void run()
{
NDRange global;
Event event;
boost::array<boost::uint32_t,4> size;
size.assign(2*256*64);
global = NDRange(256,64);
// run kernel A
_kernel.setArg(0,size);
_kernel.setArg(1,_input);
_kernel.setArg(2,_output);
_queue.enqueueNDRangeKernel( _kernel, global, &event );
_queue.flush();
_queue.waitForEvent(event);
}
void run( int workgroup_size )
{
NDRange global,local;
Event event;
local = NDRange(workgroup_size);
global = NDRange(_nr_groups*workgroup_size);
posix_time::ptime t1 = posix_time::microsec_clock::local_time();
_kernel.setArg(0,_output);
_queue.enqueueNDRangeKernel( _kernel, global, local, &event );
_queue.flush();
_queue.waitForEvent(event);
posix_time::ptime t2 = posix_time::microsec_clock::local_time();
_exec_time = posix_time::time_period(t1,t2).length().total_microseconds();
}