diff options
Diffstat (limited to 'arm_compute/runtime/CL')
13 files changed, 251 insertions, 145 deletions
diff --git a/arm_compute/runtime/CL/functions/CLComparison.h b/arm_compute/runtime/CL/functions/CLComparison.h index ac6a41ef2e..c6d61e45f2 100644 --- a/arm_compute/runtime/CL/functions/CLComparison.h +++ b/arm_compute/runtime/CL/functions/CLComparison.h @@ -79,7 +79,7 @@ public: public: /** Initialise the kernel's inputs and outputs. * - * @param[in] input1 Source tensor. Data types supported: U8/S8/QASYMM8/U16/S16/U32/S32/F16/F32. + * @param[in] input1 Source tensor. Data types supported: All. * The input1 tensor is [in, out] because its TensorInfo might be modified inside the kernel in case of broadcasting of dimension 0. * @param[in] input2 Source tensor. Data types supported: Same as @p input1. * The input2 tensor is [in, out] because its TensorInfo might be modified inside the kernel in case of broadcasting of dimension 0. @@ -92,7 +92,7 @@ public: /** Initialise the kernel's inputs and outputs. * * @param[in] compile_context The compile context to be used. - * @param[in] input1 Source tensor. Data types supported: U8/S8/QASYMM8/U16/S16/U32/S32/F16/F32. + * @param[in] input1 Source tensor. Data types supported: All. * The input1 tensor is [in, out] because its TensorInfo might be modified inside the kernel in case of broadcasting of dimension 0. * @param[in] input2 Source tensor. Data types supported: Same as @p input1. * The input2 tensor is [in, out] because its TensorInfo might be modified inside the kernel in case of broadcasting of dimension 0. @@ -101,7 +101,7 @@ public: void configure(const CLCompileContext &compile_context, ICLTensor *input1, ICLTensor *input2, ICLTensor *output); /** Static function to check if given info will lead to a valid configuration of @ref CLComparison * - * @param[in] input1 Source tensor. Data types supported: U8/S8/QASYMM8/U16/S16/U32/S32/F16/F32. + * @param[in] input1 Source tensor. Data types supported: All. * @param[in] input2 Source tensor. Data types supported: Same as @p input1. * @param[in] output Destination tensor. Data types supported: U8. * diff --git a/arm_compute/runtime/CL/functions/CLCopy.h b/arm_compute/runtime/CL/functions/CLCopy.h index ee5f31bc4c..c20d75eea8 100644 --- a/arm_compute/runtime/CL/functions/CLCopy.h +++ b/arm_compute/runtime/CL/functions/CLCopy.h @@ -38,7 +38,7 @@ class CLCopy : public ICLSimpleFunction public: /** Initialise the function's source and destination. * - * @param[in] input Source tensor. Data types supported: U8/S8/QASYMM8/U16/S16/F16/U32/S32/F32. + * @param[in] input Source tensor. Data types supported: All. * @param[out] output Output tensor. Data types supported: Same as @p input. * */ @@ -46,14 +46,14 @@ public: /** Initialise the function's source and destination. * * @param[in] compile_context The compile context to be used. - * @param[in] input Source tensor. Data types supported: U8/S8/QASYMM8/U16/S16/F16/U32/S32/F32. + * @param[in] input Source tensor. Data types supported: All. * @param[out] output Output tensor. Data types supported: Same as @p input. * */ void configure(const CLCompileContext &compile_context, ICLTensor *input, ICLTensor *output); /** Static function to check if given info will lead to a valid configuration of @ref CLCopy * - * @param[in] input Source tensor. Data types supported: U8/S8/QASYMM8/U16/S16/F16/U32/S32/F32. + * @param[in] input Source tensor. Data types supported: All. * @param[in] output Output tensor. Data types supported: Same as @p input. * * @return a status diff --git a/arm_compute/runtime/CL/functions/CLElementwiseOperations.h b/arm_compute/runtime/CL/functions/CLElementwiseOperations.h index 8cf1ca83c5..9cd3c150cc 100644 --- a/arm_compute/runtime/CL/functions/CLElementwiseOperations.h +++ b/arm_compute/runtime/CL/functions/CLElementwiseOperations.h @@ -33,7 +33,7 @@ class ICLTensor; /** Basic function to run @ref CLSaturatedArithmeticOperationKernel for addition * - * @note The tensor data type for the inputs must be U8/QASYMM8/S16/QSYMM16/S32/U32/F16/F32. + * @note The tensor data type for the inputs must be U8/QASYMM8/QASYMM8_SIGNED/S16/QSYMM16/S32/F16/F32. * @note The function performs an arithmetic addition between two tensors. */ class CLArithmeticAddition : public ICLSimpleFunction @@ -41,32 +41,74 @@ class CLArithmeticAddition : public ICLSimpleFunction public: /** Initialise the kernel's inputs, output and conversion policy. * - * @param[in, out] input1 First tensor input. Data types supported: U8/QASYMM8/S16/QSYMM16/S32/U32/F16/F32. + * Valid configurations (Input1,Input2) -> Output : + * + * - (U8,U8) -> U8 + * - (U8,U8) -> S16 + * - (S16,U8) -> S16 + * - (U8,S16) -> S16 + * - (S16,S16) -> S16 + * - (S32,S32) -> S32 + * - (F16,F16) -> F16 + * - (F32,F32) -> F32 + * - (QASYMM8,QASYMM8) -> QASYMM8 + * - (QASYMM8_SIGNED,QASYMM8_SIGNED) -> QASYMM8_SIGNED + * - (QSYMM16,QSYMM16) -> QSYMM16 + * + * @param[in, out] input1 First tensor input. Data types supported: U8/QASYMM8/QASYMM8_SIGNED/S16/QSYMM16/S32/F16/F32. * The input tensor is [in, out] because its TensorInfo might be modified inside the kernel in case of broadcasting of dimension 0. - * @param[in, out] input2 Second tensor input. Data types supported: U8, QASYMM8 (only if @p input1 is QASYMM8), QSYMM16 (only if @p input1 is QSYMM16), S16/F16/F32. + * @param[in, out] input2 Second tensor input. Data types supported: U8/QASYMM8/QASYMM8_SIGNED/S16/QSYMM16/S32/F16/F32. * The input tensor is [in, out] because its TensorInfo might be modified inside the kernel in case of broadcasting of dimension 0. - * @param[out] output Output tensor. Data types supported: U8 (Only if both inputs are U8), QASYMM8 (only if both inputs are QASYMM8), QSYMM16 (only if both inputs is QSYMM16), S16/F16/F32. + * @param[out] output Output tensor. Data types supported: U8/QASYMM8/QASYMM8_SIGNED/S16/QSYMM16/S32/F16/F32. * @param[in] policy Policy to use to handle overflow. * @param[in] act_info (Optional) Activation layer information in case of a fused activation. */ void configure(ICLTensor *input1, ICLTensor *input2, ICLTensor *output, ConvertPolicy policy, const ActivationLayerInfo &act_info = ActivationLayerInfo()); /** Initialise the kernel's inputs, output and conversion policy. * + * Valid configurations (Input1,Input2) -> Output : + * + * - (U8,U8) -> U8 + * - (U8,U8) -> S16 + * - (S16,U8) -> S16 + * - (U8,S16) -> S16 + * - (S16,S16) -> S16 + * - (S32,S32) -> S32 + * - (F16,F16) -> F16 + * - (F32,F32) -> F32 + * - (QASYMM8,QASYMM8) -> QASYMM8 + * - (QASYMM8_SIGNED,QASYMM8_SIGNED) -> QASYMM8_SIGNED + * - (QSYMM16,QSYMM16) -> QSYMM16 + * * @param[in] compile_context The compile context to be used. - * @param[in, out] input1 First tensor input. Data types supported: U8/QASYMM8/S16/QSYMM16/S32/U32/F16/F32. + * @param[in, out] input1 First tensor input. Data types supported: U8/QASYMM8/QASYMM8_SIGNED/S16/QSYMM16/S32/F16/F32. * The input tensor is [in, out] because its TensorInfo might be modified inside the kernel in case of broadcasting of dimension 0. - * @param[in, out] input2 Second tensor input. Data types supported: U8, QASYMM8 (only if @p input1 is QASYMM8), QSYMM16 (only if @p input1 is QSYMM16), S16/F16/F32. + * @param[in, out] input2 Second tensor input. Data types supported: U8/QASYMM8/QASYMM8_SIGNED/S16/QSYMM16/S32/F16/F32. * The input tensor is [in, out] because its TensorInfo might be modified inside the kernel in case of broadcasting of dimension 0. - * @param[out] output Output tensor. Data types supported: U8 (Only if both inputs are U8), QASYMM8 (only if both inputs are QASYMM8), QSYMM16 (only if both inputs is QSYMM16), S16/F16/F32. + * @param[out] output Output tensor. Data types supported: U8/QASYMM8/QASYMM8_SIGNED/S16/QSYMM16/S32/F16/F32. * @param[in] policy Policy to use to handle overflow. * @param[in] act_info (Optional) Activation layer information in case of a fused activation. */ void configure(const CLCompileContext &compile_context, ICLTensor *input1, ICLTensor *input2, ICLTensor *output, ConvertPolicy policy, const ActivationLayerInfo &act_info = ActivationLayerInfo()); /** Static function to check if given info will lead to a valid configuration of @ref CLSaturatedArithmeticOperationKernel for addition * - * @param[in] input1 First tensor input info. Data types supported: U8/QASYMM8/S16/QSYMM16/S32/U32/F16/F32. - * @param[in] input2 Second tensor input info. Data types supported: U8, QASYMM8 (only if @p input1 is QASYMM8), QSYMM16 (only if @p input1 is QSYMM16), S16/F16/F32. - * @param[in] output Output tensor info. Data types supported: U8 (Only if both inputs are U8), QASYMM8 (only if both inputs are QASYMM8), QSYMM16 (only if both inputs is QSYMM16), S16/F16/F32. + * Valid configurations (Input1,Input2) -> Output : + * + * - (U8,U8) -> U8 + * - (U8,U8) -> S16 + * - (S16,U8) -> S16 + * - (U8,S16) -> S16 + * - (S16,S16) -> S16 + * - (S32,S32) -> S32 + * - (F16,F16) -> F16 + * - (F32,F32) -> F32 + * - (QASYMM8,QASYMM8) -> QASYMM8 + * - (QASYMM8_SIGNED,QASYMM8_SIGNED) -> QASYMM8_SIGNED + * - (QSYMM16,QSYMM16) -> QSYMM16 + * + * @param[in] input1 First tensor input info. Data types supported: U8/QASYMM8/QASYMM8_SIGNED/S16/QSYMM16/S32/F16/F32. + * @param[in] input2 Second tensor input info. Data types supported: U8/QASYMM8/QASYMM8_SIGNED/S16/QSYMM16/S32/F16/F32. + * @param[in] output Output tensor info. Data types supported: U8/QASYMM8/QASYMM8_SIGNED/S16/QSYMM16/S32/F16/F32. * @param[in] policy Policy to use to handle overflow. * @param[in] act_info (Optional) Activation layer information in case of a fused activation. * @@ -77,7 +119,7 @@ public: /** Basic function to run @ref CLSaturatedArithmeticOperationKernel for subtraction * - * @note The tensor data type for the inputs must be U8/QASYMM8/S16/S32/U32/F16/F32. + * @note The tensor data type for the inputs must be U8/QASYMM8/QASYMM8_SIGNED/S16/S32/F16/F32. * @note The function performs an arithmetic subtraction between two tensors. */ class CLArithmeticSubtraction : public ICLSimpleFunction @@ -85,32 +127,74 @@ class CLArithmeticSubtraction : public ICLSimpleFunction public: /** Initialise the kernel's inputs, output and conversion policy. * - * @param[in, out] input1 First tensor input. Data types supported: U8/QASYMM8/S16/S32/U32/F16/F32. + * Valid configurations (Input1,Input2) -> Output : + * + * - (U8,U8) -> U8 + * - (U8,U8) -> S16 + * - (S16,U8) -> S16 + * - (U8,S16) -> S16 + * - (S16,S16) -> S16 + * - (S32,S32) -> S32 + * - (F16,F16) -> F16 + * - (F32,F32) -> F32 + * - (QASYMM8,QASYMM8) -> QASYMM8 + * - (QASYMM8_SIGNED,QASYMM8_SIGNED) -> QASYMM8_SIGNED + * - (QSYMM16,QSYMM16) -> QSYMM16 + * + * @param[in, out] input1 First tensor input. Data types supported: U8/QASYMM8/QASYMM8_SIGNED/S16/QSYMM16/S32/F16/F32. * The input tensor is [in, out] because its TensorInfo might be modified inside the kernel in case of broadcasting of dimension 0. - * @param[in, out] input2 Second tensor input. Data types supported: U8, QASYMM8 (only if @p input1 is QASYMM8), S16/F16/F32. + * @param[in, out] input2 Second tensor input. Data types supported: U8/QASYMM8/QASYMM8_SIGNED/S16/QSYMM16/S32/F16/F32. * The input tensor is [in, out] because its TensorInfo might be modified inside the kernel in case of broadcasting of dimension 0. - * @param[out] output Output tensor. Data types supported: U8 (Only if both inputs are U8), QASYMM8 (only if both inputs are QASYMM8), S16/F16/F32. + * @param[out] output Output tensor. Data types supported: U8/QASYMM8/QASYMM8_SIGNED/S16/QSYMM16/S32/F16/F32. * @param[in] policy Policy to use to handle overflow. * @param[in] act_info (Optional) Activation layer information in case of a fused activation. */ void configure(ICLTensor *input1, ICLTensor *input2, ICLTensor *output, ConvertPolicy policy, const ActivationLayerInfo &act_info = ActivationLayerInfo()); /** Initialise the kernel's inputs, output and conversion policy. * + * Valid configurations (Input1,Input2) -> Output : + * + * - (U8,U8) -> U8 + * - (U8,U8) -> S16 + * - (S16,U8) -> S16 + * - (U8,S16) -> S16 + * - (S16,S16) -> S16 + * - (S32,S32) -> S32 + * - (F16,F16) -> F16 + * - (F32,F32) -> F32 + * - (QASYMM8,QASYMM8) -> QASYMM8 + * - (QASYMM8_SIGNED,QASYMM8_SIGNED) -> QASYMM8_SIGNED + * - (QSYMM16,QSYMM16) -> QSYMM16 + * * @param[in] compile_context The compile context to be used. - * @param[in, out] input1 First tensor input. Data types supported: U8/QASYMM8/S16/S32/U32/F16/F32. + * @param[in, out] input1 First tensor input. Data types supported: U8/QASYMM8/QASYMM8_SIGNED/S16/QSYMM16/S32/F16/F32. * The input tensor is [in, out] because its TensorInfo might be modified inside the kernel in case of broadcasting of dimension 0. - * @param[in, out] input2 Second tensor input. Data types supported: U8, QASYMM8 (only if @p input1 is QASYMM8), S16/F16/F32. + * @param[in, out] input2 Second tensor input. Data types supported: U8/QASYMM8/QASYMM8_SIGNED/S16/QSYMM16/S32/F16/F32. * The input tensor is [in, out] because its TensorInfo might be modified inside the kernel in case of broadcasting of dimension 0. - * @param[out] output Output tensor. Data types supported: U8 (Only if both inputs are U8), QASYMM8 (only if both inputs are QASYMM8), S16/F16/F32. + * @param[out] output Output tensor. Data types supported: U8/QASYMM8/QASYMM8_SIGNED/S16/QSYMM16/S32/F16/F32. * @param[in] policy Policy to use to handle overflow. * @param[in] act_info (Optional) Activation layer information in case of a fused activation. */ void configure(const CLCompileContext &compile_context, ICLTensor *input1, ICLTensor *input2, ICLTensor *output, ConvertPolicy policy, const ActivationLayerInfo &act_info = ActivationLayerInfo()); /** Static function to check if given info will lead to a valid configuration of @ref CLSaturatedArithmeticOperationKernel for subtraction * - * @param[in] input1 First tensor input info. Data types supported: U8/QASYMM8/S16/S32/U32/F16/F32. - * @param[in] input2 Second tensor input info. Data types supported: U8, QASYMM8 (only if @p input1 is QASYMM8), S16/F16/F32. - * @param[in] output Output tensor info. Data types supported: U8 (Only if both inputs are U8), QASYMM8 ( only if both inputs are QASYMM8), S16/F16/F32. + * Valid configurations (Input1,Input2) -> Output : + * + * - (U8,U8) -> U8 + * - (U8,U8) -> S16 + * - (S16,U8) -> S16 + * - (U8,S16) -> S16 + * - (S16,S16) -> S16 + * - (S32,S32) -> S32 + * - (F16,F16) -> F16 + * - (F32,F32) -> F32 + * - (QASYMM8,QASYMM8) -> QASYMM8 + * - (QASYMM8_SIGNED,QASYMM8_SIGNED) -> QASYMM8_SIGNED + * - (QSYMM16,QSYMM16) -> QSYMM16 + * + * @param[in] input1 First tensor input info. Data types supported: U8/QASYMM8/QASYMM8_SIGNED/S16/QSYMM16/S32/F16/F32. + * @param[in] input2 Second tensor input info. Data types supported: U8/QASYMM8/QASYMM8_SIGNED/S16/QSYMM16/S32/F16/F32. + * @param[in] output Output tensor info. Data types supported: U8/QASYMM8/QASYMM8_SIGNED/S16/QSYMM16/S32/F16/F32. * @param[in] policy Policy to use to handle overflow. * @param[in] act_info (Optional) Activation layer information in case of a fused activation. * @@ -170,30 +254,30 @@ class CLElementwiseMax : public ICLSimpleFunction public: /** Initialise the kernel's inputs, output and conversion policy. * - * @param[in, out] input1 First tensor input. Data types supported: U8/QASYMM8/S16/QSYMM16/S32/U32/F16/F32. + * @param[in, out] input1 First tensor input. Data types supported: U8/QASYMM8/QASYMM8_SIGNED/S16/QSYMM16/S32/U32/F16/F32. * The input tensor is [in, out] because its TensorInfo might be modified inside the kernel in case of broadcasting of dimension 0. - * @param[in, out] input2 Second tensor input. Data types supported: U8, QASYMM8 (only if @p input1 is QASYMM8), S16, QSYMM16 (only if @p input1 is QSYMM16), F16/F32. + * @param[in, out] input2 Second tensor input. Data types supported: same as @p input1. * The input tensor is [in, out] because its TensorInfo might be modified inside the kernel in case of broadcasting of dimension 0. - * @param[out] output Output tensor. Data types supported: U8 (Only if both inputs are U8), QASYMM8 (only if both inputs are QASYMM8), S16, QSYMM16 (only if both inputs are QSYMM16), F16/F32. + * @param[out] output Output tensor. Data types supported: same as @p input1. * @param[in] act_info (Optional) Activation layer information in case of a fused activation. */ void configure(ICLTensor *input1, ICLTensor *input2, ICLTensor *output, const ActivationLayerInfo &act_info = ActivationLayerInfo()); /** Initialise the kernel's inputs, output and conversion policy. * * @param[in] compile_context The compile context to be used. - * @param[in, out] input1 First tensor input. Data types supported: U8/QASYMM8/S16/QSYMM16/S32/U32/F16/F32. + * @param[in, out] input1 First tensor input. Data types supported: U8/QASYMM8/QASYMM8_SIGNED/S16/QSYMM16/S32/U32/F16/F32. * The input tensor is [in, out] because its TensorInfo might be modified inside the kernel in case of broadcasting of dimension 0. - * @param[in, out] input2 Second tensor input. Data types supported: U8, QASYMM8 (only if @p input1 is QASYMM8), S16, QSYMM16 (only if @p input1 is QSYMM16), F16/F32. + * @param[in, out] input2 Second tensor input. Data types supported: same as @p input1. * The input tensor is [in, out] because its TensorInfo might be modified inside the kernel in case of broadcasting of dimension 0. - * @param[out] output Output tensor. Data types supported: U8 (Only if both inputs are U8), QASYMM8 (only if both inputs are QASYMM8), S16, QSYMM16 (only if both inputs are QSYMM16), F16/F32. + * @param[out] output Output tensor. Data types supported: same as @p input1. * @param[in] act_info (Optional) Activation layer information in case of a fused activation. */ void configure(const CLCompileContext &compile_context, ICLTensor *input1, ICLTensor *input2, ICLTensor *output, const ActivationLayerInfo &act_info = ActivationLayerInfo()); /** Static function to check if given info will lead to a valid configuration of @ref CLArithmeticOperationKernel for max * - * @param[in] input1 First tensor input info. Data types supported: U8/QASYMM8/S16/QSYMM16/S32/U32/F16/F32. - * @param[in] input2 Second tensor input info. Data types supported: U8, QASYMM8 (only if @p input1 is QASYMM8), S16, QSYMM16 (only if @p input1 is QSYMM16), F16/F32. - * @param[in] output Output tensor info. Data types supported: U8 (Only if both inputs are U8), QASYMM8 ( only if both inputs are QASYMM8), S16, QSYMM16 (only if both inputs are QSYMM16), F16/F32. + * @param[in] input1 First tensor input info. Data types supported: U8/QASYMM8/QASYMM8_SIGNED/S16/QSYMM16/S32/U32/F16/F32. + * @param[in] input2 Second tensor input info. Data types supported: same as @p input1. + * @param[in] output Output tensor info. Data types supported: same as @p input1. * @param[in] act_info (Optional) Activation layer information in case of a fused activation. * * @return a status @@ -211,30 +295,30 @@ class CLElementwiseMin : public ICLSimpleFunction public: /** Initialise the kernel's inputs, output and conversion policy. * - * @param[in, out] input1 First tensor input. Data types supported: U8/QASYMM8/S16/QSYMM16/S32/U32/F16/F32. + * @param[in, out] input1 First tensor input. Data types supported: U8/QASYMM8/QASYMM8_SIGNED/S16/QSYMM16/S32/U32/F16/F32. * The input tensor is [in, out] because its TensorInfo might be modified inside the kernel in case of broadcasting of dimension 0. - * @param[in, out] input2 Second tensor input. Data types supported: U8, QASYMM8 (only if @p input1 is QASYMM8), S16, QSYMM16 (only if @p input1 is QSYMM16), F16/F32. + * @param[in, out] input2 Second tensor input. Data types supported: same as @p input1. * The input tensor is [in, out] because its TensorInfo might be modified inside the kernel in case of broadcasting of dimension 0. - * @param[out] output Output tensor. Data types supported: U8 (Only if both inputs are U8), QASYMM8 (only if both inputs are QASYMM8), S16, QSYMM16 (only if both inputs are QSYMM16), F16/F32. + * @param[out] output Output tensor. Data types supported: same as @p input1. * @param[in] act_info (Optional) Activation layer information in case of a fused activation. */ void configure(ICLTensor *input1, ICLTensor *input2, ICLTensor *output, const ActivationLayerInfo &act_info = ActivationLayerInfo()); /** Initialise the kernel's inputs, output and conversion policy. * * @param[in] compile_context The compile context to be used. - * @param[in, out] input1 First tensor input. Data types supported: U8/QASYMM8/S16/QSYMM16/S32/U32/F16/F32. + * @param[in, out] input1 First tensor input. Data types supported: U8/QASYMM8/QASYMM8_SIGNED/S16/QSYMM16/S32/U32/F16/F32. * The input tensor is [in, out] because its TensorInfo might be modified inside the kernel in case of broadcasting of dimension 0. - * @param[in, out] input2 Second tensor input. Data types supported: U8, QASYMM8 (only if @p input1 is QASYMM8), S16, QSYMM16 (only if @p input1 is QSYMM16), F16/F32. + * @param[in, out] input2 Second tensor input. Data types supported: same as @p input1. * The input tensor is [in, out] because its TensorInfo might be modified inside the kernel in case of broadcasting of dimension 0. - * @param[out] output Output tensor. Data types supported: U8 (Only if both inputs are U8), QASYMM8 (only if both inputs are QASYMM8), S16, QSYMM16 (only if both inputs are QSYMM16), F16/F32. + * @param[out] output Output tensor. Data types supported: same as @p input1. * @param[in] act_info (Optional) Activation layer information in case of a fused activation. */ void configure(const CLCompileContext &compile_context, ICLTensor *input1, ICLTensor *input2, ICLTensor *output, const ActivationLayerInfo &act_info = ActivationLayerInfo()); /** Static function to check if given info will lead to a valid configuration of @ref CLArithmeticOperationKernel for min * - * @param[in] input1 First tensor input info. Data types supported: U8/QASYMM8/S16/QSYMM16/S32/U32/F16/F32. - * @param[in] input2 Second tensor input info. Data types supported: U8, QASYMM8 (only if @p input1 is QASYMM8), S16, QSYMM16 (only if @p input1 is QSYMM16), F16/F32. - * @param[in] output Output tensor info. Data types supported: U8 (Only if both inputs are U8), QASYMM8 ( only if both inputs are QASYMM8), S16, QSYMM16 (only if both inputs are QSYMM16), F16/F32. + * @param[in] input1 First tensor input info. Data types supported: U8/QASYMM8/QASYMM8_SIGNED/S16/QSYMM16/S32/U32/F16/F32. + * @param[in] input2 Second tensor input info. Data types supported: same as @p input1. + * @param[in] output Output tensor info. Data types supported: same as @p input1. * @param[in] act_info (Optional) Activation layer information in case of a fused activation. * * @return a status @@ -252,30 +336,30 @@ class CLElementwiseSquaredDiff : public ICLSimpleFunction public: /** Initialise the kernel's inputs, output and conversion policy. * - * @param[in, out] input1 First tensor input. Data types supported: U8/QASYMM8/S16/QSYMM16/F16/F32. + * @param[in, out] input1 First tensor input. Data types supported: U8/QASYMM8/QASYMM8_SIGNED/S16/QSYMM16/F16/F32. * The input tensor is [in, out] because its TensorInfo might be modified inside the kernel in case of broadcasting of dimension 0. - * @param[in, out] input2 Second tensor input. Data types supported: U8, QASYMM8 (only if @p input1 is QASYMM8), S16, QSYMM16 (only if @p input1 is QSYMM16), F16/F32. + * @param[in, out] input2 Second tensor input. Data types supported: same as @p input1. * The input tensor is [in, out] because its TensorInfo might be modified inside the kernel in case of broadcasting of dimension 0. - * @param[out] output Output tensor. Data types supported: U8 (Only if both inputs are U8), QASYMM8 (only if both inputs are QASYMM8), S16, QSYMM16 (only if both inputs are QSYMM16), F16/F32. + * @param[out] output Output tensor. Data types supported: same as @p input1. * @param[in] act_info (Optional) Activation layer information in case of a fused activation. */ void configure(ICLTensor *input1, ICLTensor *input2, ICLTensor *output, const ActivationLayerInfo &act_info = ActivationLayerInfo()); /** Initialise the kernel's inputs, output and conversion policy. * * @param[in] compile_context The compile context to be used. - * @param[in, out] input1 First tensor input. Data types supported: U8/QASYMM8/S16/QSYMM16/F16/F32. + * @param[in, out] input1 First tensor input. Data types supported: U8/QASYMM8/QASYMM8_SIGNED/S16/QSYMM16/F16/F32. * The input tensor is [in, out] because its TensorInfo might be modified inside the kernel in case of broadcasting of dimension 0. - * @param[in, out] input2 Second tensor input. Data types supported: U8, QASYMM8 (only if @p input1 is QASYMM8), S16, QSYMM16 (only if @p input1 is QSYMM16), F16/F32. + * @param[in, out] input2 Second tensor input. Data types supported: same as @p input1. * The input tensor is [in, out] because its TensorInfo might be modified inside the kernel in case of broadcasting of dimension 0. - * @param[out] output Output tensor. Data types supported: U8 (Only if both inputs are U8), QASYMM8 (only if both inputs are QASYMM8), S16, QSYMM16 (only if both inputs are QSYMM16), F16/F32. + * @param[out] output Output tensor. Data types supported: same as @p input1. * @param[in] act_info (Optional) Activation layer information in case of a fused activation. */ void configure(const CLCompileContext &compile_context, ICLTensor *input1, ICLTensor *input2, ICLTensor *output, const ActivationLayerInfo &act_info = ActivationLayerInfo()); /** Static function to check if given info will lead to a valid configuration of @ref CLArithmeticOperationKernel for squared difference * - * @param[in] input1 First tensor input info. Data types supported: U8/QASYMM8/S16/QSYMM16/F16/F32. - * @param[in] input2 Second tensor input info. Data types supported: U8, QASYMM8 (only if @p input1 is QASYMM8), S16, QSYMM16 (only if @p input1 is QSYMM16), F16/F32. - * @param[in] output Output tensor info. Data types supported: U8 (Only if both inputs are U8), QASYMM8 ( only if both inputs are QASYMM8), S16, QSYMM16 (only if both inputs are QSYMM16), F16/F32. + * @param[in] input1 First tensor input info. Data types supported: U8/QASYMM8/QASYMM8_SIGNED/S16/QSYMM16/F16/F32. + * @param[in] input2 Second tensor input info. Data types supported: same as @p input1. + * @param[in] output Output tensor info. Data types supported: same as @p input1. * @param[in] act_info (Optional) Activation layer information in case of a fused activation. * * @return a status diff --git a/arm_compute/runtime/CL/functions/CLFill.h b/arm_compute/runtime/CL/functions/CLFill.h index 99cae077fe..b79b234158 100644 --- a/arm_compute/runtime/CL/functions/CLFill.h +++ b/arm_compute/runtime/CL/functions/CLFill.h @@ -38,14 +38,14 @@ class CLFill : public ICLSimpleFunction public: /** Initialize the function * - * @param[in,out] tensor Source tensor. Data types supported: U8/S8/QASYMM8/U16/S16/F16/U32/S32/F32 + * @param[in,out] tensor Source tensor. Data types supported: All. * @param[in] constant_value Constant value to use to fill tensor. */ void configure(ICLTensor *tensor, PixelValue constant_value); /** Initialize the function * * @param[in] compile_context The compile context to be used. - * @param[in,out] tensor Source tensor. Data types supported: U8/S8/QASYMM8/U16/S16/F16/U32/S32/F32 + * @param[in,out] tensor Source tensor. Data types supported: All. * @param[in] constant_value Constant value to use to fill tensor. */ void configure(const CLCompileContext &compile_context, ICLTensor *tensor, PixelValue constant_value); diff --git a/arm_compute/runtime/CL/functions/CLFillBorder.h b/arm_compute/runtime/CL/functions/CLFillBorder.h index e552c353db..18bc20e654 100644 --- a/arm_compute/runtime/CL/functions/CLFillBorder.h +++ b/arm_compute/runtime/CL/functions/CLFillBorder.h @@ -38,7 +38,7 @@ class CLFillBorder : public ICLSimpleFunction public: /** Initialize the function * - * @param[in,out] tensor Source tensor. Data types supported: U8/S16 + * @param[in,out] tensor Source tensor. Data types supported: U8/QASYMM8/S8/QASYMM8_SIGNED/U16/S16/U32/S32/F16/F32. * @param[in] border_width The border width * @param[in] border_mode Strategy to use for borders. * @param[in] constant_border_value (Optional) Constant value to use for borders if border_mode is set to CONSTANT. @@ -47,7 +47,7 @@ public: /** Initialize the function * * @param[in] compile_context The compile context to be used. - * @param[in,out] tensor Source tensor. Data types supported: U8/S16 + * @param[in,out] tensor Source tensor. Data types supported: U8/QASYMM8/S8/QASYMM8_SIGNED/U16/S16/U32/S32/F16/F32. * @param[in] border_width The border width * @param[in] border_mode Strategy to use for borders. * @param[in] constant_border_value (Optional) Constant value to use for borders if border_mode is set to CONSTANT. diff --git a/arm_compute/runtime/CL/functions/CLGEMM.h b/arm_compute/runtime/CL/functions/CLGEMM.h index 018d88d366..8e4d3906d1 100644 --- a/arm_compute/runtime/CL/functions/CLGEMM.h +++ b/arm_compute/runtime/CL/functions/CLGEMM.h @@ -74,7 +74,7 @@ public: /** Configures the @ref CLGEMMReshapeRHSMatrixKernel kernel * - * @param[in] input Input tensor. Data types supported: U8/S8/QASYMM8/U16/S16/F16/U32/S32/F32 + * @param[in] input Input tensor. Data types supported: All * @param[in] info RHS matrix information to be used for reshaping. */ void configure(const ICLTensor *input, GEMMRHSMatrixInfo info) @@ -85,7 +85,7 @@ public: /** Configures the @ref CLGEMMReshapeRHSMatrixKernel kernel * * @param[in] compile_context The compile context to be used. - * @param[in] input Input tensor. Data types supported: U8/S8/QASYMM8/U16/S16/F16/U32/S32/F32 + * @param[in] input Input tensor. Data types supported: All * @param[in] info RHS matrix information to be used for reshaping. */ void configure(const CLCompileContext &compile_context, const ICLTensor *input, GEMMRHSMatrixInfo info) diff --git a/arm_compute/runtime/CL/functions/CLGEMMConvolutionLayer.h b/arm_compute/runtime/CL/functions/CLGEMMConvolutionLayer.h index be9283b6ec..277b27f690 100644 --- a/arm_compute/runtime/CL/functions/CLGEMMConvolutionLayer.h +++ b/arm_compute/runtime/CL/functions/CLGEMMConvolutionLayer.h @@ -98,8 +98,8 @@ class CLConvolutionLayerReshapeWeightsTransform : public ITransformWeights public: /** Configures the @ref CLConvolutionLayerReshapeWeights function * - * @param[in] input Input tensor. Data type supported: QASYMM8/QASYMM8_SIGNED/F16/F32. - * @param[in] biases Biases tensor. Data type supported: Same as @p input. + * @param[in] input Input tensor. Data type supported: QASYMM8/QASYMM8_SIGNED/QSYMM8_PER_CHANNEL/F16/F32. + * @param[in] biases Biases tensor. Data type supported: same as @p input, S32 if @p input is quantized. * @param[in] num_groups Number of groups when performing a grouped convolution. */ void configure(const ICLTensor *input, const ICLTensor *biases, unsigned int num_groups) @@ -109,8 +109,8 @@ public: /** Configures the @ref CLConvolutionLayerReshapeWeights function * * @param[in] compile_context The compile context to be used. - * @param[in] input Input tensor. Data type supported: QASYMM8/QASYMM8_SIGNED/F16/F32. - * @param[in] biases Biases tensor. Data type supported: Same as @p input. + * @param[in] input Input tensor. Data type supported: QASYMM8/QASYMM8_SIGNED/QSYMM8_PER_CHANNEL/F16/F32. + * @param[in] biases Biases tensor. Data type supported: same as @p input, S32 if @p input is quantized. * @param[in] num_groups Number of groups when performing a grouped convolution. */ void configure(const CLCompileContext &compile_context, const ICLTensor *input, const ICLTensor *biases, unsigned int num_groups) @@ -183,11 +183,11 @@ public: * * @param[in] input Source tensor. 3 lower dimensions represent a single input [width, height, IFM], * while every optional dimension from 4 and above represent a batch of inputs. - * Data types supported: QASYMM8/F16/F32. + * Data types supported: QASYMM8/QASYMM8_SIGNED/F16/F32. * @param[in] weights Weights tensor. Weights are 4D tensor with dimensions [kernel_x, kernel_y, IFM, OFM]. - * Data type supported: Same as @p input or QASYMM8/QSYMM8_PER_CHANNEL when @p input is QASYMM8. + * Data type supported: Same as @p input or QASYMM8/QSYMM8_PER_CHANNEL when @p input is QASYMM8 or QASYMM8_SIGNED/QSYMM8_PER_CHANNEL when @p input is QASYMM8_SIGNED. * @param[in] biases Biases tensor. Shared biases supported. Biases are 1D tensor with dimensions [OFM]. - * Data type supported: Should match @p input data type, except for input of QASYMM8 type where biases should be of S32 type. + * Data type supported: Should match @p input data type, except for input of quantized type where biases should be of S32 type. * @param[out] output Destination tensor. 3 lower dimensions represent a single output [width, height, OFM], while the rest represent batch of outputs. * Data types supported: Same as @p input. * @param[in] conv_info Contains padding and stride information described in @ref PadStrideInfo. @@ -204,11 +204,11 @@ public: * @param[in] compile_context The compile context to be used. * @param[in] input Source tensor. 3 lower dimensions represent a single input [width, height, IFM], * while every optional dimension from 4 and above represent a batch of inputs. - * Data types supported: QASYMM8/F16/F32. + * Data types supported: QASYMM8/QASYMM8_SIGNED/F16/F32. * @param[in] weights Weights tensor. Weights are 4D tensor with dimensions [kernel_x, kernel_y, IFM, OFM]. - * Data type supported: Same as @p input or QASYMM8/QSYMM8_PER_CHANNEL when @p input is QASYMM8. + * Data type supported: Same as @p input or QASYMM8/QSYMM8_PER_CHANNEL when @p input is QASYMM8 or QASYMM8_SIGNED/QSYMM8_PER_CHANNEL when @p input is QASYMM8_SIGNED. * @param[in] biases Biases tensor. Shared biases supported. Biases are 1D tensor with dimensions [OFM]. - * Data type supported: Should match @p input data type, except for input of QASYMM8 type where biases should be of S32 type. + * Data type supported: Should match @p input data type, except for input of quantized type where biases should be of S32 type. * @param[out] output Destination tensor. 3 lower dimensions represent a single output [width, height, OFM], while the rest represent batch of outputs. * Data types supported: Same as @p input. * @param[in] conv_info Contains padding and stride information described in @ref PadStrideInfo. @@ -225,11 +225,11 @@ public: * * @param[in] input Source tensor. 3 lower dimensions represent a single input [width, height, IFM], * while every optional dimension from 4 and above represent a batch of inputs. - * Data types supported: QASYMM8/F16/F32. + * Data types supported: QASYMM8/QASYMM8_SIGNED/F16/F32. * @param[in] weights Weights tensor. Weights are 4D tensor with dimensions [kernel_x, kernel_y, IFM, OFM]. - * Data type supported: Same as @p input or QASYMM8/QSYMM8_PER_CHANNEL when @p input is QASYMM8. + * Data type supported: Same as @p input or QASYMM8/QSYMM8_PER_CHANNEL when @p input is QASYMM8 or QASYMM8_SIGNED/QSYMM8_PER_CHANNEL when @p input is QASYMM8_SIGNED. * @param[in] biases Biases tensor. Shared biases supported. Biases are 1D tensor with dimensions [OFM]. - * Data type supported: Should match @p input data type, except for input of QASYMM8 type where biases should be of S32 type. + * Data type supported: Should match @p input data type, except for input of quantized type where biases should be of S32 type. * @param[out] output Destination tensor. 3 lower dimensions represent a single output [width, height, OFM], while the rest represent batch of outputs. * Data types supported: Same as @p input. * @param[in] conv_info Contains padding and stride information described in @ref PadStrideInfo. @@ -252,12 +252,12 @@ private: /** Configures the appropriate matrix multiply routine * * @param[in] compile_context The compile context to be used. - * @param[in] input Input tensor. Data types supported: QASYMM8/F16/F32. - * @param[in] weights Weights tensor. Data type supported: Same as @p input or QASYMM8/QSYMM8_PER_CHANNEL when @p input is QASYMM8. + * @param[in] input Input tensor. Data types supported: QASYMM8/QASYMM8_SIGNED/F16/F32. + * @param[in] weights Weights tensor. Data type supported: Same as @p input or QASYMM8/QSYMM8_PER_CHANNEL when @p input is QASYMM8 or + * QASYMM8_SIGNED/QSYMM8_PER_CHANNEL when @p input is QASYMM8_SIGNED. * @param[in] biases Biases tensor. Shared biases supported. Biases are 1D tensor with dimensions [OFM]. - * Data type supported: Should match @p input data type, except for input of QASYMM8 type where biases should be of S32 type. - * @param[in, out] output Output tensor. Data types supported: Same as @p input, - * except for input of QASYMM8 type where output should be of S32 type. + * Data type supported: Should match @p input data type, except for input of quantized type where biases should be of S32 type. + * @param[in, out] output Output tensor. Data types supported: same as @p input. * @param[in] gemmlowp_output_stage GEMMLowp output stage info * @param[in] gemm_3d_depth Depth of GEMM 3D * @param[in] act_info Activation to apply after the matrix multiplication @@ -267,12 +267,12 @@ private: int gemm_3d_depth, const ActivationLayerInfo &act_info); /** Static function to check if given info will lead to a valid configuration of @ref CLGEMMConvolutionLayer matrix multiply routines * - * @param[in] input Input tensor info. Data types supported: QASYMM8/F16/F32. - * @param[in] weights Weights tensor info. Data type supported: Same as @p input or QASYMM8/QSYMM8_PER_CHANNEL when @p input is QASYMM8. + * @param[in] input Input tensor info. Data types supported: QASYMM8/QASYMM8_SIGNED/F16/F32. + * @param[in] weights Weights tensor info. Data type supported: Same as @p input or QASYMM8/QSYMM8_PER_CHANNEL when @p input is QASYMM8 or + * QASYMM8_SIGNED/QSYMM8_PER_CHANNEL when @p input is QASYMM8_SIGNED. * @param[in] biases Biases tensor info. Shared biases supported. Biases are 1D tensor with dimensions [OFM]. - * Data type supported: Should match @p input data type, except for input of QASYMM8 type where biases should be of S32 type. - * @param[in] output Output tensor info. Data types supported: Same as @p input, - * except for input of QASYMM8 type where output should be of S32 type. + * Data type supported: Should match @p input data type, except for input of quantized type where biases should be of S32 type. + * @param[in] output Output tensor info. Data types supported: same as @p input. * @param[in] gemmlowp_output_stage GEMMLowp output stage info * @param[in] gemm_3d_depth Depth of GEMM 3D * @param[in] skip_im2col Flag which specifies if im2col has to be skipped. i.e. 1x1 convolution with NHWC data layout. diff --git a/arm_compute/runtime/CL/functions/CLGEMMLowpMatrixMultiplyCore.h b/arm_compute/runtime/CL/functions/CLGEMMLowpMatrixMultiplyCore.h index 0e822bc5a2..57b1e30df5 100644 --- a/arm_compute/runtime/CL/functions/CLGEMMLowpMatrixMultiplyCore.h +++ b/arm_compute/runtime/CL/functions/CLGEMMLowpMatrixMultiplyCore.h @@ -59,8 +59,8 @@ public: * @note GEMMLowp: low precision GEMM kernel. [A * B + C] * This kernel performs the following computations: * - * -# Convert a values from QASYMM8 to int32 and add a_offset to each of them. - * -# Convert b values from QASYMM8 to int32 and add b_offset to each of them. + * -# Convert a values from 8-bit quantized to int32 and add a_offset to each of them. + * -# Convert b values from 8-bit quantized to int32 and add b_offset to each of them. * -# Compute the matrix product of the resulting a * b in int32. * -# Quantize to uint8 if gemm_info.gemmlowp_output_stage != NONE * @@ -77,8 +77,8 @@ public: * @note GEMMLowp: low precision GEMM kernel. [A * B + C] * This kernel performs the following computations: * - * -# Convert a values from QASYMM8 to int32 and add a_offset to each of them. - * -# Convert b values from QASYMM8 to int32 and add b_offset to each of them. + * -# Convert a values from 8-bit quantized to int32 and add a_offset to each of them. + * -# Convert b values from 8-bit quantized to int32 and add b_offset to each of them. * -# Compute the matrix product of the resulting a * b in int32. * -# Quantize to uint8 if gemm_info.gemmlowp_output_stage != NONE * diff --git a/arm_compute/runtime/CL/functions/CLPixelWiseMultiplication.h b/arm_compute/runtime/CL/functions/CLPixelWiseMultiplication.h index a647281e29..b87daba1f8 100644 --- a/arm_compute/runtime/CL/functions/CLPixelWiseMultiplication.h +++ b/arm_compute/runtime/CL/functions/CLPixelWiseMultiplication.h @@ -38,19 +38,25 @@ class CLPixelWiseMultiplication : public ICLSimpleFunction public: /** Initialise the kernel's inputs, output and convertion policy. * + * Valid configurations (Input1,Input2) -> Output : + * + * - (U8,U8) -> U8 + * - (U8,U8) -> S16 + * - (U8,S16) -> S16 + * - (S16,U8) -> S16 + * - (S16,S16) -> S16 + * - (F16,F16) -> F16 + * - (F32,F32) -> F32 + * - (QASYMM8,QASYMM8) -> QASYMM8 + * - (QASYMM8_SIGNED,QASYMM8_SIGNED) -> QASYMM8_SIGNED + * - (QSYMM16,QSYMM16) -> QSYMM16 + * - (QSYMM16,QSYMM16) -> S32 + * * @param[in, out] input1 An input tensor. Data types supported: U8/QASYMM8/QASYMM8_SIGNED/S16/QSYMM16/F16/F32. * The input tensor is [in, out] because its TensorInfo might be modified inside the kernel in case of broadcasting of dimension 0. - * @param[in, out] input2 An input tensor. Data types supported: same as @p input1. + * @param[in, out] input2 An input tensor. Data types supported: U8/QASYMM8/QASYMM8_SIGNED/S16/QSYMM16/F16/F32. * The input tensor is [in, out] because its TensorInfo might be modified inside the kernel in case of broadcasting of dimension 0. - * @param[out] output The output tensor. Data types supported: - * - U8, only if both input are U8 - * - QASYMM8, only if both inputs are QASYMM8 - * - QASYMM8_SIGNED, only if both inputs are QASYMM8_SIGNED - * - S16 - * - QSYMM16, only if both inputs are QSYMM16 - * - S32, only if both inputs are QSYMM16 - * - F16 - * - F32 + * @param[out] output The output tensor. Data types supported: U8/QASYMM8/QASYMM8_SIGNED/S16/QSYMM16/F16/F32. * @param[in] scale Scale to apply after multiplication. * Scale must be positive and its value must be either 1/255 or 1/2^n where n is between 0 and 15. * @param[in] overflow_policy Overflow policy. Supported overflow policies: Wrap, Saturate @@ -61,12 +67,26 @@ public: ConvertPolicy overflow_policy, RoundingPolicy rounding_policy, const ActivationLayerInfo &act_info = ActivationLayerInfo()); /** Initialise the kernel's inputs, output and convertion policy. * + * Valid configurations (Input1,Input2) -> Output : + * + * - (U8,U8) -> U8 + * - (U8,U8) -> S16 + * - (U8,S16) -> S16 + * - (S16,U8) -> S16 + * - (S16,S16) -> S16 + * - (F16,F16) -> F16 + * - (F32,F32) -> F32 + * - (QASYMM8,QASYMM8) -> QASYMM8 + * - (QASYMM8_SIGNED,QASYMM8_SIGNED) -> QASYMM8_SIGNED + * - (QSYMM16,QSYMM16) -> QSYMM16 + * - (QSYMM16,QSYMM16) -> S32 + * * @param[in] compile_context The compile context to be used. * @param[in, out] input1 An input tensor. Data types supported: U8/QASYMM8/QASYMM8_SIGNED/S16/QSYMM16/F16/F32. * The input tensor is [in, out] because its TensorInfo might be modified inside the kernel in case of broadcasting of dimension 0. - * @param[in, out] input2 An input tensor. Data types supported: same as @p input1. + * @param[in, out] input2 An input tensor. Data types supported: U8/QASYMM8/QASYMM8_SIGNED/S16/QSYMM16/F16/F32. * The input tensor is [in, out] because its TensorInfo might be modified inside the kernel in case of broadcasting of dimension 0. - * @param[out] output The output tensor, Data types supported: same as @p input1. Note: U8 requires both inputs to be U8. + * @param[out] output The output tensor. Data types supported: U8/QASYMM8/QASYMM8_SIGNED/S16/QSYMM16/F16/F32. * @param[in] scale Scale to apply after multiplication. * Scale must be positive and its value must be either 1/255 or 1/2^n where n is between 0 and 15. * @param[in] overflow_policy Overflow policy. Supported overflow policies: Wrap, Saturate @@ -77,17 +97,24 @@ public: ConvertPolicy overflow_policy, RoundingPolicy rounding_policy, const ActivationLayerInfo &act_info = ActivationLayerInfo()); /** Static function to check if given info will lead to a valid configuration of @ref CLPixelWiseMultiplication * + * Valid configurations (Input1,Input2) -> Output : + * + * - (U8,U8) -> U8 + * - (U8,U8) -> S16 + * - (U8,S16) -> S16 + * - (S16,U8) -> S16 + * - (S16,S16) -> S16 + * - (F16,F16) -> F16 + * - (F32,F32) -> F32 + * - (QASYMM8,QASYMM8) -> QASYMM8 + * - (QASYMM8_SIGNED,QASYMM8_SIGNED) -> QASYMM8_SIGNED + * - (QSYMM16,QSYMM16) -> QSYMM16 + * - (QSYMM16,QSYMM16) -> S32 + * + * * @param[in] input1 An input tensor info. Data types supported: U8/QASYMM8/QASYMM8_SIGNED/S16/QSYMM16/F16/F32. - * @param[in] input2 An input tensor info. Data types supported: same as @p input1. - * @param[in] output The output tensor info. Data types supported: - * - U8, only if both input are U8 - * - QASYMM8, only if both inputs are QASYMM8 - * - QASYMM8_SIGNED, only if both inputs are QASYMM8_SIGNED - * - S16 - * - QSYMM16, only if both inputs are QSYMM16 - * - S32, only if both inputs are QSYMM16 - * - F16 - * - F32 + * @param[in] input2 An input tensor info. Data types supported: U8/QASYMM8/QASYMM8_SIGNED/S16/QSYMM16/F16/F32. + * @param[in] output The output tensor info. Data types supported: U8/QASYMM8/QASYMM8_SIGNED/S16/QSYMM16/F16/F32. * @param[in] scale Scale to apply after multiplication. * Scale must be positive and its value must be either 1/255 or 1/2^n where n is between 0 and 15. * @param[in] overflow_policy Overflow policy. Supported overflow policies: Wrap, Saturate diff --git a/arm_compute/runtime/CL/functions/CLReductionOperation.h b/arm_compute/runtime/CL/functions/CLReductionOperation.h index 013ba647b1..5d050d71d6 100644 --- a/arm_compute/runtime/CL/functions/CLReductionOperation.h +++ b/arm_compute/runtime/CL/functions/CLReductionOperation.h @@ -26,19 +26,17 @@ #include "arm_compute/core/CL/kernels/CLFillBorderKernel.h" #include "arm_compute/core/CL/kernels/CLReductionOperationKernel.h" -#include "arm_compute/core/Types.h" #include "arm_compute/runtime/CL/CLTensor.h" #include "arm_compute/runtime/CL/functions/CLReshapeLayer.h" #include "arm_compute/runtime/IFunction.h" #include "arm_compute/runtime/IMemoryManager.h" #include "arm_compute/runtime/MemoryGroup.h" -#include <cstdint> #include <memory> -#include <vector> namespace arm_compute { +// Forward declarations class ICLTensor; /** Perform reduction operation. @@ -54,7 +52,7 @@ public: /** Set the input and output tensors. * - * @param[in] input Source tensor. Data types supported: QASYMM8/QASYMM8_SIGNED/F16/F32. + * @param[in] input Source tensor. Data types supported: QASYMM8/QASYMM8_SIGNED/F16/F32/S32. * @param[out] output Destination tensor. Data types and data layouts supported: Same as @p input. * @param[in] axis Axis along which to reduce. Supported reduction axis : 0, 1, 2, 3 * @param[in] op Reduction operation to perform. Operations supported: MEAN_SUM, PROD, SUM_SQUARE, SUM, MIN, MAX @@ -64,7 +62,7 @@ public: /** Set the input and output tensors. * * @param[in] compile_context The compile context to be used. - * @param[in] input Source tensor. Data types supported: QASYMM8/QASYMM8_SIGNED/F16/F32. + * @param[in] input Source tensor. Data types supported: QASYMM8/QASYMM8_SIGNED/F16/F32/S32. * @param[out] output Destination tensor. Data types and data layouts supported: Same as @p input. * @param[in] axis Axis along which to reduce. Supported reduction axis : 0, 1, 2, 3 * @param[in] op Reduction operation to perform. Operations supported: MEAN_SUM, PROD, SUM_SQUARE, SUM, MIN, MAX @@ -74,7 +72,7 @@ public: /** Static function to check if given info will lead to a valid configuration of @ref CLReductionOperation. * - * @param[in] input Source tensor info. Data types supported: QASYMM8/QASYMM8_SIGNED/F16/F32. + * @param[in] input Source tensor info. Data types supported: QASYMM8/QASYMM8_SIGNED/F16/F32/S32. * @param[in] output Destination tensor info. Data types and data layouts supported: Same as @p input. * @param[in] axis Axis along which to reduce. Supported reduction axis : 0, 1, 2, 3 * @param[in] op Reduction operation to perform. Operations supported: MEAN_SUM, PROD, SUM_SQUARE, SUM, MIN, MAX @@ -95,7 +93,6 @@ private: std::vector<CLReductionOperationKernel> _reduction_kernels_vector; std::vector<CLFillBorderKernel> _border_handlers_vector; CLReshapeLayer _reshape; - ReductionOperation _op; unsigned int _num_of_stages; unsigned int _reduction_axis; bool _is_serial; diff --git a/arm_compute/runtime/CL/functions/CLSoftmaxLayer.h b/arm_compute/runtime/CL/functions/CLSoftmaxLayer.h index 93ad24e893..ec57bacf07 100644 --- a/arm_compute/runtime/CL/functions/CLSoftmaxLayer.h +++ b/arm_compute/runtime/CL/functions/CLSoftmaxLayer.h @@ -47,8 +47,6 @@ class ICLTensor; * @f[ out = (x - max(x) * beta) - log(\sum{e^{x - max(x) * beta}}) @f] * * This function runs the following kernels: - * -# @ref CLLogits1DMaxKernel - * -# @ref CLLogits1DShiftExpSumKernel * -# @ref CLLogits1DNormKernel * And if the reduce_end_axis is not 0, the function will use one of the the following kernels to reshape the input and * perform softmax on the reshaped input: @@ -63,36 +61,36 @@ public: CLSoftmaxLayerGeneric(std::shared_ptr<IMemoryManager> memory_manager = nullptr); /** Set the input and output tensors. * - * @param[in] input Source tensor. Data types supported: QASYMM8/F16/F32 for Softmax and F16/F32 for Log Softmax + * @param[in] input Source tensor. Data types supported: QASYMM8/QASYMM8_SIGNED/F16/F32 for Softmax and F16/F32 for Log Softmax * @param[out] output Destination tensor. Data types supported: same as @p input * @param[in] beta (Optional) A scaling factor for the exponent. Defaults to 1.f * @param[in] reduce_end_axis (Optional) The last axis of the first n dimensions (inclusive)to reduce. Defaults to 0. - * It has the purpose of squashing together the first n dimensions till (including) the @p reduce_end_axis. For instance, given a [2x3x4x5] image, - * when @p reduce_end_axis is 1, the reduction will be applied to axes 0 and 1, and the Softmax op will be applied on each of the [2x3] planes of the input image. - * Must be in range [0, input_num_dimensions). + * It has the purpose of squashing together the first n dimensions till (including) the @p reduce_end_axis. For instance, given a [2x3x4x5] image, + * when @p reduce_end_axis is 1, the reduction will be applied to axes 0 and 1, and the Softmax op will be applied on each of the [2x3] planes of the input image. + * Must be in range [0, input_num_dimensions). */ void configure(const ICLTensor *input, ICLTensor *output, float beta = 1.0f, size_t reduce_end_axis = 0); /** Set the input and output tensors. * * @param[in] compile_context The compile context to be used. - * @param[in] input Source tensor. Data types supported: QASYMM8/F16/F32 for Softmax and F16/F32 for Log Softmax + * @param[in] input Source tensor. Data types supported: QASYMM8/QASYMM8_SIGNED/F16/F32 for Softmax and F16/F32 for Log Softmax * @param[out] output Destination tensor. Data types supported: same as @p input * @param[in] beta (Optional) A scaling factor for the exponent. Defaults to 1.f * @param[in] reduce_end_axis (Optional) The last axis of the first n dimensions (inclusive)to reduce. Defaults to 0. - * It has the purpose of squashing together the first n dimensions till (including) the @p reduce_end_axis. For instance, given a [2x3x4x5] image, - * when @p reduce_end_axis is 1, the reduction will be applied to axes 0 and 1, and the Softmax op will be applied on each of the [2x3] planes of the input image. - * Must be in range [0, input_num_dimensions). + * It has the purpose of squashing together the first n dimensions till (including) the @p reduce_end_axis. For instance, given a [2x3x4x5] image, + * when @p reduce_end_axis is 1, the reduction will be applied to axes 0 and 1, and the Softmax op will be applied on each of the [2x3] planes of the input image. + * Must be in range [0, input_num_dimensions). */ void configure(const CLCompileContext &compile_context, const ICLTensor *input, ICLTensor *output, float beta = 1.0f, size_t reduce_end_axis = 0); /** Static function to check if given info will lead to a valid configuration of @ref CLSoftmaxLayer * - * @param[in] input Source tensor. Data types supported: QASYMM8/F16/F32 for Softmax and F16/F32 for Log Softmax + * @param[in] input Source tensor. Data types supported: QASYMM8/QASYMM8_SIGNED/F16/F32 for Softmax and F16/F32 for Log Softmax * @param[in] output Destination tensor. Data types supported: same as @p input * @param[in] beta (Optional) A scaling factor for the exponent. Defaults to 1.f * @param[in] reduce_end_axis (Optional) The last axis of the first n dimensions (inclusive)to reduce. Defaults to 0. - * It has the purpose of squashing together the first n dimensions till (including) the @p reduce_end_axis. For instance, given a [2x3x4x5] image, - * when @p reduce_end_axis is 1, the reduction will be applied to axes 0 and 1, and the Softmax op will be applied on each of the [2x3] planes of the input image. - * Must be in range [0, input_num_dimensions). + * It has the purpose of squashing together the first n dimensions till (including) the @p reduce_end_axis. For instance, given a [2x3x4x5] image, + * when @p reduce_end_axis is 1, the reduction will be applied to axes 0 and 1, and the Softmax op will be applied on each of the [2x3] planes of the input image. + * Must be in range [0, input_num_dimensions). * @return a status */ static Status validate(const ITensorInfo *input, const ITensorInfo *output, float beta = 1.0f, size_t reduce_end_axis = 0); @@ -111,9 +109,9 @@ private: * @param[in] input Original source tensor. * @param[in] output Original destination tensor. * @param[in] reduce_end_axis (Optional) The last axis of the first n dimensions (inclusive)to reduce. Defaults to 0. - * It has the purpose of squashing together the first n dimensions till (including) the @p reduce_end_axis. For instance, given a [2x3x4x5] image, - * when @p reduce_end_axis is 1, the reduction will be applied to axes 0 and 1, and the Softmax op will be applied on each of the [2x3] planes of the input image. - * Must be in range [0, input_num_dimensions). + * It has the purpose of squashing together the first n dimensions till (including) the @p reduce_end_axis. For instance, given a [2x3x4x5] image, + * when @p reduce_end_axis is 1, the reduction will be applied to axes 0 and 1, and the Softmax op will be applied on each of the [2x3] planes of the input image. + * Must be in range [0, input_num_dimensions). */ void configure_reshape_input_kernel(const ICLTensor *input, const ICLTensor *output, size_t reduce_end_axis); /** Utility method to configure the kernels needed to flatten the input @@ -127,9 +125,9 @@ private: * @param[in] input Original source tensor. * @param[in] output Original destination tensor. * @param[in] reduce_end_axis (Optional) The last axis of the first n dimensions (inclusive)to reduce. Defaults to 0. - * It has the purpose of squashing together the first n dimensions till (including) the @p reduce_end_axis. For instance, given a [2x3x4x5] image, - * when @p reduce_end_axis is 1, the reduction will be applied to axes 0 and 1, and the Softmax op will be applied on each of the [2x3] planes of the input image. - * Must be in range [0, input_num_dimensions). + * It has the purpose of squashing together the first n dimensions till (including) the @p reduce_end_axis. For instance, given a [2x3x4x5] image, + * when @p reduce_end_axis is 1, the reduction will be applied to axes 0 and 1, and the Softmax op will be applied on each of the [2x3] planes of the input image. + * Must be in range [0, input_num_dimensions). */ void configure_reshape_input_kernel(const CLCompileContext &compile_context, const ICLTensor *input, const ICLTensor *output, size_t reduce_end_axis); diff --git a/arm_compute/runtime/CL/functions/CLUnstack.h b/arm_compute/runtime/CL/functions/CLUnstack.h index bdef44fb0d..5d4d5710ab 100644 --- a/arm_compute/runtime/CL/functions/CLUnstack.h +++ b/arm_compute/runtime/CL/functions/CLUnstack.h @@ -48,8 +48,8 @@ public: CLUnstack(); /** Set the input, output and unstacking axis. * - * @param[in] input A tensor to be unstacked. Data type supported: U8/S8/QASYMM8/U16/S16/U32/S32/F16/F32. - * @param[in,out] output_vector A vector of tensors. Data types supported: Same as @p input. + * @param[in] input A tensor to be unstacked. Data type supported: All. + * @param[in,out] output_vector A vector of tensors. Data types supported: same as @p input. * Note: The number of elements of the vector will be used as the number of slices to be taken from the axis. * @param[in] axis The axis to unstack along. Valid values are [-R,R) where R is the input's rank. Negative values wrap around. * @@ -58,8 +58,8 @@ public: /** Set the input, output and unstacking axis. * * @param[in] compile_context The compile context to be used. - * @param[in] input A tensor to be unstacked. Data type supported: U8/S8/QASYMM8/U16/S16/U32/S32/F16/F32. - * @param[in,out] output_vector A vector of tensors. Data types supported: Same as @p input. + * @param[in] input A tensor to be unstacked. Data type supported: All. + * @param[in,out] output_vector A vector of tensors. Data types supported: same as @p input. * Note: The number of elements of the vector will be used as the number of slices to be taken from the axis. * @param[in] axis The axis to unstack along. Valid values are [-R,R) where R is the input's rank. Negative values wrap around. * @@ -67,8 +67,8 @@ public: void configure(const CLCompileContext &compile_context, const ICLTensor *input, const std::vector<ICLTensor *> &output_vector, int axis); /** Static function to check if given info will lead to a valid configuration of @ref CLUnstack * - * @param[in] input Input tensor info. Data type supported: U8/S8/QASYMM8/U16/S16/U32/S32/F16/F32 - * @param[in] output_vector Vector of output tensors' info. Data types supported: Same as @p input. + * @param[in] input Input tensor info. Data type supported: All. + * @param[in] output_vector Vector of output tensors' info. Data types supported: same as @p input. * @param[in] axis The axis to unstack along. Valid values are [-R,R) where R is the input's rank. Negative values wrap around. * * @return a status diff --git a/arm_compute/runtime/CL/functions/CLUpsampleLayer.h b/arm_compute/runtime/CL/functions/CLUpsampleLayer.h index dcaa13471f..07b4c8aecb 100644 --- a/arm_compute/runtime/CL/functions/CLUpsampleLayer.h +++ b/arm_compute/runtime/CL/functions/CLUpsampleLayer.h @@ -53,7 +53,7 @@ public: /** Initialize the function's source, destination, interpolation type and border_mode. * - * @param[in] input Source tensor. Data type supported: QASYMM8/QASYMM8_SIGNED/F16/F32. + * @param[in] input Source tensor. Data type supported: All. * @param[out] output Destination tensor. Data types supported: same as @p input. * @param[in] info Contains stride information described in @ref Size2D. * @param[in] upsampling_policy Defines the policy to fill the intermediate pixels. @@ -63,16 +63,16 @@ public: /** Initialize the function's source, destination, interpolation type and border_mode. * * @param[in] compile_context The compile context to be used. - * @param[in] input Source tensor. Data type supported: QASYMM8/QASYMM8_SIGNED/F16/F32. + * @param[in] input Source tensor. Data type supported: All. * @param[out] output Destination tensor. Data types supported: same as @p input. * @param[in] info Contains stride information described in @ref Size2D. * @param[in] upsampling_policy Defines the policy to fill the intermediate pixels. */ void configure(const CLCompileContext &compile_context, ICLTensor *input, ICLTensor *output, const Size2D &info, const InterpolationPolicy upsampling_policy); - /** Static function to check if given info will lead to a valid configuration of @ref CLDeconvolutionLayerUpsample + /** Static function to check if given info will lead to a valid configuration of @ref CLUpsampleLayerKernel * - * @param[in] input Source tensor info. Data types supported: QASYMM8/QASYMM8_SIGNED/F16/F32. + * @param[in] input Source tensor info. Data types supported: All. * @param[in] output Destination tensor info. Data types supported: same as @p input. * @param[in] info Contains stride information described in @ref Size2D. * @param[in] upsampling_policy Defines the policy to fill the intermediate pixels. |