| 1 |
/* { dg-do compile } */ |
| 2 |
/* { dg-options "-O3 -march=rv64gcv -mabi=lp64d -mrvv-max-lmul=m8" } */ |
| 3 |
|
| 4 |
#include "perm.h" |
| 5 |
|
| 6 |
#define MASKE_2(x, y) (x), (x + 2) |
| 7 |
#define MASKE_4(x, y) MASKE_2 (x, y), MASKE_2 (x + 4, y) |
| 8 |
#define MASKE_8(x, y) MASKE_4 (x, y), MASKE_4 (x + 8, y) |
| 9 |
#define MASKE_16(x, y) MASKE_8 (x, y), MASKE_8 (x + 16, y) |
| 10 |
#define MASKE_32(x, y) MASKE_16 (x, y), MASKE_16 (x + 32, y) |
| 11 |
#define MASKE_64(x, y) MASKE_32 (x, y), MASKE_32 (x + 64, y) |
| 12 |
|
| 13 |
#define MASKO_2(x, y) (x + 1), (x + 3) |
| 14 |
#define MASKO_4(x, y) MASKO_2 (x, y), MASKO_2 (x + 4, y) |
| 15 |
#define MASKO_8(x, y) MASKO_4 (x, y), MASKO_4 (x + 8, y) |
| 16 |
#define MASKO_16(x, y) MASKO_8 (x, y), MASKO_8 (x + 16, y) |
| 17 |
#define MASKO_32(x, y) MASKO_16 (x, y), MASKO_16 (x + 32, y) |
| 18 |
#define MASKO_64(x, y) MASKO_32 (x, y), MASKO_32 (x + 64, y) |
| 19 |
|
| 20 |
#define PERMUTE1(TYPE, NUNITS) \ |
| 21 |
__attribute__ ((noipa)) void permute1_##TYPE (TYPE values1, TYPE values2, \ |
| 22 |
TYPE *out) \ |
| 23 |
{ \ |
| 24 |
TYPE v = __builtin_shufflevector (values1, values2, \ |
| 25 |
MASKE_##NUNITS (0, NUNITS)); \ |
| 26 |
*(TYPE *) out = v; \ |
| 27 |
} |
| 28 |
|
| 29 |
#define PERMUTE2(TYPE, NUNITS) \ |
| 30 |
__attribute__ ((noipa)) void permute2_##TYPE (TYPE values1, TYPE values2, \ |
| 31 |
TYPE *out) \ |
| 32 |
{ \ |
| 33 |
TYPE v = __builtin_shufflevector (values1, values2, \ |
| 34 |
MASKO_##NUNITS (0, NUNITS)); \ |
| 35 |
*(TYPE *) out = v; \ |
| 36 |
} |
| 37 |
|
| 38 |
#define TEST_ALL(T) \ |
| 39 |
T (vnx4qi, 4) \ |
| 40 |
T (vnx8qi, 8) \ |
| 41 |
T (vnx16qi, 16) \ |
| 42 |
T (vnx32qi, 32) \ |
| 43 |
T (vnx64qi, 64) \ |
| 44 |
T (vnx4hi, 4) \ |
| 45 |
T (vnx8hi, 8) \ |
| 46 |
T (vnx16hi, 16) \ |
| 47 |
T (vnx32hi, 32) \ |
| 48 |
T (vnx64hi, 64) \ |
| 49 |
T (vnx4si, 4) \ |
| 50 |
T (vnx8si, 8) \ |
| 51 |
T (vnx16si, 16) \ |
| 52 |
T (vnx32si, 32) \ |
| 53 |
T (vnx4di, 4) \ |
| 54 |
T (vnx8di, 8) \ |
| 55 |
T (vnx16di, 16) \ |
| 56 |
T (vnx4sf, 4) \ |
| 57 |
T (vnx8sf, 8) \ |
| 58 |
T (vnx16sf, 16) \ |
| 59 |
T (vnx32sf, 32) \ |
| 60 |
T (vnx4df, 4) \ |
| 61 |
T (vnx8df, 8) \ |
| 62 |
T (vnx16df, 16) |
| 63 |
|
| 64 |
TEST_ALL (PERMUTE1) |
| 65 |
TEST_ALL (PERMUTE2) |
| 66 |
|
| 67 |
/* { dg-final { scan-assembler-times "vslideup" 48 } } */ |
| 68 |
/* { dg-final { scan-assembler-times "vcompress" 96 } } */ |