21.05
|
Namespaces | |
kernels | |
Data Structures | |
class | CpuActivation |
Basic function to run kernels::CpuActivationKernel. More... | |
class | CpuAdd |
Basic function to run kernels::CpuAddKernel. More... | |
struct | CpuCapabilities |
Structure that encodes the CPU capabilities to be used. More... | |
class | CpuComplexMul |
Basic function to run kernels::CpuComplexMulKernel. More... | |
class | CpuConcatenate |
Basic function to execute concatenate tensors along a given axis. More... | |
class | CpuContext |
CPU context implementation class. More... | |
class | CpuConvertFullyConnectedWeights |
Basic function to run kernels::CpuConvertFullyConnectedWeightsKernel. More... | |
class | CpuCopy |
Basic function to run kernels::CpuCopyKernel. More... | |
class | CpuDepthwiseConvolution |
Function to execute a depthwise convolution. More... | |
class | CpuDepthwiseConvolutionAssemblyDispatch |
Depthwise convolution assembly kernel glue. More... | |
class | CpuDequantization |
Basic function to run kernels::CpuDequantizationKernel that dequantizes an input tensor. More... | |
class | CpuDirectConvolution |
Function to run the direct convolution. More... | |
class | CpuElementwiseArithmetic |
Class to run cpu::kernels::CpuArithmeticKernel except for division and power. More... | |
class | CpuElementwiseBase |
class | CpuElementwiseComparison |
Basic function to run cpu::kernels::CpuComparisonKernel. More... | |
class | CpuElementwiseComparisonStatic |
Basic function to run cpu::kernels::CpuComparisonKernel. More... | |
class | CpuElementwiseDivision |
Basic function to run cpu::kernels::CpuArithmeticKernel for division. More... | |
class | CpuElementwisePower |
Basic function to run cpu::kernels::CpuArithmeticKernel for power. More... | |
class | CpuElementwiseUnary |
class | CpuFill |
Basic function to run kernels::CpuFillKernel. More... | |
class | CpuFloor |
Basic function to run kernels::CpuFloorKernel. More... | |
class | CpuLogits1DSoftmaxKernel |
class | CpuMul |
Basic function to run kernels::CpuMulKernel. More... | |
class | CpuPermute |
Basic function to run kernels::CpuPermuteKernel. More... | |
class | CpuPooling |
Basic function to simulate a pooling layer with the specified pooling operation. More... | |
class | CpuPoolingAssemblyDispatch |
Basic function to run pooling assembly kernels. More... | |
class | CpuQuantization |
Basic function to simulate a quantization layer. More... | |
class | CpuQueue |
CPU queue implementation class. More... | |
class | CpuReshape |
Basic function to run kernels::CpuReshapeKernel. More... | |
class | CpuScale |
Basic function to compute Scale. More... | |
class | CpuSoftmaxGeneric |
Basic function to compute a SoftmaxLayer and a Log SoftmaxLayer. More... | |
class | CpuSub |
Basic function to run kernels::CpuSubKernel. More... | |
class | CpuTensor |
CPU tensor implementation class. More... | |
class | CpuTranspose |
Basic function to run kernels::CpuTransposeKernel. More... | |
Functions | |
void | qasymm8_neon_activation (const ITensor *src, ITensor *dst, const ActivationLayerInfo &act_info, const Window &window) |
void | qasymm8_sve_activation (const ITensor *src, ITensor *dst, const ActivationLayerInfo &act_info, const Window &window) |
void | qasymm8_signed_neon_activation (const ITensor *src, ITensor *dst, const ActivationLayerInfo &act_info, const Window &window) |
void | qasymm8_signed_sve_activation (const ITensor *src, ITensor *dst, const ActivationLayerInfo &act_info, const Window &window) |
void | qsymm16_neon_activation (const ITensor *src, ITensor *dst, const ActivationLayerInfo &act_info, const Window &window) |
void | qsymm16_sve_activation (const ITensor *src, ITensor *dst, const ActivationLayerInfo &act_info, const Window &window) |
void | fp16_neon_activation (const ITensor *src, ITensor *dst, const ActivationLayerInfo &act_info, const Window &window) |
void | fp16_sve_activation (const ITensor *src, ITensor *dst, const ActivationLayerInfo &act_info, const Window &window) |
void | fp32_neon_activation (const ITensor *src, ITensor *dst, const ActivationLayerInfo &act_info, const Window &window) |
void | fp32_sve_activation (const ITensor *src, ITensor *dst, const ActivationLayerInfo &act_info, const Window &window) |
void | add_u8_u8_s16_neon (const ITensor *src0, const ITensor *src1, ITensor *dst, const ConvertPolicy &policy, const Window &window) |
void | add_s16_u8_s16_neon (const ITensor *src0, const ITensor *src1, ITensor *dst, const ConvertPolicy &policy, const Window &window) |
void | add_u8_s16_s16_neon (const ITensor *src0, const ITensor *src1, ITensor *dst, const ConvertPolicy &policy, const Window &window) |
void | add_qasymm8_neon (const ITensor *src0, const ITensor *src1, ITensor *dst, const ConvertPolicy &policy, const Window &window) |
void | add_qasymm8_signed_neon (const ITensor *src0, const ITensor *src1, ITensor *dst, const ConvertPolicy &policy, const Window &window) |
void | add_qsymm16_neon (const ITensor *src0, const ITensor *src1, ITensor *dst, const ConvertPolicy &policy, const Window &window) |
template<typename ScalarType > | |
void | add_same_neon (const ITensor *src0, const ITensor *src1, ITensor *dst, const ConvertPolicy &policy, const Window &window) |
template<typename InputScalarType , typename OutputScalarType , typename InputVectorType > | |
void | elementwise_op (const ITensor *in1, const ITensor *in2, ITensor *out, const Window &window, OutputScalarType(*scalar_func)(const InputScalarType &, const InputScalarType &), int(*broadcast_func)(int, int, int, const InputScalarType *, const InputScalarType &, OutputScalarType *, const bool), int(*neon_func)(int, int, int, const InputScalarType *, const InputScalarType *, OutputScalarType *)) |
template<ArithmeticOperation op, typename ScalarType > | |
ScalarType | elementwise_arithm_op_scalar (const ScalarType &a, const ScalarType &b) |
template<ArithmeticOperation op, typename VectorType > | |
VectorType::type | elementwise_arithm_op (const typename VectorType::type &a, const typename VectorType::type &b) |
template<> | |
int32x4_t | elementwise_arithm_op< ArithmeticOperation::DIV, typename wrapper::traits::neon_vector< int32_t, 4 > > (const int32x4_t &a, const int32x4_t &b) |
template<> | |
float32x4_t | elementwise_arithm_op< ArithmeticOperation::DIV, typename wrapper::traits::neon_vector< float, 4 > > (const float32x4_t &a, const float32x4_t &b) |
template<> | |
float32x4_t | elementwise_arithm_op< ArithmeticOperation::POWER, typename wrapper::traits::neon_vector< float, 4 > > (const float32x4_t &a, const float32x4_t &b) |
template<ArithmeticOperation op, typename ScalarType , typename VectorType > | |
VectorType::type | elementwise_arithm_op_broadcast (const typename VectorType::type &a, const ScalarType &broadcast_value, const bool reorder) |
template<ArithmeticOperation op, typename ScalarType , typename VectorType > | |
int | elementwise_arithm_op_loop (int window_start_x, int window_end_x, int window_step_x, const ScalarType *input1_ptr, const ScalarType *input2_ptr, ScalarType *output_ptr) |
template<ArithmeticOperation op, typename ScalarType , typename VectorType > | |
int | elementwise_arithm_op_broadcast_loop (int window_start_x, int window_end_x, int window_step_x, const ScalarType *non_broadcast_input_ptr, const ScalarType &broadcast_value, ScalarType *output_ptr, const bool reorder) |
template<ArithmeticOperation op, typename VectorType > | |
void | elementwise_arithm_op (const ITensor *in1, const ITensor *in2, ITensor *out, const Window &window) |
template<ComparisonOperation op, typename InputScalarType > | |
uint8_t | elementwise_comp_op_scalar (const InputScalarType &a, const InputScalarType &b) |
template<ComparisonOperation op, typename InputVectorType , typename OutputVectorType > | |
OutputVectorType | elementwise_comp_op (const InputVectorType &a, const InputVectorType &b) |
template<ComparisonOperation op, typename InputScalarType , typename InputVectorType , typename OutputVectorType > | |
OutputVectorType | elementwise_comp_op_broadcast (const InputVectorType &a, const InputScalarType &broadcast_value, const bool reorder) |
template<ComparisonOperation op, typename InputScalarType , typename InputVectorType > | |
int | elementwise_comp_op_broadcast_8_loop (int window_start_x, int window_end_x, int window_step_x, const InputScalarType *non_broadcast_input_ptr, const InputScalarType &broadcast_value, uint8_t *output_ptr, const bool reorder) |
template<ComparisonOperation op, typename InputScalarType , typename InputVectorType > | |
int | elementwise_comp_op_broadcast_16_loop (int window_start_x, int window_end_x, int window_step_x, const InputScalarType *non_broadcast_input_ptr, const InputScalarType &broadcast_value, uint8_t *output_ptr, const bool reorder) |
template<ComparisonOperation op, typename InputScalarType , typename InputVectorType > | |
int | elementwise_comp_op_broadcast_32_loop (int window_start_x, int window_end_x, int window_step_x, const InputScalarType *non_broadcast_input_ptr, const InputScalarType &broadcast_value, uint8_t *output_ptr, const bool reorder) |
template<ComparisonOperation op, typename InputScalarType , typename InputVectorType > | |
int | elementwise_comp_op_8_loop (int window_start_x, int window_end_x, int window_step_x, const InputScalarType *input1_ptr, const InputScalarType *input2_ptr, uint8_t *output_ptr) |
template<ComparisonOperation op, typename InputScalarType , typename InputVectorType > | |
int | elementwise_comp_op_16_loop (int window_start_x, int window_end_x, int window_step_x, const InputScalarType *input1_ptr, const InputScalarType *input2_ptr, uint8_t *output_ptr) |
template<ComparisonOperation op, typename InputScalarType , typename InputVectorType > | |
int | elementwise_comp_op_32_loop (int window_start_x, int window_end_x, int window_step_x, const InputScalarType *input1_ptr, const InputScalarType *input2_ptr, uint8_t *output_ptr) |
template<ComparisonOperation op, typename InputScalarType , typename InputVectorType > | |
void | elementwise_comp_op_8 (const ITensor *in1, const ITensor *in2, ITensor *out, const Window &window) |
template<ComparisonOperation op, typename InputScalarType , typename InputVectorType > | |
void | elementwise_comp_op_16 (const ITensor *in1, const ITensor *in2, ITensor *out, const Window &window) |
template<ComparisonOperation op, typename InputScalarType , typename InputVectorType > | |
void | elementwise_comp_op_32 (const ITensor *in1, const ITensor *in2, ITensor *out, const Window &window) |
float32x4x4_t | load_quantized (const uint8_t *input1_ptr, const int32x4_t &offset, const float32x4_t &scale) |
float32x4x4_t | load_quantized_signed (const int8_t *input1_ptr, const int32x4_t &offset, const float32x4_t &scale) |
void | store_quantized (uint8_t *output_ptr, const uint32x4x4_t &out) |
void | store_quantized (uint8_t *output_ptr, const int32x4x4_t &out) |
void | store_quantized (uint8_t *output_ptr, const float32x4x4_t &rf, const float32x4_t &offset, const float32x4_t &invscale) |
void | store_quantized_signed (int8_t *output_ptr, const int32x4x4_t &out) |
void | store_quantized_signed (int8_t *output_ptr, const float32x4x4_t &rf, const float32x4_t &offset, const float32x4_t &invscale) |
template<ArithmeticOperation op> | |
uint8_t | elementwise_arithm_op_quantized_scalar (const float &a, const float &b, UniformQuantizationInfo qinfo) |
template<ArithmeticOperation op> | |
int8_t | elementwise_arithm_op_quantized_signed_scalar (const float &a, const float &b, UniformQuantizationInfo qinfo) |
template<ArithmeticOperation op> | |
float32x4x4_t | elementwise_arithm_op (const float32x4x4_t &a, const float32x4x4_t &b) |
template<ComparisonOperation op> | |
uint8_t | elementwise_comp_op_quantized_scalar (const float &a, const float &b, UniformQuantizationInfo qinfo) |
template<ComparisonOperation op> | |
uint32x4x4_t | elementwise_comp_op (const float32x4x4_t &a, const float32x4x4_t &b) |
template<ArithmeticOperation op> | |
int | elementwise_arithm_op_quantized_loop (int window_start_x, int window_end_x, int window_step_x, const uint8_t *input1_ptr, const uint8_t *input2_ptr, uint8_t *output_ptr, int32x4_t voffset1, int32x4_t voffset2, float32x4_t vscale1, float32x4_t vscale2, float32x4_t voffseto, float32x4_t invvscaleo) |
template<ArithmeticOperation op> | |
int | elementwise_arithm_op_quantized_singed_loop (int window_start_x, int window_end_x, int window_step_x, const int8_t *input1_ptr, const int8_t *input2_ptr, int8_t *output_ptr, int32x4_t voffset1, int32x4_t voffset2, float32x4_t vscale1, float32x4_t vscale2, float32x4_t voffseto, float32x4_t invvscaleo) |
template<ArithmeticOperation op> | |
int | elementwise_arithm_op_quantized_broadcast_loop (int window_start_x, int window_end_x, int window_step_x, const uint8_t *non_broadcast_input_ptr, float32x4x4_t broadcast_vector, uint8_t *output_ptr, int32x4_t voffset_non_broadcast, float32x4_t vscale_non_broadcast, float32x4_t voffseto, float32x4_t invvscaleo, bool reorder) |
template<ArithmeticOperation op> | |
int | elementwise_arithm_op_quantized_signed_broadcast_loop (int window_start_x, int window_end_x, int window_step_x, const int8_t *non_broadcast_input_ptr, float32x4x4_t broadcast_vector, int8_t *output_ptr, int32x4_t voffset_non_broadcast, float32x4_t vscale_non_broadcast, float32x4_t voffseto, float32x4_t invvscaleo, bool reorder) |
template<ComparisonOperation op> | |
int | elementwise_comp_op_quantized_loop (int window_start_x, int window_end_x, int window_step_x, const uint8_t *input1_ptr, const uint8_t *input2_ptr, uint8_t *output_ptr, int32x4_t voffset1, int32x4_t voffset2, float32x4_t vscale1, float32x4_t vscale2, float32x4_t voffseto, float32x4_t invvscaleo) |
template<ComparisonOperation op> | |
int | elementwise_comp_op_quantized_signed_loop (int window_start_x, int window_end_x, int window_step_x, const int8_t *input1_ptr, const int8_t *input2_ptr, uint8_t *output_ptr, int32x4_t voffset1, int32x4_t voffset2, float32x4_t vscale1, float32x4_t vscale2, float32x4_t voffseto, float32x4_t invvscaleo) |
template<ComparisonOperation op> | |
int | elementwise_comp_op_quantized_broadcast_loop (int window_start_x, int window_end_x, int window_step_x, const uint8_t *non_broadcast_input_ptr, float32x4x4_t broadcast_vector, uint8_t *output_ptr, int32x4_t voffset_non_broadcast, float32x4_t vscale_non_broadcast, float32x4_t voffseto, float32x4_t invvscaleo, bool reorder) |
template<ComparisonOperation op> | |
int | elementwise_comp_op_quantized_signed_broadcast_loop (int window_start_x, int window_end_x, int window_step_x, const int8_t *non_broadcast_input_ptr, float32x4x4_t broadcast_vector, uint8_t *output_ptr, int32x4_t voffset_non_broadcast, float32x4_t vscale_non_broadcast, float32x4_t voffseto, float32x4_t invvscaleo, bool reorder) |
void | elementwise_op_quantized (const ITensor *in1, const ITensor *in2, ITensor *out, const Window &window, uint8_t(*scalar_func)(const float &, const float &, UniformQuantizationInfo), int(*broadcast_func)(int, int, int, const uint8_t *, float32x4x4_t, uint8_t *, int32x4_t, float32x4_t, float32x4_t, float32x4_t, const bool), int(*neon_func)(int, int, int, const uint8_t *, const uint8_t *, uint8_t *, int32x4_t, int32x4_t, float32x4_t, float32x4_t, float32x4_t, float32x4_t)) |
void | elementwise_comp_quantized_signed (const ITensor *in1, const ITensor *in2, ITensor *out, const Window &window, uint8_t(*scalar_func)(const float &, const float &, UniformQuantizationInfo), int(*broadcast_func)(int, int, int, const int8_t *, float32x4x4_t, uint8_t *, int32x4_t, float32x4_t, float32x4_t, float32x4_t, const bool), int(*neon_func)(int, int, int, const int8_t *, const int8_t *, uint8_t *, int32x4_t, int32x4_t, float32x4_t, float32x4_t, float32x4_t, float32x4_t)) |
void | elementwise_op_quantized_signed (const ITensor *in1, const ITensor *in2, ITensor *out, const Window &window, int8_t(*scalar_func)(const float &, const float &, UniformQuantizationInfo), int(*broadcast_func)(int, int, int, const int8_t *, float32x4x4_t, int8_t *, int32x4_t, float32x4_t, float32x4_t, float32x4_t, const bool), int(*neon_func)(int, int, int, const int8_t *, const int8_t *, int8_t *, int32x4_t, int32x4_t, float32x4_t, float32x4_t, float32x4_t, float32x4_t)) |
template<ArithmeticOperation op> | |
void | elementwise_arithm_op_quantized (const ITensor *in1, const ITensor *in2, ITensor *out, const Window &window) |
template<ArithmeticOperation op> | |
void | elementwise_arithm_op_quantized_signed (const ITensor *in1, const ITensor *in2, ITensor *out, const Window &window) |
template<ComparisonOperation op> | |
void | elementwise_comp_op_quantized (const ITensor *in1, const ITensor *in2, ITensor *out, const Window &window) |
template<ComparisonOperation op> | |
void | elementwise_comp_op_quantized_signed (const ITensor *in1, const ITensor *in2, ITensor *out, const Window &window) |
template<typename ScalarType > | |
ScalarType | elementwise_op_scalar_imp (ElementWiseUnary op, const ScalarType &a) |
template<typename ScalarType , typename VectorType > | |
VectorType | elementwise_op_imp (ElementWiseUnary op, const VectorType &a) |
template<typename ScalarType > | |
void | elementwise_op (const ITensor *in, ITensor *out, const Window &window, ElementWiseUnary op) |
void | fp16_neon_floor (const void *src, void *dst, int len) |
void | fp32_neon_floor (const void *src, void *dst, int len) |
void | poolingMxN_fp32_neon_nhwc (const ITensor *src, ITensor *dst0, ITensor *dst1, PoolingLayerInfo &pool_info, const Window &window_src, const Window &window) |
void | poolingMxN_qasymm8_neon_nhwc (const ITensor *src0, ITensor *dst0, ITensor *dst1, PoolingLayerInfo &, const Window &window_src, const Window &window) |
void | poolingMxN_qasymm8_signed_neon_nhwc (const ITensor *src0, ITensor *dst0, ITensor *dst1, PoolingLayerInfo &, const Window &window_src, const Window &window) |
void | poolingMxN_fp16_neon_nhwc (const ITensor *src0, ITensor *dst0, ITensor *dst1, PoolingLayerInfo &, const Window &window_src, const Window &window) |
template<typename T > | |
uint32_t | offset_no_padding (uint32_t padded_offset, const Coordinates &id, const ITensorInfo &info, int pool_stride_x, int pool_stride_y, DataLayout data_layout) |
template<typename T > | |
std::enable_if< std::is_same< T, int8_t >::value, int8_t >::type | quantize (float val, const UniformQuantizationInfo &info) |
template<typename T > | |
std::enable_if< std::is_same< T, uint8_t >::value, uint8_t >::type | quantize (float val, const UniformQuantizationInfo &info) |
template<typename T > | |
T | vcvtq_q32_f32 (float32x4_t values) |
template<> | |
uint32x4_t | vcvtq_q32_f32 (float32x4_t values) |
template<> | |
int32x4_t | vcvtq_q32_f32 (float32x4_t values) |
template<typename T > | |
float32x4_t | vcvtq_f32_q32 (T values) |
template<> | |
float32x4_t | vcvtq_f32_q32 (uint32x4_t values) |
template<> | |
float32x4_t | vcvtq_f32_q32 (int32x4_t values) |
template<typename Tout > | |
Tout | vrequantize_pooling_with_scale (const float32x4x4_t &acc, const float quant_rescale, const float scale_pooling, const int32_t new_offset) |
template<> | |
uint8x16_t | vrequantize_pooling_with_scale (const float32x4x4_t &acc, const float quant_rescale, const float scale_pooling, const int32_t new_offset) |
template<> | |
int8x16_t | vrequantize_pooling_with_scale (const float32x4x4_t &acc, const float quant_rescale, const float scale_pooling, const int32_t new_offset) |
template<typename Tin , typename Tout > | |
Tout | vrequantize_pooling (Tin vec1, Tin vec2, const UniformQuantizationInfo &requant_qinfo) |
template<> | |
uint8x16_t | vrequantize_pooling (uint8x8_t vec1, uint8x8_t vec2, const UniformQuantizationInfo &requant_qinfo) |
template<> | |
int8x16_t | vrequantize_pooling (int8x8_t vec1, int8x8_t vec2, const UniformQuantizationInfo &requant_qinfo) |
template<typename T > | |
T | vrequantize_pooling (T &vec, const UniformQuantizationInfo &requant_qinfo) |
template<> | |
uint8x8_t | vrequantize_pooling (uint8x8_t &vec, const UniformQuantizationInfo &requant_qinfo) |
template<> | |
int8x8_t | vrequantize_pooling (int8x8_t &vec, const UniformQuantizationInfo &requant_qinfo) |
float | calculate_avg_scale (bool exclude_padding, DataLayout data_layout, const Coordinates &id, const int pool_size_x, const int pool_size_y, const int upper_bound_w, const int upper_bound_h, const int pad_x, const int pad_y, const int stride_x, const int stride_y) |
template<typename T > | |
void | poolingMxN_q8_neon_nhwc (const ITensor *src, ITensor *dst0, ITensor *dst1, PoolingLayerInfo &pool_info, const Window &window_src, const Window &window) |
void | u8_neon_scale (const ITensor *src, ITensor *dst, const ITensor *offsets, const ITensor *dx, const ITensor *dy, InterpolationPolicy policy, BorderMode border_mode, PixelValue constant_border_value, float sampling_offset, bool align_corners, const Window &window) |
void | s16_neon_scale (const ITensor *src, ITensor *dst, const ITensor *offsets, const ITensor *dx, const ITensor *dy, InterpolationPolicy policy, BorderMode border_mode, PixelValue constant_border_value, float sampling_offset, bool align_corners, const Window &window) |
void | qasymm8_neon_scale (const ITensor *src, ITensor *dst, const ITensor *offsets, const ITensor *dx, const ITensor *dy, InterpolationPolicy policy, BorderMode border_mode, PixelValue constant_border_value, float sampling_offset, bool align_corners, const Window &window) |
void | qasymm8_signed_neon_scale (const ITensor *src, ITensor *dst, const ITensor *offsets, const ITensor *dx, const ITensor *dy, InterpolationPolicy policy, BorderMode border_mode, PixelValue constant_border_value, float sampling_offset, bool align_corners, const Window &window) |
template<typename T > | |
void | nearest_neon_scale (const ITensor *src, ITensor *dst, const ITensor *offsets, float sampling_offset, bool align_corners, const Window &window) |
template<typename T > | |
void | bilinear_neon_scale (const ITensor *src, ITensor *dst, const ITensor *offsets, const ITensor *dx, const ITensor *dy, BorderMode border_mode, PixelValue constant_border_value, float sampling_offset, bool align_corners, const Window &window) |
template<typename T > | |
void | common_neon_scale (const ITensor *src, ITensor *dst, const ITensor *offsets, const ITensor *dx, const ITensor *dy, InterpolationPolicy policy, BorderMode border_mode, PixelValue constant_border_value, float sampling_offset, bool align_corners, const Window &window) |
void | fp16_sve_scale (const ITensor *src, ITensor *dst, const ITensor *offsets, const ITensor *dx, const ITensor *dy, InterpolationPolicy policy, BorderMode border_mode, PixelValue constant_border_value, float sampling_offset, bool align_corners, const Window &window) |
void | fp32_sve_scale (const ITensor *src, ITensor *dst, const ITensor *offsets, const ITensor *dx, const ITensor *dy, InterpolationPolicy policy, BorderMode border_mode, PixelValue constant_border_value, float sampling_offset, bool align_corners, const Window &window) |
void | s16_sve_scale (const ITensor *src, ITensor *dst, const ITensor *offsets, const ITensor *dx, const ITensor *dy, InterpolationPolicy policy, BorderMode border_mode, PixelValue constant_border_value, float sampling_offset, bool align_corners, const Window &window) |
void | u8_sve_scale (const ITensor *src, ITensor *dst, const ITensor *offsets, const ITensor *dx, const ITensor *dy, InterpolationPolicy policy, BorderMode border_mode, PixelValue constant_border_value, float sampling_offset, bool align_corners, const Window &window) |
void | qasymm8_sve_scale (const ITensor *src, ITensor *dst, const ITensor *offsets, const ITensor *dx, const ITensor *dy, InterpolationPolicy policy, BorderMode border_mode, PixelValue constant_border_value, float sampling_offset, bool align_corners, const Window &window) |
void | qasymm8_signed_sve_scale (const ITensor *src, ITensor *dst, const ITensor *offsets, const ITensor *dx, const ITensor *dy, InterpolationPolicy policy, BorderMode border_mode, PixelValue constant_border_value, float sampling_offset, bool align_corners, const Window &window) |
template<typename T > | |
void | neon_logits_1d_max (const ITensor *in, ITensor *out, const Window &window) |
template<typename T > | |
void | neon_softmax_logits_1d_quantized (const ITensor *in, const ITensor *max, void *const tmp, ITensor *out, float beta, bool is_log, const Window &window) |
template<typename T > | |
void | neon_softmax_logits_1d_float (const ITensor *in, const ITensor *max, void *const tmp, ITensor *out, const float beta, bool is_log, const Window &window) |
void | sub_s16_u8_s16_neon (const ITensor *src0, const ITensor *src1, ITensor *dst, const ConvertPolicy &policy, const Window &window) |
void | sub_u8_s16_s16_neon (const ITensor *src0, const ITensor *src1, ITensor *dst, const ConvertPolicy &policy, const Window &window) |
void | sub_u8_u8_s16_neon (const ITensor *src0, const ITensor *src1, ITensor *dst, const ConvertPolicy &policy, const Window &window) |
void | sub_qasymm8_neon (const ITensor *src0, const ITensor *src1, ITensor *dst, const ConvertPolicy &policy, const Window &window) |
void | sub_qasymm8_signed_neon (const ITensor *src0, const ITensor *src1, ITensor *dst, const ConvertPolicy &policy, const Window &window) |
void | sub_qsymm16_neon (const ITensor *src0, const ITensor *src1, ITensor *dst, const ConvertPolicy &policy, const Window &window) |
template<typename T > | |
void | sub_same_neon (const ITensor *src0, const ITensor *src1, ITensor *dst, const ConvertPolicy &policy, const Window &window) |
void | fp16_neon_batch_normalization (ITensor *src, ITensor *dst, const ITensor *mean, const ITensor *var, const ITensor *beta, const ITensor *gamma, float epsilon, ActivationLayerInfo &act_info, const Window &window) |
void | fp16_sve_batch_normalization (ITensor *src, ITensor *dst, const ITensor *mean, const ITensor *var, const ITensor *beta, const ITensor *gamma, float epsilon, ActivationLayerInfo &act_info, const Window &window) |
void | fp32_neon_batch_normalization (ITensor *src, ITensor *dst, const ITensor *mean, const ITensor *var, const ITensor *beta, const ITensor *gamma, float epsilon, ActivationLayerInfo &act_info, const Window &window) |
void | fp32_sve_batch_normalization (ITensor *src, ITensor *dst, const ITensor *mean, const ITensor *var, const ITensor *beta, const ITensor *gamma, float epsilon, ActivationLayerInfo &act_info, const Window &window) |
Variables | |
constexpr int | step = 4 |
Class to run cpu::kernels::CpuArithmeticKernel except for maximum operation.
Definition at line 67 of file CpuElementwise.h.
Class to run cpu::kernels::CpuArithmeticKernel except for minimum operation.
Definition at line 69 of file CpuElementwise.h.
Class to run cpu::kernels::CpuArithmeticKernel except for squared difference operation.
Definition at line 71 of file CpuElementwise.h.
using CpuLogSoftmax = CpuSoftmaxGeneric<true> |
Definition at line 101 of file CpuSoftmax.h.
Class to run cpu::kernels::CpuArithmeticKernel except for PRelu operation.
Definition at line 34 of file CpuPRelu.h.
using CpuSoftmax = CpuSoftmaxGeneric<false> |
Definition at line 100 of file CpuSoftmax.h.
using ICpuKernel = arm_compute::ICPPKernel |
Definition at line 33 of file ICpuKernel.h.
Definition at line 33 of file ICpuOperator.h.
Definition at line 32 of file CpuElementwiseUnary.cpp.
Basic function to run equal comparison.
Definition at line 182 of file CpuElementwise.h.
Basic function to run greater comparison.
Definition at line 186 of file CpuElementwise.h.
Basic function to run greater-equal comparison.
Definition at line 188 of file CpuElementwise.h.
Basic function to run less comparison.
Definition at line 190 of file CpuElementwise.h.
Basic function to run less-equal comparison.
Definition at line 192 of file CpuElementwise.h.
Basic function to run not equal comparison.
Definition at line 184 of file CpuElementwise.h.
void add_qasymm8_neon | ( | const ITensor * | src0, |
const ITensor * | src1, | ||
ITensor * | dst, | ||
const ConvertPolicy & | policy, | ||
const Window & | window | ||
) |
Definition at line 35 of file qasymm8.cpp.
References ARM_COMPUTE_UNUSED, arm_compute::test::validation::b, arm_compute::graph::bfs(), Window::broadcast_if_dimension_le_one(), Window::DimX, arm_compute::test::validation::dst, Window::Dimension::end(), arm_compute::execute_window_loop(), ITensor::info(), UniformQuantizationInfo::offset, Iterator::ptr(), ITensorInfo::quantization_info(), arm_compute::quantize_qasymm8(), UniformQuantizationInfo::scale, Window::set(), Window::Dimension::start(), Window::Dimension::step(), ITensorInfo::tensor_shape(), QuantizationInfo::uniform(), Dimensions< T >::x(), and Window::x().
void add_qasymm8_signed_neon | ( | const ITensor * | src0, |
const ITensor * | src1, | ||
ITensor * | dst, | ||
const ConvertPolicy & | policy, | ||
const Window & | window | ||
) |
Definition at line 35 of file qasymm8_signed.cpp.
References ARM_COMPUTE_UNUSED, arm_compute::test::validation::b, arm_compute::graph::bfs(), Window::broadcast_if_dimension_le_one(), Window::DimX, arm_compute::test::validation::dst, Window::Dimension::end(), arm_compute::execute_window_loop(), ITensor::info(), UniformQuantizationInfo::offset, Iterator::ptr(), ITensorInfo::quantization_info(), arm_compute::quantize_qasymm8_signed(), UniformQuantizationInfo::scale, Window::set(), Window::Dimension::start(), Window::Dimension::step(), ITensorInfo::tensor_shape(), QuantizationInfo::uniform(), Dimensions< T >::x(), and Window::x().
void add_qsymm16_neon | ( | const ITensor * | src0, |
const ITensor * | src1, | ||
ITensor * | dst, | ||
const ConvertPolicy & | policy, | ||
const Window & | window | ||
) |
Definition at line 35 of file qsymm16.cpp.
References ARM_COMPUTE_UNUSED, arm_compute::test::validation::b, arm_compute::graph::bfs(), Window::broadcast_if_dimension_le_one(), Window::DimX, arm_compute::test::validation::dst, Window::Dimension::end(), arm_compute::execute_window_loop(), ITensor::info(), Iterator::ptr(), ITensorInfo::quantization_info(), arm_compute::quantize_qsymm16(), UniformQuantizationInfo::scale, Window::set(), Window::Dimension::start(), Window::Dimension::step(), ITensorInfo::tensor_shape(), QuantizationInfo::uniform(), Dimensions< T >::x(), and Window::x().
void add_s16_u8_s16_neon | ( | const ITensor * | src0, |
const ITensor * | src1, | ||
ITensor * | dst, | ||
const ConvertPolicy & | policy, | ||
const Window & | window | ||
) |
Definition at line 100 of file integer.cpp.
References arm_compute::wrapper::add_sat(), Window::broadcast_if_dimension_le_one(), Window::DimX, arm_compute::test::validation::dst, Window::Dimension::end(), arm_compute::execute_window_loop(), ITensor::info(), Iterator::ptr(), Window::set(), Window::Dimension::start(), ITensorInfo::tensor_shape(), arm_compute::wrapper::vadd(), arm_compute::wrapper::vload(), arm_compute::wrapper::vloadq(), arm_compute::wrapper::vmovl(), arm_compute::wrapper::vqadd(), arm_compute::wrapper::vstore(), arm_compute::WRAP, and Window::x().
Referenced by add_u8_s16_s16_neon().
void arm_compute::cpu::add_same_neon | ( | const ITensor * | src0, |
const ITensor * | src1, | ||
ITensor * | dst, | ||
const ConvertPolicy & | policy, | ||
const Window & | window | ||
) |
SIMD vector tag type.
Definition at line 48 of file list.h.
References arm_compute::wrapper::add_sat(), Window::broadcast_if_dimension_le_one(), Window::DimX, arm_compute::test::validation::dst, Window::Dimension::end(), arm_compute::execute_window_loop(), ITensor::info(), Iterator::ptr(), arm_compute::SATURATE, Window::set(), Window::Dimension::start(), Window::Dimension::step(), ITensorInfo::tensor_shape(), arm_compute::wrapper::vadd(), arm_compute::wrapper::vdup_n(), arm_compute::wrapper::vloadq(), arm_compute::wrapper::vqadd(), arm_compute::wrapper::vstore(), Dimensions< T >::x(), and Window::x().
void add_u8_s16_s16_neon | ( | const ITensor * | src0, |
const ITensor * | src1, | ||
ITensor * | dst, | ||
const ConvertPolicy & | policy, | ||
const Window & | window | ||
) |
Definition at line 164 of file integer.cpp.
References add_s16_u8_s16_neon(), and arm_compute::test::validation::dst.
void add_u8_u8_s16_neon | ( | const ITensor * | src0, |
const ITensor * | src1, | ||
ITensor * | dst, | ||
const ConvertPolicy & | policy, | ||
const Window & | window | ||
) |
Definition at line 35 of file integer.cpp.
References arm_compute::wrapper::add_sat(), Window::broadcast_if_dimension_le_one(), Window::DimX, arm_compute::test::validation::dst, Window::Dimension::end(), arm_compute::execute_window_loop(), ITensor::info(), Iterator::ptr(), Window::set(), Window::Dimension::start(), ITensorInfo::tensor_shape(), arm_compute::wrapper::vadd(), arm_compute::wrapper::vload(), arm_compute::wrapper::vmovl(), arm_compute::wrapper::vqadd(), arm_compute::wrapper::vstore(), arm_compute::WRAP, and Window::x().
void arm_compute::cpu::bilinear_neon_scale | ( | const ITensor * | src, |
ITensor * | dst, | ||
const ITensor * | offsets, | ||
const ITensor * | dx, | ||
const ITensor * | dy, | ||
BorderMode | border_mode, | ||
PixelValue | constant_border_value, | ||
float | sampling_offset, | ||
bool | align_corners, | ||
const Window & | window | ||
) |
Definition at line 94 of file list.h.
References ARM_COMPUTE_ERROR, arm_compute::scale_utils::calculate_resize_ratio(), arm_compute::CONSTANT, arm_compute::scale_helpers::delta_bilinear(), Window::DimY, Window::DimZ, arm_compute::test::validation::dst, arm_compute::execute_window_loop(), PixelValue::get(), offset(), Iterator::ptr(), ITensor::ptr_to_element(), arm_compute::REPLICATE, arm_compute::test::validation::src, and type.
|
inline |
Definition at line 162 of file quantized.h.
References arm_compute::test::validation::data_layout, arm_compute::get_data_layout_dimension_index(), arm_compute::HEIGHT, arm_compute::test::validation::idx_height, arm_compute::test::validation::idx_width, and arm_compute::WIDTH.
Referenced by poolingMxN_fp32_neon_nhwc(), and poolingMxN_q8_neon_nhwc().
void arm_compute::cpu::common_neon_scale | ( | const ITensor * | src, |
ITensor * | dst, | ||
const ITensor * | offsets, | ||
const ITensor * | dx, | ||
const ITensor * | dy, | ||
InterpolationPolicy | policy, | ||
BorderMode | border_mode, | ||
PixelValue | constant_border_value, | ||
float | sampling_offset, | ||
bool | align_corners, | ||
const Window & | window | ||
) |
Definition at line 169 of file list.h.
References arm_compute::BILINEAR, arm_compute::test::validation::dst, arm_compute::NEAREST_NEIGHBOR, and arm_compute::test::validation::src.
|
inline |
Definition at line 125 of file elementwise_quantized_list.h.
References arm_compute::test::validation::b.
|
inline |
Definition at line 160 of file elementwise_list.h.
References ARM_COMPUTE_ERROR, arm_compute::test::validation::b, arm_compute::MAX, arm_compute::MIN, arm_compute::PRELU, arm_compute::SQUARED_DIFF, type, arm_compute::wrapper::vbsl(), arm_compute::wrapper::vcgt(), arm_compute::wrapper::vdup_n(), arm_compute::wrapper::vmax(), arm_compute::wrapper::vmin(), arm_compute::wrapper::vmul(), and arm_compute::wrapper::vsub().
void arm_compute::cpu::elementwise_arithm_op | ( | const ITensor * | in1, |
const ITensor * | in2, | ||
ITensor * | out, | ||
const Window & | window | ||
) |
Definition at line 269 of file elementwise_list.h.
|
inline |
Definition at line 206 of file elementwise_list.h.
References arm_compute::test::validation::b, and arm_compute::wrapper::vdiv().
|
inline |
Definition at line 200 of file elementwise_list.h.
References arm_compute::test::validation::b, arm_compute::wrapper::vdiv(), and arm_compute::vfloorq_f32().
|
inline |
Definition at line 212 of file elementwise_list.h.
References arm_compute::test::validation::b, and arm_compute::wrapper::vpow().
|
inline |
Definition at line 232 of file elementwise_list.h.
References type, and arm_compute::wrapper::vdup_n().
|
inline |
Definition at line 256 of file elementwise_list.h.
References arm_compute::wrapper::vloadq(), and arm_compute::wrapper::vstore().
|
inline |
Definition at line 242 of file elementwise_list.h.
References arm_compute::test::validation::b, arm_compute::wrapper::vloadq(), and arm_compute::wrapper::vstore().
void arm_compute::cpu::elementwise_arithm_op_quantized | ( | const ITensor * | in1, |
const ITensor * | in2, | ||
ITensor * | out, | ||
const Window & | window | ||
) |
Definition at line 622 of file elementwise_quantized_list.h.
References elementwise_op_quantized().
|
inline |
Definition at line 199 of file elementwise_quantized_list.h.
References load_quantized(), and store_quantized().
|
inline |
Definition at line 163 of file elementwise_quantized_list.h.
References load_quantized(), and store_quantized().
|
inline |
Definition at line 113 of file elementwise_quantized_list.h.
References arm_compute::test::validation::b, arm_compute::test::validation::qinfo, and arm_compute::quantize_qasymm8().
void arm_compute::cpu::elementwise_arithm_op_quantized_signed | ( | const ITensor * | in1, |
const ITensor * | in2, | ||
ITensor * | out, | ||
const Window & | window | ||
) |
Definition at line 629 of file elementwise_quantized_list.h.
References elementwise_op_quantized_signed().
|
inline |
Definition at line 214 of file elementwise_quantized_list.h.
References load_quantized_signed(), and store_quantized_signed().
|
inline |
Definition at line 119 of file elementwise_quantized_list.h.
References arm_compute::test::validation::b, arm_compute::test::validation::qinfo, and arm_compute::quantize_qasymm8_signed().
|
inline |
Definition at line 181 of file elementwise_quantized_list.h.
References load_quantized_signed(), and store_quantized_signed().
|
inline |
Definition at line 113 of file elementwise_list.h.
References ARM_COMPUTE_ERROR, arm_compute::test::validation::b, arm_compute::DIV, arm_compute::MAX, arm_compute::MIN, arm_compute::POWER, arm_compute::PRELU, and arm_compute::SQUARED_DIFF.
|
inline |
Definition at line 148 of file elementwise_quantized_list.h.
References arm_compute::test::validation::b.
|
inline |
Definition at line 311 of file elementwise_list.h.
References ARM_COMPUTE_ERROR, arm_compute::test::validation::b, arm_compute::Equal, arm_compute::Greater, arm_compute::GreaterEqual, arm_compute::Less, arm_compute::LessEqual, arm_compute::NotEqual, arm_compute::wrapper::vceq(), arm_compute::wrapper::vcge(), arm_compute::wrapper::vcgt(), and arm_compute::wrapper::vnot().
void arm_compute::cpu::elementwise_comp_op_16 | ( | const ITensor * | in1, |
const ITensor * | in2, | ||
ITensor * | out, | ||
const Window & | window | ||
) |
Definition at line 467 of file elementwise_list.h.
|
inline |
Definition at line 414 of file elementwise_list.h.
References arm_compute::test::validation::b, arm_compute::wrapper::vloadq(), arm_compute::wrapper::vmovn(), and arm_compute::wrapper::vstore().
void arm_compute::cpu::elementwise_comp_op_32 | ( | const ITensor * | in1, |
const ITensor * | in2, | ||
ITensor * | out, | ||
const Window & | window | ||
) |
Definition at line 476 of file elementwise_list.h.
|
inline |
Definition at line 429 of file elementwise_list.h.
References arm_compute::test::validation::b, arm_compute::wrapper::vcombine(), arm_compute::wrapper::vgetlane(), arm_compute::wrapper::vloadq(), arm_compute::wrapper::vmovn(), and arm_compute::wrapper::vstore().
void arm_compute::cpu::elementwise_comp_op_8 | ( | const ITensor * | in1, |
const ITensor * | in2, | ||
ITensor * | out, | ||
const Window & | window | ||
) |
Definition at line 458 of file elementwise_list.h.
|
inline |
Definition at line 399 of file elementwise_list.h.
References arm_compute::test::validation::b, arm_compute::wrapper::vloadq(), and arm_compute::wrapper::vstore().
|
inline |
Definition at line 343 of file elementwise_list.h.
References arm_compute::wrapper::vdup_n().
|
inline |
Definition at line 363 of file elementwise_list.h.
References arm_compute::wrapper::vloadq(), arm_compute::wrapper::vmovn(), and arm_compute::wrapper::vstore().
|
inline |
Definition at line 376 of file elementwise_list.h.
References arm_compute::test::validation::b, arm_compute::wrapper::vcombine(), arm_compute::wrapper::vgetlane(), arm_compute::wrapper::vloadq(), arm_compute::wrapper::vmovn(), and arm_compute::wrapper::vstore().
|
inline |
Definition at line 350 of file elementwise_list.h.
References arm_compute::wrapper::vloadq(), and arm_compute::wrapper::vstore().
void arm_compute::cpu::elementwise_comp_op_quantized | ( | const ITensor * | in1, |
const ITensor * | in2, | ||
ITensor * | out, | ||
const Window & | window | ||
) |
Definition at line 637 of file elementwise_quantized_list.h.
References elementwise_op_quantized().
|
inline |
Definition at line 266 of file elementwise_quantized_list.h.
References ARM_COMPUTE_UNUSED, load_quantized(), and store_quantized().
|
inline |
Definition at line 230 of file elementwise_quantized_list.h.
References ARM_COMPUTE_UNUSED, load_quantized(), and store_quantized().
|
inline |
Definition at line 141 of file elementwise_quantized_list.h.
References ARM_COMPUTE_UNUSED, arm_compute::test::validation::b, and arm_compute::test::validation::qinfo.
void arm_compute::cpu::elementwise_comp_op_quantized_signed | ( | const ITensor * | in1, |
const ITensor * | in2, | ||
ITensor * | out, | ||
const Window & | window | ||
) |
Definition at line 645 of file elementwise_quantized_list.h.
References elementwise_comp_quantized_signed().
|
inline |
Definition at line 283 of file elementwise_quantized_list.h.
References ARM_COMPUTE_UNUSED, load_quantized_signed(), and store_quantized().
|
inline |
Definition at line 248 of file elementwise_quantized_list.h.
References ARM_COMPUTE_UNUSED, load_quantized_signed(), and store_quantized().
|
inline |
Definition at line 280 of file elementwise_list.h.
References ARM_COMPUTE_ERROR, arm_compute::test::validation::b, arm_compute::Equal, arm_compute::Greater, arm_compute::GreaterEqual, arm_compute::Less, arm_compute::LessEqual, and arm_compute::NotEqual.
void arm_compute::cpu::elementwise_comp_quantized_signed | ( | const ITensor * | in1, |
const ITensor * | in2, | ||
ITensor * | out, | ||
const Window & | window, | ||
uint8_t(*)(const float &, const float &, UniformQuantizationInfo) | scalar_func, | ||
int(*)(int, int, int, const int8_t *, float32x4x4_t, uint8_t *, int32x4_t, float32x4_t, float32x4_t, float32x4_t, const bool) | broadcast_func, | ||
int(*)(int, int, int, const int8_t *, const int8_t *, uint8_t *, int32x4_t, int32x4_t, float32x4_t, float32x4_t, float32x4_t, float32x4_t) | neon_func | ||
) |
Definition at line 407 of file elementwise_quantized_list.h.
References arm_compute::graph::bfs(), Window::broadcast_if_dimension_le_one(), arm_compute::dequantize_qasymm8_signed(), Window::DimX, Window::Dimension::end(), arm_compute::execute_window_loop(), ITensor::info(), UniformQuantizationInfo::offset, Iterator::ptr(), ITensorInfo::quantization_info(), UniformQuantizationInfo::scale, Window::set(), Window::Dimension::start(), Window::Dimension::step(), ITensorInfo::tensor_shape(), QuantizationInfo::uniform(), arm_compute::vdequantize(), Dimensions< T >::x(), and Window::x().
Referenced by elementwise_comp_op_quantized_signed().
void arm_compute::cpu::elementwise_op | ( | const ITensor * | in1, |
const ITensor * | in2, | ||
ITensor * | out, | ||
const Window & | window, | ||
OutputScalarType(*)(const InputScalarType &, const InputScalarType &) | scalar_func, | ||
int(*)(int, int, int, const InputScalarType *, const InputScalarType &, OutputScalarType *, const bool) | broadcast_func, | ||
int(*)(int, int, int, const InputScalarType *, const InputScalarType *, OutputScalarType *) | neon_func | ||
) |
Definition at line 36 of file elementwise_list.h.
References arm_compute::test::validation::b, Window::broadcast_if_dimension_le_one(), Window::DimX, Window::Dimension::end(), arm_compute::execute_window_loop(), ITensor::info(), Iterator::ptr(), Window::set(), Window::Dimension::start(), Window::Dimension::step(), ITensorInfo::tensor_shape(), Dimensions< T >::x(), and Window::x().
void arm_compute::cpu::elementwise_op | ( | const ITensor * | in, |
ITensor * | out, | ||
const Window & | window, | ||
ElementWiseUnary | op | ||
) |
Definition at line 83 of file elementwise_unary_list.h.
References Window::DimX, elementwise_op_scalar_imp(), Window::Dimension::end(), arm_compute::execute_window_loop(), arm_compute::test::validation::input, Iterator::ptr(), Window::set(), Window::Dimension::start(), arm_compute::wrapper::vloadq(), arm_compute::wrapper::vstore(), and Window::x().
|
inline |
Definition at line 59 of file elementwise_unary_list.h.
References arm_compute::ABS, ARM_COMPUTE_ERROR, arm_compute::EXP, arm_compute::LOG, arm_compute::NEG, arm_compute::ROUND, arm_compute::RSQRT, arm_compute::SIN, arm_compute::wrapper::vabs(), arm_compute::wrapper::vexpq(), arm_compute::wrapper::vinvsqrt(), arm_compute::wrapper::vlog(), arm_compute::wrapper::vneg(), arm_compute::wrapper::vround(), and arm_compute::wrapper::vsin().
void arm_compute::cpu::elementwise_op_quantized | ( | const ITensor * | in1, |
const ITensor * | in2, | ||
ITensor * | out, | ||
const Window & | window, | ||
uint8_t(*)(const float &, const float &, UniformQuantizationInfo) | scalar_func, | ||
int(*)(int, int, int, const uint8_t *, float32x4x4_t, uint8_t *, int32x4_t, float32x4_t, float32x4_t, float32x4_t, const bool) | broadcast_func, | ||
int(*)(int, int, int, const uint8_t *, const uint8_t *, uint8_t *, int32x4_t, int32x4_t, float32x4_t, float32x4_t, float32x4_t, float32x4_t) | neon_func | ||
) |
Definition at line 299 of file elementwise_quantized_list.h.
References arm_compute::graph::bfs(), Window::broadcast_if_dimension_le_one(), arm_compute::dequantize_qasymm8(), Window::DimX, Window::Dimension::end(), arm_compute::execute_window_loop(), ITensor::info(), UniformQuantizationInfo::offset, Iterator::ptr(), ITensorInfo::quantization_info(), UniformQuantizationInfo::scale, Window::set(), Window::Dimension::start(), Window::Dimension::step(), ITensorInfo::tensor_shape(), QuantizationInfo::uniform(), arm_compute::vdequantize(), Dimensions< T >::x(), and Window::x().
Referenced by elementwise_arithm_op_quantized(), and elementwise_comp_op_quantized().
void arm_compute::cpu::elementwise_op_quantized_signed | ( | const ITensor * | in1, |
const ITensor * | in2, | ||
ITensor * | out, | ||
const Window & | window, | ||
int8_t(*)(const float &, const float &, UniformQuantizationInfo) | scalar_func, | ||
int(*)(int, int, int, const int8_t *, float32x4x4_t, int8_t *, int32x4_t, float32x4_t, float32x4_t, float32x4_t, const bool) | broadcast_func, | ||
int(*)(int, int, int, const int8_t *, const int8_t *, int8_t *, int32x4_t, int32x4_t, float32x4_t, float32x4_t, float32x4_t, float32x4_t) | neon_func | ||
) |
Definition at line 514 of file elementwise_quantized_list.h.
References arm_compute::graph::bfs(), Window::broadcast_if_dimension_le_one(), arm_compute::dequantize_qasymm8_signed(), Window::DimX, Window::Dimension::end(), arm_compute::execute_window_loop(), ITensor::info(), UniformQuantizationInfo::offset, Iterator::ptr(), ITensorInfo::quantization_info(), UniformQuantizationInfo::scale, Window::set(), Window::Dimension::start(), Window::Dimension::step(), ITensorInfo::tensor_shape(), QuantizationInfo::uniform(), arm_compute::vdequantize(), Dimensions< T >::x(), and Window::x().
Referenced by elementwise_arithm_op_quantized_signed().
|
inline |
Definition at line 35 of file elementwise_unary_list.h.
References arm_compute::ABS, ARM_COMPUTE_ERROR, arm_compute::EXP, arm_compute::LOG, arm_compute::support::cpp11::nearbyint(), arm_compute::NEG, arm_compute::ROUND, arm_compute::RSQRT, and arm_compute::SIN.
Referenced by elementwise_op().
void arm_compute::cpu::fp16_neon_activation | ( | const ITensor * | src, |
ITensor * | dst, | ||
const ActivationLayerInfo & | act_info, | ||
const Window & | window | ||
) |
void arm_compute::cpu::fp16_neon_batch_normalization | ( | ITensor * | src, |
ITensor * | dst, | ||
const ITensor * | mean, | ||
const ITensor * | var, | ||
const ITensor * | beta, | ||
const ITensor * | gamma, | ||
float | epsilon, | ||
ActivationLayerInfo & | act_info, | ||
const Window & | window | ||
) |
void arm_compute::cpu::fp16_neon_floor | ( | const void * | src, |
void * | dst, | ||
int | len | ||
) |
void arm_compute::cpu::fp16_sve_activation | ( | const ITensor * | src, |
ITensor * | dst, | ||
const ActivationLayerInfo & | act_info, | ||
const Window & | window | ||
) |
void arm_compute::cpu::fp16_sve_batch_normalization | ( | ITensor * | src, |
ITensor * | dst, | ||
const ITensor * | mean, | ||
const ITensor * | var, | ||
const ITensor * | beta, | ||
const ITensor * | gamma, | ||
float | epsilon, | ||
ActivationLayerInfo & | act_info, | ||
const Window & | window | ||
) |
void arm_compute::cpu::fp16_sve_scale | ( | const ITensor * | src, |
ITensor * | dst, | ||
const ITensor * | offsets, | ||
const ITensor * | dx, | ||
const ITensor * | dy, | ||
InterpolationPolicy | policy, | ||
BorderMode | border_mode, | ||
PixelValue | constant_border_value, | ||
float | sampling_offset, | ||
bool | align_corners, | ||
const Window & | window | ||
) |
void fp32_neon_activation | ( | const ITensor * | src, |
ITensor * | dst, | ||
const ActivationLayerInfo & | act_info, | ||
const Window & | window | ||
) |
SIMD vector tag type.
SIMD vector tag type.
Definition at line 49 of file fp32.cpp.
References ActivationLayerInfo::a(), ActivationLayerInfo::ABS, ActivationLayerInfo::activation(), ARM_COMPUTE_ERROR, arm_compute::test::validation::b, ActivationLayerInfo::b(), ActivationLayerInfo::BOUNDED_RELU, Window::collapse_if_possible(), Window::DimX, Window::DimZ, arm_compute::test::validation::dst, ActivationLayerInfo::ELU, Window::Dimension::end(), arm_compute::execute_window_loop(), ActivationLayerInfo::HARD_SWISH, ActivationLayerInfo::IDENTITY, arm_compute::test::validation::input, ActivationLayerInfo::LEAKY_RELU, ActivationLayerInfo::LINEAR, ActivationLayerInfo::LOGISTIC, ActivationLayerInfo::LU_BOUNDED_RELU, Iterator::ptr(), ActivationLayerInfo::RELU, Window::set(), ActivationLayerInfo::SOFT_RELU, ActivationLayerInfo::SQRT, ActivationLayerInfo::SQUARE, arm_compute::test::validation::src, Window::Dimension::start(), ActivationLayerInfo::TANH, arm_compute::wrapper::vabs(), arm_compute::wrapper::vadd(), arm_compute::wrapper::vbsl(), arm_compute::wrapper::vceq(), arm_compute::wrapper::vcge(), arm_compute::wrapper::vcgt(), arm_compute::wrapper::vdup_n(), arm_compute::wrapper::vexpq(), arm_compute::wrapper::vinv(), arm_compute::wrapper::vinvsqrt(), arm_compute::wrapper::vloadq(), arm_compute::wrapper::vlog(), arm_compute::wrapper::vmax(), arm_compute::wrapper::vmin(), arm_compute::wrapper::vmla(), arm_compute::wrapper::vmul(), arm_compute::wrapper::vneg(), arm_compute::wrapper::vnot(), arm_compute::wrapper::vstore(), arm_compute::wrapper::vsub(), arm_compute::wrapper::vtanh(), and Window::x().
void fp32_neon_batch_normalization | ( | ITensor * | src, |
ITensor * | dst, | ||
const ITensor * | mean, | ||
const ITensor * | var, | ||
const ITensor * | beta, | ||
const ITensor * | gamma, | ||
float | epsilon, | ||
ActivationLayerInfo & | act_info, | ||
const Window & | window | ||
) |
Definition at line 135 of file fp32.cpp.
References ActivationLayerInfo::activation(), arm_compute::test::validation::dst, ActivationLayerInfo::enabled(), arm_compute::quantization::epsilon, and arm_compute::test::validation::src.
void fp32_neon_floor | ( | const void * | src, |
void * | dst, | ||
int | len | ||
) |
Definition at line 37 of file fp32.cpp.
References ARM_COMPUTE_ASSERT, ARM_COMPUTE_ASSERT_NOT_NULLPTR, arm_compute::test::validation::dst, arm_compute::test::validation::src, step, and arm_compute::vfloorq_f32().
void arm_compute::cpu::fp32_sve_activation | ( | const ITensor * | src, |
ITensor * | dst, | ||
const ActivationLayerInfo & | act_info, | ||
const Window & | window | ||
) |
void arm_compute::cpu::fp32_sve_batch_normalization | ( | ITensor * | src, |
ITensor * | dst, | ||
const ITensor * | mean, | ||
const ITensor * | var, | ||
const ITensor * | beta, | ||
const ITensor * | gamma, | ||
float | epsilon, | ||
ActivationLayerInfo & | act_info, | ||
const Window & | window | ||
) |
void arm_compute::cpu::fp32_sve_scale | ( | const ITensor * | src, |
ITensor * | dst, | ||
const ITensor * | offsets, | ||
const ITensor * | dx, | ||
const ITensor * | dy, | ||
InterpolationPolicy | policy, | ||
BorderMode | border_mode, | ||
PixelValue | constant_border_value, | ||
float | sampling_offset, | ||
bool | align_corners, | ||
const Window & | window | ||
) |
float32x4x4_t arm_compute::cpu::load_quantized | ( | const uint8_t * | input1_ptr, |
const int32x4_t & | offset, | ||
const float32x4_t & | scale | ||
) |
Definition at line 33 of file elementwise_quantized_list.h.
References offset(), and arm_compute::test::validation::scale.
Referenced by elementwise_arithm_op_quantized_broadcast_loop(), elementwise_arithm_op_quantized_loop(), elementwise_comp_op_quantized_broadcast_loop(), and elementwise_comp_op_quantized_loop().
float32x4x4_t arm_compute::cpu::load_quantized_signed | ( | const int8_t * | input1_ptr, |
const int32x4_t & | offset, | ||
const float32x4_t & | scale | ||
) |
Definition at line 48 of file elementwise_quantized_list.h.
References offset(), and arm_compute::test::validation::scale.
Referenced by elementwise_arithm_op_quantized_signed_broadcast_loop(), elementwise_arithm_op_quantized_singed_loop(), elementwise_comp_op_quantized_signed_broadcast_loop(), and elementwise_comp_op_quantized_signed_loop().
void arm_compute::cpu::nearest_neon_scale | ( | const ITensor * | src, |
ITensor * | dst, | ||
const ITensor * | offsets, | ||
float | sampling_offset, | ||
bool | align_corners, | ||
const Window & | window | ||
) |
Definition at line 51 of file list.h.
References arm_compute::scale_utils::calculate_resize_ratio(), Window::DimX, arm_compute::test::validation::dst, Window::Dimension::end(), arm_compute::execute_window_loop(), offset(), Iterator::ptr(), ITensor::ptr_to_element(), arm_compute::utils::rounding::round_half_away_from_zero(), Window::set(), arm_compute::test::validation::src, Window::Dimension::start(), arm_compute::wrapper::vloadq(), arm_compute::wrapper::vstore(), and Window::x().
void arm_compute::cpu::neon_logits_1d_max | ( | const ITensor * | in, |
ITensor * | out, | ||
const Window & | window | ||
) |
SIMD vector tag type.
Definition at line 37 of file list.h.
References Window::DimX, Window::Dimension::end(), arm_compute::execute_window_loop(), arm_compute::test::validation::input, Iterator::ptr(), Window::set(), Window::Dimension::start(), arm_compute::wrapper::vdup_n(), arm_compute::wrapper::vgethigh(), arm_compute::wrapper::vgetlane(), arm_compute::wrapper::vgetlow(), arm_compute::wrapper::vloadq(), arm_compute::wrapper::vmax(), arm_compute::wrapper::vpmax(), and Window::x().
void arm_compute::cpu::neon_softmax_logits_1d_float | ( | const ITensor * | in, |
const ITensor * | max, | ||
void *const | tmp, | ||
ITensor * | out, | ||
const float | beta, | ||
bool | is_log, | ||
const Window & | window | ||
) |
SIMD vector tag type.
Definition at line 260 of file list.h.
References ValidRegion::anchor, arm_compute::execute_window_loop(), ITensor::info(), input_width, Iterator::ptr(), ValidRegion::shape, arm_compute::wrapper::vadd(), ITensorInfo::valid_region(), arm_compute::wrapper::vdup_n(), arm_compute::wrapper::vexpq(), arm_compute::wrapper::vgethigh(), arm_compute::wrapper::vgetlane(), arm_compute::wrapper::vgetlow(), arm_compute::wrapper::vloadq(), arm_compute::wrapper::vmul(), arm_compute::wrapper::vpadd(), arm_compute::wrapper::vstore(), arm_compute::wrapper::vsub(), and Dimensions< T >::x().
void arm_compute::cpu::neon_softmax_logits_1d_quantized | ( | const ITensor * | in, |
const ITensor * | max, | ||
void *const | tmp, | ||
ITensor * | out, | ||
float | beta, | ||
bool | is_log, | ||
const Window & | window | ||
) |
Definition at line 87 of file list.h.
References ValidRegion::anchor, arm_compute::execute_window_loop(), ITensor::info(), input_width, Iterator::ptr(), ITensorInfo::quantization_info(), UniformQuantizationInfo::scale, ValidRegion::shape, QuantizationInfo::uniform(), ITensorInfo::valid_region(), arm_compute::wrapper::vdup_n(), arm_compute::vexpq_f32(), arm_compute::wrapper::vgetlane(), arm_compute::wrapper::vloadq(), arm_compute::wrapper::vqsub(), arm_compute::wrapper::vstore(), arm_compute::wrapper::vsub(), and Dimensions< T >::x().
|
inline |
Definition at line 62 of file list.h.
References arm_compute::test::validation::data_layout, arm_compute::test::validation::info, arm_compute::NCHW, and pool_stride_x.
void arm_compute::cpu::poolingMxN_fp16_neon_nhwc | ( | const ITensor * | src0, |
ITensor * | dst0, | ||
ITensor * | dst1, | ||
PoolingLayerInfo & | , | ||
const Window & | window_src, | ||
const Window & | window | ||
) |
void poolingMxN_fp32_neon_nhwc | ( | const ITensor * | src, |
ITensor * | dst0, | ||
ITensor * | dst1, | ||
PoolingLayerInfo & | pool_info, | ||
const Window & | window_src, | ||
const Window & | window | ||
) |
Definition at line 146 of file fp32.cpp.
References calculate_avg_scale(), Window::DimX, Window::Dimension::end(), PoolingLayerInfo::exclude_padding, arm_compute::execute_window_loop(), Size2D::height, arm_compute::test::validation::idx_height, arm_compute::test::validation::idx_width, PoolingLayerInfo::is_global_pooling, arm_compute::L2, arm_compute::support::cpp11::lowest(), arm_compute::MAX, arm_compute::NHWC, PadStrideInfo::pad_bottom(), PadStrideInfo::pad_left(), PadStrideInfo::pad_right(), PoolingLayerInfo::pad_stride_info, PadStrideInfo::pad_top(), PoolingLayerInfo::pool_size, pool_stride_x, PoolingLayerInfo::pool_type, Iterator::ptr(), arm_compute::test::validation::scale, Window::set(), arm_compute::test::validation::src, Window::Dimension::start(), PadStrideInfo::stride(), Size2D::width, Window::x(), Window::y(), and Window::z().
void arm_compute::cpu::poolingMxN_q8_neon_nhwc | ( | const ITensor * | src, |
ITensor * | dst0, | ||
ITensor * | dst1, | ||
PoolingLayerInfo & | pool_info, | ||
const Window & | window_src, | ||
const Window & | window | ||
) |
Definition at line 182 of file quantized.h.
References ARM_COMPUTE_UNUSED, calculate_avg_scale(), Window::DimX, Window::Dimension::end(), PoolingLayerInfo::exclude_padding, arm_compute::execute_window_loop(), Size2D::height, arm_compute::test::validation::idx_height, arm_compute::test::validation::idx_width, ITensor::info(), PoolingLayerInfo::is_global_pooling, arm_compute::MAX, arm_compute::NHWC, UniformQuantizationInfo::offset, PadStrideInfo::pad_bottom(), PadStrideInfo::pad_left(), PadStrideInfo::pad_right(), PoolingLayerInfo::pad_stride_info, PadStrideInfo::pad_top(), PoolingLayerInfo::pool_size, pool_stride_x, PoolingLayerInfo::pool_type, Iterator::ptr(), ITensorInfo::quantization_info(), UniformQuantizationInfo::scale, arm_compute::test::validation::scale, Window::set(), arm_compute::test::validation::src, Window::Dimension::start(), PadStrideInfo::stride(), QuantizationInfo::uniform(), arm_compute::wrapper::vadd(), arm_compute::wrapper::vcombine(), vcvtq_f32_q32(), arm_compute::wrapper::vdup_n(), arm_compute::wrapper::vgethigh(), arm_compute::wrapper::vgetlow(), arm_compute::wrapper::vload(), arm_compute::wrapper::vloadq(), arm_compute::wrapper::vmax(), arm_compute::wrapper::vmla(), arm_compute::wrapper::vmovl(), arm_compute::wrapper::vmovn(), arm_compute::wrapper::vstore(), Size2D::width, Window::x(), Window::y(), and Window::z().
void poolingMxN_qasymm8_neon_nhwc | ( | const ITensor * | src0, |
ITensor * | dst0, | ||
ITensor * | dst1, | ||
PoolingLayerInfo & | pool_info, | ||
const Window & | window_src, | ||
const Window & | window | ||
) |
void poolingMxN_qasymm8_signed_neon_nhwc | ( | const ITensor * | src0, |
ITensor * | dst0, | ||
ITensor * | dst1, | ||
PoolingLayerInfo & | pool_info, | ||
const Window & | window_src, | ||
const Window & | window | ||
) |
void qasymm8_neon_activation | ( | const ITensor * | src, |
ITensor * | dst, | ||
const ActivationLayerInfo & | act_info, | ||
const Window & | window | ||
) |
Definition at line 39 of file qasymm8.cpp.
References ActivationLayerInfo::a(), ActivationLayerInfo::activation(), ARM_COMPUTE_ERROR, arm_compute::test::validation::b, ActivationLayerInfo::b(), ActivationLayerInfo::BOUNDED_RELU, Window::collapse_if_possible(), arm_compute::dequantize_qasymm8(), Window::DimX, Window::DimZ, arm_compute::test::validation::dst, Window::Dimension::end(), arm_compute::execute_window_loop(), ActivationLayerInfo::HARD_SWISH, arm_compute::test::validation::input, ActivationLayerInfo::LEAKY_RELU, ActivationLayerInfo::LOGISTIC, ActivationLayerInfo::LU_BOUNDED_RELU, UniformQuantizationInfo::offset, Iterator::ptr(), arm_compute::quantize_qasymm8(), ActivationLayerInfo::RELU, UniformQuantizationInfo::scale, Window::set(), arm_compute::test::validation::src, Window::Dimension::start(), ActivationLayerInfo::TANH, arm_compute::wrapper::vadd(), arm_compute::wrapper::vbsl(), arm_compute::wrapper::vcgt(), arm_compute::vdequantize(), arm_compute::wrapper::vdiv(), arm_compute::wrapper::vexpq(), arm_compute::wrapper::vloadq(), arm_compute::wrapper::vmax(), arm_compute::wrapper::vmin(), arm_compute::vmlaq_qasymm8(), arm_compute::wrapper::vmul(), arm_compute::wrapper::vneg(), arm_compute::vquantize(), arm_compute::wrapper::vstore(), arm_compute::wrapper::vtanh(), and Window::x().
void qasymm8_neon_scale | ( | const ITensor * | src, |
ITensor * | dst, | ||
const ITensor * | offsets, | ||
const ITensor * | dx, | ||
const ITensor * | dy, | ||
InterpolationPolicy | policy, | ||
BorderMode | border_mode, | ||
PixelValue | constant_border_value, | ||
float | sampling_offset, | ||
bool | align_corners, | ||
const Window & | window | ||
) |
Definition at line 131 of file qasymm8.cpp.
References arm_compute::BILINEAR, arm_compute::test::validation::dst, arm_compute::NEAREST_NEIGHBOR, and arm_compute::test::validation::src.
void qasymm8_signed_neon_activation | ( | const ITensor * | src, |
ITensor * | dst, | ||
const ActivationLayerInfo & | act_info, | ||
const Window & | window | ||
) |
Definition at line 38 of file qasymm8_signed.cpp.
References ActivationLayerInfo::a(), ActivationLayerInfo::activation(), ARM_COMPUTE_ERROR, arm_compute::test::validation::b, ActivationLayerInfo::b(), ActivationLayerInfo::BOUNDED_RELU, Window::collapse_if_possible(), arm_compute::dequantize_qasymm8_signed(), Window::DimX, Window::DimZ, arm_compute::test::validation::dst, Window::Dimension::end(), arm_compute::execute_window_loop(), ActivationLayerInfo::HARD_SWISH, arm_compute::test::validation::input, ActivationLayerInfo::LEAKY_RELU, ActivationLayerInfo::LOGISTIC, ActivationLayerInfo::LU_BOUNDED_RELU, UniformQuantizationInfo::offset, Iterator::ptr(), arm_compute::quantize_qasymm8_signed(), ActivationLayerInfo::RELU, UniformQuantizationInfo::scale, Window::set(), arm_compute::test::validation::src, Window::Dimension::start(), ActivationLayerInfo::TANH, arm_compute::wrapper::vadd(), arm_compute::wrapper::vbsl(), arm_compute::wrapper::vcgt(), arm_compute::vdequantize(), arm_compute::wrapper::vdiv(), arm_compute::wrapper::vexpq(), arm_compute::wrapper::vloadq(), arm_compute::wrapper::vmax(), arm_compute::wrapper::vmin(), arm_compute::vmlaq_qasymm8_signed(), arm_compute::wrapper::vmul(), arm_compute::wrapper::vneg(), arm_compute::vquantize_signed(), arm_compute::wrapper::vstore(), arm_compute::wrapper::vtanh(), and Window::x().
void qasymm8_signed_neon_scale | ( | const ITensor * | src, |
ITensor * | dst, | ||
const ITensor * | offsets, | ||
const ITensor * | dx, | ||
const ITensor * | dy, | ||
InterpolationPolicy | policy, | ||
BorderMode | border_mode, | ||
PixelValue | constant_border_value, | ||
float | sampling_offset, | ||
bool | align_corners, | ||
const Window & | window | ||
) |
Definition at line 131 of file qasymm8_signed.cpp.
References arm_compute::BILINEAR, arm_compute::test::validation::dst, arm_compute::NEAREST_NEIGHBOR, and arm_compute::test::validation::src.
void arm_compute::cpu::qasymm8_signed_sve_activation | ( | const ITensor * | src, |
ITensor * | dst, | ||
const ActivationLayerInfo & | act_info, | ||
const Window & | window | ||
) |
void arm_compute::cpu::qasymm8_signed_sve_scale | ( | const ITensor * | src, |
ITensor * | dst, | ||
const ITensor * | offsets, | ||
const ITensor * | dx, | ||
const ITensor * | dy, | ||
InterpolationPolicy | policy, | ||
BorderMode | border_mode, | ||
PixelValue | constant_border_value, | ||
float | sampling_offset, | ||
bool | align_corners, | ||
const Window & | window | ||
) |
void arm_compute::cpu::qasymm8_sve_activation | ( | const ITensor * | src, |
ITensor * | dst, | ||
const ActivationLayerInfo & | act_info, | ||
const Window & | window | ||
) |
void arm_compute::cpu::qasymm8_sve_scale | ( | const ITensor * | src, |
ITensor * | dst, | ||
const ITensor * | offsets, | ||
const ITensor * | dx, | ||
const ITensor * | dy, | ||
InterpolationPolicy | policy, | ||
BorderMode | border_mode, | ||
PixelValue | constant_border_value, | ||
float | sampling_offset, | ||
bool | align_corners, | ||
const Window & | window | ||
) |
void qsymm16_neon_activation | ( | const ITensor * | src, |
ITensor * | dst, | ||
const ActivationLayerInfo & | act_info, | ||
const Window & | window | ||
) |
Definition at line 40 of file qsymm16.cpp.
References ActivationLayerInfo::a(), ActivationLayerInfo::activation(), ARM_COMPUTE_ERROR, ARM_COMPUTE_UNUSED, ActivationLayerInfo::b(), Window::collapse_if_possible(), arm_compute::dequantize_qsymm16(), Window::DimX, Window::DimZ, arm_compute::test::validation::dst, Window::Dimension::end(), arm_compute::execute_window_loop(), arm_compute::test::validation::input, ActivationLayerInfo::LOGISTIC, Iterator::ptr(), arm_compute::quantize_qsymm16(), UniformQuantizationInfo::scale, Window::set(), arm_compute::test::validation::src, Window::Dimension::start(), ActivationLayerInfo::TANH, arm_compute::wrapper::vadd(), arm_compute::vdequantize_int16(), arm_compute::wrapper::vdiv(), arm_compute::wrapper::vexpq(), arm_compute::wrapper::vloadq(), arm_compute::wrapper::vmul(), arm_compute::wrapper::vneg(), arm_compute::vquantize_int16(), arm_compute::wrapper::vstore(), arm_compute::wrapper::vtanh(), and Window::x().
void arm_compute::cpu::qsymm16_sve_activation | ( | const ITensor * | src, |
ITensor * | dst, | ||
const ActivationLayerInfo & | act_info, | ||
const Window & | window | ||
) |
|
inline |
Definition at line 41 of file quantized.h.
References arm_compute::test::validation::info, and arm_compute::quantize_qasymm8_signed().
|
inline |
Definition at line 48 of file quantized.h.
References arm_compute::test::validation::info, and arm_compute::quantize_qasymm8().
void arm_compute::cpu::s16_neon_scale | ( | const ITensor * | src, |
ITensor * | dst, | ||
const ITensor * | offsets, | ||
const ITensor * | dx, | ||
const ITensor * | dy, | ||
InterpolationPolicy | policy, | ||
BorderMode | border_mode, | ||
PixelValue | constant_border_value, | ||
float | sampling_offset, | ||
bool | align_corners, | ||
const Window & | window | ||
) |
Definition at line 279 of file integer.cpp.
References arm_compute::BILINEAR, arm_compute::test::validation::dst, arm_compute::NEAREST_NEIGHBOR, and arm_compute::test::validation::src.
void arm_compute::cpu::s16_sve_scale | ( | const ITensor * | src, |
ITensor * | dst, | ||
const ITensor * | offsets, | ||
const ITensor * | dx, | ||
const ITensor * | dy, | ||
InterpolationPolicy | policy, | ||
BorderMode | border_mode, | ||
PixelValue | constant_border_value, | ||
float | sampling_offset, | ||
bool | align_corners, | ||
const Window & | window | ||
) |
void arm_compute::cpu::store_quantized | ( | uint8_t * | output_ptr, |
const uint32x4x4_t & | out | ||
) |
Definition at line 63 of file elementwise_quantized_list.h.
Referenced by elementwise_arithm_op_quantized_broadcast_loop(), elementwise_arithm_op_quantized_loop(), elementwise_comp_op_quantized_broadcast_loop(), elementwise_comp_op_quantized_loop(), elementwise_comp_op_quantized_signed_broadcast_loop(), elementwise_comp_op_quantized_signed_loop(), and store_quantized().
void arm_compute::cpu::store_quantized | ( | uint8_t * | output_ptr, |
const int32x4x4_t & | out | ||
) |
Definition at line 70 of file elementwise_quantized_list.h.
void arm_compute::cpu::store_quantized | ( | uint8_t * | output_ptr, |
const float32x4x4_t & | rf, | ||
const float32x4_t & | offset, | ||
const float32x4_t & | invscale | ||
) |
Definition at line 77 of file elementwise_quantized_list.h.
References offset(), and store_quantized().
void arm_compute::cpu::store_quantized_signed | ( | int8_t * | output_ptr, |
const int32x4x4_t & | out | ||
) |
Definition at line 91 of file elementwise_quantized_list.h.
Referenced by elementwise_arithm_op_quantized_signed_broadcast_loop(), elementwise_arithm_op_quantized_singed_loop(), and store_quantized_signed().
void arm_compute::cpu::store_quantized_signed | ( | int8_t * | output_ptr, |
const float32x4x4_t & | rf, | ||
const float32x4_t & | offset, | ||
const float32x4_t & | invscale | ||
) |
Definition at line 98 of file elementwise_quantized_list.h.
References offset(), and store_quantized_signed().
void sub_qasymm8_neon | ( | const ITensor * | src0, |
const ITensor * | src1, | ||
ITensor * | dst, | ||
const ConvertPolicy & | policy, | ||
const Window & | window | ||
) |
Definition at line 35 of file qasymm8.cpp.
References ARM_COMPUTE_UNUSED, arm_compute::test::validation::b, arm_compute::graph::bfs(), Window::broadcast_if_dimension_le_one(), Window::DimX, arm_compute::test::validation::dst, Window::Dimension::end(), arm_compute::execute_window_loop(), ITensor::info(), UniformQuantizationInfo::offset, Iterator::ptr(), ITensorInfo::quantization_info(), arm_compute::quantize_qasymm8(), UniformQuantizationInfo::scale, Window::set(), Window::Dimension::start(), Window::Dimension::step(), ITensorInfo::tensor_shape(), QuantizationInfo::uniform(), arm_compute::wrapper::vcombine(), arm_compute::wrapper::vdup_n(), arm_compute::wrapper::vgethigh(), arm_compute::wrapper::vgetlow(), arm_compute::wrapper::vloadq(), arm_compute::wrapper::vmovl(), arm_compute::wrapper::vreinterpret(), arm_compute::wrapper::vstore(), Dimensions< T >::x(), and Window::x().
void sub_qasymm8_signed_neon | ( | const ITensor * | src0, |
const ITensor * | src1, | ||
ITensor * | dst, | ||
const ConvertPolicy & | policy, | ||
const Window & | window | ||
) |
Definition at line 35 of file qasymm8_signed.cpp.
References ARM_COMPUTE_UNUSED, arm_compute::test::validation::b, arm_compute::graph::bfs(), Window::broadcast_if_dimension_le_one(), Window::DimX, arm_compute::test::validation::dst, Window::Dimension::end(), arm_compute::execute_window_loop(), ITensor::info(), UniformQuantizationInfo::offset, Iterator::ptr(), ITensorInfo::quantization_info(), arm_compute::quantize_qasymm8_signed(), UniformQuantizationInfo::scale, Window::set(), Window::Dimension::start(), Window::Dimension::step(), ITensorInfo::tensor_shape(), QuantizationInfo::uniform(), arm_compute::wrapper::vcombine(), arm_compute::wrapper::vdup_n(), arm_compute::wrapper::vgethigh(), arm_compute::wrapper::vgetlow(), arm_compute::wrapper::vloadq(), arm_compute::wrapper::vmovl(), arm_compute::wrapper::vreinterpret(), arm_compute::wrapper::vstore(), Dimensions< T >::x(), and Window::x().
void sub_qsymm16_neon | ( | const ITensor * | src0, |
const ITensor * | src1, | ||
ITensor * | dst, | ||
const ConvertPolicy & | policy, | ||
const Window & | window | ||
) |
Definition at line 35 of file qsymm16.cpp.
References ARM_COMPUTE_UNUSED, arm_compute::test::validation::b, arm_compute::graph::bfs(), Window::broadcast_if_dimension_le_one(), Window::DimX, arm_compute::test::validation::dst, Window::Dimension::end(), arm_compute::execute_window_loop(), ITensor::info(), Iterator::ptr(), ITensorInfo::quantization_info(), arm_compute::quantize_qsymm16(), UniformQuantizationInfo::scale, Window::set(), Window::Dimension::start(), Window::Dimension::step(), ITensorInfo::tensor_shape(), QuantizationInfo::uniform(), Dimensions< T >::x(), and Window::x().
void sub_s16_u8_s16_neon | ( | const ITensor * | src0, |
const ITensor * | src1, | ||
ITensor * | dst, | ||
const ConvertPolicy & | policy, | ||
const Window & | window | ||
) |
void arm_compute::cpu::sub_same_neon | ( | const ITensor * | src0, |
const ITensor * | src1, | ||
ITensor * | dst, | ||
const ConvertPolicy & | policy, | ||
const Window & | window | ||
) |
SIMD vector tag type.
Definition at line 48 of file list.h.
References Window::broadcast_if_dimension_le_one(), Window::DimX, arm_compute::test::validation::dst, Window::Dimension::end(), arm_compute::execute_window_loop(), ITensor::info(), Iterator::ptr(), arm_compute::SATURATE, Window::set(), Window::Dimension::start(), Window::Dimension::step(), arm_compute::wrapper::sub_sat(), ITensorInfo::tensor_shape(), arm_compute::wrapper::vdup_n(), arm_compute::wrapper::vloadq(), arm_compute::wrapper::vmul(), arm_compute::wrapper::vqsub(), arm_compute::wrapper::vstore(), arm_compute::wrapper::vsub(), Dimensions< T >::x(), and Window::x().
void sub_u8_s16_s16_neon | ( | const ITensor * | src0, |
const ITensor * | src1, | ||
ITensor * | dst, | ||
const ConvertPolicy & | policy, | ||
const Window & | window | ||
) |
void sub_u8_u8_s16_neon | ( | const ITensor * | src0, |
const ITensor * | src1, | ||
ITensor * | dst, | ||
const ConvertPolicy & | policy, | ||
const Window & | window | ||
) |
Definition at line 117 of file integer.cpp.
References Window::broadcast_if_dimension_le_one(), Window::DimX, arm_compute::test::validation::dst, Window::Dimension::end(), arm_compute::execute_window_loop(), ITensor::info(), Iterator::ptr(), Window::set(), Window::Dimension::start(), arm_compute::wrapper::sub_sat(), ITensorInfo::tensor_shape(), arm_compute::wrapper::vload(), arm_compute::wrapper::vmovl(), arm_compute::wrapper::vqsub(), arm_compute::wrapper::vstore(), arm_compute::wrapper::vsub(), arm_compute::WRAP, and Window::x().
void arm_compute::cpu::u8_neon_scale | ( | const ITensor * | src, |
ITensor * | dst, | ||
const ITensor * | offsets, | ||
const ITensor * | dx, | ||
const ITensor * | dy, | ||
InterpolationPolicy | policy, | ||
BorderMode | border_mode, | ||
PixelValue | constant_border_value, | ||
float | sampling_offset, | ||
bool | align_corners, | ||
const Window & | window | ||
) |
Definition at line 265 of file integer.cpp.
References arm_compute::BILINEAR, arm_compute::test::validation::dst, arm_compute::NEAREST_NEIGHBOR, and arm_compute::test::validation::src.
void arm_compute::cpu::u8_sve_scale | ( | const ITensor * | src, |
ITensor * | dst, | ||
const ITensor * | offsets, | ||
const ITensor * | dx, | ||
const ITensor * | dy, | ||
InterpolationPolicy | policy, | ||
BorderMode | border_mode, | ||
PixelValue | constant_border_value, | ||
float | sampling_offset, | ||
bool | align_corners, | ||
const Window & | window | ||
) |
|
inline |
Referenced by poolingMxN_q8_neon_nhwc().
|
inline |
Definition at line 72 of file quantized.h.
|
inline |
Definition at line 78 of file quantized.h.
|
inline |
Definition at line 57 of file quantized.h.
|
inline |
Definition at line 57 of file quantized.h.
|
inline |
Definition at line 63 of file quantized.h.
|
inline |
|
inline |
Definition at line 104 of file quantized.h.
References arm_compute::vquantize().
|
inline |
Definition at line 119 of file quantized.h.
References arm_compute::vquantize_signed().
|
inline |
|
inline |
Definition at line 137 of file quantized.h.
References arm_compute::vquantize().
|
inline |
Definition at line 150 of file quantized.h.
References arm_compute::vquantize_signed().
|
inline |
Definition at line 87 of file quantized.h.
References arm_compute::vquantize(), and arm_compute::vquantize_signed().
|
inline |
Definition at line 87 of file quantized.h.
References arm_compute::vquantize().
|
inline |
Definition at line 94 of file quantized.h.
References arm_compute::vquantize_signed().
constexpr int step = 4 |
Definition at line 35 of file fp32.cpp.
Referenced by arm_compute::adjust_down(), arm_compute::adjust_up(), Window::collapse_if_possible(), CLRangeKernel::configure(), NERangeKernel::configure(), CLRange::configure(), NERange::configure(), arm_compute::convert_window_coord_to_position(), arm_compute::enqueue(), arm_compute::error_on_invalid_subwindow(), arm_compute::error_on_mismatching_windows(), arm_compute::error_on_window_dimensions_gte(), arm_compute::execute_window_loop(), fp32_neon_floor(), Dimensions< size_t >::increment(), arm_compute::test::framework::dataset::make(), arm_compute::num_of_elements_in_range(), arm_compute::test::validation::reference::range(), ClDirectConvolutionKernel::run_op(), Window::set_dimension_step(), WindowIterator< L >::set_step(), Window::shape(), TensorShape::shift_right(), Window::split_window(), NERangeKernel::validate(), CLRangeKernel::validate(), CLRange::validate(), NERange::validate(), and Window::validate().