2 // Copyright (c) 2016-2019 Intel Corporation
4 // Licensed under the Apache License, Version 2.0 (the "License");
5 // you may not use this file except in compliance with the License.
6 // You may obtain a copy of the License at
8 // http://www.apache.org/licenses/LICENSE-2.0
10 // Unless required by applicable law or agreed to in writing, software
11 // distributed under the License is distributed on an "AS IS" BASIS,
12 // WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
13 // See the License for the specific language governing permissions and
14 // limitations under the License.
17 #include "permute_kernel_ref.h"
18 #include "kernel_selector_utils.h"
20 namespace kernel_selector
22 ParamsKey PermuteKernelRef::GetSupportedKey() const
25 k.EnableInputDataType(Datatype::F16);
26 k.EnableInputDataType(Datatype::F32);
27 k.EnableInputDataType(Datatype::INT8);
28 k.EnableInputDataType(Datatype::INT32);
29 k.EnableInputDataType(Datatype::INT64);
30 k.EnableOutputDataType(Datatype::F16);
31 k.EnableOutputDataType(Datatype::F32);
32 k.EnableOutputDataType(Datatype::INT8);
33 k.EnableOutputDataType(Datatype::INT32);
34 k.EnableOutputDataType(Datatype::INT64);
35 k.EnableAllInputLayout();
36 k.EnableAllOutputLayout();
37 k.EnableTensorOffset();
38 k.EnableTensorPitches();
43 inline JitConstants MakePermuteJitConstants(const permute_params& params)
45 JitConstants jit = MakeBaseParamsJitConstants(params);;
46 jit.AddConstant(MakeJitConstant("PERMUTE_ORDER", params.order));
50 KernelsData PermuteKernelRef::GetKernelsData(const Params& params, const optional_params& options) const
52 assert(params.GetType() == KernelType::PERMUTE);
54 KernelData kd = KernelData::Default<permute_params>(params);
55 permute_params& newParams = *static_cast<permute_params*>(kd.params.get());
58 auto entry_point = GetEntryPoint(kernelName, newParams.layerID, options);
59 auto cldnn_jit = MakePermuteJitConstants(newParams);
60 std::string jit = CreateJit(kernelName, cldnn_jit, entry_point);
62 const auto& in = newParams.inputs[0];
63 auto& kernel = kd.kernels[0];
65 kernel.workGroups.global = { in.Y().v, in.X().v, in.Feature().v * in.Batch().v};
66 kernel.workGroups.local = GetOptimalLocalWorkGroupSizes(kernel.workGroups.global);
67 kernel.kernelString = GetKernelString(kernelName, jit, entry_point, params.engineInfo, DEFAULT);
68 kernel.arguments = GetArgsDesc(1, false, false);
70 kd.estimatedTime = DONT_USE_IF_HAVE_SOMETHING_ELSE;