QUDA
v1.1.0
A library for QCD on GPUs
|
#include <cub_helper.cuh>
Go to the source code of this file.
Classes | |
struct | quda::atomic_type< T > |
The atomic word size we use for a given reduction type. This type should be lock-free to guarantee correct behaviour on platforms that are not coherent with respect to the host. More... | |
struct | quda::atomic_type< float > |
struct | quda::ReduceArg< T > |
Namespaces | |
quda | |
quda::reducer | |
Typedefs | |
using | device_reduce_t = double |
using | count_t = unsigned int |
Functions | |
size_t | quda::reducer::buffer_size () |
void * | quda::reducer::get_device_buffer () |
void * | quda::reducer::get_mapped_buffer () |
void * | quda::reducer::get_host_buffer () |
count_t * | quda::reducer::get_count () |
cudaEvent_t & | quda::reducer::get_event () |
constexpr int | quda::max_n_reduce () |
template<typename T > | |
constexpr T | quda::init_value () |
The initialization value we used to check for completion. More... | |
template<typename T > | |
constexpr T | quda::terminate_value () |
The termination value we use to prevent a possible hang in case the computed reduction is equal to the initialization. More... | |
using count_t = unsigned int |
Definition at line 15 of file reduce_helper.h.
using device_reduce_t = double |
Definition at line 8 of file reduce_helper.h.