diff options
| author | Gian Marco Iodice <gianmarco.iodice@arm.com> | 2017-06-22 12:09:49 +0100 |
|---|---|---|
| committer | Anthony Barbier <anthony.barbier@arm.com> | 2018-09-17 14:14:20 +0100 |
| commit | 9f89baebb81e6a01ec06fe916564da45eb204f34 (patch) | |
| tree | edd21a058a2701f0e89456717e10011ab44d473a /src/core/CL/kernels/CLGEMMInterleave4x4Kernel.cpp | |
| parent | 10c672c2e21bb77b7234d9d3611267400dce7ae0 (diff) | |
| download | ComputeLibrary-9f89baebb81e6a01ec06fe916564da45eb204f34.tar.gz | |
COMPMID-411 - Ported CLGEMMInterleave4x4Kernel and CLGEMMTranspose1xWKernel to support 8 bit fixed point
Change-Id: If236c9047ed536e808a0ed26e97e1799ca938e03
Reviewed-on: http://mpd-gerrit.cambridge.arm.com/78529
Tested-by: Kaizen <jeremy.johnson+kaizengerrit@arm.com>
Reviewed-by: Moritz Pflanzer <moritz.pflanzer@arm.com>
Reviewed-by: Georgios Pinitas <georgios.pinitas@arm.com>
Diffstat (limited to 'src/core/CL/kernels/CLGEMMInterleave4x4Kernel.cpp')
| -rw-r--r-- | src/core/CL/kernels/CLGEMMInterleave4x4Kernel.cpp | 16 |
1 files changed, 12 insertions, 4 deletions
diff --git a/src/core/CL/kernels/CLGEMMInterleave4x4Kernel.cpp b/src/core/CL/kernels/CLGEMMInterleave4x4Kernel.cpp index 71d42c5606..7312cc25cb 100644 --- a/src/core/CL/kernels/CLGEMMInterleave4x4Kernel.cpp +++ b/src/core/CL/kernels/CLGEMMInterleave4x4Kernel.cpp @@ -43,11 +43,19 @@ 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::U16, DataType::S16, DataType::U32, DataType::S32, DataType::F16, DataType::F32); - ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(output, 1, DataType::U8, DataType::S8, DataType::U16, DataType::S16, 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::U16, DataType::S16, DataType::U32, DataType::S32, DataType::F16, DataType::F32); + ARM_COMPUTE_ERROR_ON_NULLPTR(output); + + TensorShape output_shape = input->info()->tensor_shape(); + output_shape.set(0, input->info()->dimension(0) * 4); + 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()); + + ARM_COMPUTE_ERROR_ON_MISMATCHING_DIMENSIONS(output->info()->tensor_shape(), output_shape); ARM_COMPUTE_ERROR_ON_MISMATCHING_DATA_TYPES(input, output); - ARM_COMPUTE_ERROR_ON(output->info()->dimension(0) != input->info()->dimension(0) * 4); - ARM_COMPUTE_ERROR_ON(output->info()->dimension(1) != std::ceil(static_cast<float>(input->info()->dimension(1)) / 4.0f)); + ARM_COMPUTE_ERROR_ON_MISMATCHING_FIXED_POINT(input, output); _input = input; _output = output; |