We read every piece of feedback, and take your input very seriously.
To see all available qualifiers, see our documentation.
There was an error while loading. Please reload this page.
1 parent b65d7e3 commit 7e83e39Copy full SHA for 7e83e39
1 file changed
modules/dnn/src/opencl/conv_spatial_helper.cl
@@ -84,5 +84,7 @@ __kernel void TEMPLATE(copyWeightsSwizzled, Dtype)
84
int idxOut = FP*(kernel_w*kernel_h*channels*swizzleFactor) + kernel_C*(kernel_w*kernel_h*swizzleFactor) + kernel_Y*(kernel_w*swizzleFactor) + kernel_X*swizzleFactor + F1;
85
int idxIn = filter*(kernel_w*kernel_h*channels) + kernel_C*(kernel_w*kernel_h) + kernel_Y*kernel_w + kernel_X;
86
87
- weightOut[idxOut] = weightIn[idxIn];
+ // idxIn is not valid if (filter >= outputs) - no data for these elements. Output gaps are filled by zeros
88
+ Dtype v = (filter < outputs) ? weightIn[idxIn] : (Dtype)0;
89
+ weightOut[idxOut] = v;
90
}
0 commit comments