/* SPDX-License-Identifier: MIT * * Permission is hereby granted, free of charge, to any person * obtaining a copy of this software and associated documentation * files (the "Software"), to deal in the Software without * restriction, including without limitation the rights to use, copy, * modify, merge, publish, distribute, sublicense, and/or sell copies * of the Software, and to permit persons to whom the Software is * furnished to do so, subject to the following conditions: * * The above copyright notice and this permission notice shall be * included in all copies or substantial portions of the Software. * * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE * SOFTWARE. * * Copyright: * 2020 Evan Nemerson * 2020 Sean Maher (Copyright owned by Google, LLC) */ #if !defined(SIMDE_ARM_NEON_DOT_H) #define SIMDE_ARM_NEON_DOT_H #include "types.h" #include "add.h" #include "combine.h" #include "dup_n.h" #include "get_low.h" #include "get_high.h" #include "paddl.h" #include "movn.h" #include "mull.h" HEDLEY_DIAGNOSTIC_PUSH SIMDE_DISABLE_UNWANTED_DIAGNOSTICS SIMDE_BEGIN_DECLS_ SIMDE_FUNCTION_ATTRIBUTES simde_int32x2_t simde_vdot_s32(simde_int32x2_t r, simde_int8x8_t a, simde_int8x8_t b) { #if defined(SIMDE_ARM_NEON_A32V8_NATIVE) && defined(__ARM_FEATURE_DOTPROD) return vdot_s32(r, a, b); #elif defined(SIMDE_ARM_NEON_A32V7_NATIVE) return simde_vadd_s32(r, simde_vmovn_s64(simde_vpaddlq_s32(simde_vpaddlq_s16(simde_vmull_s8(a, b))))); #else simde_int32x2_private r_; simde_int8x8_private a_ = simde_int8x8_to_private(a), b_ = simde_int8x8_to_private(b); for (int i = 0 ; i < 2 ; i++) { int32_t acc = 0; SIMDE_VECTORIZE_REDUCTION(+:acc) for (int j = 0 ; j < 4 ; j++) { const int idx = j + (i << 2); acc += HEDLEY_STATIC_CAST(int32_t, a_.values[idx]) * HEDLEY_STATIC_CAST(int32_t, b_.values[idx]); } r_.values[i] = acc; } return simde_vadd_s32(r, simde_int32x2_from_private(r_)); #endif } #if defined(SIMDE_ARM_NEON_A32V8_ENABLE_NATIVE_ALIASES) || (defined(SIMDE_ENABLE_NATIVE_ALIASES) && !defined(__ARM_FEATURE_DOTPROD)) #undef vdot_s32 #define vdot_s32(r, a, b) simde_vdot_s32((r), (a), (b)) #endif SIMDE_FUNCTION_ATTRIBUTES simde_uint32x2_t simde_vdot_u32(simde_uint32x2_t r, simde_uint8x8_t a, simde_uint8x8_t b) { #if defined(SIMDE_ARM_NEON_A32V8_NATIVE) && defined(__ARM_FEATURE_DOTPROD) return vdot_u32(r, a, b); #elif defined(SIMDE_ARM_NEON_A32V7_NATIVE) return simde_vadd_u32(r, simde_vmovn_u64(simde_vpaddlq_u32(simde_vpaddlq_u16(simde_vmull_u8(a, b))))); #else simde_uint32x2_private r_; simde_uint8x8_private a_ = simde_uint8x8_to_private(a), b_ = simde_uint8x8_to_private(b); for (int i = 0 ; i < 2 ; i++) { uint32_t acc = 0; SIMDE_VECTORIZE_REDUCTION(+:acc) for (int j = 0 ; j < 4 ; j++) { const int idx = j + (i << 2); acc += HEDLEY_STATIC_CAST(uint32_t, a_.values[idx]) * HEDLEY_STATIC_CAST(uint32_t, b_.values[idx]); } r_.values[i] = acc; } return simde_vadd_u32(r, simde_uint32x2_from_private(r_)); #endif } #if defined(SIMDE_ARM_NEON_A32V8_ENABLE_NATIVE_ALIASES) || (defined(SIMDE_ENABLE_NATIVE_ALIASES) && !defined(__ARM_FEATURE_DOTPROD)) #undef vdot_u32 #define vdot_u32(r, a, b) simde_vdot_u32((r), (a), (b)) #endif SIMDE_FUNCTION_ATTRIBUTES simde_int32x4_t simde_vdotq_s32(simde_int32x4_t r, simde_int8x16_t a, simde_int8x16_t b) { #if defined(SIMDE_ARM_NEON_A32V8_NATIVE) && defined(__ARM_FEATURE_DOTPROD) return vdotq_s32(r, a, b); #elif defined(SIMDE_ARM_NEON_A32V7_NATIVE) return simde_vaddq_s32(r, simde_vcombine_s32(simde_vmovn_s64(simde_vpaddlq_s32(simde_vpaddlq_s16(simde_vmull_s8(simde_vget_low_s8(a), simde_vget_low_s8(b))))), simde_vmovn_s64(simde_vpaddlq_s32(simde_vpaddlq_s16(simde_vmull_s8(simde_vget_high_s8(a), simde_vget_high_s8(b))))))); #else simde_int32x4_private r_; simde_int8x16_private a_ = simde_int8x16_to_private(a), b_ = simde_int8x16_to_private(b); for (int i = 0 ; i < 4 ; i++) { int32_t acc = 0; SIMDE_VECTORIZE_REDUCTION(+:acc) for (int j = 0 ; j < 4 ; j++) { const int idx = j + (i << 2); acc += HEDLEY_STATIC_CAST(int32_t, a_.values[idx]) * HEDLEY_STATIC_CAST(int32_t, b_.values[idx]); } r_.values[i] = acc; } return simde_vaddq_s32(r, simde_int32x4_from_private(r_)); #endif } #if defined(SIMDE_ARM_NEON_A32V8_ENABLE_NATIVE_ALIASES) || (defined(SIMDE_ENABLE_NATIVE_ALIASES) && !defined(__ARM_FEATURE_DOTPROD)) #undef vdotq_s32 #define vdotq_s32(r, a, b) simde_vdotq_s32((r), (a), (b)) #endif SIMDE_FUNCTION_ATTRIBUTES simde_uint32x4_t simde_vdotq_u32(simde_uint32x4_t r, simde_uint8x16_t a, simde_uint8x16_t b) { #if defined(SIMDE_ARM_NEON_A32V8_NATIVE) && defined(__ARM_FEATURE_DOTPROD) return vdotq_u32(r, a, b); #elif defined(SIMDE_ARM_NEON_A32V7_NATIVE) return simde_vaddq_u32(r, simde_vcombine_u32(simde_vmovn_u64(simde_vpaddlq_u32(simde_vpaddlq_u16(simde_vmull_u8(simde_vget_low_u8(a), simde_vget_low_u8(b))))), simde_vmovn_u64(simde_vpaddlq_u32(simde_vpaddlq_u16(simde_vmull_u8(simde_vget_high_u8(a), simde_vget_high_u8(b))))))); #else simde_uint32x4_private r_; simde_uint8x16_private a_ = simde_uint8x16_to_private(a), b_ = simde_uint8x16_to_private(b); for (int i = 0 ; i < 4 ; i++) { uint32_t acc = 0; SIMDE_VECTORIZE_REDUCTION(+:acc) for (int j = 0 ; j < 4 ; j++) { const int idx = j + (i << 2); acc += HEDLEY_STATIC_CAST(uint32_t, a_.values[idx]) * HEDLEY_STATIC_CAST(uint32_t, b_.values[idx]); } r_.values[i] = acc; } return simde_vaddq_u32(r, simde_uint32x4_from_private(r_)); #endif } #if defined(SIMDE_ARM_NEON_A32V8_ENABLE_NATIVE_ALIASES) || (defined(SIMDE_ENABLE_NATIVE_ALIASES) && !defined(__ARM_FEATURE_DOTPROD)) #undef vdotq_u32 #define vdotq_u32(r, a, b) simde_vdotq_u32((r), (a), (b)) #endif SIMDE_END_DECLS_ HEDLEY_DIAGNOSTIC_POP #endif /* !defined(SIMDE_ARM_NEON_DOT_H) */