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
|
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc -march=r600 -mcpu=cayman < %s | FileCheck -check-prefixes=CM %s
; RUN: llc -march=r600 -mcpu=redwood < %s | FileCheck -check-prefixes=EG %s
define amdgpu_kernel void @test_umul24_i32(ptr addrspace(1) %out, i32 %a, i32 %b) {
; CM-LABEL: test_umul24_i32:
; CM: ; %bb.0: ; %entry
; CM-NEXT: ALU 7, @4, KC0[CB0:0-32], KC1[]
; CM-NEXT: MEM_RAT_CACHELESS STORE_DWORD T1.X, T0.X
; CM-NEXT: CF_END
; CM-NEXT: PAD
; CM-NEXT: ALU clause starting at 4:
; CM-NEXT: LSHR T0.X, KC0[2].Y, literal.x,
; CM-NEXT: AND_INT T0.Z, KC0[2].W, literal.y,
; CM-NEXT: AND_INT * T0.W, KC0[2].Z, literal.y,
; CM-NEXT: 2(2.802597e-45), 16777215(2.350989e-38)
; CM-NEXT: MULLO_INT T1.X, T0.W, T0.Z,
; CM-NEXT: MULLO_INT T1.Y (MASKED), T0.W, T0.Z,
; CM-NEXT: MULLO_INT T1.Z (MASKED), T0.W, T0.Z,
; CM-NEXT: MULLO_INT * T1.W (MASKED), T0.W, T0.Z,
;
; EG-LABEL: test_umul24_i32:
; EG: ; %bb.0: ; %entry
; EG-NEXT: ALU 5, @4, KC0[CB0:0-32], KC1[]
; EG-NEXT: MEM_RAT_CACHELESS STORE_RAW T1.X, T0.X, 1
; EG-NEXT: CF_END
; EG-NEXT: PAD
; EG-NEXT: ALU clause starting at 4:
; EG-NEXT: AND_INT T0.W, KC0[2].W, literal.x,
; EG-NEXT: AND_INT * T1.W, KC0[2].Z, literal.x,
; EG-NEXT: 16777215(2.350989e-38), 0(0.000000e+00)
; EG-NEXT: LSHR T0.X, KC0[2].Y, literal.x,
; EG-NEXT: MULLO_INT * T1.X, PS, PV.W,
; EG-NEXT: 2(2.802597e-45), 0(0.000000e+00)
entry:
%0 = shl i32 %a, 8
%a_24 = lshr i32 %0, 8
%1 = shl i32 %b, 8
%b_24 = lshr i32 %1, 8
%2 = mul i32 %a_24, %b_24
store i32 %2, ptr addrspace(1) %out
ret void
}
; The result must be sign-extended.
define amdgpu_kernel void @test_umul24_i16_sext(ptr addrspace(1) %out, i16 %a, i16 %b) {
; CM-LABEL: test_umul24_i16_sext:
; CM: ; %bb.0: ; %entry
; CM-NEXT: ALU 0, @10, KC0[], KC1[]
; CM-NEXT: TEX 1 @6
; CM-NEXT: ALU 7, @11, KC0[CB0:0-32], KC1[]
; CM-NEXT: MEM_RAT_CACHELESS STORE_DWORD T0.X, T1.X
; CM-NEXT: CF_END
; CM-NEXT: PAD
; CM-NEXT: Fetch clause starting at 6:
; CM-NEXT: VTX_READ_16 T1.X, T0.X, 40, #3
; CM-NEXT: VTX_READ_16 T0.X, T0.X, 42, #3
; CM-NEXT: ALU clause starting at 10:
; CM-NEXT: MOV * T0.X, 0.0,
; CM-NEXT: ALU clause starting at 11:
; CM-NEXT: MULLO_INT T0.X, T1.X, T0.X,
; CM-NEXT: MULLO_INT T0.Y (MASKED), T1.X, T0.X,
; CM-NEXT: MULLO_INT T0.Z (MASKED), T1.X, T0.X,
; CM-NEXT: MULLO_INT * T0.W (MASKED), T1.X, T0.X,
; CM-NEXT: BFE_INT * T0.X, PV.X, 0.0, literal.x,
; CM-NEXT: 16(2.242078e-44), 0(0.000000e+00)
; CM-NEXT: LSHR * T1.X, KC0[2].Y, literal.x,
; CM-NEXT: 2(2.802597e-45), 0(0.000000e+00)
;
; EG-LABEL: test_umul24_i16_sext:
; EG: ; %bb.0: ; %entry
; EG-NEXT: ALU 0, @10, KC0[], KC1[]
; EG-NEXT: TEX 1 @6
; EG-NEXT: ALU 3, @11, KC0[CB0:0-32], KC1[]
; EG-NEXT: MEM_RAT_CACHELESS STORE_RAW T0.X, T1.X, 1
; EG-NEXT: CF_END
; EG-NEXT: PAD
; EG-NEXT: Fetch clause starting at 6:
; EG-NEXT: VTX_READ_16 T1.X, T0.X, 40, #3
; EG-NEXT: VTX_READ_16 T0.X, T0.X, 42, #3
; EG-NEXT: ALU clause starting at 10:
; EG-NEXT: MOV * T0.X, 0.0,
; EG-NEXT: ALU clause starting at 11:
; EG-NEXT: MULLO_INT * T0.X, T1.X, T0.X,
; EG-NEXT: BFE_INT T0.X, PS, 0.0, literal.x,
; EG-NEXT: LSHR * T1.X, KC0[2].Y, literal.y,
; EG-NEXT: 16(2.242078e-44), 2(2.802597e-45)
entry:
%mul = mul i16 %a, %b
%ext = sext i16 %mul to i32
store i32 %ext, ptr addrspace(1) %out
ret void
}
; The result must be sign-extended.
define amdgpu_kernel void @test_umul24_i8(ptr addrspace(1) %out, i8 %a, i8 %b) {
; CM-LABEL: test_umul24_i8:
; CM: ; %bb.0: ; %entry
; CM-NEXT: ALU 0, @10, KC0[], KC1[]
; CM-NEXT: TEX 1 @6
; CM-NEXT: ALU 7, @11, KC0[CB0:0-32], KC1[]
; CM-NEXT: MEM_RAT_CACHELESS STORE_DWORD T0.X, T1.X
; CM-NEXT: CF_END
; CM-NEXT: PAD
; CM-NEXT: Fetch clause starting at 6:
; CM-NEXT: VTX_READ_8 T1.X, T0.X, 40, #3
; CM-NEXT: VTX_READ_8 T0.X, T0.X, 41, #3
; CM-NEXT: ALU clause starting at 10:
; CM-NEXT: MOV * T0.X, 0.0,
; CM-NEXT: ALU clause starting at 11:
; CM-NEXT: MULLO_INT T0.X, T1.X, T0.X,
; CM-NEXT: MULLO_INT T0.Y (MASKED), T1.X, T0.X,
; CM-NEXT: MULLO_INT T0.Z (MASKED), T1.X, T0.X,
; CM-NEXT: MULLO_INT * T0.W (MASKED), T1.X, T0.X,
; CM-NEXT: BFE_INT * T0.X, PV.X, 0.0, literal.x,
; CM-NEXT: 8(1.121039e-44), 0(0.000000e+00)
; CM-NEXT: LSHR * T1.X, KC0[2].Y, literal.x,
; CM-NEXT: 2(2.802597e-45), 0(0.000000e+00)
;
; EG-LABEL: test_umul24_i8:
; EG: ; %bb.0: ; %entry
; EG-NEXT: ALU 0, @10, KC0[], KC1[]
; EG-NEXT: TEX 1 @6
; EG-NEXT: ALU 3, @11, KC0[CB0:0-32], KC1[]
; EG-NEXT: MEM_RAT_CACHELESS STORE_RAW T0.X, T1.X, 1
; EG-NEXT: CF_END
; EG-NEXT: PAD
; EG-NEXT: Fetch clause starting at 6:
; EG-NEXT: VTX_READ_8 T1.X, T0.X, 40, #3
; EG-NEXT: VTX_READ_8 T0.X, T0.X, 41, #3
; EG-NEXT: ALU clause starting at 10:
; EG-NEXT: MOV * T0.X, 0.0,
; EG-NEXT: ALU clause starting at 11:
; EG-NEXT: MULLO_INT * T0.X, T1.X, T0.X,
; EG-NEXT: BFE_INT T0.X, PS, 0.0, literal.x,
; EG-NEXT: LSHR * T1.X, KC0[2].Y, literal.y,
; EG-NEXT: 8(1.121039e-44), 2(2.802597e-45)
entry:
%mul = mul i8 %a, %b
%ext = sext i8 %mul to i32
store i32 %ext, ptr addrspace(1) %out
ret void
}
define amdgpu_kernel void @test_umulhi24_i32_i64(ptr addrspace(1) %out, i32 %a, i32 %b) {
; CM-LABEL: test_umulhi24_i32_i64:
; CM: ; %bb.0: ; %entry
; CM-NEXT: ALU 5, @4, KC0[CB0:0-32], KC1[]
; CM-NEXT: MEM_RAT_CACHELESS STORE_DWORD T1.X, T0.X
; CM-NEXT: CF_END
; CM-NEXT: PAD
; CM-NEXT: ALU clause starting at 4:
; CM-NEXT: LSHR * T0.X, KC0[2].Y, literal.x,
; CM-NEXT: 2(2.802597e-45), 0(0.000000e+00)
; CM-NEXT: MULHI_UINT24 T1.X, KC0[2].Z, KC0[2].W,
; CM-NEXT: MULHI_UINT24 T1.Y (MASKED), KC0[2].Z, KC0[2].W,
; CM-NEXT: MULHI_UINT24 T1.Z (MASKED), KC0[2].Z, KC0[2].W,
; CM-NEXT: MULHI_UINT24 * T1.W (MASKED), KC0[2].Z, KC0[2].W,
;
; EG-LABEL: test_umulhi24_i32_i64:
; EG: ; %bb.0: ; %entry
; EG-NEXT: ALU 2, @4, KC0[CB0:0-32], KC1[]
; EG-NEXT: MEM_RAT_CACHELESS STORE_RAW T1.X, T0.X, 1
; EG-NEXT: CF_END
; EG-NEXT: PAD
; EG-NEXT: ALU clause starting at 4:
; EG-NEXT: LSHR * T0.X, KC0[2].Y, literal.x,
; EG-NEXT: 2(2.802597e-45), 0(0.000000e+00)
; EG-NEXT: MULHI_UINT24 * T1.X, KC0[2].Z, KC0[2].W,
entry:
%a.24 = and i32 %a, 16777215
%b.24 = and i32 %b, 16777215
%a.24.i64 = zext i32 %a.24 to i64
%b.24.i64 = zext i32 %b.24 to i64
%mul48 = mul i64 %a.24.i64, %b.24.i64
%mul48.hi = lshr i64 %mul48, 32
%mul24hi = trunc i64 %mul48.hi to i32
store i32 %mul24hi, ptr addrspace(1) %out
ret void
}
define amdgpu_kernel void @test_umulhi24(ptr addrspace(1) %out, i64 %a, i64 %b) {
; CM-LABEL: test_umulhi24:
; CM: ; %bb.0: ; %entry
; CM-NEXT: ALU 5, @4, KC0[CB0:0-32], KC1[]
; CM-NEXT: MEM_RAT_CACHELESS STORE_DWORD T1.X, T0.X
; CM-NEXT: CF_END
; CM-NEXT: PAD
; CM-NEXT: ALU clause starting at 4:
; CM-NEXT: LSHR * T0.X, KC0[2].Y, literal.x,
; CM-NEXT: 2(2.802597e-45), 0(0.000000e+00)
; CM-NEXT: MULHI_UINT24 T1.X, KC0[2].W, KC0[3].Y,
; CM-NEXT: MULHI_UINT24 T1.Y (MASKED), KC0[2].W, KC0[3].Y,
; CM-NEXT: MULHI_UINT24 T1.Z (MASKED), KC0[2].W, KC0[3].Y,
; CM-NEXT: MULHI_UINT24 * T1.W (MASKED), KC0[2].W, KC0[3].Y,
;
; EG-LABEL: test_umulhi24:
; EG: ; %bb.0: ; %entry
; EG-NEXT: ALU 2, @4, KC0[CB0:0-32], KC1[]
; EG-NEXT: MEM_RAT_CACHELESS STORE_RAW T1.X, T0.X, 1
; EG-NEXT: CF_END
; EG-NEXT: PAD
; EG-NEXT: ALU clause starting at 4:
; EG-NEXT: LSHR * T0.X, KC0[2].Y, literal.x,
; EG-NEXT: 2(2.802597e-45), 0(0.000000e+00)
; EG-NEXT: MULHI_UINT24 * T1.X, KC0[2].W, KC0[3].Y,
entry:
%a.24 = and i64 %a, 16777215
%b.24 = and i64 %b, 16777215
%mul48 = mul i64 %a.24, %b.24
%mul48.hi = lshr i64 %mul48, 32
%mul24.hi = trunc i64 %mul48.hi to i32
store i32 %mul24.hi, ptr addrspace(1) %out
ret void
}
; Multiply with 24-bit inputs and 64-bit output.
define amdgpu_kernel void @test_umul24_i64(ptr addrspace(1) %out, i64 %a, i64 %b) {
; CM-LABEL: test_umul24_i64:
; CM: ; %bb.0: ; %entry
; CM-NEXT: ALU 12, @4, KC0[CB0:0-32], KC1[]
; CM-NEXT: MEM_RAT_CACHELESS STORE_DWORD T1, T0.X
; CM-NEXT: CF_END
; CM-NEXT: PAD
; CM-NEXT: ALU clause starting at 4:
; CM-NEXT: LSHR T0.X, KC0[2].Y, literal.x,
; CM-NEXT: AND_INT * T0.Z, KC0[3].Y, literal.y,
; CM-NEXT: 2(2.802597e-45), 16777215(2.350989e-38)
; CM-NEXT: AND_INT * T0.W, KC0[2].W, literal.x,
; CM-NEXT: 16777215(2.350989e-38), 0(0.000000e+00)
; CM-NEXT: MULLO_INT T1.X, T0.W, T0.Z,
; CM-NEXT: MULLO_INT T1.Y (MASKED), T0.W, T0.Z,
; CM-NEXT: MULLO_INT T1.Z (MASKED), T0.W, T0.Z,
; CM-NEXT: MULLO_INT * T1.W (MASKED), T0.W, T0.Z,
; CM-NEXT: MULHI_UINT24 T1.X (MASKED), KC0[2].W, KC0[3].Y,
; CM-NEXT: MULHI_UINT24 T1.Y, KC0[2].W, KC0[3].Y,
; CM-NEXT: MULHI_UINT24 T1.Z (MASKED), KC0[2].W, KC0[3].Y,
; CM-NEXT: MULHI_UINT24 * T1.W (MASKED), KC0[2].W, KC0[3].Y,
;
; EG-LABEL: test_umul24_i64:
; EG: ; %bb.0: ; %entry
; EG-NEXT: ALU 6, @4, KC0[CB0:0-32], KC1[]
; EG-NEXT: MEM_RAT_CACHELESS STORE_RAW T1.XY, T0.X, 1
; EG-NEXT: CF_END
; EG-NEXT: PAD
; EG-NEXT: ALU clause starting at 4:
; EG-NEXT: AND_INT T0.W, KC0[3].Y, literal.x,
; EG-NEXT: AND_INT * T1.W, KC0[2].W, literal.x,
; EG-NEXT: 16777215(2.350989e-38), 0(0.000000e+00)
; EG-NEXT: LSHR T0.X, KC0[2].Y, literal.x,
; EG-NEXT: MULLO_INT * T1.X, PS, PV.W,
; EG-NEXT: 2(2.802597e-45), 0(0.000000e+00)
; EG-NEXT: MULHI_UINT24 * T1.Y, KC0[2].W, KC0[3].Y,
entry:
%tmp0 = shl i64 %a, 40
%a_24 = lshr i64 %tmp0, 40
%tmp1 = shl i64 %b, 40
%b_24 = lshr i64 %tmp1, 40
%tmp2 = mul i64 %a_24, %b_24
store i64 %tmp2, ptr addrspace(1) %out
ret void
}
|
