diff options
| -rw-r--r-- | src/core/CL/kernels/CLDirectConvolutionLayerKernel.cpp | 53 |
1 files changed, 53 insertions, 0 deletions
diff --git a/src/core/CL/kernels/CLDirectConvolutionLayerKernel.cpp b/src/core/CL/kernels/CLDirectConvolutionLayerKernel.cpp index e1901af217..3c5799f7ba 100644 --- a/src/core/CL/kernels/CLDirectConvolutionLayerKernel.cpp +++ b/src/core/CL/kernels/CLDirectConvolutionLayerKernel.cpp @@ -131,6 +131,23 @@ void CLDirectConvolutionLayerKernel::configure(const ICLTensor *input, const ICL unsigned int num_elems_written_per_iteration_x = 0; unsigned int num_elems_written_per_iteration_y = 0; + // Through extensive experimentation with over 30 representative tensor + // shapes, we found a small number of local work size configurations + // that result in nearly optimal execution times. Selecting the right + // lws for a given shape, however, required a complex decision tree, + // until we constructed a simple feature as described below. + // + // We started from the number of multiply-accumulate operations for a + // convolution layer, which is equal to the product of the input + // dimensions 0..2 and the weights dimensions 0..2. Unfortunately, + // this resulted in ties between distinct shapes that required distinct + // lws configurations. Replacing the width of the input with the kernel + // size, however, resulted in nearly optimal predictions. We use underscores + // in variable names to indicate when they are intentionally misleading. + const size_t product_of_weights_dimensions = weights->info()->dimension(0) * weights->info()->dimension(1) * weights->info()->dimension(2); + const size_t product_of_input_dimensions_ = input->info()->dimension(0) * weights->info()->dimension(1) * input->info()->dimension(2); + const float mega_ops_ = 1e-6 * product_of_weights_dimensions * product_of_input_dimensions_; + switch(kernel_size) { case 1: @@ -139,6 +156,18 @@ void CLDirectConvolutionLayerKernel::configure(const ICLTensor *input, const ICL num_elems_read_per_iteration_y = 4; num_elems_written_per_iteration_x = 4; num_elems_written_per_iteration_y = 4; + if(mega_ops_ < 1.f) + { + _lws_hint = cl::NDRange(1, 1, 8); + } + else if(mega_ops_ < 7.f) + { + _lws_hint = cl::NDRange(1, 1, 4); + } + else + { + _lws_hint = cl::NDRange(1, 1, 2); + } break; } case 3: @@ -147,6 +176,22 @@ void CLDirectConvolutionLayerKernel::configure(const ICLTensor *input, const ICL num_elems_read_per_iteration_y = 5; num_elems_written_per_iteration_x = 4; num_elems_written_per_iteration_y = 3; + if(mega_ops_ < 1.f) + { + _lws_hint = cl::NDRange(1, 1, 8); + } + else if(mega_ops_ < 13.f) + { + _lws_hint = cl::NDRange(2, 1, 4); + } + else if(mega_ops_ < 50.f) + { + _lws_hint = cl::NDRange(3, 1, 4); + } + else + { + _lws_hint = cl::NDRange(2, 1, 6); + } break; } case 5: @@ -155,6 +200,14 @@ void CLDirectConvolutionLayerKernel::configure(const ICLTensor *input, const ICL num_elems_read_per_iteration_y = 6; num_elems_written_per_iteration_x = 4; num_elems_written_per_iteration_y = 2; + if(mega_ops_ < 2.f || mega_ops_ > 80.f) + { + _lws_hint = cl::NDRange(2, 1, 4); + } + else + { + _lws_hint = cl::NDRange(2, 1, 8); + } break; } default: |