diff options
author | Giorgio Arena <giorgio.arena@arm.com> | 2018-04-04 17:44:26 +0100 |
---|---|---|
committer | Anthony Barbier <anthony.barbier@arm.com> | 2018-11-02 16:50:48 +0000 |
commit | 7657224de2b697a8a92cccf26d98e53ccd7c1a03 (patch) | |
tree | 1dcfa4541dbaf753854a628c93991652158d373e /src/core/CL/cl_kernels/depthwise_convolution_quantized.cl | |
parent | e74b201ca1abca040ca9f30837fdf19aa610e7c4 (diff) | |
download | ComputeLibrary-7657224de2b697a8a92cccf26d98e53ccd7c1a03.tar.gz |
COMPMID-926 Add depth multiplier support to NEON/CL/GLES depthwise convolution
Change-Id: I03f32c62350e5ea43e77bb15fc5a832d83719e3b
Reviewed-on: https://eu-gerrit-1.euhpc.arm.com/126657
Tested-by: Jenkins <bsgcomp@arm.com>
Reviewed-by: Michele DiGiorgio <michele.digiorgio@arm.com>
Reviewed-by: Georgios Pinitas <georgios.pinitas@arm.com>
Diffstat (limited to 'src/core/CL/cl_kernels/depthwise_convolution_quantized.cl')
-rw-r--r-- | src/core/CL/cl_kernels/depthwise_convolution_quantized.cl | 2 |
1 files changed, 2 insertions, 0 deletions
diff --git a/src/core/CL/cl_kernels/depthwise_convolution_quantized.cl b/src/core/CL/cl_kernels/depthwise_convolution_quantized.cl index a0c0a8b1fb..ccb3a1ffe2 100644 --- a/src/core/CL/cl_kernels/depthwise_convolution_quantized.cl +++ b/src/core/CL/cl_kernels/depthwise_convolution_quantized.cl @@ -126,6 +126,8 @@ __kernel void depthwise_convolution_3x3_quantized_nchw( int bias_value = *((__global int *)(vector_offset(&biases, get_global_id(2)))); #endif //defined(HAS_BIAS) + src.ptr -= (get_global_id(2) - get_global_id(2) / DEPTH_MULTIPLIER) * src_step_z; + uchar3 w0 = vload3(0, weights.ptr + 0 * weights_stride_y); uchar3 w1 = vload3(0, weights.ptr + 1 * weights_stride_y); uchar3 w2 = vload3(0, weights.ptr + 2 * weights_stride_y); |