| /external/ComputeLibrary/src/cpu/kernels/crop/generic/neon/ |
| D | impl.cpp | 37 …int32_t window_step_x, int32_t output_width_start, int32_t output_width_limit, bool input_has_sing… in in_bounds_crop_window() argument
47 negative_offset.set(1, negative_offset[1] - window_step_x + 1); in in_bounds_crop_window()
48 …for(; x <= output_width_limit - window_step_x; x += window_step_x, negative_offset[1] -= window_st… in in_bounds_crop_window()
57 input_offset[1] = negative_offset[1] + window_step_x - 1; in in_bounds_crop_window()
69 …st<int32_t>(input->info()->dimension(0)) - window_step_x; c += window_step_x, input_offset[0] += w… in in_bounds_crop_window()
95 for(; x <= limit - window_step_x; x += window_step_x, input_offset[0] += window_step_x) in in_bounds_crop_window()
110 …int32_t window_step_x, int32_t output_width_start, int32_t output_width_limit, bool input_has_sing…
113 …int32_t window_step_x, int32_t output_width_start, int32_t output_width_limit, bool input_has_sing…
115 …int32_t window_step_x, int32_t output_width_start, int32_t output_width_limit, bool input_has_sing…
117 …int32_t window_step_x, int32_t output_width_start, int32_t output_width_limit, bool input_has_sing…
[all …]
|
| D | integer.cpp | 33 …int32_t window_step_x, int32_t output_width_start, int32_t output_width_limit, bool input_has_sing… in u8_in_bounds_crop_window() argument
36 …window_step_x, output_width_start, output_width_limit, input_has_single_channel, is_width_flipped); in u8_in_bounds_crop_window()
40 …int32_t window_step_x, int32_t output_width_start, int32_t output_width_limit, bool input_has_sing… in u16_in_bounds_crop_window() argument
43 …window_step_x, output_width_start, output_width_limit, input_has_single_channel, is_width_flipped); in u16_in_bounds_crop_window()
47 …int32_t window_step_x, int32_t output_width_start, int32_t output_width_limit, bool input_has_sing… in u32_in_bounds_crop_window() argument
50 …window_step_x, output_width_start, output_width_limit, input_has_single_channel, is_width_flipped); in u32_in_bounds_crop_window()
54 …int32_t window_step_x, int32_t output_width_start, int32_t output_width_limit, bool input_has_sing… in s8_in_bounds_crop_window() argument
57 …window_step_x, output_width_start, output_width_limit, input_has_single_channel, is_width_flipped); in s8_in_bounds_crop_window()
61 …int32_t window_step_x, int32_t output_width_start, int32_t output_width_limit, bool input_has_sing… in s16_in_bounds_crop_window() argument
64 …window_step_x, output_width_start, output_width_limit, input_has_single_channel, is_width_flipped); in s16_in_bounds_crop_window()
[all …]
|
| /external/ComputeLibrary/src/cpu/kernels/cast/generic/neon/ |
| D | fp16.cpp | 42 const int window_step_x = 16; in neon_qasymm8_signed_to_fp16_cast() local
60 for(; x <= (window_end_x - window_step_x); x += window_step_x) in neon_qasymm8_signed_to_fp16_cast()
91 const int window_step_x = 16; in neon_s32_to_fp16_cast() local
110 for(; x <= (window_end_x - window_step_x); x += window_step_x) in neon_s32_to_fp16_cast()
142 const int window_step_x = 16; in neon_fp32_to_fp16_cast() local
161 for(; x <= (window_end_x - window_step_x); x += window_step_x) in neon_fp32_to_fp16_cast()
193 const int window_step_x = 16; in neon_fp16_to_other_dt_cast() local
216 for(; x <= (window_end_x - window_step_x); x += window_step_x) in neon_fp16_to_other_dt_cast()
248 for(; x <= (window_end_x - window_step_x); x += window_step_x) in neon_fp16_to_other_dt_cast()
280 for(; x <= (window_end_x - window_step_x); x += window_step_x) in neon_fp16_to_other_dt_cast()
[all …]
|
| D | bfloat16.cpp | 43 const int window_step_x = 16; in neon_fp32_to_bfloat16_cast() local
63 for(; x <= (window_end_x - window_step_x); x += window_step_x) in neon_fp32_to_bfloat16_cast()
86 const int window_step_x = 16; in neon_bfloat16_to_fp32_cast() local
109 for(; x <= (window_end_x - window_step_x); x += window_step_x) in neon_bfloat16_to_fp32_cast()
|
| /external/ComputeLibrary/src/cpu/kernels/ |
| D | CpuCastKernel.cpp | 186 const int window_step_x = 16; in run_op() local
219 for(; x <= (window_end_x - window_step_x); x += window_step_x) in run_op()
253 for(; x <= (window_end_x - window_step_x); x += window_step_x) in run_op()
289 for(; x <= (window_end_x - window_step_x); x += window_step_x) in run_op()
342 for(; x <= (window_end_x - window_step_x); x += window_step_x) in run_op()
376 for(; x <= (window_end_x - window_step_x); x += window_step_x) in run_op()
412 for(; x <= (window_end_x - window_step_x); x += window_step_x) in run_op()
454 for(; x <= (window_end_x - window_step_x); x += window_step_x) in run_op()
499 for(; x <= (window_end_x - window_step_x); x += window_step_x) in run_op()
528 for(; x <= (window_end_x - window_step_x); x += window_step_x) in run_op()
[all …]
|
| D | CpuDequantizeKernel.cpp | 115 const int window_step_x = 16; in run_dequantization_qasymm8() local
133 for(; x <= (window_end_x - window_step_x); x += window_step_x) in run_dequantization_qasymm8()
156 const int window_step_x = 16; in run_dequantization_qsymm8_per_channel_nchw() local
174 for(; x <= (window_end_x - window_step_x); x += window_step_x) in run_dequantization_qsymm8_per_channel_nchw()
197 const int window_step_x = 16; in run_dequantization_qsymm8_per_channel_nhwc() local
215 for(; x <= (window_end_x - window_step_x); x += window_step_x) in run_dequantization_qsymm8_per_channel_nhwc()
248 const int window_step_x = 16; in run_dequantization_qsymm8() local
266 for(; x <= (window_end_x - window_step_x); x += window_step_x) in run_dequantization_qsymm8()
290 const int window_step_x = 8; in run_dequantization_qsymm16() local
308 for(; x <= (window_end_x - window_step_x); x += window_step_x) in run_dequantization_qsymm16()
|
| D | CpuConcatenateWidthKernel.cpp | 99 constexpr int window_step_x = 16; in run_op() local
115 for(; x <= (window_end_x - window_step_x); x += window_step_x) in run_op()
133 for(; x <= (window_end_x - window_step_x); x += window_step_x) in run_op()
154 for(; x <= (window_end_x - window_step_x); x += window_step_x) in run_op()
|
| D | CpuConcatenateHeightKernel.cpp | 98 const int window_step_x = 16; in run_op() local
116 for(; x <= (window_end_x - window_step_x); x += window_step_x) in run_op()
135 for(; x <= (window_end_x - window_step_x); x += window_step_x) in run_op()
157 for(; x <= (window_end_x - window_step_x); x += window_step_x) in run_op()
|
| D | CpuConcatenateDepthKernel.cpp | 60 const int window_step_x = 16 / dst->info()->element_size(); in depth_concat() local
79 for(; x <= (window_end_x - window_step_x); x += window_step_x) in depth_concat()
99 for(; x <= (window_end_x - window_step_x); x += window_step_x) in depth_concat()
119 for(; x <= (window_end_x - window_step_x); x += window_step_x) in depth_concat()
|
| D | CpuConcatenateBatchKernel.cpp | 57 const int window_step_x = 16 / dst->info()->element_size(); in batch_concat() local
77 for(; x <= (window_end_x - window_step_x); x += window_step_x) in batch_concat()
97 for(; x <= (window_end_x - window_step_x); x += window_step_x) in batch_concat()
117 for(; x <= (window_end_x - window_step_x); x += window_step_x) in batch_concat()
|
| D | CpuMulKernel.cpp | 165 const int window_step_x = 16 / sizeof(T); in mul_saturate_quantized_8() local
190 using ExactTagType = typename wrapper::traits::neon_vector<T, window_step_x>::tag_type; in mul_saturate_quantized_8()
203 for(; x <= (window_end_x - window_step_x); x += window_step_x) in mul_saturate_quantized_8()
262 for(; x <= (window_end_x - window_step_x); x += window_step_x) in mul_saturate_quantized_8()
347 constexpr int window_step_x = 16; in mul_q8_neon_fixedpoint() local
410 for(; x <= (window_end_x - window_step_x); x += window_step_x) in mul_q8_neon_fixedpoint()
492 for(; x <= (window_end_x - window_step_x); x += window_step_x) in mul_q8_neon_fixedpoint()
579 const int window_step_x = 16; in mul_saturate_QSYMM16_QSYMM16_QSYMM16() local
594 for(; x <= (window_end_x - window_step_x); x += window_step_x) in mul_saturate_QSYMM16_QSYMM16_QSYMM16()
663 const int window_step_x = 16; in mul_QSYMM16_QSYMM16_S32() local
[all …]
|
| D | CpuDirectConv2dOutputStageKernel.cpp | 110 const int window_step_x = 16 / src->info()->element_size(); in output_stage_nchw() local
119 for(; x <= (window_end_x - window_step_x); x += window_step_x) in output_stage_nchw()
174 const int window_step_x = 16 / src->info()->element_size(); in output_stage_nhwc() local
185 for(; x <= (window_end_x - window_step_x); x += window_step_x) in output_stage_nhwc()
238 const int window_step_x = 16 / src->info()->element_size(); in output_stage_nchw() local
249 for(; x <= (window_end_x - window_step_x); x += window_step_x) in output_stage_nchw()
324 const int window_step_x = 16 / src->info()->element_size(); in output_stage_nhwc() local
335 for(; x <= (window_end_x - window_step_x); x += window_step_x) in output_stage_nhwc()
|
| D | CpuGemmLowpOffsetContributionOutputStageKernel.cpp | 313 …int window_step_x, int window_start_x, int window_end_x, bool has_a_offset, bool has_b_offset, boo… in run_offset_contribution_output_stage_window() argument
331 for(; x <= (window_end_x - window_step_x); x += window_step_x) in run_offset_contribution_output_stage_window()
404 …int window_step_x, int window_start_x, int window_end_x, bool has_a_offset, bool has_bias, bool is… in run_offset_contribution_output_stage_window_symm() argument
414 for(; x <= (window_end_x - window_step_x); x += window_step_x) in run_offset_contribution_output_stage_window_symm()
507 const int window_step_x = 16; in run_offset_contribution_output_stage() local
547 … window_step_x, window_start_x, window_end_x, true, true, true, is_bounded_relu, is_fixed_point); in run_offset_contribution_output_stage()
563 … window_step_x, window_start_x, window_end_x, true, true, false, is_bounded_relu, is_fixed_point); in run_offset_contribution_output_stage()
589 … window_step_x, window_start_x, window_end_x, false, true, true, is_bounded_relu, is_fixed_point); in run_offset_contribution_output_stage()
604 … window_step_x, window_start_x, window_end_x, false, true, false, is_bounded_relu, is_fixed_point); in run_offset_contribution_output_stage()
630 … window_step_x, window_start_x, window_end_x, true, false, true, is_bounded_relu, is_fixed_point); in run_offset_contribution_output_stage()
[all …]
|
| D | CpuGemmLowpQuantizeDownInt32ToInt16ScaleByFixedPointKernel.cpp | 82 const int window_step_x = 8; in run_internal() local
102 for(; x <= (window_end_x - window_step_x); x += window_step_x) in run_internal()
148 for(; x <= (window_end_x - window_step_x); x += window_step_x) in run_internal()
|
| /external/ComputeLibrary/src/cpu/kernels/elementwise_binary/generic/neon/ |
| D | impl.h | 97 const int window_step_x = std::min(16 / static_cast<int>(sizeof(OutputScalarType)), 8); in elementwise_op() local
123 …int x = (*broadcast_func)(window_start_x, window_end_x, window_step_x, non_broadcast_input_ptr, br… in elementwise_op()
148 …int x = (*neon_func)(window_start_x, window_end_x, window_step_x, input1_ptr, input2_ptr, output_p… in elementwise_op()
240 inline int elementwise_arithm_op_loop(int window_start_x, int window_end_x, int window_step_x, in elementwise_arithm_op_loop() argument
244 for(; x <= (window_end_x - window_step_x); x += window_step_x) in elementwise_arithm_op_loop()
254 …e int elementwise_arithm_op_broadcast_loop(int window_start_x, int window_end_x, int window_step_x, in elementwise_arithm_op_broadcast_loop() argument
258 for(; x <= (window_end_x - window_step_x); x += window_step_x) in elementwise_arithm_op_broadcast_loop()
348 …e int elementwise_comp_op_broadcast_8_loop(int window_start_x, int window_end_x, int window_step_x, in elementwise_comp_op_broadcast_8_loop() argument
352 for(; x <= (window_end_x - window_step_x); x += window_step_x) in elementwise_comp_op_broadcast_8_loop()
361 … int elementwise_comp_op_broadcast_16_loop(int window_start_x, int window_end_x, int window_step_x, in elementwise_comp_op_broadcast_16_loop() argument
[all …]
|
| /external/ComputeLibrary/src/cpu/kernels/meanstddevnorm/generic/neon/ |
| D | impl.cpp | 41 const int window_step_x = size; in mean_stddev_normalization() local
57 for(; x <= (window_end_x - window_step_x); x += window_step_x) in mean_stddev_normalization()
89 for(x = window_start_x; x <= (window_end_x - window_step_x); x += window_step_x) in mean_stddev_normalization()
113 const int window_step_x = 8; in mean_stddev_normalization() local
129 for(; x <= (window_end_x - window_step_x); x += window_step_x) in mean_stddev_normalization()
165 for(x = window_start_x; x <= (window_end_x - window_step_x); x += window_step_x) in mean_stddev_normalization()
|
| D | qasymm8.cpp | 56 const int window_step_x = 16; in neon_qasymm8_meanstddevnorm() local
81 for(; x <= (window_end_x - window_step_x); x += window_step_x) in neon_qasymm8_meanstddevnorm()
120 for(x = window_start_x; x <= (window_end_x - window_step_x); x += window_step_x) in neon_qasymm8_meanstddevnorm()
|
| /external/ComputeLibrary/src/cpu/kernels/select/generic/neon/ |
| D | impl.cpp | 39 …const int window_step_x, const int window_start_x, const int window_end_x, const int limit, Vector… in select_op() argument
57 for(; x <= limit; x += window_step_x) in select_op()
78 const auto window_step_x = 16 / sizeof(ScalarType); in select_op_8() local
82 …ectorType>(cond, in1, in2, out, window, window_step_x, window_start_x, window_end_x, window_end_x … in select_op_8()
92 const auto window_step_x = 16 / sizeof(ScalarType); in select_op_16() local
96 …ectorType>(cond, in1, in2, out, window, window_step_x, window_start_x, window_end_x, window_end_x … in select_op_16()
106 const auto window_step_x = 16 / sizeof(ScalarType); in select_op_32() local
110 …ectorType>(cond, in1, in2, out, window, window_step_x, window_start_x, window_end_x, window_end_x … in select_op_32()
|
| /external/ComputeLibrary/src/cpu/kernels/gemm_matrix_add/generic/neon/ |
| D | impl.cpp | 37 constexpr int window_step_x = 16; in matrix_addition_f32() local
53 for(; x < (window_end_x - window_step_x); x += window_step_x) in matrix_addition_f32()
82 constexpr int window_step_x = 16; in matrix_addition_f16() local
98 for(; x < (window_end_x - window_step_x); x += window_step_x) in matrix_addition_f16()
|
| /external/ComputeLibrary/src/cpu/kernels/pool3d/neon/ |
| D | impl.cpp | 40 … const int window_start_x, const int window_end_x, const int window_step_x) in max_poolingMxNxD_fp_neon_ndhwc() argument
97 for(; x_off <= (window_end_x - window_step_x); x_off += window_step_x) // C in max_poolingMxNxD_fp_neon_ndhwc()
146 …onst Window &window_out, const int window_start_x, const int window_end_x, const int window_step_x) in avg_poolingMxNxD_fp_neon_ndhwc() argument
215 for(; x_off <= (window_end_x - window_step_x); x_off += window_step_x) // C in avg_poolingMxNxD_fp_neon_ndhwc()
273 …onst Window &window_out, const int window_start_x, const int window_end_x, const int window_step_x) in l2_poolingMxNxD_fp_neon_ndhwc() argument
341 for(; x_off <= (window_end_x - window_step_x); x_off += window_step_x) // C in l2_poolingMxNxD_fp_neon_ndhwc()
411 constexpr int window_step_x = 16 / sizeof(T); in poolingMxNxD_fp_neon_ndhwc() local
420 …xD_fp_neon_ndhwc<T>(src, dst0, pool_info, window_out, window_start_x, window_end_x, window_step_x); in poolingMxNxD_fp_neon_ndhwc()
423 …xD_fp_neon_ndhwc<T>(src, dst0, pool_info, window_out, window_start_x, window_end_x, window_step_x); in poolingMxNxD_fp_neon_ndhwc()
426 …xD_fp_neon_ndhwc<T>(src, dst0, pool_info, window_out, window_start_x, window_end_x, window_step_x); in poolingMxNxD_fp_neon_ndhwc()
[all …]
|
| /external/ComputeLibrary/src/cpu/kernels/l2normlayer/generic/neon/ |
| D | impl.cpp | 42 const int window_step_x = 16 / data_size_from_type(in->info()->data_type()); in l2_normalize_x() local
64 for(; x <= (window_end_x - window_step_x); x += window_step_x) in l2_normalize_x()
83 const int window_step_x = 16 / data_size_from_type(in->info()->data_type()); in l2_normalize_yz() local
107 for(; x <= (window_end_x - window_step_x); x += window_step_x) in l2_normalize_yz()
|
| /external/ComputeLibrary/src/core/NEON/kernels/ |
| D | NECropKernel.cpp | 123 … int32_t window_step_x, int32_t output_width_start, int32_t output_width_limit) in out_of_bounds_crop_window() argument
129 for(; x <= limit - window_step_x; x += window_step_x) in out_of_bounds_crop_window()
144 const int window_step_x = 16 / sizeof(float); in execute_window() local
162 …out_of_bounds_crop_window(output, output_ptr, extrapolation_value, window_step_x, 0, rows_out_of_b… in execute_window()
172 …out_of_bounds_crop_window(output, output_ptr, extrapolation_value, window_step_x, 0, cols_out_of_b… in execute_window()
177 …(*in_bounds_crop_function)(input, output, output_ptr, input_offset, window_step_x, cols_out_of_bou… in execute_window()
183 …out_of_bounds_crop_window(output, output_ptr, extrapolation_value, window_step_x, output->info()->… in execute_window()
188 …out_of_bounds_crop_window(output, output_ptr, extrapolation_value, window_step_x, 0, rows_out_of_b… in execute_window()
|
| /external/ComputeLibrary/src/cpu/kernels/range/generic/neon/ |
| D | impl.cpp | 47 const int window_step_x = 16 / sizeof(T); in neon_range_function() local
57 for(; x <= (window_end_x - window_step_x); x += window_step_x) in neon_range_function()
59 for(int count = 0; count < window_step_x; ++count) in neon_range_function()
|
| /external/ComputeLibrary/src/cpu/kernels/add/generic/neon/ |
| D | impl.cpp | 47 constexpr int window_step_x = 16 / sizeof(ScalarType); in add_same_neon() local
78 for(; x <= (window_end_x - window_step_x); x += window_step_x) in add_same_neon()
113 for(; x <= (window_end_x - window_step_x); x += window_step_x) in add_same_neon()
196 constexpr int window_step_x = 16; in add_sub_q8_neon_fixedpoint() local
259 for(; x <= (window_end_x - window_step_x); x += window_step_x) in add_sub_q8_neon_fixedpoint()
327 for(; x <= (window_end_x - window_step_x); x += window_step_x) in add_sub_q8_neon_fixedpoint()
394 constexpr int window_step_x = 16; in add_sub_qasymm8_neon() local
438 for(; x <= (window_end_x - window_step_x); x += window_step_x) in add_sub_qasymm8_neon()
508 for(; x <= (window_end_x - window_step_x); x += window_step_x) in add_sub_qasymm8_neon()
577 constexpr int window_step_x = 16; in add_sub_qasymm8_signed_neon() local
[all …]
|
| /external/ComputeLibrary/src/cpu/kernels/sub/neon/ |
| D | list.h | 62 constexpr int window_step_x = 16 / sizeof(T); in sub_same_neon() local
97 for(; x <= (window_end_x - window_step_x); x += window_step_x) in sub_same_neon()
142 for(; x <= (window_end_x - window_step_x); x += window_step_x) in sub_same_neon()
|