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
|
/*
* Copyright (C) 2014, The University of Texas at Austin
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions are
* met:
* - Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* - Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
* - Neither the name of The University of Texas at Austin nor the names
* of its contributors may be used to endorse or promote products
* derived from this software without specific prior written permission.
*
* THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
* "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
* LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
* A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
* HOLDER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
* SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
* LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
* DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
* THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
* (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
* OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*
*/
/*
* Copyright (C) 2014-2015, Michael Lehn
*
* ulmBLAS adopted general ideas from BLIS. Using micro kernels from BLIS
* only requires minor modifications,
*
*/
#ifndef ULMBLAS_IMPL_LEVEL3_UKERNEL_SSE_UTRUSM_TCC
#define ULMBLAS_IMPL_LEVEL3_UKERNEL_SSE_UTRUSM_TCC 1
#include <iostream>
#include <ulmblas/impl/level3/ukernel/ugemm.h>
#include <ulmblas/impl/level1extensions/gecopy.h>
#include <ulmblas/impl/level3/ukernel/sse/utrusm.h>
namespace ulmBLAS { namespace sse {
template <typename IndexType>
static typename std::enable_if<std::is_convertible<IndexType,long>::value,
void>::type
utrusm(const double *A,
const double *B,
double *C,
IndexType incRowC_,
IndexType incColC_)
{
long incRowC = incRowC_;
long incColC = incColC_;
__asm__ volatile
(
"movq %0, %%rax \n\t" // Address of A stored in %rax
"movq %1, %%rbx \n\t" // Address of B stored in %rbx
"movq %2, %%rcx \n\t" // Address of C stored in %rcx
"movq %3, %%rsi \n\t" // load incRowC
"movq %4, %%rdi \n\t" // load incColC
"leaq (,%%rsi,8), %%rsi \n\t" // incRowC *= sizeof(double)
"leaq (,%%rdi,8), %%rdi \n\t" // incColC *= sizeof(double)
"leaq (%%rcx,%%rsi,4), %%rcx \n\t" // %rcx = &C[4*incRowC]
"leaq (%%rcx,%%rdi,2), %%rdx \n\t" // %rdx = &C[4*incRowC+2*incColC]
"subq $1, %%rsi \n\t"
"notq %%rsi \n\t" // negate incRow
"movaps 0 * 16(%%rbx), %%xmm8 \n\t" // load xmm8 = (b00, b01 )
"movaps 1 * 16(%%rbx), %%xmm12 \n\t" // load xmm12 = (b02, b03 )
"movaps 2 * 16(%%rbx), %%xmm9 \n\t" // load xmm9 = (b10, b11 )
"movaps 3 * 16(%%rbx), %%xmm13 \n\t" // load xmm13 = (b12, b13 )
"movaps 4 * 16(%%rbx), %%xmm10 \n\t" // load xmm10 = (b20, b21 )
"movaps 5 * 16(%%rbx), %%xmm14 \n\t" // load xmm14 = (b22, b23 )
"movaps 6 * 16(%%rbx), %%xmm11 \n\t" // load xmm11 = (b30, b31 )
"movaps 7 * 16(%%rbx), %%xmm15 \n\t" // load xmm15 = (b32, b33 )
"leaq (%%rcx,%%rsi), %%rcx \n\t" // Move %rcx to previous row of C
"leaq (%%rdx,%%rsi), %%rdx \n\t" // Move %rdx to previous row of C
//
// Compute (c30, c31, c32, c33)
//
"movddup (3+3*4)*8(%%rax), %%xmm3 \n\t" // load xmm3 = (a33, a33)
"mulpd %%xmm3, %%xmm11 \n\t" // (c30,c31) = (b30*a33, b31*a33)
"mulpd %%xmm3, %%xmm15 \n\t" // (c32,c33) = (b32*a33, b33*a33)
"movlpd %%xmm11, (%%rcx) \n\t" // store c30
"movhpd %%xmm11, (%%rcx,%%rdi) \n\t" // store c31
"movlpd %%xmm15, (%%rdx) \n\t" // store c32
"movhpd %%xmm15, (%%rdx,%%rdi) \n\t" // store c33
"leaq (%%rcx,%%rsi), %%rcx \n\t" // Move %rcx to previous row of C
"leaq (%%rdx,%%rsi), %%rdx \n\t" // Move %rdx to previous row of C
//
// Compute (c20, c21, c22, c23)
//
"movddup (2+2*4)*8(%%rax), %%xmm2 \n\t" // load xmm3 = (a22, a22)
"movddup (2+3*4)*8(%%rax), %%xmm3 \n\t" // load xmm2 = (a23, a23)
"movaps %%xmm3, %%xmm7 \n\t" // xmm7 = xmm3
"mulpd %%xmm11, %%xmm3 \n\t" // (a23*c30, a23*c31)
"mulpd %%xmm15, %%xmm7 \n\t" // (a23*c32, a23*c33)
"subpd %%xmm3, %%xmm10 \n\t" // (b20-a23*c30, b21-a23*c31)
"subpd %%xmm7, %%xmm14 \n\t" // (b22-a23*c32, b23-a23*c33)
"mulpd %%xmm2, %%xmm10 \n\t" // (c20, c21)
"mulpd %%xmm2, %%xmm14 \n\t" // (c22, c23)
"movlpd %%xmm10, (%%rcx) \n\t" // store c20
"movhpd %%xmm10, (%%rcx,%%rdi) \n\t" // store c21
"movlpd %%xmm14, (%%rdx) \n\t" // store c22
"movhpd %%xmm14, (%%rdx,%%rdi) \n\t" // store c23
"leaq (%%rcx,%%rsi), %%rcx \n\t" // Move %rcx to previous row of C
"leaq (%%rdx,%%rsi), %%rdx \n\t" // Move %rdx to previous row of C
//
// Compute (c10, c11, c12, c13)
//
"movddup (1+1*4)*8(%%rax), %%xmm1 \n\t" // load xmm1 = (a11, a11)
"movddup (1+2*4)*8(%%rax), %%xmm2 \n\t" // load xmm2 = (a12, a12)
"movddup (1+3*4)*8(%%rax), %%xmm3 \n\t" // load xmm3 = (a13, a13)
"movaps %%xmm2, %%xmm6 \n\t" // xmm6 = xmm2
"movaps %%xmm3, %%xmm7 \n\t" // xmm7 = xmm3
"mulpd %%xmm11, %%xmm3 \n\t" // (a13*c30, a13*c31)
"mulpd %%xmm15, %%xmm7 \n\t" // (a13*c32, a13*c33)
"mulpd %%xmm10, %%xmm2 \n\t" // (a12*c20, a12*c21)
"mulpd %%xmm14, %%xmm6 \n\t" // (a12*c22, a12*c23)
"subpd %%xmm3, %%xmm9 \n\t" // (b10,b11) -= (a13*c30,a13*c31)
"subpd %%xmm7, %%xmm13 \n\t" // (b12,b13) -= (a13*c32,a13*c33)
"subpd %%xmm2, %%xmm9 \n\t" // (b10,b11) -= (a12*c20,a12*c21)
"subpd %%xmm6, %%xmm13 \n\t" // (b12,b13) -= (a12*c22,a12*c23)
"mulpd %%xmm1, %%xmm9 \n\t" // (c10, c11)
"mulpd %%xmm1, %%xmm13 \n\t" // (c12, c13)
"movlpd %%xmm9, (%%rcx) \n\t" // store c10
"movhpd %%xmm9, (%%rcx,%%rdi) \n\t" // store c11
"movlpd %%xmm13, (%%rdx) \n\t" // store c12
"movhpd %%xmm13, (%%rdx,%%rdi) \n\t" // store c13
"leaq (%%rcx,%%rsi), %%rcx \n\t" // Move %rcx to next row of C
"leaq (%%rdx,%%rsi), %%rdx \n\t" // Move %rdx to next row of C
//
// Compute (c00, c01, c02, c03)
//
"movddup (0+0*4)*8(%%rax), %%xmm0 \n\t" // load xmm0 = (a00, a00)
"movddup (0+1*4)*8(%%rax), %%xmm1 \n\t" // load xmm1 = (a01, a01)
"movddup (0+2*4)*8(%%rax), %%xmm2 \n\t" // load xmm2 = (a02, a02)
"movddup (0+3*4)*8(%%rax), %%xmm3 \n\t" // load xmm3 = (a03, a03)
"movaps %%xmm1, %%xmm5 \n\t" // xmm5 = xmm1
"movaps %%xmm2, %%xmm6 \n\t" // xmm6 = xmm2
"movaps %%xmm3, %%xmm7 \n\t" // xmm7 = xmm3
"mulpd %%xmm11, %%xmm3 \n\t" // (a03*c30, a03*c31)
"mulpd %%xmm15, %%xmm7 \n\t" // (a03*c32, a03*c33)
"mulpd %%xmm10, %%xmm2 \n\t" // (a02*c10, a03*c11)
"mulpd %%xmm14, %%xmm6 \n\t" // (a02*c12, a03*c13)
"mulpd %%xmm9, %%xmm1 \n\t" // (a01*c10, a01*c11)
"mulpd %%xmm13, %%xmm5 \n\t" // (a01*c12, a01*c13)
"subpd %%xmm3, %%xmm8 \n\t" // (b00,b01) -= (a03*c30,a03*c31)
"subpd %%xmm7, %%xmm12 \n\t" // (b02,b03) -= (a03*c32,a03*c33)
"subpd %%xmm2, %%xmm8 \n\t" // (b00,b01) -= (a02*c20,a02*c21)
"subpd %%xmm6, %%xmm12 \n\t" // (b02,b03) -= (a02*c22,a02*c23)
"subpd %%xmm1, %%xmm8 \n\t" // (b00,b01) -= (a01*c10,a01*c11)
"subpd %%xmm5, %%xmm12 \n\t" // (b02,b03) -= (a01*c12,a01*c13)
"mulpd %%xmm0, %%xmm8 \n\t" // (c00, c01)
"mulpd %%xmm0, %%xmm12 \n\t" // (c02, c03)
"movlpd %%xmm8, (%%rcx) \n\t" // store c20
"movhpd %%xmm8, (%%rcx,%%rdi) \n\t" // store c21
"movlpd %%xmm12, (%%rdx) \n\t" // store c22
"movhpd %%xmm12, (%%rdx,%%rdi) \n\t" // store c23
: // output
: // input
"m" (A), // 0
"m" (B), // 1
"m" (C), // 2
"m" (incRowC), // 3
"m" (incColC) // 4
: // register clobber list
"rax", "rbx", "rcx", "rdx", "rsi", "rdi",
"xmm0", "xmm1", "xmm2", "xmm3",
"xmm5", "xmm6", "xmm7",
"xmm8", "xmm9", "xmm10", "xmm11",
"xmm12", "xmm13", "xmm14", "xmm15"
);
}
} } // namespace sse, ulmBLAS
#endif // ULMBLAS_IMPL_LEVEL3_UKERNEL_SSE_UTRUSM_TCC 1
|