ConfigurationHeader.md

Configuration Header for cupla

cupla provides configuration header to use the library without CMake, e.g in a cling environment.

Available Pre-compiler Definitions

Pre-compiler definitions can be used to control the behavior of cupla/alpaka. The definitions must be passed via a compiler flag or be defined before the accelerator header is included. The default value will be used if the configuration header is included without defining any of the following options.

• CUPLA_STREAM_ASYNC_ENABLED: 0 use synchronous streams (default), 1 use asynchronous streams
• CUPLA_HEADER_ONLY: 1 cupla will be used as header-only library (default), otherwise you must compile all .cpp files in src/

linker dependencies

Depending of the used accelerator you must link the library pthread and/or activate OpenMP support.

Example

To select an accelerator you must include the corresponding accelerator header from cupla/config/

// use synchronous streams for the accelerator
#define CUPLA_STREAM_ASYNC_ENABLED 0
// use cupla as header-only library
#define CUPLA_HEADER_ONLY 1
#include <cupla/config/CpuSerial.hpp>


int main()
{
    int* res_ptr_d = nullptr;
    cudaMalloc( (void**)&res_ptr_d, sizeof( int ) );

    return 0;
}