32 0, i32 0, i32 0, i32 0)
  ret i32 %v
}

define i32 @lds_then_dma.buffer.wg(<4 x i32> inreg %rsrc, ptr addrspace(3) inreg %lds) #0 {
; CHECK-LABEL: lds_then_dma.buffer.wg:
; CHECK:       ; %bb.0:
; CHECK-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; CHECK-NEXT:    v_mov_b32_e32 v0, s16
; CHECK-NEXT:    s_mov_b32 m0, s16
; CHECK-NEXT:    ds_read_b32 v0, v0
; CHECK-NEXT:    s_waitcnt lgkmcnt(0)
; CHECK-NEXT:    buffer_load_dword off, s[0:3], 0 lds
; CHECK-NEXT:    s_waitcnt vmcnt(0)
; CHECK-NEXT:    s_setpc_b64 s[30:31]
  %v = load i32, ptr addrspace(3) %lds, align 4
  fence syncscope("workgroup") release, !mmra !0
  call void @llvm.amdgcn.raw.buffer.load.lds(<4 x i32> %rsrc, ptr addrspace(3) %lds, i32 4, i32 0, i32 0, i32 0, i32 0)
  ret i32 %v
}

define i32 @lds_then_dma.buffer.agent(<4 x i32> inreg %rsrc, ptr addrspace(3) inreg %lds) #0 {
; CHECK-LABEL: lds_then_dma.buffer.agent:
; CHECK:       ; %bb.0:
; CHECK-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; CHECK-NEXT:    v_mov_b32_e32 v0, s16
; CHECK-NEXT:    ds_read_b32 v0, v0
; CHECK-NEXT:    s_mov_b32 m0, s16
; CHECK-NEXT:    s_waitcnt lgkmcnt(0)
; CHECK-NEXT:    buffer_load_dword off, s[0:3], 0 lds
; CHECK-NEXT:    s_waitcnt vmcnt(0)
; CHECK-NEXT:    s_setpc_b64 s[30:31]
  %v = load i32, ptr addrspace(3) %lds, align 4
  fence syncscope("agent") release, !mmra !0
  call void @llvm.amdgcn.raw.buffer.load.lds(<4 x i32> %rsrc, ptr addrspace(3) %lds, i32 4, i32 0, i32 0, i32 0, i32 0)
  ret i32 %v
}

define i32 @lds_then_dma.buffer.system(<4 x i32> inreg %rsrc, ptr addrspace(3) inreg %lds) #0 {
; CHECK-LABEL: lds_then_dma.buffer.system:
; CHECK:       ; %bb.0:
; CHECK-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; CHECK-NEXT:    v_mov_b32_e32 v0, s16
; CHECK-NEXT:    ds_read_b32 v0, v0
; CHECK-NEXT:    s_mov_b32 m0, s16
; CHECK-NEXT:    s_waitcnt lgkmcnt(0)
; CHECK-NEXT:    buffer_load_dword off, s[0:3], 0 lds
; CHECK-NEXT:    s_waitcnt vmcnt(0)
; CHECK-NEXT:    s_setpc_b64 s[30:31]
  %v = load i32, ptr addrspace(3) %lds, align 4
  fence release, !mmra !0
  call void @llvm.amdgcn.raw.buffer.load.lds(<4 x i32> %rsrc, ptr addrspace(3) %lds, i32 4, i32 0, i32 0, i32 0, i32 0)
  ret i32 %v
}

define i32 @lds_then_dma.buffer.singlethread.partial(<4 x i32> inreg %rsrc, ptr addrspace(3) inreg %lds, ptr addrspace(3) inreg %lds2) #0 {
; CHECK-LABEL: lds_then_dma.buffer.singlethread.partial:
; CHECK:       ; %bb.0:
; CHECK-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; CHECK-NEXT:    s_mov_b32 m0, s16
; CHECK-NEXT:    v_mov_b32_e32 v0, s16
; CHECK-NEXT:    v_mov_b32_e32 v1, s17
; CHECK-NEXT:    ds_read_b32 v0, v0
; CHECK-NEXT:    ds_read_b32 v1, v1
; CHECK-NEXT:    buffer_load_dword off, s[0:3], 0 lds
; CHECK-NEXT:    s_waitcnt lgkmcnt(0)
; CHECK-NEXT:    v_add_u32_e32 v0, v0, v1
; CHECK-NEXT:    s_waitcnt vmcnt(0)
; CHECK-NEXT:    s_setpc_b64 s[30:31]
  %v1 = load i32, ptr addrspace(3) %lds, align 4
  fence syncscope("singlethread") release, !mmra !0
  %v2 = load i32, ptr addrspace(3) %lds2, align 4
  call void @llvm.amdgcn.raw.buffer.load.lds(<4 x i32> %rsrc, ptr addrspace(3) %lds, i32 4, i32 0, i32 0, i32 0, i32 0)
  %s = add i32 %v1, %v2
  ret i32 %s
}

define i32 @lds_then_dma.buffer.wave.partial(<4 x i32> inreg %rsrc, ptr addrspace(3) inreg %lds, ptr addrspace(3) inreg %lds2) #0 {
; CHECK-LABEL: lds_then_dma.buffer.wave.partial:
; CHECK:       ; %bb.0:
; CHECK-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; CHECK-NEXT:    s_mov_b32 m0, s16
; CHECK-NEXT:    v_mov_b32_e32 v0, s16
; CHECK-NEXT:    v_mov_b32_e32 v1, s17
; CHECK-NEXT:    ds_read_b32 v0, v0
; CHECK-NEXT:    ds_read_b32 v1, v1
; CHECK-NEXT:    buffer_load_dword off, s[0:3], 0 lds
; CHECK-NEXT:    s_waitcnt lgkmcnt(0)
; CHECK-NEXT:    v_add_u32_e32 v0, v0, v1
; CHECK-NEXT:    s_waitcnt vmcnt(0)
; CHECK-NEXT:    s_setpc_b64 s[30:31]
  %v1 = load i32, ptr addrspace(3) %lds, align 4
  fence syncscope("wavefront") release, !mmra !0
  %v2 = load i32, ptr addrspace(3) %lds2, align 4
  call void @llvm.amdgcn.raw.buffer.load.lds(<4 x i32> %rsrc, ptr addrspace(3) %lds, i32 4, i32 0, i32 0, i32 0, i32 0)
  %s = add i32 %v1, %v2
  ret i32 %s
}

define i32 @lds_then_dma.buffer.wg.partial(<4 x i32> inreg %rsrc, ptr addrspace(3) inreg %lds, ptr addrspace(3) inreg %lds2) #0 {
; CHECK-LABEL: lds_then_dma.buffer.wg.partial:
; CHECK:       ; %bb.0:
; CHECK-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; CHECK-NEXT:    s_mov_b32 m0, s16
; CHECK-NEXT:    v_mov_b32_e32 v0, s16
; CHECK-NEXT:    v_mov_b32_e32 v1, s17
; CHECK-NEXT:    ds_read_b32 v0, v0
; CHECK-NEXT:    s_waitcnt lgkmcnt(0)
; CHECK-NEXT:    ds_read_b32 v1, v1
; CHECK-NEXT:    buffer_load_dword off, s[0:3], 0 lds
; CHECK-NEXT:    s_waitcnt lgkmcnt(0)
; CHECK-NEXT:    v_add_u32_e32 v0, v0, v1
; CHECK-NEXT:    s_waitcnt vmcnt(0)
; CHECK-NEXT:    s_setpc_b64 s[30:31]
  %v1 = load i32, ptr addrspace(3) %lds, align 4
  fence syncscope("workgroup") release, !mmra !0
  %v2 = load i32, ptr addrspace(3) %lds2, align 4
  call void @llvm.amdgcn.raw.buffer.load.lds(<4 x i32> %rsrc, ptr addrspace(3) %lds, i32 4, i32 0, i32 0, i32 0, i32 0)
  %s = add i32 %v1, %v2
  ret i32 %s
}

define i32 @lds_then_dma.buffer.agent.partial(<4 x i32> inreg %rsrc, ptr addrspace(3) inreg %lds, ptr addrspace(3) inreg %lds2) #0 {
; CHECK-LABEL: lds_then_dma.buffer.agent.partial:
; CHECK:       ; %bb.0:
; CHECK-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; CHECK-NEXT:    v_mov_b32_e32 v0, s16
; CHECK-NEXT:    ds_read_b32 v0, v0
; CHECK-NEXT:    s_mov_b32 m0, s16
; CHECK-NEXT:    v_mov_b32_e32 v1, s17
; CHECK-NEXT:    s_waitcnt lgkmcnt(0)
; CHECK-NEXT:    ds_read_b32 v1, v1
; CHECK-NEXT:    buffer_load_dword off, s[0:3], 0 lds
; CHECK-NEXT:    s_waitcnt lgkmcnt(0)
; CHECK-NEXT:    v_add_u32_e32 v0, v0, v1
; CHECK-NEXT:    s_waitcnt vmcnt(0)
; CHECK-NEXT:    s_setpc_b64 s[30:31]
  %v1 = load i32, ptr addrspace(3) %lds, align 4
  fence syncscope("agent") release, !mmra !0
  %v2 = load i32, ptr addrspace(3) %lds2, align 4
  call void @llvm.amdgcn.raw.buffer.load.lds(<4 x i32> %rsrc, ptr addrspace(3) %lds, i32 4, i32 0, i32 0, i32 0, i32 0)
  %s = add i32 %v1, %v2
  ret i32 %s
}

define i32 @lds_then_dma.buffer.system.partial(<4 x i32> inreg %rsrc, ptr addrspace(3) inreg %lds, ptr addrspace(3) inreg %lds2) #0 {
; CHECK-LABEL: lds_then_dma.buffer.system.partial:
; CHECK:       ; %bb.0:
; CHECK-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; CHECK-NEXT:    v_mov_b32_e32 v0, s16
; CHECK-NEXT:    ds_read_b32 v0, v0
; CHECK-NEXT:    s_mov_b32 m0, s16
; CHECK-NEXT:    v_mov_b32_e32 v1, s17
; CHECK-NEXT:    s_waitcnt lgkmcnt(0)
; CHECK-NEXT:    ds_read_b32 v1, v1
; CHECK-NEXT:    buffer_load_dword off, s[0:3], 0 lds
; CHECK-NEXT:    s_waitcnt lgkmcnt(0)
; CHECK-NEXT:    v_add_u32_e32 v0, v0, v1
; CHECK-NEXT:    s_waitcnt vmcnt(0)
; CHECK-NEXT:    s_setpc_b64 s[30:31]
  %v1 = load i32, ptr addrspace(3) %lds, align 4
  fence release, !mmra !0
  %v2 = load i32, ptr addrspace(3) %lds2, align 4
  call void @llvm.amdgcn.raw.buffer.load.lds(<4 x i32> %rsrc, ptr addrspace(3) %lds, i32 4, i32 0, i32 0, i32 0, i32 0)
  %s = add i32 %v1, %v2
  ret i32 %s
}

attributes #0 = { nounwind "amdgpu-flat-work-group-size"="1,32" }

!0 = !{!"amdgpu-synchronize-as", !"local"}