potok/opencv
1
0
Fork 0
mirror of https://github.com/opencv/opencv.git synced 2026-07-29 15:23:05 +04:00
Code Packages Releases Wiki Activity

fp16 ocl support for more layers

Browse Source 
Signed-off-by: Li Peng <peng.li@intel.com>
This commit is contained in:
Li Peng
2018-04-26 19:36:19 +08:00
parent 3dd916882a
commit ba5e8befa9
14 changed files with 121 additions and 50 deletions
Download Patch File Download Diff File Expand all files Collapse all files
modules/dnn/src/layers/permute_layer.cpp
+4 -2
View File
@@ -288,9 +288,11 @@ public:
if (!_needsPermute)
return false;
bool use_half = (inps.depth() == CV_16S);
String opts = format("-DDtype=%s", use_half ? "half" : "float");
for (size_t i = 0; i < inputs.size(); i++)
{
ocl::Kernel kernel("permute", ocl::dnn::permute_oclsrc);
ocl::Kernel kernel("permute", ocl::dnn::permute_oclsrc, opts);
kernel.set(0, (int)_count);
kernel.set(1, ocl::KernelArg::PtrReadOnly(inputs[i]));
@@ -313,7 +315,7 @@ public:
CV_TRACE_FUNCTION();
CV_TRACE_ARG_VALUE(name, "name", name.c_str());
CV_OCL_RUN((preferableTarget == DNN_TARGET_OPENCL) &&
CV_OCL_RUN(IS_DNN_OPENCL_TARGET(preferableTarget) &&
OCL_PERFORMANCE_CHECK(ocl::Device::getDefault().isIntel()),
forward_ocl(inputs_arr, outputs_arr, internals_arr))