1; RUN: llc -mtriple=amdgcn-- -amdgpu-atomic-optimizations=true -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefixes=GCN,GCN64,GFX7LESS %s
2; RUN: llc  -mtriple=amdgcn-- -mcpu=tonga -mattr=-flat-for-global -amdgpu-atomic-optimizations=true -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefixes=GCN,GCN64,GFX8MORE,GFX8MORE64,DPPCOMB %s
3; RUN: llc -mtriple=amdgcn-- -mcpu=gfx900 -mattr=-flat-for-global -amdgpu-atomic-optimizations=true -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefixes=GCN,GCN64,GFX8MORE,GFX8MORE64,DPPCOMB %s
4; RUN: llc -mtriple=amdgcn-- -mcpu=gfx1010 -mattr=-wavefrontsize32,+wavefrontsize64 -mattr=-flat-for-global -amdgpu-atomic-optimizations=true -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefixes=GCN,GCN64,GFX8MORE,GFX8MORE64 %s
5; RUN: llc -mtriple=amdgcn-- -mcpu=gfx1010 -mattr=+wavefrontsize32,-wavefrontsize64 -mattr=-flat-for-global -amdgpu-atomic-optimizations=true -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefixes=GCN,GCN32,GFX8MORE,GFX8MORE32 %s
6
7declare i1 @llvm.amdgcn.wqm.vote(i1)
8declare i32 @llvm.amdgcn.buffer.atomic.add(i32, <4 x i32>, i32, i32, i1)
9declare void @llvm.amdgcn.buffer.store.f32(float, <4 x i32>, i32, i32, i1, i1)
10
11; Show that what the atomic optimization pass will do for raw buffers.
12
13; GCN-LABEL: add_i32_constant:
14; GCN-LABEL: BB0_1:
15; GCN32: v_cmp_ne_u32_e64 s[[exec_lo:[0-9]+]], 1, 0
16; GCN64: v_cmp_ne_u32_e64 s{{\[}}[[exec_lo:[0-9]+]]:[[exec_hi:[0-9]+]]{{\]}}, 1, 0
17; GCN: v_mbcnt_lo_u32_b32{{(_e[0-9]+)?}} v[[mbcnt:[0-9]+]], s[[exec_lo]], 0
18; GCN64: v_mbcnt_hi_u32_b32{{(_e[0-9]+)?}} v[[mbcnt]], s[[exec_hi]], v[[mbcnt]]
19; GCN: v_cmp_eq_u32{{(_e[0-9]+)?}} vcc{{(_lo)?}}, 0, v[[mbcnt]]
20; GCN32: s_bcnt1_i32_b32 s[[popcount:[0-9]+]], s[[exec_lo]]
21; GCN64: s_bcnt1_i32_b64 s[[popcount:[0-9]+]], s{{\[}}[[exec_lo]]:[[exec_hi]]{{\]}}
22; GCN: v_mul_u32_u24{{(_e[0-9]+)?}} v[[value:[0-9]+]], s[[popcount]], 5
23; GCN: buffer_atomic_add v[[value]]
24; GCN: v_readfirstlane_b32 s{{[0-9]+}}, v[[value]]
25define amdgpu_ps void @add_i32_constant(<4 x i32> inreg %out, <4 x i32> inreg %inout) {
26entry:
27  %cond1 = call i1 @llvm.amdgcn.wqm.vote(i1 true)
28  %old = call i32 @llvm.amdgcn.buffer.atomic.add(i32 5, <4 x i32> %inout, i32 0, i32 0, i1 0)
29  %cond2 = call i1 @llvm.amdgcn.wqm.vote(i1 true)
30  %cond = and i1 %cond1, %cond2
31  br i1 %cond, label %if, label %else
32if:
33  %bitcast = bitcast i32 %old to float
34  call void @llvm.amdgcn.buffer.store.f32(float %bitcast, <4 x i32> %out, i32 0, i32 0, i1 0, i1 0)
35  ret void
36else:
37  ret void
38}
39
40; GCN-LABEL: add_i32_varying:
41; GFX7LESS-NOT: v_mbcnt_lo_u32_b32
42; GFX7LESS-NOT: v_mbcnt_hi_u32_b32
43; GFX8MORE32: v_cmp_ne_u32_e64 s[[exec_lo:[0-9]+]], 1, 0
44; GFX8MORE64: v_cmp_ne_u32_e64 s{{\[}}[[exec_lo:[0-9]+]]:[[exec_hi:[0-9]+]]{{\]}}, 1, 0
45; GFX8MORE: v_mbcnt_lo_u32_b32{{(_e[0-9]+)?}} v[[mbcnt:[0-9]+]], s[[exec_lo]], 0
46; GFX8MORE64: v_mbcnt_hi_u32_b32{{(_e[0-9]+)?}} v[[mbcnt]], s[[exec_hi]], v[[mbcnt]]
47; DPPCOMB: v_add_u32_dpp
48; DPPCOMB: v_add_u32_dpp
49; GFX8MORE32: v_readlane_b32 s[[scalar_value:[0-9]+]], v{{[0-9]+}}, 31
50; GFX8MORE64: v_readlane_b32 s[[scalar_value:[0-9]+]], v{{[0-9]+}}, 63
51; GFX8MORE: v_cmp_eq_u32{{(_e[0-9]+)?}} vcc{{(_lo)?}}, 0, v[[mbcnt]]
52; GFX8MORE: v_mov_b32{{(_e[0-9]+)?}} v[[value:[0-9]+]], s[[scalar_value]]
53; GFX8MORE: buffer_atomic_add v[[value]]
54; GFX8MORE: v_readfirstlane_b32 s{{[0-9]+}}, v[[value]]
55define amdgpu_ps void @add_i32_varying(<4 x i32> inreg %out, <4 x i32> inreg %inout, i32 %val) {
56entry:
57  %cond1 = call i1 @llvm.amdgcn.wqm.vote(i1 true)
58  %old = call i32 @llvm.amdgcn.buffer.atomic.add(i32 %val, <4 x i32> %inout, i32 0, i32 0, i1 0)
59  %cond2 = call i1 @llvm.amdgcn.wqm.vote(i1 true)
60  %cond = and i1 %cond1, %cond2
61  br i1 %cond, label %if, label %else
62if:
63  %bitcast = bitcast i32 %old to float
64  call void @llvm.amdgcn.buffer.store.f32(float %bitcast, <4 x i32> %out, i32 0, i32 0, i1 0, i1 0)
65  ret void
66else:
67  ret void
68}
69