Searched refs:scaled_v_low (Results 1 – 4 of 4) sorted by relevance
/external/tensorflow/tensorflow/lite/kernels/internal/optimized/ |
D | avx2_quantization_utils.h | 105 __m256i scaled_v_low = _mm256_mul_epi32( in MultiplyByQuantizedMultiplier() local 112 scaled_v_low = _mm256_srlv_epi64(scaled_v_low, _mm256_set1_epi64x(31)); in MultiplyByQuantizedMultiplier() 116 __m256i result = _mm256_blend_epi32(scaled_v_low, scaled_v_high, 0xaa); in MultiplyByQuantizedMultiplier() 140 __m256i scaled_v_low = _mm256_mul_epi32( in MultiplyByQuantizedMultiplier() local 147 scaled_v_low = _mm256_srlv_epi64(scaled_v_low, _mm256_set1_epi64x(31)); in MultiplyByQuantizedMultiplier() 151 __m256i result = _mm256_blend_epi32(scaled_v_low, scaled_v_high, 0xaa); in MultiplyByQuantizedMultiplier()
|
/external/ruy/ruy/ |
D | kernel_avx512.cc | 395 __m512i scaled_v_low = _mm512_mul_epi32( 402 scaled_v_low = _mm512_add_epi64(scaled_v_low, offset_vector_val); 405 scaled_v_low = 406 _mm512_srav_epi64(scaled_v_low, final_right_shift_val); 410 rounding_right_shift(scaled_v_low, right_shift_val); 413 accum = _mm512_castsi256_si512(_mm512_cvtepi64_epi32(scaled_v_low)); 437 __m512i scaled_v_low = _mm512_mul_epi32( 444 scaled_v_low = _mm512_add_epi64(scaled_v_low, offset_vector); 447 scaled_v_low = 448 _mm512_srav_epi64(scaled_v_low, final_right_shift_low); [all …]
|
D | kernel_avx2_fma.cc | 486 __m256i scaled_v_low = _mm256_mul_epi32( 493 scaled_v_low = _mm256_add_epi64(scaled_v_low, offset_vector); 496 scaled_v_low = _mm256_srlv_epi64(scaled_v_low, final_right_shift_low); 502 _mm256_blend_epi32(scaled_v_low, scaled_v_high, 0xaa); 900 __m256i scaled_v_low = _mm256_mul_epi32( 907 scaled_v_low = _mm256_add_epi64(scaled_v_low, offset_vector); 910 scaled_v_low = _mm256_srlv_epi64(scaled_v_low, final_right_shift_low); 915 __m256i results = _mm256_blend_epi32(scaled_v_low, scaled_v_high, 0xaa);
|
D | kernel_avx.cc | 872 __m256i scaled_v_low = intrin_utils::mm256_mul_epi32<path>( 880 scaled_v_low = intrin_utils::mm256_add_epi64<path>(scaled_v_low, 885 scaled_v_low = intrin_utils::mm256_srlv_epi64<path>( 886 scaled_v_low, final_right_shift_low); 892 mm256_blend_epi32(results, scaled_v_low, scaled_v_high, 0xaa); 1351 __m256i scaled_v_low = intrin_utils::mm256_mul_epi32<path>( 1360 scaled_v_low = intrin_utils::mm256_add_epi64<path>(scaled_v_low, 1365 scaled_v_low = intrin_utils::mm256_srlv_epi64<path>( 1366 scaled_v_low, final_right_shift_low); 1372 mm256_blend_epi32(results, scaled_v_low, scaled_v_high, 0xaa);
|