static void AddGreenToBlueAndRed(uint32_t* argb_data, int num_pixels) {
const uint32_t* const end = argb_data + (num_pixels & ~3);
const uint8x8_t shuffle = vld1_u8(kGreenShuffle);
for (; argb_data < end; argb_data += 4) {
const uint8x16_t argb = vld1q_u8((uint8_t*)argb_data);
const uint8x16_t greens =
vcombine_u8(vtbl1_u8(vget_low_u8(argb), shuffle),
vtbl1_u8(vget_high_u8(argb), shuffle));
vst1q_u8((uint8_t*)argb_data, vaddq_u8(argb, greens));
}
// fallthrough and finish off with plain-C
VP8LAddGreenToBlueAndRed_C(argb_data, num_pixels & 3);
}
void test_vtbl1u8 (void)
{
uint8x8_t out_uint8x8_t;
uint8x8_t arg0_uint8x8_t;
uint8x8_t arg1_uint8x8_t;
out_uint8x8_t = vtbl1_u8 (arg0_uint8x8_t, arg1_uint8x8_t);
}
void vpx_d45_predictor_8x8_neon(uint8_t *dst, ptrdiff_t stride,
const uint8_t *above, const uint8_t *left) {
static const uint8_t shuffle1[8] = { 1, 2, 3, 4, 5, 6, 7, 7 };
static const uint8_t shuffle2[8] = { 2, 3, 4, 5, 6, 7, 7, 7 };
const uint8x8_t sh_12345677 = vld1_u8(shuffle1);
const uint8x8_t sh_23456777 = vld1_u8(shuffle2);
const uint8x8_t A0 = vld1_u8(above); // top row
const uint8x8_t A1 = vtbl1_u8(A0, sh_12345677);
const uint8x8_t A2 = vtbl1_u8(A0, sh_23456777);
const uint8x8_t avg1 = vhadd_u8(A0, A2);
uint8x8_t row = vrhadd_u8(avg1, A1);
int i;
(void)left;
for (i = 0; i < 7; ++i) {
vst1_u8(dst + i * stride, row);
row = vtbl1_u8(row, sh_12345677);
}
vst1_u8(dst + i * stride, row);
}
static void ConvertBGRAToRGBA(const uint32_t* src,
int num_pixels, uint8_t* dst) {
const uint32_t* const end = src + (num_pixels & ~1);
const uint8x8_t shuffle = vld1_u8(kRGBAShuffle);
for (; src < end; src += 2) {
const uint8x8_t pixels = vld1_u8((uint8_t*)src);
vst1_u8(dst, vtbl1_u8(pixels, shuffle));
dst += 8;
}
VP8LConvertBGRAToRGBA_C(src, num_pixels & 1, dst); // left-overs
}
inline void copyOverlap8Shuffle(UInt8 * op, const UInt8 *& match, const size_t offset)
{
static constexpr UInt8 __attribute__((__aligned__(8))) masks[] =
{
0, 1, 2, 2, 4, 3, 2, 1, /* offset = 0, not used as mask, but for shift amount instead */
0, 0, 0, 0, 0, 0, 0, 0, /* offset = 1 */
0, 1, 0, 1, 0, 1, 0, 1,
0, 1, 2, 0, 1, 2, 0, 1,
0, 1, 2, 3, 0, 1, 2, 3,
0, 1, 2, 3, 4, 0, 1, 2,
0, 1, 2, 3, 4, 5, 0, 1,
0, 1, 2, 3, 4, 5, 6, 0,
};
unalignedStore(op, vtbl1_u8(unalignedLoad<uint8x8_t>(match), unalignedLoad<uint8x8_t>(masks + 8 * offset)));
match += masks[offset];
}
static WEBP_INLINE uint8x16_t DoGreenShuffle(const uint8x16_t argb,
const uint8x8_t shuffle) {
return vcombine_u8(vtbl1_u8(vget_low_u8(argb), shuffle),
vtbl1_u8(vget_high_u8(argb), shuffle));
}
uint8x8_t
tbl_testu8_ (uint8x8_t tab, uint8x8_t idx)
{
return vtbl1_u8 (tab, idx);
}
uint8x8_t test_vtbl1_u8(uint8x8_t a, uint8x8_t b) {
// CHECK: test_vtbl1_u8
return vtbl1_u8(a, b);
// CHECK: tbl {{v[0-9]+}}.8b, {{{v[0-9]+}}.16b}, {{v[0-9]+}}.8b
}
