| 1 |
/* { dg-do run { target { rv64 } } } */ |
| 2 |
/* { dg-require-effective-target rv64 } */ |
| 3 |
/* { dg-require-effective-target riscv_v_ok } */ |
| 4 |
/* { dg-require-effective-target riscv_zvfh_ok } */ |
| 5 |
/* { dg-options " -march=rv64gcv_zvfh -mabi=lp64d -O2" } */ |
| 6 |
|
| 7 |
#include "riscv_vector.h" |
| 8 |
|
| 9 |
int8_t a[1]; |
| 10 |
uint16_t b[1]; |
| 11 |
float c[1], n[1]; |
| 12 |
uint16_t d[1]; |
| 13 |
uint8_t e[1]; |
| 14 |
uint16_t f[1]; |
| 15 |
_Float16 g[1], k[1], m[1], p[1]; |
| 16 |
uint16_t i[1]; |
| 17 |
int8_t j[1]; |
| 18 |
uint8_t o[1]; |
| 19 |
uint32_t l[1]; |
| 20 |
uint16_t q[1]; |
| 21 |
uint32_t r[1]; |
| 22 |
uint32_t s[1]; |
| 23 |
int16_t t[1]; |
| 24 |
int main() |
| 25 |
{ |
| 26 |
int u = 25; |
| 27 |
int8_t *v = a; |
| 28 |
uint32_t *w; |
| 29 |
uint16_t *aa = b; |
| 30 |
float *ab = c, *as = n; |
| 31 |
uint32_t *ad; |
| 32 |
uint16_t *ah = f; |
| 33 |
_Float16 *ai = g, *aj = k, *an = m, *au = p; |
| 34 |
int32_t *ak; |
| 35 |
int16_t *al; |
| 36 |
uint16_t *am = i; |
| 37 |
int8_t *ao = j; |
| 38 |
uint8_t *ap = o; |
| 39 |
uint32_t *aq = l; |
| 40 |
uint16_t *ar = q; |
| 41 |
uint32_t *at = r; |
| 42 |
uint32_t *av = s; |
| 43 |
int32_t *ax; |
| 44 |
int16_t *ay = t; |
| 45 |
for (size_t az; u; u -= az) |
| 46 |
{ |
| 47 |
az = __riscv_vsetvl_e32m8(u); |
| 48 |
vint8m2_t ba = __riscv_vle8_v_i8m2(v, az); |
| 49 |
vbool4_t bb = __riscv_vmseq_vx_i8m2_b4(ba, 1, az); |
| 50 |
vuint16m4_t bc = __riscv_vsll_vx_u16m4(__riscv_vid_v_u16m4(az), 2, az); |
| 51 |
vuint32m8_t bd = __riscv_vsll_vx_u32m8(__riscv_vid_v_u32m8(az), 1, az); |
| 52 |
vuint32m8_t be = __riscv_vluxei16_v_u32m8_m(bb, w, bc, az); |
| 53 |
vuint16m4_t bf; |
| 54 |
__riscv_vsuxei16_v_u32m8_m(bb, aq, bf, be, az); |
| 55 |
vuint8m2_t bg = __riscv_vsll_vx_u8m2(__riscv_vid_v_u8m2(az), 1, az); |
| 56 |
vuint16m4_t bh = __riscv_vloxei8_v_u16m4(aa, bg, az); |
| 57 |
vfloat16m4_t bi; |
| 58 |
vuint16m4_t bj = __riscv_vsll_vx_u16m4(__riscv_vid_v_u16m4(az), 1, az); |
| 59 |
vint16m4_t bk = __riscv_vloxei32_v_i16m4_m(bb, al, bd, az); |
| 60 |
__riscv_vsse16_v_u16m4(ar, 2, bh, az); |
| 61 |
vuint16m4_t bl = __riscv_vloxei16_v_u16m4(d, bj, az); |
| 62 |
vfloat16m4_t bm = __riscv_vle16_v_f16m4(ai, az); |
| 63 |
vuint16m4_t bn = __riscv_vlse16_v_u16m4(ah, 2, az); |
| 64 |
vint32m8_t bo = __riscv_vle32_v_i32m8_m(bb, ak, az); |
| 65 |
vfloat16m1_t bp = __riscv_vle16_v_f16m1(aj, az); |
| 66 |
vuint16m4_t bq = __riscv_vrgatherei16_vv_u16m4(bl, bn, az); |
| 67 |
__riscv_vse16_v_u16m4(am, bq, az); |
| 68 |
vfloat16m1_t br = __riscv_vfredusum_vs_f16m4_f16m1_m(bb, bm, bp, az); |
| 69 |
vuint8m2_t bs; |
| 70 |
vuint32m8_t bt; |
| 71 |
__riscv_vse16_v_f16m1(an, br, az); |
| 72 |
vfloat32m8_t bu = __riscv_vloxei8_v_f32m8(ab, bs, az); |
| 73 |
__riscv_vse16_v_i16m4(ay, bk, az); |
| 74 |
bi = __riscv_vfmv_s_f_f16m4(1, az); |
| 75 |
__riscv_vse16_v_f16m4(au, bi, az); |
| 76 |
vuint16m4_t bw = __riscv_vsll_vx_u16m4(__riscv_vid_v_u16m4(az), 0, az); |
| 77 |
vuint32m8_t by = __riscv_vle32_v_u32m8_m(bb, ad, az); |
| 78 |
bt = __riscv_vmv_s_x_u32m8(3090659, az); |
| 79 |
__riscv_vse32_v_u32m8(at, bt, az); |
| 80 |
vuint8m2_t bz = __riscv_vloxei16_v_u8m2(e, bw, az); |
| 81 |
__riscv_vse32_v_u32m8(av, by, az); |
| 82 |
vint8m2_t cd; |
| 83 |
__riscv_vse8_v_i8m2(ao, cd, az); |
| 84 |
__riscv_vsse32_v_i32m8_m(bb, ax, 4, bo, az); |
| 85 |
__riscv_vse32_v_f32m8(as, bu, az); |
| 86 |
vuint16m4_t cf; |
| 87 |
__riscv_vsoxei16_v_u32m8(aq, cf, be, az); |
| 88 |
vuint8m2_t cg = __riscv_vmulhu_vx_u8m2(bz, 0, az); |
| 89 |
vuint32m8_t ch = __riscv_vsll_vx_u32m8(__riscv_vid_v_u32m8(az), 0, az); |
| 90 |
__riscv_vsoxei32_v_u8m2(ap, ch, cg, az); |
| 91 |
} |
| 92 |
return 0; |
| 93 |
} |