simd

问题 This question is really a curiosity. I was converting a routine into SIMD instructions (and I am quite new to SIMD programming), and had trouble with the following bit of code: // args: uint32_t phase_current; uint32_t phase_increment; uint32_t phase_increment_step; for (int i = 0; i < blockSize; ++i) { USEFUL_FUNC(phase_current); phase_increment += phase_increment_step; phase_current += phase_increment; } The Question: Assuming that USEFUL_FUNC has a SIMD implementation and I am simply

问题 This question is really a curiosity. I was converting a routine into SIMD instructions (and I am quite new to SIMD programming), and had trouble with the following bit of code: // args: uint32_t phase_current; uint32_t phase_increment; uint32_t phase_increment_step; for (int i = 0; i < blockSize; ++i) { USEFUL_FUNC(phase_current); phase_increment += phase_increment_step; phase_current += phase_increment; } The Question: Assuming that USEFUL_FUNC has a SIMD implementation and I am simply

问题 This question is really a curiosity. I was converting a routine into SIMD instructions (and I am quite new to SIMD programming), and had trouble with the following bit of code: // args: uint32_t phase_current; uint32_t phase_increment; uint32_t phase_increment_step; for (int i = 0; i < blockSize; ++i) { USEFUL_FUNC(phase_current); phase_increment += phase_increment_step; phase_current += phase_increment; } The Question: Assuming that USEFUL_FUNC has a SIMD implementation and I am simply

问题 I am trying to do left rotation of a 128 bit number in AVX2. Since there is no direct method of doing this, I have tried using left shift and right shift to accomplish my task. Here is a snippet of my code to do the same. l = 4; r = 4; targetrotate = _mm_set_epi64x (l, r); targetleftrotate = _mm_sllv_epi64 (target, targetrotate); The above c ode snippet rotates target by 4 to the left. When I tested the above code with a sample input, I could see the result is not rotated correctly. Here is

问题 I am trying to do left rotation of a 128 bit number in AVX2. Since there is no direct method of doing this, I have tried using left shift and right shift to accomplish my task. Here is a snippet of my code to do the same. l = 4; r = 4; targetrotate = _mm_set_epi64x (l, r); targetleftrotate = _mm_sllv_epi64 (target, targetrotate); The above c ode snippet rotates target by 4 to the left. When I tested the above code with a sample input, I could see the result is not rotated correctly. Here is

问题 I'm looking for some SSE/AVX advice to optimize a routine that premultiplies RGB channel with its alpha channel: RGB * alpha / 255 (+ we keep the original alpha channel). for (int i = 0, max = width * height * 4; i < max; i+=4) { data[i] = static_cast<uint16_t>(data[i] * data[i+3]) / 255; data[i+1] = static_cast<uint16_t>(data[i+1] * data[i+3]) / 255; data[i+2] = static_cast<uint16_t>(data[i+2] * data[i+3]) / 255; } You will find below my current implementation but I think it could be much

问题 I have two __m256i vectors, filled with 32 8-bit integers. Something like this: __int8 *a0 = new __int8[32] {2}; __int8 *a1 = new __int8[32] {3}; __m256i v0 = _mm256_loadu_si256((__m256i*)a0); __m256i v1 = _mm256_loadu_si256((__m256i*)a1); How can i multiply these vectors, using something like _mm256_mul_epi8(v0, v1) (which does not exist) or any another way? I want 2 vectors of results, because the output element width is twice the input element width. Or something that works similarly to

问题 I have two __m256i vectors, filled with 32 8-bit integers. Something like this: __int8 *a0 = new __int8[32] {2}; __int8 *a1 = new __int8[32] {3}; __m256i v0 = _mm256_loadu_si256((__m256i*)a0); __m256i v1 = _mm256_loadu_si256((__m256i*)a1); How can i multiply these vectors, using something like _mm256_mul_epi8(v0, v1) (which does not exist) or any another way? I want 2 vectors of results, because the output element width is twice the input element width. Or something that works similarly to

问题 I'm trying to pack 16 bits data to 8 bits by using _mm256_shuffle_epi8 but the result i have is not what i'm expecting. auto srcData = _mm256_setr_epi8(1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32); __m256i vperm = _mm256_setr_epi8( 0, 2, 4, 6, 8, 10, 12, 14, 16, 18, 20, 22, 24, 26, 28, 30, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1); auto result = _mm256_shuffle_epi8(srcData, vperm); I'm expecting

问题 I'm trying to pack 16 bits data to 8 bits by using _mm256_shuffle_epi8 but the result i have is not what i'm expecting. auto srcData = _mm256_setr_epi8(1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32); __m256i vperm = _mm256_setr_epi8( 0, 2, 4, 6, 8, 10, 12, 14, 16, 18, 20, 22, 24, 26, 28, 30, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1); auto result = _mm256_shuffle_epi8(srcData, vperm); I'm expecting