diff options
author | Felix Thomasmathibalan <felixjohnny.thomasmathibalan@arm.com> | 2023-09-27 17:46:17 +0100 |
---|---|---|
committer | felixjohnny.thomasmathibalan <felixjohnny.thomasmathibalan@arm.com> | 2023-09-28 12:08:05 +0000 |
commit | afd38f0c617d6f89b2b4532c6c44f116617e2b6f (patch) | |
tree | 03bc7d5a762099989b16a656fa8d397b490ed70e /src/core/NEON/kernels/assembly/depthwise_common.hpp | |
parent | bdcb4c148ee2fdeaaddf4cf1e57bbb0de02bb894 (diff) | |
download | ComputeLibrary-afd38f0c617d6f89b2b4532c6c44f116617e2b6f.tar.gz |
Apply clang-format on repository
Code is formatted as per a revised clang format configuration
file(not part of this delivery). Version 14.0.6 is used.
Exclusion List:
- files with .cl extension
- files that are not strictly C/C++ (e.g. Android.bp, Sconscript ...)
And the following directories
- compute_kernel_writer/validation/
- tests/
- include/
- src/core/NEON/kernels/convolution/
- src/core/NEON/kernels/arm_gemm/
- src/core/NEON/kernels/arm_conv/
- data/
There will be a follow up for formatting of .cl files and the
files under tests/ and compute_kernel_writer/validation/.
Signed-off-by: Felix Thomasmathibalan <felixjohnny.thomasmathibalan@arm.com>
Change-Id: Ib7eb1fcf4e7537b9feaefcfc15098a804a3fde0a
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/10391
Benchmark: Arm Jenkins <bsgcomp@arm.com>
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: Gunes Bayir <gunes.bayir@arm.com>
Diffstat (limited to 'src/core/NEON/kernels/assembly/depthwise_common.hpp')
-rw-r--r-- | src/core/NEON/kernels/assembly/depthwise_common.hpp | 106 |
1 files changed, 48 insertions, 58 deletions
diff --git a/src/core/NEON/kernels/assembly/depthwise_common.hpp b/src/core/NEON/kernels/assembly/depthwise_common.hpp index a5db793b3d..5ff848e281 100644 --- a/src/core/NEON/kernels/assembly/depthwise_common.hpp +++ b/src/core/NEON/kernels/assembly/depthwise_common.hpp @@ -49,11 +49,7 @@ struct KernelDescription bool is_default = false; uint64_t cycle_estimate = 0; - KernelDescription( - DepthwiseMethod method, - std::string name, - bool is_default, - uint64_t cycle_estimate) + KernelDescription(DepthwiseMethod method, std::string name, bool is_default, uint64_t cycle_estimate) : method(method), name(name), is_default(is_default), cycle_estimate(cycle_estimate) { } @@ -78,58 +74,51 @@ public: // pointer the bias vector (which may be nullptr in the case of no bias) and // a pointer to the array of weights (stored in HWIO order). virtual void pack_parameters( - void *buffer, - const void *biases, - const void *weights, - size_t ld_weight_col = 0, - size_t ld_weight_row = 0) = 0; + void *buffer, const void *biases, const void *weights, size_t ld_weight_col = 0, size_t ld_weight_row = 0) = 0; // Determine the amount of working space required virtual size_t get_working_size(unsigned int n_threads) const = 0; // Execute the convolution over the specified area of memory. - virtual void execute( - const void *input, // Pointer to input tensor - const void *parameters, // Packed parameters buffer - void *output, - void *working_space, - unsigned int thread_id, - unsigned int n_threads) const = 0; - - virtual void execute( - const void *input, - size_t ld_input_col, - size_t ld_input_row, - size_t ld_input_batch, - const void *parameters, - void *output, - size_t ld_output_col, - size_t ld_output_row, - size_t ld_output_batch, - void *working_space, - unsigned int thread_id, - unsigned int n_threads) const = 0; - - virtual void execute( - unsigned int batches, - unsigned int input_height, - unsigned int input_width, - unsigned int channels, - const PaddingValues &, - const void *input, - size_t ld_input_col, - size_t ld_input_row, - size_t ld_input_batch, - const void *parameters, - unsigned int output_height, - unsigned int output_width, - void *output, - size_t ld_output_col, - size_t ld_output_row, - size_t ld_output_batch, - void *working_space, - unsigned int thread_id, - unsigned int n_threads) const = 0; + virtual void execute(const void *input, // Pointer to input tensor + const void *parameters, // Packed parameters buffer + void *output, + void *working_space, + unsigned int thread_id, + unsigned int n_threads) const = 0; + + virtual void execute(const void *input, + size_t ld_input_col, + size_t ld_input_row, + size_t ld_input_batch, + const void *parameters, + void *output, + size_t ld_output_col, + size_t ld_output_row, + size_t ld_output_batch, + void *working_space, + unsigned int thread_id, + unsigned int n_threads) const = 0; + + virtual void execute(unsigned int batches, + unsigned int input_height, + unsigned int input_width, + unsigned int channels, + const PaddingValues &, + const void *input, + size_t ld_input_col, + size_t ld_input_row, + size_t ld_input_batch, + const void *parameters, + unsigned int output_height, + unsigned int output_width, + void *output, + size_t ld_output_col, + size_t ld_output_row, + size_t ld_output_batch, + void *working_space, + unsigned int thread_id, + unsigned int n_threads) const = 0; }; // To handle a dilation factor of D execute the kernel once for each d in @@ -145,12 +134,13 @@ public: // - Number of valid input pixels corresponding to `d` // - Offset of the first pixel corresponding to `d` // - Amount of padding in the view for `d` -std::tuple<size_t, size_t, size_t, size_t, size_t> -get_reduced_view_for_dilation( - size_t out_size, size_t in_size, - size_t d, size_t dilation_factor, - size_t kernel_size, size_t stride, - size_t pad_before); +std::tuple<size_t, size_t, size_t, size_t, size_t> get_reduced_view_for_dilation(size_t out_size, + size_t in_size, + size_t d, + size_t dilation_factor, + size_t kernel_size, + size_t stride, + size_t pad_before); } // namespace depthwise } // namespace arm_conv |