aboutsummaryrefslogtreecommitdiff
diff options
context:
space:
mode:
authorPablo Tello <pablo.tello@arm.com>2019-02-27 13:32:51 +0000
committerGiuseppe Rossini <giuseppe.rossini@arm.com>2019-03-11 12:55:11 +0000
commitda0e5139cbd76a7077ee553abc34b09b655d00a9 (patch)
treefdece4673a0c40fb7b1d1321931fea04f1b8d6e1
parent471043616a869f0e696c8db4e1d0a62b45b4decf (diff)
downloadComputeLibrary-da0e5139cbd76a7077ee553abc34b09b655d00a9.tar.gz
COMPMID-1882: Improve memory coalescence when reshaping the weights for CLDepthwiseConvolution
Change-Id: Ib226fd3e7d2715847594eb5bafa46c938672cf9b Signed-off-by: giuros01 <giuseppe.rossini@arm.com> Reviewed-on: https://review.mlplatform.org/c/839 Reviewed-by: Georgios Pinitas <georgios.pinitas@arm.com> Tested-by: Arm Jenkins <bsgcomp@arm.com>
-rw-r--r--src/core/CL/kernels/CLDepthwiseConvolutionLayer3x3NCHWKernel.cpp2
1 files changed, 1 insertions, 1 deletions
diff --git a/src/core/CL/kernels/CLDepthwiseConvolutionLayer3x3NCHWKernel.cpp b/src/core/CL/kernels/CLDepthwiseConvolutionLayer3x3NCHWKernel.cpp
index 770740d180..83fac16d88 100644
--- a/src/core/CL/kernels/CLDepthwiseConvolutionLayer3x3NCHWKernel.cpp
+++ b/src/core/CL/kernels/CLDepthwiseConvolutionLayer3x3NCHWKernel.cpp
@@ -173,7 +173,7 @@ std::pair<Status, Window> validate_and_configure_window(ITensorInfo *input, ITen
kernel_name = is_qasymm ? "dwc_3x3_native_qasymm8" : "depthwise_convolution_3x3";
kernel_name += (is_qasymm && is_dot8_supported ? "_dot8" : "");
- kernel_name += "_nchw";
+ kernel_name += (is_qasymm ? "_nchw" : "");
num_elems_written_per_iteration_x = 8 / data_size_from_type(input->data_type());
num_elems_written_per_iteration_y = (is_qasymm && conv_stride_y == 1) ? 2 : 1;