// Auto-generated file. Do not edit! // Template: src/qs8-f32-vcvt/avx.c.in // Generator: tools/xngen // // Copyright 2021 Google LLC // // This source code is licensed under the BSD-style license found in the // LICENSE file in the root directory of this source tree. #include #include #include "xnnpack/common.h" #include "xnnpack/unaligned.h" #include "xnnpack/vcvt.h" void xnn_qu8_f32_vcvt_ukernel__avx_u8( size_t batch, const uint8_t* input, float* output, const struct xnn_qu8_f32_cvt_params params[restrict XNN_MIN_ELEMENTS(1)]) XNN_OOB_READS { assert(batch != 0); assert(batch % sizeof(uint8_t) == 0); assert(input != NULL); assert(output != NULL); const __m128i vzero_point = _mm_set1_epi32(params->scalar.zero_point); const __m256 vscale = _mm256_set1_ps(params->scalar.scale); XNN_FORCE_REALIZATION(vzero_point); XNN_FORCE_REALIZATION(vscale); for (; batch >= 8 * sizeof(uint8_t); batch -= 8 * sizeof(uint8_t)) { __m128i vx0123 = _mm_cvtepu8_epi32(_mm_cvtsi32_si128((int) unaligned_load_s32(input))); __m128i vx4567 = _mm_cvtepu8_epi32(_mm_cvtsi32_si128((int) unaligned_load_s32(input + 4))); input += 8; vx0123 = _mm_sub_epi32(vx0123, vzero_point); vx4567 = _mm_sub_epi32(vx4567, vzero_point); const __m256i vx01234567 = _mm256_insertf128_si256(_mm256_castsi128_si256(vx0123), vx4567, 1); __m256 vy01234567 = _mm256_cvtepi32_ps(vx01234567); vy01234567 = _mm256_mul_ps(vy01234567, vscale); _mm256_storeu_ps(output, vy01234567); output += 8; } for (; batch >= 4 * sizeof(uint8_t); batch -= 4 * sizeof(uint8_t)) { __m128i vx = _mm_cvtepu8_epi32(_mm_cvtsi32_si128((int) unaligned_load_s32(input))); vx = _mm_sub_epi32(vx, vzero_point); input += 4; __m128 vy = _mm_cvtepi32_ps(vx); vy = _mm_mul_ps(vy, _mm256_castps256_ps128(vscale)); _mm_storeu_ps(output, vy); output += 4; } if XNN_UNLIKELY(batch != 0) { assert(batch >= 1 * sizeof(uint8_t)); assert(batch <= 3 * sizeof(uint8_t)); __m128i vx = _mm_cvtepu8_epi32(_mm_cvtsi32_si128((int) unaligned_load_s32(input))); vx = _mm_sub_epi32(vx, vzero_point); __m128 vy = _mm_cvtepi32_ps(vx); vy = _mm_mul_ps(vy, _mm256_castps256_ps128(vscale)); if (batch & (2 * sizeof(uint8_t))) { _mm_storel_pi((__m64*) output, vy); vy = _mm_movehl_ps(vy, vy); output += 2; } if (batch & (1 * sizeof(uint8_t))) { _mm_store_ss(output, vy); } } }