COMPMID-631: Merge branches/gles_compute branch
Last commit:
commit b25c5f68042b0c81bf611d59a1bb8535e1c42497
Author: Xinghang Zhou <xinghang.zhou@arm.com>
Date: Wed Oct 25 18:48:10 2017 +0800
Synced validation's tolerances of GCSoftmax from cl side
Change-Id: Ibe72054205c1c8721845d679a31af7ed0a7c5cf6
Reviewed-on: http://mpd-gerrit.cambridge.arm.com/93283
Reviewed-by: Anthony Barbier <anthony.barbier@arm.com>
Tested-by: Kaizen <jeremy.johnson+kaizengerrit@arm.com>
diff --git a/src/core/GLES_COMPUTE/kernels/GCDropoutKernel.cpp b/src/core/GLES_COMPUTE/kernels/GCDropoutKernel.cpp
new file mode 100644
index 0000000..6244fbe
--- /dev/null
+++ b/src/core/GLES_COMPUTE/kernels/GCDropoutKernel.cpp
@@ -0,0 +1,110 @@
+/*
+ * Copyright (c) 2017 ARM Limited.
+ *
+ * SPDX-License-Identifier: MIT
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a copy
+ * of this software and associated documentation files (the "Software"), to
+ * deal in the Software without restriction, including without limitation the
+ * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
+ * sell copies of the Software, and to permit persons to whom the Software is
+ * furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in all
+ * copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+ * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+ * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#include "arm_compute/core/GLES_COMPUTE/kernels/GCDropoutKernel.h"
+
+#include "arm_compute/core/Error.h"
+#include "arm_compute/core/GLES_COMPUTE/GCHelpers.h"
+#include "arm_compute/core/GLES_COMPUTE/GCKernelLibrary.h"
+#include "arm_compute/core/GLES_COMPUTE/IGCTensor.h"
+#include "arm_compute/core/GLES_COMPUTE/OpenGLES.h"
+#include "arm_compute/core/Helpers.h"
+#include "arm_compute/core/Types.h"
+#include "arm_compute/core/Validate.h"
+#include "support/ToolchainSupport.h"
+
+#include <cmath>
+#include <random>
+#include <tuple>
+
+using namespace arm_compute;
+
+GCDropoutKernel::GCDropoutKernel()
+ : _input(nullptr), _mask(nullptr), _output(nullptr), _num_elems_processed_per_iteration(0)
+{
+}
+
+void GCDropoutKernel::configure(const IGCTensor *input, IGCTensor *mask, IGCTensor *output, float ratio, bool forward)
+{
+ ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input, 1, DataType::F16, DataType::F32);
+ ARM_COMPUTE_ERROR_ON_MISMATCHING_DATA_TYPES(input, mask, output);
+
+ _input = input;
+ _mask = mask;
+ _output = output;
+ _kernel.clear_params();
+
+ std::set<std::string> build_opts;
+ std::string dt_name = (input->info()->data_type() == DataType::F32) ? "DATA_TYPE_FP32" : "DATA_TYPE_FP16";
+ std::string fporbp = forward ? "FORWARD" : "BACKWARD";
+ std::random_device rd;
+ std::mt19937 mt(rd());
+ std::uniform_real_distribution<float> dist(0.f, 1.f);
+
+ build_opts.emplace("#define LOCAL_SIZE_X " + support::cpp11::to_string(1));
+ build_opts.emplace("#define LOCAL_SIZE_Y " + support::cpp11::to_string(1));
+ build_opts.emplace("#define LOCAL_SIZE_Z " + support::cpp11::to_string(1));
+ build_opts.emplace("#define RATIO " + support::cpp11::to_string(ratio));
+ build_opts.emplace("#define SCALE " + support::cpp11::to_string(1. / (1. - ratio)));
+ build_opts.emplace("#define SEED " + support::cpp11::to_string(dist(mt)));
+ build_opts.emplace("#define " + dt_name);
+ build_opts.emplace("#define " + fporbp);
+
+ _num_elems_processed_per_iteration = 4 / input->info()->element_size();
+
+ // Create kernel
+ _kernel = static_cast<GCKernel>(GCKernelLibrary::get().create_kernel("dropout", build_opts));
+
+ // Configure kernel window
+ Window win = calculate_max_window(*input->info(), Steps(_num_elems_processed_per_iteration));
+
+ output->info()->set_valid_region(ValidRegion(Coordinates(), output->info()->tensor_shape()));
+
+ // set shader params binding point
+ _kernel.set_shader_params_binding_point(0);
+ IGCKernel::configure(win);
+}
+
+void GCDropoutKernel::run(const Window &window)
+{
+ ARM_COMPUTE_ERROR_ON_UNCONFIGURED_KERNEL(this);
+ ARM_COMPUTE_ERROR_ON_MISMATCHING_WINDOWS(IGCKernel::window(), window);
+
+ _kernel.use();
+
+ Window slice = window.first_slice_window_3D();
+
+ do
+ {
+ unsigned int idx = 0;
+
+ add_3D_tensor_argument(idx, _input, BufferParam(1, 2), slice);
+ add_3D_tensor_argument(idx, _mask, BufferParam(2, 2), slice);
+ add_3D_tensor_argument(idx, _output, BufferParam(3, 2), slice);
+
+ _kernel.update_shader_params();
+ enqueue(*this, slice);
+ }
+ while(window.slide_window_slice_3D(slice));
+}