diff options
Diffstat (limited to 'arm_compute/runtime/CL/functions/CLGEMMLowpOutputStage.h')
-rw-r--r-- | arm_compute/runtime/CL/functions/CLGEMMLowpOutputStage.h | 51 |
1 files changed, 41 insertions, 10 deletions
diff --git a/arm_compute/runtime/CL/functions/CLGEMMLowpOutputStage.h b/arm_compute/runtime/CL/functions/CLGEMMLowpOutputStage.h index 564135eed8..b6619da5d2 100644 --- a/arm_compute/runtime/CL/functions/CLGEMMLowpOutputStage.h +++ b/arm_compute/runtime/CL/functions/CLGEMMLowpOutputStage.h @@ -64,7 +64,7 @@ public: * @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[out] output Output tensor. Data type supported: Data type supported: QASYMM8 + * @param[out] output Output tensor. Data type supported: QASYMM8 * @param[in] result_offset Offset to be added to each element of the input matrix * @param[in] result_mult_int Value to be multiplied to each element of the input matrix when once the result_offset has been add * @param[in] result_shift Number of bits to shift right the result before converting back to QASYMM8 @@ -79,7 +79,7 @@ public: * @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] output Output tensor. Data type supported: QASYMM8 * @param[in] min (Optional) Min value used to saturate down the output result before converting back to QASYMM8. Defaults to the minimum possible 32-bit signed integer. * @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. Defaults to the maximum possible 32-bit signed integer. @@ -125,7 +125,7 @@ public: * @param[in] input Input tensor. Data type supported: S32 * @param[in] bias Biases tensor. Only shared biases supported and it can be a nullptr if the biases addition is not required. * Biases are 1D tensor with dimensions [OFM]. Data type supported: Same as @p input. - * @param[out] output Output tensor. Data type supported: Data type supported: QASYMM8 + * @param[out] output Output tensor. Data type supported: QASYMM8 * @param[in] result_fixedpoint_multiplier Fixed point value to be multiplied to each element of the input matrix when once the result_offset has been add * @param[in] result_shift Number of bits to shift right the result after the fixed point multiplication * @param[in] result_offset_after_shift Offset to be applied to result before converting it back to QASYMM8 @@ -140,7 +140,7 @@ public: * @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] output Output tensor. Data type supported: QASYMM8 * @param[in] min (Optional) Min value used to saturate down the output result before converting back to QASYMM8. Defaults to the minimum possible 32-bit signed integer. * @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. Defaults to the maximum possible 32-bit signed integer. @@ -186,7 +186,7 @@ public: * @param[in] input Input tensor. Data type supported: S32 * @param[in] bias Biases tensor. Only shared biases supported and it can be a nullptr if the biases addition is not required. * Biases are 1D tensor with dimensions [OFM]. Data type supported: Same as @p input. - * @param[out] output Output tensor. Data type supported: Data type supported: QASYMM8_SIGNED + * @param[out] output Output tensor. Data type supported: QASYMM8_SIGNED * @param[in] result_fixedpoint_multiplier Fixed point value to be multiplied to each element of the input matrix when once the result_offset has been add * @param[in] result_shift Number of bits to shift right the result after the fixed point multiplication * @param[in] result_offset_after_shift Offset to be applied to result before converting it back to QASYMM8_SIGNED @@ -201,7 +201,7 @@ public: * @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_SIGNED + * @param[in] output Output tensor. Data type supported: QASYMM8_SIGNED * @param[in] min (Optional) Min value used to saturate down the output result before converting back to QASYMM8_SIGNED. Defaults to the minimum possible 32-bit signed integer. * @param[in] max (Optional) Max value used to saturate up the output result before converting back to QASYMM8_SIGNED. Defaults to 0 * Along with @p min, this value can be used to implement "rectified linear unit" activation functions. Defaults to the maximum possible 32-bit signed integer. @@ -228,7 +228,7 @@ public: * @param[in] input Input tensor. Data type supported: S32 * @param[in] bias Biases tensor. Only shared biases supported and it can be a nullptr if the biases addition is not required. * Biases are 1D tensor with dimensions [OFM]. Data type supported: Same as @p input. - * @param[out] output Output tensor. Data type supported: Data type supported: QASYMM8 + * @param[out] output Output tensor. Data type supported: QASYMM8 * @param[in] multiplier Float multiplier to be multiplied to each element of the input matrix * @param[in] offset Offset to be applied to result before converting it back to QASYMM8 * @param[in] min (Optional) Min value used to saturate down the output result before converting back to QASYMM8. Defaults to the minimum possible 32-bit signed integer. @@ -242,7 +242,7 @@ public: * @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] output Output tensor. Data type supported: QASYMM8 * @param[in] min (Optional) Min value used to saturate down the output result before converting back to QASYMM8. Defaults to the minimum possible 32-bit signed integer. * @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. Defaults to the maximum possible 32-bit signed integer. @@ -287,7 +287,7 @@ public: * @param[in] input Input tensor. Data type supported: S32 * @param[in] bias Biases tensor. Only shared biases supported and it can be a nullptr if the biases addition is not required. * Biases are 1D tensor with dimensions [OFM]. Data type supported: Same as @p input. - * @param[out] output Output tensor. Data type supported: Data type supported: QSYMM16 + * @param[out] output Output tensor. Data type supported: QSYMM16 * @param[in] result_fixedpoint_multiplier Fixed point value to be multiplied to each element of the input matrix when once the result_offset has been add * @param[in] result_shift Number of bits to shift right the result after the fixed point multiplication * @param[in] min (Optional) Min value used to saturate down the output result before converting back to QSYMM16. Defaults to the minimum possible 32-bit signed integer. @@ -301,7 +301,7 @@ public: * @param[in] input Input tensor info. It is the output of @ref CLGEMMLowpMatrixMultiplyCore function. Data type supported: S32 * @param[in] bias Biases tensor info. 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 info. Data type supported: Data type supported: QSYMM16 + * @param[in] output Output tensor info. Data type supported: QSYMM16 * @param[in] min (Optional) Min value used to saturate down the output result before converting back to QSYMM16. Defaults to the minimum possible 32-bit signed integer. * @param[in] max (Optional) Max value used to saturate up the output result before converting back to QSYMM16, * Along with @p min, this value can be used to implement "rectified linear unit" activation functions. Defaults to the maximum possible 32-bit signed integer. @@ -310,5 +310,36 @@ public: */ static Status validate(const ITensorInfo *input, const ITensorInfo *bias, const ITensorInfo *output, int min = std::numeric_limits<int32_t>::lowest(), int max = std::numeric_limits<int32_t>::max()); }; +/** Basic function to execute GEMMLowpQuantizeDown kernels on CL. + * + * This function calls the following CL kernels: + * + * -# @ref CLGEMMLowpQuantizeDownInt32ToUint8ScaleByFixedPointKernel + * -# @ref CLGEMMLowpQuantizeDownInt32ToInt8ScaleByFixedPointKernel +*/ +class CLGEMMLowpOutputStage : public ICLSimpleFunction +{ +public: + /** Initialise the kernel's inputs, output + * + * @param[in] input Input tensor. Data type supported: S32 + * @param[in] bias Biases tensor. Only shared biases supported and it can be a nullptr if the biases addition is not required. + * Biases are 1D tensor with dimensions [OFM]. Data type supported: Same as @p input. + * @param[out] output Output tensor. Data type supported: QASYMM8/QASYMM8_SIGNED + * @param[in] info GEMMLowp output stage metadata. + */ + void configure(const ICLTensor *input, const ICLTensor *bias, ICLTensor *output, const GEMMLowpOutputStageInfo &info); + /** Static function to check if given info will lead to a valid configuration of @ref CLGEMMLowpQuantizeDownInt32ToUint8ScaleByFixedPointKernel + * + * @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: QASYMM8/QASYMM8_SIGNED + * @param[in] info GEMMLowp output stage metadata. + * + * @return a status + */ + static Status validate(const ITensorInfo *input, const ITensorInfo *bias, const ITensorInfo *output, const GEMMLowpOutputStageInfo &info); +}; } // namespace arm_compute #endif /*ARM_COMPUTE_CLGEMMLOWPOUTPUTSTAGE_H */
\ No newline at end of file |