diff options
| author | Georgios Pinitas <georgios.pinitas@arm.com> | 2017-11-29 11:06:49 +0000 |
|---|---|---|
| committer | Anthony Barbier <anthony.barbier@arm.com> | 2018-11-02 16:41:58 +0000 |
| commit | 45bcc3a1c287a208098ae99288273a5129ddd5eb (patch) | |
| tree | f4f957dbc76f8e8e9a4871b16652e1033bcd4c73 /src/core/CL/kernels/CLGEMMInterleave4x4Kernel.cpp | |
| parent | 303be90ee1f03f75309b421297ba16428ea98ea5 (diff) | |
| download | ComputeLibrary-45bcc3a1c287a208098ae99288273a5129ddd5eb.tar.gz | |
COMPMID-661: QASYMM8 support for fully connected layer.
Change-Id: I70e04d3a175ba366432ada98e9ca893c9f81b260
Reviewed-on: https://eu-gerrit-1.euhpc.arm.com/111094
Reviewed-by: Gian Marco Iodice <gianmarco.iodice@arm.com>
Tested-by: BSG Visual Compute Jenkins server to access repositories on http://mpd-gerrit.cambridge.arm.com <bsgcomp@arm.com>
Reviewed-by: Anthony Barbier <anthony.barbier@arm.com>
Diffstat (limited to 'src/core/CL/kernels/CLGEMMInterleave4x4Kernel.cpp')
| -rw-r--r-- | src/core/CL/kernels/CLGEMMInterleave4x4Kernel.cpp | 14 |
1 files changed, 7 insertions, 7 deletions
diff --git a/src/core/CL/kernels/CLGEMMInterleave4x4Kernel.cpp b/src/core/CL/kernels/CLGEMMInterleave4x4Kernel.cpp index 984121c5bc..7741f12900 100644 --- a/src/core/CL/kernels/CLGEMMInterleave4x4Kernel.cpp +++ b/src/core/CL/kernels/CLGEMMInterleave4x4Kernel.cpp @@ -43,9 +43,10 @@ CLGEMMInterleave4x4Kernel::CLGEMMInterleave4x4Kernel() void CLGEMMInterleave4x4Kernel::configure(const ICLTensor *input, ICLTensor *output) { - ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input, 1, DataType::U8, DataType::S8, DataType::QS8, DataType::QASYMM8, DataType::U16, DataType::S16, DataType::QS16, DataType::U32, DataType::S32, - DataType::F16, - DataType::F32); + ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input, 1, DataType::U8, DataType::S8, DataType::QS8, DataType::QASYMM8, + DataType::U16, DataType::S16, DataType::QS16, + DataType::U32, DataType::S32, + DataType::F16, DataType::F32); ARM_COMPUTE_ERROR_ON_NULLPTR(output); TensorShape output_shape = input->info()->tensor_shape(); @@ -53,7 +54,7 @@ void CLGEMMInterleave4x4Kernel::configure(const ICLTensor *input, ICLTensor *out output_shape.set(1, std::ceil(input->info()->dimension(1) / 4.0f)); // Output auto inizialitation if not yet initialized - auto_init_if_empty(*output->info(), output_shape, 1, input->info()->data_type(), input->info()->fixed_point_position(), input->info()->quantization_info()); + auto_init_if_empty(*output->info(), input->info()->clone()->set_tensor_shape(output_shape)); ARM_COMPUTE_ERROR_ON_MISMATCHING_DIMENSIONS(output->info()->tensor_shape(), output_shape); ARM_COMPUTE_ERROR_ON_MISMATCHING_DATA_TYPES(input, output); @@ -63,9 +64,8 @@ void CLGEMMInterleave4x4Kernel::configure(const ICLTensor *input, ICLTensor *out _output = output; // Create kernel - std::string data_type_name; - data_type_name = support::cpp11::to_string(input->info()->element_size() * 8) + "bit"; - _kernel = static_cast<cl::Kernel>(CLKernelLibrary::get().create_kernel("gemm_interleave4x4_" + data_type_name)); + std::string kernel_name = "gemm_interleave4x4_" + support::cpp11::to_string(input->info()->element_size() * 8) + "bit"; + _kernel = static_cast<cl::Kernel>(CLKernelLibrary::get().create_kernel(kernel_name)); // Configure kernel window const unsigned int num_elems_processed_per_iteration_x = max_cl_vector_width / data_size_from_type(input->info()->data_type()); |