From 9637b2e4fc33b2264aa5586dd6b2ed1045db5075 Mon Sep 17 00:00:00 2001 From: Michele Di Giorgio Date: Mon, 23 Sep 2019 16:49:49 +0100 Subject: COMPMID-2671: Change ArgMinMax NEON/CL output type to Signed32 Change-Id: I718f3884928271c5b0afb259d5bfe9df284f18e6 Signed-off-by: Michele Di Giorgio Reviewed-on: https://review.mlplatform.org/c/1995 Reviewed-by: Georgios Pinitas Comments-Addressed: Arm Jenkins Tested-by: Arm Jenkins --- arm_compute/core/CL/kernels/CLReductionOperationKernel.h | 12 +++++++++--- 1 file changed, 9 insertions(+), 3 deletions(-) (limited to 'arm_compute/core/CL/kernels/CLReductionOperationKernel.h') diff --git a/arm_compute/core/CL/kernels/CLReductionOperationKernel.h b/arm_compute/core/CL/kernels/CLReductionOperationKernel.h index aba11e1ad1..18a1bab2dc 100644 --- a/arm_compute/core/CL/kernels/CLReductionOperationKernel.h +++ b/arm_compute/core/CL/kernels/CLReductionOperationKernel.h @@ -31,7 +31,13 @@ namespace arm_compute { class ICLTensor; -/** Interface for the reduction operation kernel */ +/** Interface for the reduction operation kernel + * + * @note For ARG_MIN/ARG_MAX reduction, the indices are computed in unsigned + * 32-bit (U32). It is the user's responsibility to check that the + * results do not overflow in case the output data type is set to signed + * 32-bit integer (S32). + */ class CLReductionOperationKernel : public ICLKernel { public: @@ -51,7 +57,7 @@ public: /** Set the input and output tensors. * * @param[in] input Source tensor. Data types supported: QASYMM8/S32/F16/F32. - * @param[out] output Destination tensor. Data types and data layouts supported: Same as @p input. + * @param[out] output Destination tensor. Data types and data layouts supported: Same as @p input, U32/S32 for ARG_MIX/ARG_MAX. * Output will have the same number of dimensions as input. * @param[in] axis Axis along which to reduce. Supported reduction axis : 0,1,2,3 * @param[in] op Reduction operation to perform. @@ -62,7 +68,7 @@ public: /** Static function to check if given info will lead to a valid configuration of @ref CLReductionOperationKernel. * * @param[in] input Source tensor info. Data types supported: QASYMM8/S32/F16/F32. - * @param[in] output Destination tensor info. Data types and data layouts supported: Same as @p input. + * @param[in] output Destination tensor info. Data types and data layouts supported: Same as @p input, U32/S32 for ARG_MIX/ARG_MAX. * Output will have the same number of dimensions as input. * @param[in] axis Axis along which to reduce. Supported reduction axis : 0,1,2,3 * @param[in] op Reduction operation to perform. -- cgit v1.2.1