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
|
/*
* Copyright (c) 2021, 2023-2024 Arm Limited.
*
* SPDX-License-Identifier: MIT
*
* Permission is hereby granted, free of charge, to any person obtaining a copy
* of this software and associated documentation files (the "Software"), to
* deal in the Software without restriction, including without limitation the
* rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
* sell copies of the Software, and to permit persons to whom the Software is
* furnished to do so, subject to the following conditions:
*
* The above copyright notice and this permission notice shall be included in all
* copies or substantial portions of the Software.
*
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
* AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
* LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
* OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
* SOFTWARE.
*/
#pragma once
#if defined(ARM_COMPUTE_ENABLE_SVE)
namespace {
void sve_transpose_interleave_6VL_2x4_fp32bf16(bfloat16 *out, const float *in, size_t width, size_t in_stride, size_t height)
{
float *pad_row = reinterpret_cast<float *>(alloca(width * sizeof(float)));
if (height % 4) {
memset(pad_row, 0, width * sizeof(float));
}
size_t out_stride = 6 * roundup<size_t>(height, 4) * get_vector_length<uint32_t>();
__asm__ __volatile__(
"ptrue p2.b\n"
"1:" // Main row loop: Head
"mov x26, %x[in]\n"
"mov x25, %x[width]\n"
"cnth x20, ALL, MUL #3\n"
"cmp %x[height], #0x3\n"
"mov x24, %x[out]\n"
"add x23, x26, %x[in_stride]\n"
"add x22, x23, %x[in_stride]\n"
"add x21, x22, %x[in_stride]\n"
"add %x[in], x21, %x[in_stride]\n"
"csel x21, x21, %x[pad_row], GT\n"
"csel x22, x22, %x[pad_row], GE\n"
"cmp %x[height], #0x1\n"
"csel x23, x23, %x[pad_row], GT\n"
"cmp x25, x20\n"
"sub %x[height], %x[height], #0x4\n"
"blt 3f\n"
"2:" // Main row loop: Unroll column loop
"ld1w { z17.s }, p2/Z, [x26]\n"
"ld1w { z24.s }, p2/Z, [x26, #1, MUL VL]\n"
"sub x25, x25, x20\n"
"ld1w { z23.s }, p2/Z, [x26, #2, MUL VL]\n"
"ld1w { z16.s }, p2/Z, [x22]\n"
"cmp x25, x20\n"
"ld1w { z22.s }, p2/Z, [x22, #1, MUL VL]\n"
"ld1w { z21.s }, p2/Z, [x22, #2, MUL VL]\n"
"ld1w { z20.s }, p2/Z, [x26, #3, MUL VL]\n"
"ld1w { z19.s }, p2/Z, [x26, #4, MUL VL]\n"
"ld1w { z5.s }, p2/Z, [x26, #5, MUL VL]\n"
"ld1w { z18.s }, p2/Z, [x22, #3, MUL VL]\n"
"zip1 z4.s, z17.s, z16.s\n"
"zip2 z3.s, z17.s, z16.s\n"
"ld1w { z17.s }, p2/Z, [x22, #4, MUL VL]\n"
"ld1w { z16.s }, p2/Z, [x22, #5, MUL VL]\n"
"zip1 z2.s, z24.s, z22.s\n"
"zip2 z1.s, z24.s, z22.s\n"
"ld1w { z0.s }, p2/Z, [x23]\n"
"ld1w { z31.s }, p2/Z, [x23, #1, MUL VL]\n"
"zip1 z30.s, z23.s, z21.s\n"
"zip2 z29.s, z23.s, z21.s\n"
"ld1w { z28.s }, p2/Z, [x23, #2, MUL VL]\n"
"ld1w { z27.s }, p2/Z, [x21]\n"
"zip1 z26.s, z20.s, z18.s\n"
"zip2 z25.s, z20.s, z18.s\n"
"ld1w { z24.s }, p2/Z, [x21, #1, MUL VL]\n"
"ld1w { z23.s }, p2/Z, [x21, #2, MUL VL]\n"
"zip1 z22.s, z19.s, z17.s\n"
"zip2 z10.s, z19.s, z17.s\n"
"ld1w { z21.s }, p2/Z, [x23, #3, MUL VL]\n"
"ld1w { z20.s }, p2/Z, [x23, #4, MUL VL]\n"
"zip1 z19.s, z5.s, z16.s\n"
"zip2 z9.s, z5.s, z16.s\n"
"ld1w { z8.s }, p2/Z, [x23, #5, MUL VL]\n"
"ld1w { z18.s }, p2/Z, [x21, #3, MUL VL]\n"
".inst 0x658aa887 // bfcvt z7.h, p2/M, z4.s\n"
"zip1 z6.s, z0.s, z27.s\n"
"ld1w { z17.s }, p2/Z, [x21, #4, MUL VL]\n"
"ld1w { z16.s }, p2/Z, [x21, #5, MUL VL]\n"
".inst 0x658aa865 // bfcvt z5.h, p2/M, z3.s\n"
"zip2 z4.s, z0.s, z27.s\n"
".inst 0x658aa843 // bfcvt z3.h, p2/M, z2.s\n"
"zip1 z2.s, z31.s, z24.s\n"
"addvl x26, x26, #6\n"
"addvl x23, x23, #6\n"
".inst 0x658aa821 // bfcvt z1.h, p2/M, z1.s\n"
"zip2 z0.s, z31.s, z24.s\n"
"addvl x22, x22, #6\n"
"addvl x21, x21, #6\n"
".inst 0x658aabdf // bfcvt z31.h, p2/M, z30.s\n"
"zip1 z30.s, z28.s, z23.s\n"
".inst 0x658aabbd // bfcvt z29.h, p2/M, z29.s\n"
"zip2 z28.s, z28.s, z23.s\n"
".inst 0x658aab5b // bfcvt z27.h, p2/M, z26.s\n"
"zip1 z26.s, z21.s, z18.s\n"
".inst 0x658aab39 // bfcvt z25.h, p2/M, z25.s\n"
"zip2 z24.s, z21.s, z18.s\n"
".inst 0x658aaad7 // bfcvt z23.h, p2/M, z22.s\n"
"zip1 z22.s, z20.s, z17.s\n"
".inst 0x658aa955 // bfcvt z21.h, p2/M, z10.s\n"
"zip2 z20.s, z20.s, z17.s\n"
".inst 0x658aaa73 // bfcvt z19.h, p2/M, z19.s\n"
"zip1 z18.s, z8.s, z16.s\n"
".inst 0x658aa931 // bfcvt z17.h, p2/M, z9.s\n"
"zip2 z16.s, z8.s, z16.s\n"
".inst 0x648aa8c7 // bfcvtnt z7.h, p2/M, z6.s\n"
".inst 0x648aa885 // bfcvtnt z5.h, p2/M, z4.s\n"
".inst 0x648aa843 // bfcvtnt z3.h, p2/M, z2.s\n"
".inst 0x648aa801 // bfcvtnt z1.h, p2/M, z0.s\n"
".inst 0x648aabdf // bfcvtnt z31.h, p2/M, z30.s\n"
".inst 0x648aab9d // bfcvtnt z29.h, p2/M, z28.s\n"
"st1h { z7.h }, p2, [x24]\n"
"st1h { z5.h }, p2, [x24, #1, MUL VL]\n"
".inst 0x648aab5b // bfcvtnt z27.h, p2/M, z26.s\n"
".inst 0x648aab19 // bfcvtnt z25.h, p2/M, z24.s\n"
"st1h { z3.h }, p2, [x24, #2, MUL VL]\n"
".inst 0x648aaad7 // bfcvtnt z23.h, p2/M, z22.s\n"
".inst 0x648aaa95 // bfcvtnt z21.h, p2/M, z20.s\n"
"st1h { z1.h }, p2, [x24, #3, MUL VL]\n"
".inst 0x648aaa53 // bfcvtnt z19.h, p2/M, z18.s\n"
".inst 0x648aaa11 // bfcvtnt z17.h, p2/M, z16.s\n"
"st1h { z31.h }, p2, [x24, #4, MUL VL]\n"
"st1h { z29.h }, p2, [x24, #5, MUL VL]\n"
"add x24, x24, %x[out_stride]\n"
"st1h { z27.h }, p2, [x24]\n"
"st1h { z25.h }, p2, [x24, #1, MUL VL]\n"
"st1h { z23.h }, p2, [x24, #2, MUL VL]\n"
"st1h { z21.h }, p2, [x24, #3, MUL VL]\n"
"st1h { z19.h }, p2, [x24, #4, MUL VL]\n"
"st1h { z17.h }, p2, [x24, #5, MUL VL]\n"
"add x24, x24, %x[out_stride]\n"
"bge 2b\n"
"3:" // Main row loop: Unroll column loop skip
"cbz x25, 5f\n"
"4:" // Main row loop: Column loop
"mov x20, x25\n"
"decd x25, ALL, MUL #6\n"
"whilelt p0.s, XZR, x20\n"
"decw x20\n"
"whilelt p1.s, XZR, x20\n"
"decw x20\n"
"ld1w { z17.s }, p0/Z, [x26]\n"
"ld1w { z16.s }, p0/Z, [x22]\n"
"ld1w { z23.s }, p0/Z, [x23]\n"
"ld1w { z19.s }, p0/Z, [x21]\n"
"whilelt p0.s, XZR, x20\n"
"cmp x25, #0x0\n"
"ld1w { z22.s }, p1/Z, [x26, #1, MUL VL]\n"
"ld1w { z18.s }, p1/Z, [x22, #1, MUL VL]\n"
"ld1w { z31.s }, p1/Z, [x23, #1, MUL VL]\n"
"ld1w { z30.s }, p1/Z, [x21, #1, MUL VL]\n"
"zip1 z21.s, z17.s, z16.s\n"
"zip2 z17.s, z17.s, z16.s\n"
"ld1w { z20.s }, p0/Z, [x26, #2, MUL VL]\n"
"ld1w { z16.s }, p0/Z, [x22, #2, MUL VL]\n"
"zip1 z29.s, z23.s, z19.s\n"
"zip2 z28.s, z23.s, z19.s\n"
"ld1w { z27.s }, p0/Z, [x23, #2, MUL VL]\n"
"ld1w { z26.s }, p0/Z, [x21, #2, MUL VL]\n"
"zip1 z19.s, z22.s, z18.s\n"
"zip2 z18.s, z22.s, z18.s\n"
".inst 0x658aaab9 // bfcvt z25.h, p2/M, z21.s\n"
".inst 0x658aaa38 // bfcvt z24.h, p2/M, z17.s\n"
"addvl x26, x26, #3\n"
"addvl x23, x23, #3\n"
"zip1 z17.s, z20.s, z16.s\n"
"zip2 z16.s, z20.s, z16.s\n"
"addvl x22, x22, #3\n"
"addvl x21, x21, #3\n"
".inst 0x658aaa77 // bfcvt z23.h, p2/M, z19.s\n"
"zip1 z22.s, z31.s, z30.s\n"
".inst 0x658aaa55 // bfcvt z21.h, p2/M, z18.s\n"
"zip2 z20.s, z31.s, z30.s\n"
".inst 0x658aaa33 // bfcvt z19.h, p2/M, z17.s\n"
"zip1 z18.s, z27.s, z26.s\n"
".inst 0x658aaa11 // bfcvt z17.h, p2/M, z16.s\n"
"zip2 z16.s, z27.s, z26.s\n"
".inst 0x648aabb9 // bfcvtnt z25.h, p2/M, z29.s\n"
".inst 0x648aab98 // bfcvtnt z24.h, p2/M, z28.s\n"
".inst 0x648aaad7 // bfcvtnt z23.h, p2/M, z22.s\n"
".inst 0x648aaa95 // bfcvtnt z21.h, p2/M, z20.s\n"
".inst 0x648aaa53 // bfcvtnt z19.h, p2/M, z18.s\n"
".inst 0x648aaa11 // bfcvtnt z17.h, p2/M, z16.s\n"
"st1h { z25.h }, p2, [x24]\n"
"st1h { z24.h }, p2, [x24, #1, MUL VL]\n"
"st1h { z23.h }, p2, [x24, #2, MUL VL]\n"
"st1h { z21.h }, p2, [x24, #3, MUL VL]\n"
"st1h { z19.h }, p2, [x24, #4, MUL VL]\n"
"st1h { z17.h }, p2, [x24, #5, MUL VL]\n"
"add x24, x24, %x[out_stride]\n"
"bgt 4b\n"
"5:" // Main row loop: Column loop skip
"cmp %x[height], #0x1\n"
"addvl %x[out], %x[out], #6\n"
"bge 1b\n"
: [height] "+&r" (height), [in] "+&r" (in), [out] "+&r" (out)
: [in_stride] "r" (in_stride), [out_stride] "r" (out_stride), [pad_row] "r" (pad_row), [width] "r" (width)
: "cc", "memory", "p0", "p1", "p2", "x20", "x21", "x22", "x23", "x24", "x25", "x26", "z0", "z1", "z2", "z3", "z4", "z5", "z6", "z7", "z8", "z9", "z10", "z16", "z17", "z18", "z19", "z20", "z21", "z22", "z23", "z24", "z25", "z26", "z27", "z28", "z29", "z30", "z31"
);
}
} // anonymous namespace
template<>
void Transform<6, 4, true, VLType::SVE>(
bfloat16 *out, const float *in, int stride, int x0, int xmax, int k0, int kmax)
{
sve_transpose_interleave_6VL_2x4_fp32bf16(
out,
in + k0 * stride + x0,
(xmax-x0),
stride * sizeof(float),
(kmax-k0)
);
}
#endif // defined(ARM_COMPUTE_ENABLE_SVE)
|
