1 /* { dg-do compile } */
2 /* { dg-options "-march=rv32gcv -mabi=ilp32d -O3" } */
3 /* { dg-final { check-function-bodies "**" "" } } */
4 #include "riscv_vector.h"
5
6 /*
7 ** f1:
8 ** vsetivli\tzero,4,e32,m1,tu,ma
9 ** vle32\.v\tv[0-9]+,0\([a-x0-9]+\)
10 ** vle32\.v\tv[0-9]+,0\([a-x0-9]+\)
11 ** vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16
12 ** vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16
13 ** vse32\.v\tv[0-9]+,0\([a-x0-9]+\)
14 ** ret
15 */
16 void f1 (void * in, void *out, int32_t x)
17 {
18 vint32m1_t v = __riscv_vle32_v_i32m1 (in, 4);
19 vint32m1_t v2 = __riscv_vle32_v_i32m1_tu (v, in, 4);
20 vint32m1_t v3 = __riscv_vsub_vx_i32m1 (v2, 16, 4);
21 vint32m1_t v4 = __riscv_vsub_vx_i32m1_tu (v3, v2, 16, 4);
22 __riscv_vse32_v_i32m1 (out, v4, 4);
23 }
24
25 /*
26 ** f2:
27 ** vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma
28 ** vlm.v\tv[0-9]+,0\([a-x0-9]+\)
29 ** vsetivli\tzero,4,e32,m1,ta,ma
30 ** vle32.v\tv[0-9]+,0\([a-x0-9]+\),v0.t
31 ** vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16
32 ** vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t
33 ** vse32.v\tv[0-9]+,0\([a-x0-9]+\)
34 ** ret
35 */
36 void f2 (void * in, void *out, int32_t x)
37 {
38 vbool32_t mask = *(vbool32_t*)in;
39 asm volatile ("":::"memory");
40 vint32m1_t v = __riscv_vle32_v_i32m1 (in, 4);
41 vint32m1_t v2 = __riscv_vle32_v_i32m1_m (mask, in, 4);
42 vint32m1_t v3 = __riscv_vsub_vx_i32m1 (v2, 16, 4);
43 vint32m1_t v4 = __riscv_vsub_vx_i32m1_m (mask, v3, 16, 4);
44 __riscv_vse32_v_i32m1 (out, v4, 4);
45 }
46
47 /*
48 ** f3:
49 ** vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma
50 ** vlm.v\tv[0-9]+,0\([a-x0-9]+\)
51 ** vsetivli\tzero,4,e32,m1,tu,mu
52 ** vle32\.v\tv[0-9]+,0\([a-x0-9]+\)
53 ** vle32.v\tv[0-9]+,0\([a-x0-9]+\),v0.t
54 ** vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16
55 ** vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t
56 ** vse32.v\tv[0-9]+,0\([a-x0-9]+\)
57 ** ret
58 */
59 void f3 (void * in, void *out, int32_t x)
60 {
61 vbool32_t mask = *(vbool32_t*)in;
62 asm volatile ("":::"memory");
63 vint32m1_t v = __riscv_vle32_v_i32m1 (in, 4);
64 vint32m1_t v2 = __riscv_vle32_v_i32m1_tumu (mask, v, in, 4);
65 vint32m1_t v3 = __riscv_vsub_vx_i32m1 (v2, 16, 4);
66 vint32m1_t v4 = __riscv_vsub_vx_i32m1_tumu (mask, v3, v2, 16, 4);
67 __riscv_vse32_v_i32m1 (out, v4, 4);
68 }
69
70 /*
71 ** f4:
72 ** vsetivli\tzero,4,e8,mf8,tu,ma
73 ** vle8\.v\tv[0-9]+,0\([a-x0-9]+\)
74 ** vle8\.v\tv[0-9]+,0\([a-x0-9]+\)
75 ** vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16
76 ** vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16
77 ** vse8\.v\tv[0-9]+,0\([a-x0-9]+\)
78 ** ret
79 */
80 void f4 (void * in, void *out, int8_t x)
81 {
82 vint8mf8_t v = __riscv_vle8_v_i8mf8 (in, 4);
83 vint8mf8_t v2 = __riscv_vle8_v_i8mf8_tu (v, in, 4);
84 vint8mf8_t v3 = __riscv_vsub_vx_i8mf8 (v2, 16, 4);
85 vint8mf8_t v4 = __riscv_vsub_vx_i8mf8_tu (v3, v2, 16, 4);
86 __riscv_vse8_v_i8mf8 (out, v4, 4);
87 }
88
89 /*
90 ** f5:
91 ** vsetvli\t[a-x0-9]+,zero,e8,mf8,ta,ma
92 ** vlm.v\tv[0-9]+,0\([a-x0-9]+\)
93 ** vsetivli\tzero,4,e8,mf8,ta,ma
94 ** vle8.v\tv[0-9]+,0\([a-x0-9]+\),v0.t
95 ** vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16
96 ** vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t
97 ** vse8.v\tv[0-9]+,0\([a-x0-9]+\)
98 ** ret
99 */
100 void f5 (void * in, void *out, int8_t x)
101 {
102 vbool64_t mask = *(vbool64_t*)in;
103 asm volatile ("":::"memory");
104 vint8mf8_t v = __riscv_vle8_v_i8mf8 (in, 4);
105 vint8mf8_t v2 = __riscv_vle8_v_i8mf8_m (mask, in, 4);
106 vint8mf8_t v3 = __riscv_vsub_vx_i8mf8 (v2, 16, 4);
107 vint8mf8_t v4 = __riscv_vsub_vx_i8mf8_m (mask, v3, 16, 4);
108 __riscv_vse8_v_i8mf8 (out, v4, 4);
109 }
110
111 /*
112 ** f6:
113 ** vsetvli\t[a-x0-9]+,zero,e8,mf8,ta,ma
114 ** vlm.v\tv[0-9]+,0\([a-x0-9]+\)
115 ** vsetivli\tzero,4,e8,mf8,tu,mu
116 ** vle8\.v\tv[0-9]+,0\([a-x0-9]+\)
117 ** vle8.v\tv[0-9]+,0\([a-x0-9]+\),v0.t
118 ** vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16
119 ** vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t
120 ** vse8.v\tv[0-9]+,0\([a-x0-9]+\)
121 ** ret
122 */
123 void f6 (void * in, void *out, int8_t x)
124 {
125 vbool64_t mask = *(vbool64_t*)in;
126 asm volatile ("":::"memory");
127 vint8mf8_t v = __riscv_vle8_v_i8mf8 (in, 4);
128 vint8mf8_t v2 = __riscv_vle8_v_i8mf8_tumu (mask, v, in, 4);
129 vint8mf8_t v3 = __riscv_vsub_vx_i8mf8 (v2, 16, 4);
130 vint8mf8_t v4 = __riscv_vsub_vx_i8mf8_tumu (mask, v3, v2, 16, 4);
131 __riscv_vse8_v_i8mf8 (out, v4, 4);
132 }