1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
|
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx900 -denormal-fp-math-f32=preserve-sign -verify-machineinstrs < %s | FileCheck -check-prefix=GFX9 %s
; Make sure that AMDGPUCodeGenPrepare introduces mul24 intrinsics
; after SLSR, as the intrinsics would interfere. It's unclear if these
; should be introduced before LSR or not. It seems to help in some
; cases, and hurt others.
define void @lsr_order_mul24_0(i32 %arg, i32 %arg2, i32 %arg6, i32 %arg13, i32 %arg16) #0 {
; GFX9-LABEL: lsr_order_mul24_0:
; GFX9: ; %bb.0: ; %bb
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT: global_load_dword v5, v[0:1], off
; GFX9-NEXT: v_and_b32_e32 v2, 0xffffff, v2
; GFX9-NEXT: v_sub_u32_e32 v4, v4, v1
; GFX9-NEXT: s_mov_b64 s[4:5], 0
; GFX9-NEXT: s_waitcnt vmcnt(0)
; GFX9-NEXT: ds_write_b32 v0, v5
; GFX9-NEXT: .LBB0_1: ; %bb23
; GFX9-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX9-NEXT: v_mul_u32_u24_e32 v5, v0, v2
; GFX9-NEXT: v_add_u32_e32 v0, v0, v1
; GFX9-NEXT: v_sub_u32_e32 v5, v4, v5
; GFX9-NEXT: v_add_u32_e32 v5, v5, v0
; GFX9-NEXT: v_cmp_ge_u32_e32 vcc, v5, v3
; GFX9-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
; GFX9-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX9-NEXT: s_cbranch_execnz .LBB0_1
; GFX9-NEXT: ; %bb.2: ; %.loopexit
; GFX9-NEXT: s_or_b64 exec, exec, s[4:5]
; GFX9-NEXT: s_waitcnt lgkmcnt(0)
; GFX9-NEXT: s_setpc_b64 s[30:31]
bb:
%tmp22 = and i32 %arg6, 16777215
br label %bb23
.loopexit: ; preds = %bb23
ret void
bb23: ; preds = %bb23, %bb
%tmp24 = phi i32 [ %arg, %bb ], [ %tmp47, %bb23 ]
%tmp28 = and i32 %tmp24, 16777215
%tmp29 = mul i32 %tmp28, %tmp22
%tmp30 = sub i32 %tmp24, %tmp29
%tmp31 = add i32 %tmp30, %arg16
%tmp37 = icmp ult i32 %tmp31, %arg13
%tmp44 = load float, ptr addrspace(1) undef, align 4
store float %tmp44, ptr addrspace(3) undef, align 4
%tmp47 = add i32 %tmp24, %arg2
br i1 %tmp37, label %bb23, label %.loopexit
}
define void @lsr_order_mul24_1(i32 %arg, i32 %arg1, i32 %arg2, ptr addrspace(3) nocapture %arg3, i32 %arg4, i32 %arg5, i32 %arg6, i32 %arg7, i32 %arg8, i32 %arg9, ptr addrspace(1) nocapture readonly %arg10, i32 %arg11, i32 %arg12, i32 %arg13, i32 %arg14, i32 %arg15, i32 %arg16, i1 zeroext %arg17, i1 zeroext %arg18) #0 {
; GFX9-LABEL: lsr_order_mul24_1:
; GFX9: ; %bb.0: ; %bb
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT: v_and_b32_e32 v5, 1, v18
; GFX9-NEXT: v_cmp_eq_u32_e32 vcc, 1, v5
; GFX9-NEXT: v_cmp_lt_u32_e64 s[4:5], v0, v1
; GFX9-NEXT: s_and_saveexec_b64 s[8:9], s[4:5]
; GFX9-NEXT: s_cbranch_execz .LBB1_3
; GFX9-NEXT: ; %bb.1: ; %bb19
; GFX9-NEXT: v_cvt_f32_u32_e32 v7, v6
; GFX9-NEXT: v_add_u32_e32 v4, v4, v0
; GFX9-NEXT: v_and_b32_e32 v5, 0xffffff, v6
; GFX9-NEXT: v_lshl_add_u32 v6, v4, 2, v3
; GFX9-NEXT: v_rcp_iflag_f32_e32 v7, v7
; GFX9-NEXT: v_lshlrev_b32_e32 v8, 2, v2
; GFX9-NEXT: v_add_u32_e32 v9, v17, v12
; GFX9-NEXT: s_mov_b64 s[10:11], 0
; GFX9-NEXT: v_mov_b32_e32 v4, 0
; GFX9-NEXT: ; implicit-def: $vgpr3
; GFX9-NEXT: .LBB1_2: ; %bb23
; GFX9-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX9-NEXT: v_cvt_f32_u32_e32 v3, v0
; GFX9-NEXT: v_add_u32_e32 v18, v9, v0
; GFX9-NEXT: v_add_u32_e32 v12, v17, v0
; GFX9-NEXT: v_add_u32_e32 v0, v0, v2
; GFX9-NEXT: v_madak_f32 v3, v3, v7, 0x3727c5ac
; GFX9-NEXT: v_cvt_u32_f32_e32 v3, v3
; GFX9-NEXT: v_mul_u32_u24_e32 v19, v3, v5
; GFX9-NEXT: v_add_u32_e32 v20, v3, v16
; GFX9-NEXT: v_sub_u32_e32 v3, v18, v19
; GFX9-NEXT: v_sub_u32_e32 v12, v12, v19
; GFX9-NEXT: v_mad_u64_u32 v[18:19], s[6:7], v20, v15, v[3:4]
; GFX9-NEXT: v_cmp_lt_u32_e64 s[4:5], v20, v13
; GFX9-NEXT: v_cmp_lt_u32_e64 s[6:7], v12, v14
; GFX9-NEXT: s_and_b64 s[4:5], s[4:5], s[6:7]
; GFX9-NEXT: s_and_b64 s[4:5], s[4:5], vcc
; GFX9-NEXT: v_cndmask_b32_e64 v3, 0, v18, s[4:5]
; GFX9-NEXT: v_lshlrev_b64 v[18:19], 2, v[3:4]
; GFX9-NEXT: v_add_co_u32_e64 v18, s[6:7], v10, v18
; GFX9-NEXT: v_addc_co_u32_e64 v19, s[6:7], v11, v19, s[6:7]
; GFX9-NEXT: global_load_dword v3, v[18:19], off
; GFX9-NEXT: v_cmp_ge_u32_e64 s[6:7], v0, v1
; GFX9-NEXT: s_or_b64 s[10:11], s[6:7], s[10:11]
; GFX9-NEXT: s_waitcnt vmcnt(0)
; GFX9-NEXT: v_cndmask_b32_e64 v3, 0, v3, s[4:5]
; GFX9-NEXT: ds_write_b32 v6, v3
; GFX9-NEXT: v_add_u32_e32 v6, v6, v8
; GFX9-NEXT: s_andn2_b64 exec, exec, s[10:11]
; GFX9-NEXT: s_cbranch_execnz .LBB1_2
; GFX9-NEXT: .LBB1_3: ; %Flow2
; GFX9-NEXT: s_or_b64 exec, exec, s[8:9]
; GFX9-NEXT: s_waitcnt lgkmcnt(0)
; GFX9-NEXT: s_setpc_b64 s[30:31]
bb:
%tmp = icmp ult i32 %arg, %arg1
br i1 %tmp, label %bb19, label %.loopexit
bb19: ; preds = %bb
%tmp20 = uitofp i32 %arg6 to float
%tmp21 = fdiv float 1.000000e+00, %tmp20, !fpmath !0
%tmp22 = and i32 %arg6, 16777215
br label %bb23
.loopexit: ; preds = %bb23, %bb
ret void
bb23: ; preds = %bb19, %bb23
%tmp24 = phi i32 [ %arg, %bb19 ], [ %tmp47, %bb23 ]
%tmp25 = uitofp i32 %tmp24 to float
%tmp26 = tail call float @llvm.fmuladd.f32(float %tmp25, float %tmp21, float 0x3EE4F8B580000000) #2
%tmp27 = fptoui float %tmp26 to i32
%tmp28 = and i32 %tmp27, 16777215
%tmp29 = mul i32 %tmp28, %tmp22
%tmp30 = sub i32 %tmp24, %tmp29
%tmp31 = add i32 %tmp30, %arg16
%tmp32 = add i32 %tmp27, %arg15
%tmp33 = mul i32 %tmp32, %arg14
%tmp34 = add i32 %tmp33, %arg11
%tmp35 = add i32 %tmp34, %tmp31
%tmp36 = add i32 %tmp24, %arg4
%tmp37 = icmp ult i32 %tmp31, %arg13
%tmp38 = icmp ult i32 %tmp32, %arg12
%tmp39 = and i1 %tmp38, %tmp37
%tmp40 = and i1 %tmp39, %arg17
%tmp41 = zext i32 %tmp35 to i64
%tmp42 = select i1 %tmp40, i64 %tmp41, i64 0
%tmp43 = getelementptr inbounds float, ptr addrspace(1) %arg10, i64 %tmp42
%tmp44 = load float, ptr addrspace(1) %tmp43, align 4
%tmp45 = select i1 %tmp40, float %tmp44, float 0.000000e+00
%tmp46 = getelementptr inbounds float, ptr addrspace(3) %arg3, i32 %tmp36
store float %tmp45, ptr addrspace(3) %tmp46, align 4
%tmp47 = add i32 %tmp24, %arg2
%tmp48 = icmp ult i32 %tmp47, %arg1
br i1 %tmp48, label %bb23, label %.loopexit
}
define void @slsr1_0(i32 %b.arg, i32 %s.arg) #0 {
; GFX9-LABEL: slsr1_0:
; GFX9: ; %bb.0:
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT: v_and_b32_e32 v2, 0xffffff, v1
; GFX9-NEXT: v_mul_u32_u24_e32 v3, v0, v1
; GFX9-NEXT: global_store_dword v[0:1], v3, off
; GFX9-NEXT: s_waitcnt vmcnt(0)
; GFX9-NEXT: v_mad_u32_u24 v0, v0, v1, v2
; GFX9-NEXT: global_store_dword v[0:1], v0, off
; GFX9-NEXT: s_waitcnt vmcnt(0)
; GFX9-NEXT: v_add_u32_e32 v0, v0, v2
; GFX9-NEXT: global_store_dword v[0:1], v0, off
; GFX9-NEXT: s_waitcnt vmcnt(0)
; GFX9-NEXT: s_setpc_b64 s[30:31]
%b = and i32 %b.arg, 16777215
%s = and i32 %s.arg, 16777215
; CHECK-LABEL: @slsr1(
; foo(b * s);
%mul0 = mul i32 %b, %s
; CHECK: mul i32
; CHECK-NOT: mul i32
store volatile i32 %mul0, ptr addrspace(1) undef
; foo((b + 1) * s);
%b1 = add i32 %b, 1
%mul1 = mul i32 %b1, %s
store volatile i32 %mul1, ptr addrspace(1) undef
; foo((b + 2) * s);
%b2 = add i32 %b, 2
%mul2 = mul i32 %b2, %s
store volatile i32 %mul2, ptr addrspace(1) undef
ret void
}
define void @slsr1_1(i32 %b.arg, i32 %s.arg) #0 {
; GFX9-LABEL: slsr1_1:
; GFX9: ; %bb.0:
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT: s_mov_b32 s4, s33
; GFX9-NEXT: s_mov_b32 s33, s32
; GFX9-NEXT: s_or_saveexec_b64 s[6:7], -1
; GFX9-NEXT: buffer_store_dword v43, off, s[0:3], s33 offset:12 ; 4-byte Folded Spill
; GFX9-NEXT: s_mov_b64 exec, s[6:7]
; GFX9-NEXT: v_writelane_b32 v43, s4, 5
; GFX9-NEXT: v_writelane_b32 v43, s30, 0
; GFX9-NEXT: v_writelane_b32 v43, s31, 1
; GFX9-NEXT: s_addk_i32 s32, 0x800
; GFX9-NEXT: v_writelane_b32 v43, s34, 2
; GFX9-NEXT: v_writelane_b32 v43, s36, 3
; GFX9-NEXT: s_getpc_b64 s[4:5]
; GFX9-NEXT: s_add_u32 s4, s4, foo@gotpcrel32@lo+4
; GFX9-NEXT: s_addc_u32 s5, s5, foo@gotpcrel32@hi+12
; GFX9-NEXT: v_writelane_b32 v43, s37, 4
; GFX9-NEXT: s_load_dwordx2 s[36:37], s[4:5], 0x0
; GFX9-NEXT: buffer_store_dword v40, off, s[0:3], s33 offset:8 ; 4-byte Folded Spill
; GFX9-NEXT: buffer_store_dword v41, off, s[0:3], s33 offset:4 ; 4-byte Folded Spill
; GFX9-NEXT: buffer_store_dword v42, off, s[0:3], s33 ; 4-byte Folded Spill
; GFX9-NEXT: v_mov_b32_e32 v40, v1
; GFX9-NEXT: v_mov_b32_e32 v41, v0
; GFX9-NEXT: v_mul_u32_u24_e32 v0, v41, v40
; GFX9-NEXT: s_mov_b32 s34, s15
; GFX9-NEXT: v_and_b32_e32 v42, 0xffffff, v40
; GFX9-NEXT: s_waitcnt lgkmcnt(0)
; GFX9-NEXT: s_swappc_b64 s[30:31], s[36:37]
; GFX9-NEXT: v_mad_u32_u24 v40, v41, v40, v42
; GFX9-NEXT: s_mov_b32 s15, s34
; GFX9-NEXT: v_mov_b32_e32 v0, v40
; GFX9-NEXT: s_swappc_b64 s[30:31], s[36:37]
; GFX9-NEXT: v_add_u32_e32 v0, v40, v42
; GFX9-NEXT: s_mov_b32 s15, s34
; GFX9-NEXT: s_swappc_b64 s[30:31], s[36:37]
; GFX9-NEXT: buffer_load_dword v42, off, s[0:3], s33 ; 4-byte Folded Reload
; GFX9-NEXT: buffer_load_dword v41, off, s[0:3], s33 offset:4 ; 4-byte Folded Reload
; GFX9-NEXT: buffer_load_dword v40, off, s[0:3], s33 offset:8 ; 4-byte Folded Reload
; GFX9-NEXT: v_readlane_b32 s37, v43, 4
; GFX9-NEXT: v_readlane_b32 s36, v43, 3
; GFX9-NEXT: v_readlane_b32 s34, v43, 2
; GFX9-NEXT: v_readlane_b32 s31, v43, 1
; GFX9-NEXT: v_readlane_b32 s30, v43, 0
; GFX9-NEXT: v_readlane_b32 s4, v43, 5
; GFX9-NEXT: s_or_saveexec_b64 s[6:7], -1
; GFX9-NEXT: buffer_load_dword v43, off, s[0:3], s33 offset:12 ; 4-byte Folded Reload
; GFX9-NEXT: s_mov_b64 exec, s[6:7]
; GFX9-NEXT: s_addk_i32 s32, 0xf800
; GFX9-NEXT: s_mov_b32 s33, s4
; GFX9-NEXT: s_waitcnt vmcnt(0)
; GFX9-NEXT: s_setpc_b64 s[30:31]
%b = and i32 %b.arg, 16777215
%s = and i32 %s.arg, 16777215
; CHECK-LABEL: @slsr1(
; foo(b * s);
%mul0 = mul i32 %b, %s
; CHECK: mul i32
; CHECK-NOT: mul i32
call void @foo(i32 %mul0)
; foo((b + 1) * s);
%b1 = add i32 %b, 1
%mul1 = mul i32 %b1, %s
call void @foo(i32 %mul1)
; foo((b + 2) * s);
%b2 = add i32 %b, 2
%mul2 = mul i32 %b2, %s
call void @foo(i32 %mul2)
ret void
}
declare void @foo(i32) #2
declare float @llvm.fmuladd.f32(float, float, float) #1
attributes #0 = { nounwind willreturn "denormal-fp-math-f32"="preserve-sign,preserve-sign" }
attributes #1 = { nounwind readnone speculatable }
attributes #2 = { nounwind "amdgpu-no-dispatch-id" "amdgpu-no-dispatch-ptr" "amdgpu-no-implicitarg-ptr" "amdgpu-no-queue-ptr" "amdgpu-no-workgroup-id-x" "amdgpu-no-workgroup-id-y" "amdgpu-no-workgroup-id-z" "amdgpu-no-workitem-id-x" "amdgpu-no-workitem-id-y" "amdgpu-no-workitem-id-z" "uniform-work-group-size"="false" "denormal-fp-math-f32"="preserve-sign,preserve-sign" }
!0 = !{float 2.500000e+00}
|
