gpu_utils Directory Reference

Directory dependency graph for gpu_utils:

Directories
directory	tests

Files
file	clfftinitializer.cpp
	Implements ClfftInitializer class.
file	clfftinitializer.h
	Declares ClfftInitializer class, which initializes and tears down the clFFT library resources in OpenCL builds, and does nothing in other builds, and a factory function for it.
file	devicebuffer.h
	Implements the logic for handling of DeviceBuffer types in OpenCL/CUDA. Can only be included on GPU build paths.
file	devicebuffer_datatype.h
	Declares the DeviceBuffer data type.
file	devicebuffer_ocl.h
	Implements the DeviceBuffer type and routines for OpenCL. Should only be included directly by the main DeviceBuffer file devicebuffer.h. TODO: the intent is for DeviceBuffer to become a class.
file	gmxopencl.h
	Wraps the complexity of including OpenCL in Gromacs.
file	gpu_macros.h
file	gpu_testutils.cpp
	Function definitions for GPU detection, specific for tests.
file	gpu_testutils.h
	Declare functions for detection of GPU devices, specific for tests.
file	gpu_utils.cpp
	Function definitions for non-GPU builds.
file	gpu_utils.h
	Declare functions for detection and initialization for GPU devices.
file	gpu_utils_ocl.cpp
	Define functions for detection and initialization for OpenCL devices.
file	gpueventsynchronizer_ocl.h
	Implements a GpuEventSynchronizer class for OpenCL.
file	gpuregiontimer.h
	Defines the GPU region timer implementation/wrapper classes. The implementations live in gpuregiontimer.cuh for CUDA and gpuregiontimer_ocl.h for OpenCL.
file	gpuregiontimer_ocl.h
	Implements the GPU region timer for OpenCL.
file	gputraits.h
	Declares the GPU type traits for non-GPU builds.
file	gputraits_ocl.h
	Declares the OpenCL type traits.
file	hostallocator.cpp
	Implements gmx::HostAllocationPolicy for allocating memory suitable for e.g. GPU transfers on CUDA.
file	hostallocator.h
	Declares gmx::HostAllocationPolicy, gmx::HostAllocator, gmx::HostVector and gmx::PaddedHostVector, which are used to make/be standard library containers that can allocate memory suitable for transfers. Currently the only supported transfers using pinned memory are to CUDA GPUs, but other possibilities exist in future.
file	ocl_caching.cpp
	Define infrastructure for OpenCL JIT compilation for Gromacs.
file	ocl_caching.h
	Declare infrastructure for managing caching of OpenCL JIT-ted binaries.
file	ocl_compiler.cpp
	Define infrastructure for OpenCL JIT compilation for Gromacs.
file	ocl_compiler.h
	Declare infrastructure for OpenCL JIT compilation.
file	oclraii.h
	Declare RAII helpers for OpenCL types, along with supporting type traits.
file	oclutils.cpp
	Define utility routines for OpenCL.
file	oclutils.h
	Declare utility routines for OpenCL.
file	pinning.h
	Declares functions for pinning memory to be suitable for efficient GPU transfers on CUDA.
file	pmalloc_cuda.h
	Declare functions for host-side memory handling when using CUDA devices.