Debian Codegraph @

gcc/testsuite/gcc.target/riscv/rvv/base/pr112431-22.c

raw
1 /* { dg-do compile } */
2 /* { dg-options "-march=rv64gcv -mabi=lp64d -O3" } */
3
4 #include "riscv_vector.h"
5
6 size_t __attribute__ ((noinline))
7 sumation (size_t sum0, size_t sum1, size_t sum2, size_t sum3, size_t sum4,
8 size_t sum5, size_t sum6, size_t sum7, size_t sum8, size_t sum9,
9 size_t sum10, size_t sum11, size_t sum12, size_t sum13, size_t sum14,
10 size_t sum15)
11 {
12 return sum0 + sum1 + sum2 + sum3 + sum4 + sum5 + sum6 + sum7 + sum8 + sum9
13 + sum10 + sum11 + sum12 + sum13 + sum14 + sum15;
14 }
15
16 size_t
17 foo (char const *buf, size_t len)
18 {
19 size_t sum = 0;
20 size_t vl = __riscv_vsetvlmax_e8m8 ();
21 size_t step = vl * 4;
22 const char *it = buf, *end = buf + len;
23 for (; it + step <= end;)
24 {
25 vint8m1_t v0 = __riscv_vle8_v_i8m1 ((void *) it, vl);
26 it += vl;
27 vint8m1_t v1 = __riscv_vle8_v_i8m1 ((void *) it, vl);
28 it += vl;
29 vint8m1_t v2 = __riscv_vle8_v_i8m1 ((void *) it, vl);
30 it += vl;
31 vint8m1_t v3 = __riscv_vle8_v_i8m1 ((void *) it, vl);
32 it += vl;
33 vint8m1_t v4 = __riscv_vle8_v_i8m1 ((void *) it, vl);
34 it += vl;
35 vint8m1_t v5 = __riscv_vle8_v_i8m1 ((void *) it, vl);
36 it += vl;
37 vint8m1_t v6 = __riscv_vle8_v_i8m1 ((void *) it, vl);
38 it += vl;
39 vint8m1_t v7 = __riscv_vle8_v_i8m1 ((void *) it, vl);
40 it += vl;
41 vint8m1_t v8 = __riscv_vle8_v_i8m1 ((void *) it, vl);
42 it += vl;
43 vint8m1_t v9 = __riscv_vle8_v_i8m1 ((void *) it, vl);
44 it += vl;
45 vint8m1_t v10 = __riscv_vle8_v_i8m1 ((void *) it, vl);
46 it += vl;
47 vint8m1_t v11 = __riscv_vle8_v_i8m1 ((void *) it, vl);
48 it += vl;
49 vint8m1_t v12 = __riscv_vle8_v_i8m1 ((void *) it, vl);
50 it += vl;
51 vint8m1_t v13 = __riscv_vle8_v_i8m1 ((void *) it, vl);
52 it += vl;
53 vint8m1_t v14 = __riscv_vle8_v_i8m1 ((void *) it, vl);
54 it += vl;
55 vint8m1_t v15 = __riscv_vle8_v_i8m1 ((void *) it, vl);
56 it += vl;
57
58 asm volatile("nop" ::: "memory");
59 vint16m2_t vw0 = __riscv_vwadd_vx_i16m2 (v0, 33, vl);
60 vint16m2_t vw1 = __riscv_vwadd_vx_i16m2 (v1, 33, vl);
61 vint16m2_t vw2 = __riscv_vwadd_vx_i16m2 (v2, 33, vl);
62 vint16m2_t vw3 = __riscv_vwadd_vx_i16m2 (v3, 33, vl);
63 vint16m2_t vw4 = __riscv_vwadd_vx_i16m2 (v4, 33, vl);
64 vint16m2_t vw5 = __riscv_vwadd_vx_i16m2 (v5, 33, vl);
65 vint16m2_t vw6 = __riscv_vwadd_vx_i16m2 (v6, 33, vl);
66 vint16m2_t vw7 = __riscv_vwadd_vx_i16m2 (v7, 33, vl);
67 vint16m2_t vw8 = __riscv_vwadd_vx_i16m2 (v8, 33, vl);
68 vint16m2_t vw9 = __riscv_vwadd_vx_i16m2 (v9, 33, vl);
69 vint16m2_t vw10 = __riscv_vwadd_vx_i16m2 (v10, 33, vl);
70 vint16m2_t vw11 = __riscv_vwadd_vx_i16m2 (v11, 33, vl);
71 vint16m2_t vw12 = __riscv_vwadd_vx_i16m2 (v12, 33, vl);
72 vint16m2_t vw13 = __riscv_vwadd_vx_i16m2 (v13, 33, vl);
73 vint16m2_t vw14 = __riscv_vwadd_vx_i16m2 (v14, 33, vl);
74 vint16m2_t vw15 = __riscv_vwadd_vx_i16m2 (v15, 33, vl);
75
76 asm volatile("nop" ::: "memory");
77 size_t sum0 = __riscv_vmv_x_s_i16m2_i16 (vw0);
78 size_t sum1 = __riscv_vmv_x_s_i16m2_i16 (vw1);
79 size_t sum2 = __riscv_vmv_x_s_i16m2_i16 (vw2);
80 size_t sum3 = __riscv_vmv_x_s_i16m2_i16 (vw3);
81 size_t sum4 = __riscv_vmv_x_s_i16m2_i16 (vw4);
82 size_t sum5 = __riscv_vmv_x_s_i16m2_i16 (vw5);
83 size_t sum6 = __riscv_vmv_x_s_i16m2_i16 (vw6);
84 size_t sum7 = __riscv_vmv_x_s_i16m2_i16 (vw7);
85 size_t sum8 = __riscv_vmv_x_s_i16m2_i16 (vw8);
86 size_t sum9 = __riscv_vmv_x_s_i16m2_i16 (vw9);
87 size_t sum10 = __riscv_vmv_x_s_i16m2_i16 (vw10);
88 size_t sum11 = __riscv_vmv_x_s_i16m2_i16 (vw11);
89 size_t sum12 = __riscv_vmv_x_s_i16m2_i16 (vw12);
90 size_t sum13 = __riscv_vmv_x_s_i16m2_i16 (vw13);
91 size_t sum14 = __riscv_vmv_x_s_i16m2_i16 (vw14);
92 size_t sum15 = __riscv_vmv_x_s_i16m2_i16 (vw15);
93
94 sum += sumation (sum0, sum1, sum2, sum3, sum4, sum5, sum6, sum7, sum8,
95 sum9, sum10, sum11, sum12, sum13, sum14, sum15);
96 }
97 return sum;
98 }
99
100 size_t
101 foo2 (char const *buf, size_t len)
102 {
103 size_t sum = 0;
104 size_t vl = __riscv_vsetvlmax_e8m8 ();
105 size_t step = vl * 4;
106 const char *it = buf, *end = buf + len;
107 for (; it + step <= end;)
108 {
109 vint8m1_t v0 = __riscv_vle8_v_i8m1 ((void *) it, vl);
110 it += vl;
111 vint8m1_t v1 = __riscv_vle8_v_i8m1 ((void *) it, vl);
112 it += vl;
113 vint8m1_t v2 = __riscv_vle8_v_i8m1 ((void *) it, vl);
114 it += vl;
115 vint8m1_t v3 = __riscv_vle8_v_i8m1 ((void *) it, vl);
116 it += vl;
117 vint8m1_t v4 = __riscv_vle8_v_i8m1 ((void *) it, vl);
118 it += vl;
119 vint8m1_t v5 = __riscv_vle8_v_i8m1 ((void *) it, vl);
120 it += vl;
121 vint8m1_t v6 = __riscv_vle8_v_i8m1 ((void *) it, vl);
122 it += vl;
123 vint8m1_t v7 = __riscv_vle8_v_i8m1 ((void *) it, vl);
124 it += vl;
125 vint8m1_t v8 = __riscv_vle8_v_i8m1 ((void *) it, vl);
126 it += vl;
127 vint8m1_t v9 = __riscv_vle8_v_i8m1 ((void *) it, vl);
128 it += vl;
129 vint8m1_t v10 = __riscv_vle8_v_i8m1 ((void *) it, vl);
130 it += vl;
131 vint8m1_t v11 = __riscv_vle8_v_i8m1 ((void *) it, vl);
132 it += vl;
133 vint8m1_t v12 = __riscv_vle8_v_i8m1 ((void *) it, vl);
134 it += vl;
135 vint8m1_t v13 = __riscv_vle8_v_i8m1 ((void *) it, vl);
136 it += vl;
137 vint8m1_t v14 = __riscv_vle8_v_i8m1 ((void *) it, vl);
138 it += vl;
139 vint8m1_t v15 = __riscv_vle8_v_i8m1 ((void *) it, vl);
140 it += vl;
141
142 asm volatile("nop" ::: "memory");
143 vint16m2_t vw0 = __riscv_vwmulsu_vx_i16m2 (v0, 33, vl);
144 vint16m2_t vw1 = __riscv_vwmulsu_vx_i16m2 (v1, 33, vl);
145 vint16m2_t vw2 = __riscv_vwmulsu_vx_i16m2 (v2, 33, vl);
146 vint16m2_t vw3 = __riscv_vwmulsu_vx_i16m2 (v3, 33, vl);
147 vint16m2_t vw4 = __riscv_vwmulsu_vx_i16m2 (v4, 33, vl);
148 vint16m2_t vw5 = __riscv_vwmulsu_vx_i16m2 (v5, 33, vl);
149 vint16m2_t vw6 = __riscv_vwmulsu_vx_i16m2 (v6, 33, vl);
150 vint16m2_t vw7 = __riscv_vwmulsu_vx_i16m2 (v7, 33, vl);
151 vint16m2_t vw8 = __riscv_vwmulsu_vx_i16m2 (v8, 33, vl);
152 vint16m2_t vw9 = __riscv_vwmulsu_vx_i16m2 (v9, 33, vl);
153 vint16m2_t vw10 = __riscv_vwmulsu_vx_i16m2 (v10, 33, vl);
154 vint16m2_t vw11 = __riscv_vwmulsu_vx_i16m2 (v11, 33, vl);
155 vint16m2_t vw12 = __riscv_vwmulsu_vx_i16m2 (v12, 33, vl);
156 vint16m2_t vw13 = __riscv_vwmulsu_vx_i16m2 (v13, 33, vl);
157 vint16m2_t vw14 = __riscv_vwmulsu_vx_i16m2 (v14, 33, vl);
158 vint16m2_t vw15 = __riscv_vwmulsu_vx_i16m2 (v15, 33, vl);
159
160 asm volatile("nop" ::: "memory");
161 size_t sum0 = __riscv_vmv_x_s_i16m2_i16 (vw0);
162 size_t sum1 = __riscv_vmv_x_s_i16m2_i16 (vw1);
163 size_t sum2 = __riscv_vmv_x_s_i16m2_i16 (vw2);
164 size_t sum3 = __riscv_vmv_x_s_i16m2_i16 (vw3);
165 size_t sum4 = __riscv_vmv_x_s_i16m2_i16 (vw4);
166 size_t sum5 = __riscv_vmv_x_s_i16m2_i16 (vw5);
167 size_t sum6 = __riscv_vmv_x_s_i16m2_i16 (vw6);
168 size_t sum7 = __riscv_vmv_x_s_i16m2_i16 (vw7);
169 size_t sum8 = __riscv_vmv_x_s_i16m2_i16 (vw8);
170 size_t sum9 = __riscv_vmv_x_s_i16m2_i16 (vw9);
171 size_t sum10 = __riscv_vmv_x_s_i16m2_i16 (vw10);
172 size_t sum11 = __riscv_vmv_x_s_i16m2_i16 (vw11);
173 size_t sum12 = __riscv_vmv_x_s_i16m2_i16 (vw12);
174 size_t sum13 = __riscv_vmv_x_s_i16m2_i16 (vw13);
175 size_t sum14 = __riscv_vmv_x_s_i16m2_i16 (vw14);
176 size_t sum15 = __riscv_vmv_x_s_i16m2_i16 (vw15);
177
178 sum += sumation (sum0, sum1, sum2, sum3, sum4, sum5, sum6, sum7, sum8,
179 sum9, sum10, sum11, sum12, sum13, sum14, sum15);
180 }
181 return sum;
182 }
183
184 /* { dg-final { scan-assembler-not {vmv1r} } } */
185 /* { dg-final { scan-assembler-not {vmv2r} } } */
186 /* { dg-final { scan-assembler-not {vmv4r} } } */
187 /* { dg-final { scan-assembler-not {vmv8r} } } */
188 /* { dg-final { scan-assembler-not {csrr} { xfail riscv*-*-* } } } */