| 1 |
/* { dg-do compile } */ |
| 2 |
/* { dg-options "-march=rv64gcv -mabi=lp64d -O3" } */ |
| 3 |
|
| 4 |
#include "riscv_vector.h" |
| 5 |
|
| 6 |
void |
| 7 |
foo (void *in, void *out, int n) |
| 8 |
{ |
| 9 |
for (int i = 0; i < n; i++) |
| 10 |
{ |
| 11 |
asm volatile("nop" ::: "memory"); |
| 12 |
vint16m8_t v0 = __riscv_vle16_v_i16m8 (in, 4);in+=100; |
| 13 |
v0 = __riscv_vwsub_wv_i16m8_tu (v0, v0, __riscv_vreinterpret_v_i16m4_i8m4 (__riscv_vget_v_i16m8_i16m4 (v0, 1)), 4); |
| 14 |
asm volatile("nop" ::: "memory"); |
| 15 |
vint16m8_t v1 = __riscv_vle16_v_i16m8 (in, 4);in+=100; |
| 16 |
v1 = __riscv_vwsub_wv_i16m8_tu (v1, v1, __riscv_vreinterpret_v_i16m4_i8m4 (__riscv_vget_v_i16m8_i16m4 (v1, 1)), 4); |
| 17 |
asm volatile("nop" ::: "memory"); |
| 18 |
vint16m8_t v2 = __riscv_vle16_v_i16m8 (in, 4);in+=100; |
| 19 |
v2 = __riscv_vwsub_wv_i16m8_tu (v2, v2, __riscv_vreinterpret_v_i16m4_i8m4 (__riscv_vget_v_i16m8_i16m4 (v2, 1)), 4); |
| 20 |
asm volatile("nop" ::: "memory"); |
| 21 |
vint8m4_t v3_n = __riscv_vle8_v_i8m4 (in, 4);in+=100; |
| 22 |
vint16m8_t v3 = __riscv_vwcvt_x_x_v_i16m8 (v3_n, 4); |
| 23 |
|
| 24 |
asm volatile("nop" ::: "memory"); |
| 25 |
__riscv_vsse16_v_i16m8 (out, 4, v0, 4);out+=100; |
| 26 |
__riscv_vsse16_v_i16m8 (out, 4, v1, 4);out+=100; |
| 27 |
__riscv_vsse16_v_i16m8 (out, 4, v2, 4);out+=100; |
| 28 |
__riscv_vsse16_v_i16m8 (out, 4, v3, 4);out+=100; |
| 29 |
} |
| 30 |
} |
| 31 |
|
| 32 |
void |
| 33 |
foo2 (void *in, void *out, int n) |
| 34 |
{ |
| 35 |
for (int i = 0; i < n; i++) |
| 36 |
{ |
| 37 |
asm volatile("nop" ::: "memory"); |
| 38 |
vint16m8_t v0 = __riscv_vle16_v_i16m8 (in, 4);in+=100; |
| 39 |
v0 = __riscv_vwadd_wv_i16m8_tu (v0, v0, __riscv_vreinterpret_v_i16m4_i8m4 (__riscv_vget_v_i16m8_i16m4 (v0, 1)), 4); |
| 40 |
asm volatile("nop" ::: "memory"); |
| 41 |
vint16m8_t v1 = __riscv_vle16_v_i16m8 (in, 4);in+=100; |
| 42 |
v1 = __riscv_vwadd_wv_i16m8_tu (v1, v1, __riscv_vreinterpret_v_i16m4_i8m4 (__riscv_vget_v_i16m8_i16m4 (v1, 1)), 4); |
| 43 |
asm volatile("nop" ::: "memory"); |
| 44 |
vint16m8_t v2 = __riscv_vle16_v_i16m8 (in, 4);in+=100; |
| 45 |
v2 = __riscv_vwadd_wv_i16m8_tu (v2, v2, __riscv_vreinterpret_v_i16m4_i8m4 (__riscv_vget_v_i16m8_i16m4 (v2, 1)), 4); |
| 46 |
asm volatile("nop" ::: "memory"); |
| 47 |
vint8m4_t v3_n = __riscv_vle8_v_i8m4 (in, 4);in+=100; |
| 48 |
vint16m8_t v3 = __riscv_vwcvt_x_x_v_i16m8 (v3_n, 4); |
| 49 |
|
| 50 |
asm volatile("nop" ::: "memory"); |
| 51 |
__riscv_vsse16_v_i16m8 (out, 4, v0, 4);out+=100; |
| 52 |
__riscv_vsse16_v_i16m8 (out, 4, v1, 4);out+=100; |
| 53 |
__riscv_vsse16_v_i16m8 (out, 4, v2, 4);out+=100; |
| 54 |
__riscv_vsse16_v_i16m8 (out, 4, v3, 4);out+=100; |
| 55 |
} |
| 56 |
} |
| 57 |
|
| 58 |
/* { dg-final { scan-assembler-not {vmv1r} } } */ |
| 59 |
/* { dg-final { scan-assembler-not {vmv2r} } } */ |
| 60 |
/* { dg-final { scan-assembler-not {vmv4r} } } */ |
| 61 |
/* { dg-final { scan-assembler-not {vmv8r} { xfail riscv*-*-* } } } */ |
| 62 |
/* { dg-final { scan-assembler-not {csrr} { xfail riscv*-*-* } } } */ |