hashirama/uvg266
1
0
Fork 0
mirror of https://github.com/ultravideo/uvg266.git synced 2024-11-28 03:34:06 +00:00
Code Issues Actions Projects Releases Wiki Activity

Added part of the functions 16x16 DCT needs

Browse source
This commit is contained in:
Ari Lemmetti 2014-09-17 16:28:12 +03:00 committed by Ari Koivula
parent aba3acdfff
commit d6af146a2e
1 changed files with 24 additions and 0 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

24
src/strategies/avx2/dct-avx2.c
View file

@ -109,6 +109,9 @@ static void transpose_8x8_16bit(const int16_t *src, int16_t *dst)
_mm_storeu_si128((__m128i*)dst + 7, _mm256_extracti128_si256(tmp[3], 1)); _mm_storeu_si128((__m128i*)dst + 7, _mm256_extracti128_si256(tmp[3], 1));
} }
static void transpose_16x16_16bit(const int16_t *src, int16_t *dst)
{
}
static void mul_matrix_4x4_avx2(const int16_t *first, const int16_t *second, int16_t *dst, int32_t shift) static void mul_matrix_4x4_avx2(const int16_t *first, const int16_t *second, int16_t *dst, int32_t shift)
{ {
__m256i b[2], a, result, even[2], odd[2]; __m256i b[2], a, result, even[2], odd[2];
@ -202,6 +205,11 @@ static void mul_matrix_8x8_avx2(const int16_t *first, const int16_t *second, int
} }
} }
static void mul_matrix_16x16_avx2(const int16_t *first, const int16_t *second, int16_t *dst, const int32_t shift)
{
}
static void matrix_transform_2d_4x4_avx2(const int16_t *src, int16_t *dst, const int16_t *transform, const int16_t shift0, const int16_t shift1) static void matrix_transform_2d_4x4_avx2(const int16_t *src, int16_t *dst, const int16_t *transform, const int16_t shift0, const int16_t shift1)
{ {
int16_t tmp[4 * 4]; int16_t tmp[4 * 4];
@ -242,6 +250,15 @@ static void matrix_itransform_2d_8x8_avx2(const int16_t *src, int16_t *dst, cons
mul_matrix_8x8_avx2(tmp, transform, dst, shift1); mul_matrix_8x8_avx2(tmp, transform, dst, shift1);
} }
static void matrix_transform_2d_16x16_avx2(const int16_t *src, int16_t *dst, const int16_t *transform, const int16_t shift0, const int16_t shift1)
{
int16_t tmp[16 * 16];
int16_t transposed[16 * 16];
transpose_16x16_16bit(transform, transposed);
mul_matrix_16x16_avx2(src, transposed, tmp, shift0);
mul_matrix_16x16_avx2(transform, tmp, dst, shift1);
}
static void partial_butterfly_16_avx2(short *src, short *dst, static void partial_butterfly_16_avx2(short *src, short *dst,
int32_t shift) int32_t shift)
@ -512,6 +529,13 @@ static void matrix_idct_8x8_avx2(int8_t bitdepth, int16_t *dst, int16_t *src)
matrix_itransform_2d_8x8_avx2(src, dst, (const int16_t*)g_t8, shift_1st, shift_2nd); matrix_itransform_2d_8x8_avx2(src, dst, (const int16_t*)g_t8, shift_1st, shift_2nd);
} }
static void matrix_dct_16x16_avx2(int8_t bitdepth, int16_t *src, int16_t *dst)
{
int32_t shift_1st = g_convert_to_bit[16] + 1 + (bitdepth - 8);
int32_t shift_2nd = g_convert_to_bit[16] + 8;
matrix_transform_2d_16x16_avx2(src, dst, (const int16_t*)g_t16, shift_1st, shift_2nd);
}
#endif //COMPILE_INTEL_AVX2 #endif //COMPILE_INTEL_AVX2
int strategy_register_dct_avx2(void* opaque) int strategy_register_dct_avx2(void* opaque)