| 1 |
/* { dg-do compile } */ |
| 2 |
/* { dg-options "-O3 -march=rv64gcv_zvl512b -mabi=lp64d -mstrict-align -mvector-strict-align" } */ |
| 3 |
|
| 4 |
/* Make sure we do not construct the vector element-wise despite |
| 5 |
slow misaligned scalar and vector accesses. */ |
| 6 |
|
| 7 |
typedef unsigned char uint8_t; |
| 8 |
typedef unsigned short uint16_t; |
| 9 |
typedef unsigned int uint32_t; |
| 10 |
|
| 11 |
#define HADAMARD4(d0, d1, d2, d3, s0, s1, s2, s3) \ |
| 12 |
{ \ |
| 13 |
int t0 = s0 + s1; \ |
| 14 |
int t1 = s0 - s1; \ |
| 15 |
int t2 = s2 + s3; \ |
| 16 |
int t3 = s2 - s3; \ |
| 17 |
d0 = t0 + t2; \ |
| 18 |
d2 = t0 - t2; \ |
| 19 |
d1 = t1 + t3; \ |
| 20 |
d3 = t1 - t3; \ |
| 21 |
} |
| 22 |
|
| 23 |
uint32_t |
| 24 |
abs2 (uint32_t a) |
| 25 |
{ |
| 26 |
uint32_t s = ((a >> 15) & 0x10001) * 0xffff; |
| 27 |
return (a + s) ^ s; |
| 28 |
} |
| 29 |
|
| 30 |
int |
| 31 |
x264_pixel_satd_8x4 (uint8_t *pix1, int i_pix1, uint8_t *pix2, int i_pix2) |
| 32 |
{ |
| 33 |
uint32_t tmp[4][4]; |
| 34 |
uint32_t a0, a1, a2, a3; |
| 35 |
int sum = 0; |
| 36 |
for (int i = 0; i < 4; i++, pix1 += i_pix1, pix2 += i_pix2) |
| 37 |
{ |
| 38 |
a0 = (pix1[0] - pix2[0]) + ((pix1[4] - pix2[4]) << 16); |
| 39 |
a1 = (pix1[1] - pix2[1]) + ((pix1[5] - pix2[5]) << 16); |
| 40 |
a2 = (pix1[2] - pix2[2]) + ((pix1[6] - pix2[6]) << 16); |
| 41 |
a3 = (pix1[3] - pix2[3]) + ((pix1[7] - pix2[7]) << 16); |
| 42 |
HADAMARD4 (tmp[i][0], tmp[i][1], tmp[i][2], tmp[i][3], a0, a1, a2, a3); |
| 43 |
} |
| 44 |
for (int i = 0; i < 4; i++) |
| 45 |
{ |
| 46 |
HADAMARD4 (a0, a1, a2, a3, tmp[0][i], tmp[1][i], tmp[2][i], tmp[3][i]); |
| 47 |
sum += abs2 (a0) + abs2 (a1) + abs2 (a2) + abs2 (a3); |
| 48 |
} |
| 49 |
return (((uint16_t) sum) + ((uint32_t) sum >> 16)) >> 1; |
| 50 |
} |
| 51 |
|
| 52 |
/* { dg-final { scan-assembler-not "lbu" } } */ |