diff options
| author | Gian Marco Iodice <gianmarco.iodice@arm.com> | 2021-05-11 11:13:20 +0100 |
|---|---|---|
| committer | Georgios Pinitas <georgios.pinitas@arm.com> | 2021-05-14 09:34:57 +0000 |
| commit | 186fe683da63dea2dac06e46a412e354d33cd9c2 (patch) | |
| tree | e085313b7cfbb98f1446f74dd8cef937726b02b4 | |
| parent | 4f7693d8757cf12c33f049c61c63bc689379ab84 (diff) | |
| download | ComputeLibrary-186fe683da63dea2dac06e46a412e354d33cd9c2.tar.gz | |
Fix the OpenCL convolution layer heuristic for int8
- Call GEMM-based convolution rather than Direct convolution for
quantized data types
Resolves COMPMID-4497
Change-Id: Idde5808b8a9de69c5196c56538520c0e3d52ba76
Signed-off-by: Gian Marco Iodice <gianmarco.iodice@arm.com>
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/5618
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: Georgios Pinitas <georgios.pinitas@arm.com>
| -rw-r--r-- | src/runtime/CL/functions/CLConvolutionLayer.cpp | 2 |
1 files changed, 1 insertions, 1 deletions
diff --git a/src/runtime/CL/functions/CLConvolutionLayer.cpp b/src/runtime/CL/functions/CLConvolutionLayer.cpp index 1082a475b2..96d7cc72c2 100644 --- a/src/runtime/CL/functions/CLConvolutionLayer.cpp +++ b/src/runtime/CL/functions/CLConvolutionLayer.cpp @@ -232,7 +232,7 @@ ConvolutionMethod CLConvolutionLayer::get_convolution_method(const ITensorInfo * return ConvolutionMethod::FFT; } } - else + else if(is_data_type_float(input->data_type())) { if((weights->dimension(idx_h) >= 5) && (input->dimension(idx_c) >= output->dimension(idx_c)) && (CLDirectConvolutionLayer::validate(input, weights, nullptr, output, conv_info, act_info))) { |