1 /*******************************************************************************
2 Copyright (c) 2016, The OpenBLAS Project
4 Redistribution and use in source and binary forms, with or without
5 modification, are permitted provided that the following conditions are
7 1. Redistributions of source code must retain the above copyright
8 notice, this list of conditions and the following disclaimer.
9 2. Redistributions in binary form must reproduce the above copyright
10 notice, this list of conditions and the following disclaimer in
11 the documentation and/or other materials provided with the
13 3. Neither the name of the OpenBLAS project nor the names of
14 its contributors may be used to endorse or promote products
15 derived from this software without specific prior written permission.
16 THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
17 AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
18 IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
19 ARE DISCLAIMED. IN NO EVENT SHALL THE OPENBLAS PROJECT OR CONTRIBUTORS BE
20 LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
21 DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
22 SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
23 CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
24 OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE
25 USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
26 *******************************************************************************/
29 #include "macros_msa.h"
31 int CNAME(BLASLONG m, BLASLONG n, FLOAT *src, BLASLONG lda, FLOAT *dst)
34 FLOAT *psrc0, *psrc1, *psrc2, *pdst0;
35 FLOAT ctemp01, ctemp02, ctemp03, ctemp04;
36 v4f32 src0, src1, src2, src3, src4, src5, src6, src7;
37 v4f32 src8, src9, src10, src11, src12, src13, src14, src15;
43 for (j = (n >> 3); j--;)
49 for (i = (m >> 2); i--;)
51 LD_SP4(psrc1, 4, src0, src1, src2, src3);
52 LD_SP4(psrc2, 4, src4, src5, src6, src7);
53 LD_SP4(psrc1 + 2 * lda, 4, src8, src9, src10, src11);
54 LD_SP4(psrc2 + 2 * lda, 4, src12, src13, src14, src15);
55 ST_SP8_INC(src0, src1, src2, src3, src4, src5, src6, src7, pdst0, 4);
56 ST_SP8_INC(src8, src9, src10, src11, src12, src13, src14, src15, pdst0, 4);
63 LD_SP4(psrc1, 4, src0, src1, src2, src3);
64 LD_SP4(psrc2, 4, src4, src5, src6, src7);
65 ST_SP8_INC(src0, src1, src2, src3, src4, src5, src6, src7, pdst0, 4);
72 LD_SP4(psrc1, 4, src0, src1, src2, src3);
73 ST_SP4_INC(src0, src1, src2, src3, pdst0, 4);
83 for (i = (m >> 2); i--;)
85 LD_SP2(psrc1, 4, src0, src1);
86 LD_SP2(psrc2, 4, src2, src3);
87 LD_SP2(psrc1 + 2 * lda, 4, src4, src5);
88 LD_SP2(psrc2 + 2 * lda, 4, src6, src7);
90 ST_SP4_INC(src0, src1, src2, src3, pdst0, 4);
91 ST_SP4_INC(src4, src5, src6, src7, pdst0, 4);
98 LD_SP2(psrc1, 4, src0, src1);
99 LD_SP2(psrc2, 4, src2, src3);
100 ST_SP4_INC(src0, src1, src2, src3, pdst0, 4);
107 LD_SP2(psrc1, 4, src0, src1);
108 ST_SP2_INC(src0, src1, pdst0, 4);
118 for (i = (m >> 2); i--;)
122 src2 = LD_SP(psrc1 + 2 * lda);
123 src3 = LD_SP(psrc2 + 2 * lda);
124 ST_SP4_INC(src0, src1, src2, src3, pdst0, 4);
134 ST_SP2_INC(src0, src1, pdst0, 4);
154 for (i = (m >> 2); i--;)
156 ctemp01 = *(psrc1 + 0);
157 ctemp02 = *(psrc1 + 1);
158 ctemp03 = *(psrc2 + 0);
159 ctemp04 = *(psrc2 + 1);
161 *(pdst0 + 0) = ctemp01;
162 *(pdst0 + 1) = ctemp02;
163 *(pdst0 + 2) = ctemp03;
164 *(pdst0 + 3) = ctemp04;
170 ctemp01 = *(psrc1 + 0);
171 ctemp02 = *(psrc1 + 1);
172 ctemp03 = *(psrc2 + 0);
173 ctemp04 = *(psrc2 + 1);
175 *(pdst0 + 0) = ctemp01;
176 *(pdst0 + 1) = ctemp02;
177 *(pdst0 + 2) = ctemp03;
178 *(pdst0 + 3) = ctemp04;
187 ctemp01 = *(psrc1 + 0);
188 ctemp02 = *(psrc1 + 1);
189 ctemp03 = *(psrc2 + 0);
190 ctemp04 = *(psrc2 + 1);
192 *(pdst0 + 0) = ctemp01;
193 *(pdst0 + 1) = ctemp02;
194 *(pdst0 + 2) = ctemp03;
195 *(pdst0 + 3) = ctemp04;
204 ctemp01 = *(psrc1 + 0);
205 ctemp02 = *(psrc1 + 1);
207 *(pdst0 + 0) = ctemp01;
208 *(pdst0 + 1) = ctemp02;