diff options
author | Vidhya Sudhan Loganathan <vidhyasudhan.loganathan@arm.com> | 2018-05-25 13:21:26 +0100 |
---|---|---|
committer | Anthony Barbier <anthony.barbier@arm.com> | 2018-11-02 16:52:54 +0000 |
commit | f1f490634f3273e4669f16e663071554df291bea (patch) | |
tree | 3e4e39f3aa5a7ef26cd0d01079a971ad949738e4 /src/core/CL/kernels/CLArithmeticAdditionKernel.cpp | |
parent | 2345f4349761d6c78200f43e1a78289d215825c4 (diff) | |
download | ComputeLibrary-f1f490634f3273e4669f16e663071554df291bea.tar.gz |
COMPMID-655 : Check FP16 is supported by the GPU
Change-Id: I507b04680a4e88426b682bd0be03bccb560ec78d
Reviewed-on: https://eu-gerrit-1.euhpc.arm.com/132589
Tested-by: Jenkins <bsgcomp@arm.com>
Reviewed-by: Anthony Barbier <anthony.barbier@arm.com>
Diffstat (limited to 'src/core/CL/kernels/CLArithmeticAdditionKernel.cpp')
-rw-r--r-- | src/core/CL/kernels/CLArithmeticAdditionKernel.cpp | 5 |
1 files changed, 4 insertions, 1 deletions
diff --git a/src/core/CL/kernels/CLArithmeticAdditionKernel.cpp b/src/core/CL/kernels/CLArithmeticAdditionKernel.cpp index c4904ecbe1..b0177ab8b6 100644 --- a/src/core/CL/kernels/CLArithmeticAdditionKernel.cpp +++ b/src/core/CL/kernels/CLArithmeticAdditionKernel.cpp @@ -24,6 +24,7 @@ #include "arm_compute/core/CL/kernels/CLArithmeticAdditionKernel.h" #include "arm_compute/core/CL/CLHelpers.h" +#include "arm_compute/core/CL/CLValidate.h" #include "arm_compute/core/CL/ICLTensor.h" using namespace arm_compute; @@ -35,8 +36,9 @@ constexpr unsigned int num_elems_processed_per_iteration = 16; Status validate_arguments(const ITensorInfo &input1, const ITensorInfo &input2, const ITensorInfo &output, ConvertPolicy policy) { ARM_COMPUTE_UNUSED(policy); - + ARM_COMPUTE_RETURN_ERROR_ON_F16_UNSUPPORTED(&input1); ARM_COMPUTE_RETURN_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(&input1, 1, DataType::U8, DataType::QS8, DataType::QS16, DataType::S16, DataType::F16, DataType::F32); + ARM_COMPUTE_RETURN_ERROR_ON_F16_UNSUPPORTED(&input2); ARM_COMPUTE_RETURN_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(&input2, 1, DataType::U8, DataType::QS8, DataType::QS16, DataType::S16, DataType::F16, DataType::F32); const TensorShape out_shape = TensorShape::broadcast_shape(input1.tensor_shape(), input2.tensor_shape()); @@ -47,6 +49,7 @@ Status validate_arguments(const ITensorInfo &input1, const ITensorInfo &input2, // Validate in case of configured output if(output.total_size() > 0) { + ARM_COMPUTE_RETURN_ERROR_ON_F16_UNSUPPORTED(&output); ARM_COMPUTE_RETURN_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(&output, 1, DataType::U8, DataType::QS8, DataType::QS16, DataType::S16, DataType::F16, DataType::F32); ARM_COMPUTE_RETURN_ERROR_ON_MSG((output.data_type() == DataType::U8) && ((input1.data_type() != DataType::U8) || (input2.data_type() != DataType::U8)), "Output can only be U8 if both inputs are U8"); |