Optimize Transposed Convolution for CL backend (FP32/16)
This patch optimizes transposed convolution for CL backend by rewriting it in a single kernel instead of three (flip_kernel + upsample + conv). The new kernel skips the upsampling step which reduces the input space of convolution by stride_x * stride_y, resulting in significant performance improvement. It also skips the kernel flipping by traversing the weights accordingly, thus reduces the memory footprint.
Resolves: COMPMID-5676
Signed-off-by: Gunes Bayir <gunes.bayir@arm.com>
Change-Id: I8a333212dc7c5f7f0597aa58b0d56d44814baa14
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/8588
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: Gian Marco Iodice <gianmarco.iodice@arm.com>
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
Benchmark: Arm Jenkins <bsgcomp@arm.com>
diff --git a/src/gpu/cl/kernels/ClTransposedConvolutionKernel.h b/src/gpu/cl/kernels/ClTransposedConvolutionKernel.h
new file mode 100644
index 0000000..d4350dd
--- /dev/null
+++ b/src/gpu/cl/kernels/ClTransposedConvolutionKernel.h
@@ -0,0 +1,66 @@
+/*
+ * Copyright (c) 2022 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.
+ */
+#ifndef ARM_COMPUTE_CL_TRANSPOSED_CONVOLUTION_KERNEL_H
+#define ARM_COMPUTE_CL_TRANSPOSED_CONVOLUTION_KERNEL_H
+
+#include "src/core/common/Macros.h"
+#include "src/gpu/cl/ClCompileContext.h"
+#include "src/gpu/cl/IClKernel.h"
+
+namespace arm_compute
+{
+namespace opencl
+{
+namespace kernels
+{
+/** OpenCL kernel for transposed convolution. */
+class ClTransposedConvolutionKernel : public IClKernel
+{
+public:
+ ClTransposedConvolutionKernel() = default;
+ ARM_COMPUTE_DISALLOW_COPY_ALLOW_MOVE(ClTransposedConvolutionKernel);
+ /** Set the input, weights, biases and output tensors.
+ *
+ * Similar to @ref ClTransposedConvolution::configure()
+ *
+ */
+ void configure(const CLCompileContext &compile_context, const ITensorInfo *input, const ITensorInfo *weights,
+ const ITensorInfo *biases, ITensorInfo *output, const PadStrideInfo &deconv_info);
+ /** Static function to check if given info will lead to a valid configuration
+ *
+ * Similar to @ref ClTransposedConvolution::configure()
+ *
+ * @return a status
+ */
+ static Status validate(const ITensorInfo *input, const ITensorInfo *weights, const ITensorInfo *biases,
+ const ITensorInfo *output, const PadStrideInfo &deconv_info);
+
+ // Inherited methods overridden:
+ void run_op(ITensorPack &tensors, const Window &window, cl::CommandQueue &queue) override;
+};
+} // namespace kernels
+} // namespace opencl
+} // namespace arm_compute
+
+#endif /* ARM_COMPUTE_CL_TRANSPOSED_CONVOLUTION_KERNEL_H */
\ No newline at end of file