commit | b609c93e81b1ff60764c5b724f4f5511215c84d6 | [log] [tgz] |
---|---|---|
author | Annop Wongwathanarat <annop.wongwathanarat@arm.com> | Mon Jan 16 14:36:45 2023 +0000 |
committer | Annop Wongwathanarat <annop.wongwathanarat@arm.com> | Mon Jan 30 15:59:43 2023 +0000 |
tree | 3822ed89e9b99778994362cadb1f62b97aef6764 | |
parent | fbbfa5388d7e430ff7e9a3fb0bff8045574706e4 [diff] [blame] |
Skip upsampling for deconvolution when not needed If the input tensor's stride is 1 and the kernel size is 1x1, skip upsampling step and pass the input tensor pointer for convolution directly. Partially resolve: [ONCPUML-1137] Change-Id: I9de9444ff99cf35d44a51ccbe0fa6facc1035d27 Signed-off-by: Annop Wongwathanarat <annop.wongwathanarat@arm.com> Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/8994 Tested-by: Arm Jenkins <bsgcomp@arm.com> Comments-Addressed: Arm Jenkins <bsgcomp@arm.com> Reviewed-by: Gunes Bayir <gunes.bayir@arm.com> Benchmark: Arm Jenkins <bsgcomp@arm.com>
diff --git a/arm_compute/runtime/NEON/functions/NEDeconvolutionLayer.h b/arm_compute/runtime/NEON/functions/NEDeconvolutionLayer.h index 15124d6..869df69 100644 --- a/arm_compute/runtime/NEON/functions/NEDeconvolutionLayer.h +++ b/arm_compute/runtime/NEON/functions/NEDeconvolutionLayer.h
@@ -148,6 +148,7 @@ ITensor *_input; PadStrideInfo _info; bool _is_prepared; + bool _do_upsampling; }; } // arm_compute #endif /* ARM_COMPUTE_NEDECONVOLUTIONLAYER_H */