1 /***************************************************************************
2 Copyright (c) 2020, The OpenBLAS Project
3 All rights reserved.
4 Redistribution and use in source and binary forms, with or without
5 modification, are permitted provided that the following conditions are
6 met:
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
12 distribution.
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 *****************************************************************************/
27
28 #include "common.h"
29
30 #if !defined(DOUBLE)
31 #define VSETVL(n) vsetvl_e32m4(n)
32 #define FLOAT_V_T vfloat32m4_t
33 #define VLEV_FLOAT vle_v_f32m4
34 #define VLSEV_FLOAT vlse_v_f32m4
35 #define VSEV_FLOAT vse_v_f32m4
36 #define VSSEV_FLOAT vsse_v_f32m4
37 #define VFMACCVF_FLOAT vfmacc_vf_f32m4
38 #else
39 #define VSETVL(n) vsetvl_e64m4(n)
40 #define FLOAT_V_T vfloat64m4_t
41 #define VLEV_FLOAT vle_v_f64m4
42 #define VLSEV_FLOAT vlse_v_f64m4
43 #define VSEV_FLOAT vse_v_f64m4
44 #define VSSEV_FLOAT vsse_v_f64m4
45 #define VFMACCVF_FLOAT vfmacc_vf_f64m4
46 #endif
47
CNAME(BLASLONG n,BLASLONG dummy0,BLASLONG dummy1,FLOAT da,FLOAT * x,BLASLONG inc_x,FLOAT * y,BLASLONG inc_y,FLOAT * dummy,BLASLONG dummy2)48 int CNAME(BLASLONG n, BLASLONG dummy0, BLASLONG dummy1, FLOAT da, FLOAT *x, BLASLONG inc_x, FLOAT *y, BLASLONG inc_y, FLOAT *dummy, BLASLONG dummy2)
49 {
50 BLASLONG i=0, j=0, jx=0, jy=0;
51 unsigned int gvl = 0;
52 FLOAT_V_T vx0, vx1;
53 FLOAT_V_T vy0, vy1;
54 BLASLONG stride_x, stride_y;
55
56 if (n < 0) return(0);
57 if (da == 0.0) return(0);
58
59 if (inc_x == 1 && inc_y == 1) {
60
61 gvl = VSETVL(n);
62
63 if (gvl <= n/2) {
64 for (i = 0, j=0; i < n/(2*gvl); i++, j+=2*gvl) {
65 vx0 = VLEV_FLOAT(&x[j], gvl);
66 vy0 = VLEV_FLOAT(&y[j], gvl);
67 vy0 = VFMACCVF_FLOAT(vy0, da, vx0, gvl);
68 VSEV_FLOAT(&y[j], vy0, gvl);
69
70 vx1 = VLEV_FLOAT(&x[j+gvl], gvl);
71 vy1 = VLEV_FLOAT(&y[j+gvl], gvl);
72 vy1 = VFMACCVF_FLOAT(vy1, da, vx1, gvl);
73 VSEV_FLOAT(&y[j+gvl], vy1, gvl);
74 }
75 }
76 //tail
77 for (; j < n; ) {
78 gvl = VSETVL(n - j);
79 vx0 = VLEV_FLOAT(&x[j], gvl);
80 vy0 = VLEV_FLOAT(&y[j], gvl);
81 vy0 = VFMACCVF_FLOAT(vy0, da, vx0, gvl);
82 VSEV_FLOAT(&y[j], vy0, gvl);
83
84 j += gvl;
85 }
86 }else if (inc_y == 1) {
87 stride_x = inc_x * sizeof(FLOAT);
88 gvl = VSETVL(n);
89 if(gvl <= n/2){
90 BLASLONG inc_xv = inc_x * gvl;
91 for(i=0,j=0; i<n/(2*gvl); i++){
92 vx0 = VLSEV_FLOAT(&x[jx], stride_x, gvl);
93 vy0 = VLEV_FLOAT(&y[j], gvl);
94 vy0 = VFMACCVF_FLOAT(vy0, da, vx0, gvl);
95 VSEV_FLOAT(&y[j], vy0, gvl);
96
97 vx1 = VLSEV_FLOAT(&x[jx+inc_xv], stride_x, gvl);
98 vy1 = VLEV_FLOAT(&y[j+gvl], gvl);
99 vy1 = VFMACCVF_FLOAT(vy1, da, vx1, gvl);
100 VSEV_FLOAT(&y[j+gvl], vy1, gvl);
101
102 j += gvl * 2;
103 jx += inc_xv * 2;
104 }
105 }
106 for (; j<n; ) {
107 gvl = VSETVL(n - j);
108 vx0 = VLSEV_FLOAT(&x[j*inc_x], stride_x, gvl);
109 vy0 = VLEV_FLOAT(&y[j], gvl);
110 vy0 = VFMACCVF_FLOAT(vy0, da, vx0, gvl);
111 VSEV_FLOAT(&y[j], vy0, gvl);
112 j += gvl;
113 }
114 }else if(inc_x == 1){
115 stride_y = inc_y * sizeof(FLOAT);
116 gvl = VSETVL(n);
117 if(gvl <= n/2){
118 BLASLONG inc_yv = inc_y * gvl;
119 for(i=0,j=0; i<n/(2*gvl); i++){
120 vx0 = VLEV_FLOAT(&x[j], gvl);
121 vy0 = VLSEV_FLOAT(&y[jy], stride_y, gvl);
122 vy0 = VFMACCVF_FLOAT(vy0, da, vx0, gvl);
123 VSSEV_FLOAT(&y[jy], stride_y, vy0, gvl);
124
125 vx1 = VLEV_FLOAT(&x[j+gvl], gvl);
126 vy1 = VLSEV_FLOAT(&y[jy+inc_yv], stride_y, gvl);
127 vy1 = VFMACCVF_FLOAT(vy1, da, vx1, gvl);
128 VSSEV_FLOAT(&y[jy+inc_yv], stride_y, vy1, gvl);
129
130 j += gvl * 2;
131 jy += inc_yv * 2;
132 }
133 }
134 for (; j<n; ) {
135 gvl = VSETVL(n - j);
136 vx0 = VLEV_FLOAT(&x[j], gvl);
137 vy0 = VLSEV_FLOAT(&y[j*inc_y], stride_y, gvl);
138 vy0 = VFMACCVF_FLOAT(vy0, da, vx0, gvl);
139 VSSEV_FLOAT(&y[j*inc_y], stride_y, vy0, gvl);
140 j += gvl;
141 }
142 }else{
143 stride_x = inc_x * sizeof(FLOAT);
144 stride_y = inc_y * sizeof(FLOAT);
145 gvl = VSETVL(n);
146 if(gvl <= n/2){
147 BLASLONG inc_xv = inc_x * gvl;
148 BLASLONG inc_yv = inc_y * gvl;
149 for(i=0,j=0; i<n/(2*gvl); i++){
150 vx0 = VLSEV_FLOAT(&x[jx], stride_x, gvl);
151 vy0 = VLSEV_FLOAT(&y[jy], stride_y, gvl);
152 vy0 = VFMACCVF_FLOAT(vy0, da, vx0, gvl);
153 VSSEV_FLOAT(&y[jy], stride_y, vy0, gvl);
154
155 vx1 = VLSEV_FLOAT(&x[jx+inc_xv], stride_x, gvl);
156 vy1 = VLSEV_FLOAT(&y[jy+inc_yv], stride_y, gvl);
157 vy1 = VFMACCVF_FLOAT(vy1, da, vx1, gvl);
158 VSSEV_FLOAT(&y[jy+inc_yv], stride_y, vy1, gvl);
159
160 j += gvl * 2;
161 jx += inc_xv * 2;
162 jy += inc_yv * 2;
163 }
164 }
165 for (; j<n; ) {
166 gvl = VSETVL(n - j);
167 vx0 = VLSEV_FLOAT(&x[j*inc_x], stride_x, gvl);
168 vy0 = VLSEV_FLOAT(&y[j*inc_y], stride_y, gvl);
169 vy0 = VFMACCVF_FLOAT(vy0, da, vx0, gvl);
170 VSSEV_FLOAT(&y[j*inc_y], stride_y, vy0, gvl);
171 j += gvl;
172 }
173 }
174 return(0);
175 }
176
177
178