diff options
Diffstat (limited to 'arm_compute/runtime')
4 files changed, 28 insertions, 25 deletions
diff --git a/arm_compute/runtime/CL/functions/CLGEMMConvolutionLayer.h b/arm_compute/runtime/CL/functions/CLGEMMConvolutionLayer.h index c6ae535713..958e70fca4 100644 --- a/arm_compute/runtime/CL/functions/CLGEMMConvolutionLayer.h +++ b/arm_compute/runtime/CL/functions/CLGEMMConvolutionLayer.h @@ -91,7 +91,7 @@ private: * -# @ref CLGEMMLowpMatrixMultiplyCore (if the data type is QASYMM8) * -# @ref CLGEMMLowpQuantizeDownInt32ToUint8ScaleByFixedPoint (if the data type is QASYMM8) * -# @ref CLArithmeticAdditionKernel (if biases != nullptr and we have a 1x1 convolution with the NHWC data layout) - * -# @ref CLCol2ImKernel (if NCHW data layout) or @ref CLReshapeLayer (if NHWC with QASYMM8) + * -# @ref CLCol2ImKernel (if NCHW data layout) */ class CLGEMMConvolutionLayer : public IFunction { @@ -187,7 +187,6 @@ private: CLCol2ImKernel _col2im_kernel; CLActivationLayer _activationlayer_function; CLArithmeticAdditionKernel _add_bias_kernel; - CLReshapeLayer _reshape_layer; const ICLTensor *_original_weights; @@ -200,6 +199,7 @@ private: bool _append_bias; bool _skip_im2col; + bool _skip_col2im; bool _is_quantized; bool _is_activationlayer_enabled; bool _is_prepared; diff --git a/arm_compute/runtime/CL/functions/CLGEMMLowpOutputStage.h b/arm_compute/runtime/CL/functions/CLGEMMLowpOutputStage.h index 7446ff4b3f..dca00f027e 100644 --- a/arm_compute/runtime/CL/functions/CLGEMMLowpOutputStage.h +++ b/arm_compute/runtime/CL/functions/CLGEMMLowpOutputStage.h @@ -1,5 +1,5 @@ /* - * Copyright (c) 2017 ARM Limited. + * Copyright (c) 2017-2018 ARM Limited. * * SPDX-License-Identifier: MIT * @@ -131,21 +131,24 @@ public: * @param[in] min (Optional) Min value used to saturate down the output result before converting back to QASYMM8 * @param[in] max (Optional) Max value used to saturate up the output result before converting back to QASYMM8, * Along with @p min, this value can be used to implement "rectified linear unit" activation functions + * @param[in] output_3d_depth (Optional) Depth of output in 3D (Defaults to 1) */ - void configure(const ICLTensor *input, const ICLTensor *bias, ICLTensor *output, int result_fixedpoint_multiplier, int result_shift, int result_offset_after_shift, int min = 0, int max = 0); + void configure(const ICLTensor *input, const ICLTensor *bias, ICLTensor *output, int result_fixedpoint_multiplier, int result_shift, int result_offset_after_shift, + int min = 0, int max = 0, unsigned int output_3d_depth = 1); /** Static function to check if given info will lead to a valid configuration of @ref CLGEMMLowpQuantizeDownInt32ToUint8ScaleByFixedPoint * - * @param[in] input Input tensor. It is the output of @ref CLGEMMLowpMatrixMultiplyCore function. Data type supported: S32 - * @param[in] bias Biases tensor. Only shared biases supported and it can be a nullptr if the addition of biases is not required. - * Biases are 1D tensor with dimensions [OFM]. Data type supported: Same as @p input. - * @param[in] output Output tensor. Data type supported: Data type supported: QASYMM8 - * @param[in] min (Optional) Min value used to saturate down the output result before converting back to QASYMM8 - * @param[in] max (Optional) Max value used to saturate up the output result before converting back to QASYMM8, - * Along with @p min, this value can be used to implement "rectified linear unit" activation functions + * @param[in] input Input tensor. It is the output of @ref CLGEMMLowpMatrixMultiplyCore function. Data type supported: S32 + * @param[in] bias Biases tensor. Only shared biases supported and it can be a nullptr if the addition of biases is not required. + * Biases are 1D tensor with dimensions [OFM]. Data type supported: Same as @p input. + * @param[in] output Output tensor. Data type supported: Data type supported: QASYMM8 + * @param[in] min (Optional) Min value used to saturate down the output result before converting back to QASYMM8 + * @param[in] max (Optional) Max value used to saturate up the output result before converting back to QASYMM8, + * Along with @p min, this value can be used to implement "rectified linear unit" activation functions + * @param[in] output_3d_depth (Optional) Depth of output in 3D (Defaults to 1) * * @return a status */ - static Status validate(const ITensorInfo *input, const ITensorInfo *bias, const ITensorInfo *output, int min = 0, int max = 0); + static Status validate(const ITensorInfo *input, const ITensorInfo *bias, const ITensorInfo *output, int min = 0, int max = 0, unsigned int output_3d_depth = 1); }; -} +} // namespace arm_compute #endif /*__ARM_COMPUTE_CLGEMMLOWPOUTPUTSTAGE_H__ */
\ No newline at end of file diff --git a/arm_compute/runtime/NEON/functions/NEGEMMConvolutionLayer.h b/arm_compute/runtime/NEON/functions/NEGEMMConvolutionLayer.h index ae2563cfc7..cc0b2416bd 100644 --- a/arm_compute/runtime/NEON/functions/NEGEMMConvolutionLayer.h +++ b/arm_compute/runtime/NEON/functions/NEGEMMConvolutionLayer.h @@ -85,7 +85,7 @@ private: * -# @ref NEGEMMLowpMatrixMultiplyCore (if the data type is QASYMM8) * -# @ref NEGEMMLowpQuantizeDownInt32ToUint8ScaleByFixedPoint (if the data type is QASYMM8) * -# @ref NEArithmeticAdditionKernel (if biases != nullptr and we have a 1x1 convolution with the NHWC data layout) - * -# @ref NECol2ImKernel or @ref NEReshapeLayer (if NHWC and GEMM3D is not supported) + * -# @ref NECol2ImKernel (if NCHW data layout) * */ class NEGEMMConvolutionLayer : public IFunction diff --git a/arm_compute/runtime/NEON/functions/NEGEMMLowpOutputStage.h b/arm_compute/runtime/NEON/functions/NEGEMMLowpOutputStage.h index d006929996..f38ecdbbd2 100644 --- a/arm_compute/runtime/NEON/functions/NEGEMMLowpOutputStage.h +++ b/arm_compute/runtime/NEON/functions/NEGEMMLowpOutputStage.h @@ -131,24 +131,24 @@ public: * @param[in] min (Optional) Min value used to saturate down the output result before converting back to QASYMM8 * @param[in] max (Optional) Max value used to saturate up the output result before converting back to QASYMM8, * Along with @p min, this value can be used to implement "rectified linear unit" activation functions - * @param[in] gemm_3d_depth (Optional) Depth of GEMM 3D (Defaults to 1) + * @param[in] output_3d_depth (Optional) Depth of output in 3D (Defaults to 1) */ void configure(const ITensor *input, const ITensor *bias, ITensor *output, int result_fixedpoint_multiplier, int result_shift, int result_offset_after_shift, - int min = 0, int max = 0, unsigned int gemm_3d_depth = 1); + int min = 0, int max = 0, unsigned int output_3d_depth = 1); /** Static function to check if given info will lead to a valid configuration of @ref NEGEMMLowpQuantizeDownInt32ToUint8ScaleByFixedPoint * - * @param[in] input Input tensor. It is the output of @ref NEGEMMLowpMatrixMultiplyCore function. Data type supported: S32 - * @param[in] bias Biases tensor. Only shared biases supported and it can be a nullptr if the addition of biases is not required. - * Biases are 1D tensor with dimensions [OFM]. Data type supported: Same as @p input. - * @param[in] output Output tensor. Data type supported: Data type supported: QASYMM8 - * @param[in] min (Optional) Min value used to saturate down the output result before converting back to QASYMM8 - * @param[in] max (Optional) Max value used to saturate up the output result before converting back to QASYMM8, - * Along with @p min, this value can be used to implement "rectified linear unit" activation functions - * @param[in] gemm_3d_depth (Optional) Depth of GEMM 3D (Defaults to 1) + * @param[in] input Input tensor. It is the output of @ref NEGEMMLowpMatrixMultiplyCore function. Data type supported: S32 + * @param[in] bias Biases tensor. Only shared biases supported and it can be a nullptr if the addition of biases is not required. + * Biases are 1D tensor with dimensions [OFM]. Data type supported: Same as @p input. + * @param[in] output Output tensor. Data type supported: Data type supported: QASYMM8 + * @param[in] min (Optional) Min value used to saturate down the output result before converting back to QASYMM8 + * @param[in] max (Optional) Max value used to saturate up the output result before converting back to QASYMM8, + * Along with @p min, this value can be used to implement "rectified linear unit" activation functions + * @param[in] output_3d_depth (Optional) Depth of output in 3D (Defaults to 1) * * @return a status */ - static Status validate(const ITensorInfo *input, const ITensorInfo *bias, const ITensorInfo *output, int min = 0, int max = 0, unsigned int gemm_3d_depth = 1); + static Status validate(const ITensorInfo *input, const ITensorInfo *bias, const ITensorInfo *output, int min = 0, int max = 0, unsigned int output_3d_depth = 1); }; } // namespace arm_compute #endif /*__ARM_COMPUTE_NEGEMMLOWPOUTPUTSTAGE_H__ */
\ No newline at end of file |