// ================================================================================================= // This file is part of the CLBlast project. The project is licensed under Apache Version 2.0. This // project loosely follows the Google C++ styleguide and uses a tab-size of two spaces and a max- // width of 100 characters per line. // // Author(s): // Cedric Nugteren // // This file implements the common routine functions (see the header for more information). // // ================================================================================================= #include #include #include "routines/common.hpp" namespace clblast { // ================================================================================================= // Enqueues a kernel, waits for completion, and checks for errors void RunKernel(Kernel &kernel, Queue &queue, const Device &device, std::vector global, const std::vector &local, EventPointer event, const std::vector &waitForEvents) { if (!local.empty()) { // Tests for validity of the local thread sizes if (local.size() > device.MaxWorkItemDimensions()) { throw RuntimeErrorCode(StatusCode::kInvalidLocalNumDimensions); } const auto max_work_item_sizes = device.MaxWorkItemSizes(); for (auto i=size_t{0}; i max_work_item_sizes[i]) { throw RuntimeErrorCode(StatusCode::kInvalidLocalThreadsDim); } } auto local_size = size_t{1}; for (auto &item: local) { local_size *= item; } if (local_size > device.MaxWorkGroupSize()) { throw RuntimeErrorCode(StatusCode::kInvalidLocalThreadsTotal); } // Make sure the global thread sizes are at least equal to the local sizes for (auto i=size_t{0}; i(elapsed_time).count(); printf("[DEBUG] Completed kernel in %.2lf ms\n", timing); #endif } // ================================================================================================= } // namespace clblast