Hello every one (new to opencl),
i have a problem with async_work_group_copy.
I have to multiple access data within a work-group and i tried to copy it to local memory for optimization.
It works fine as long as i copy only one part of the needed data into local memory (pM_edge1 or pM_edge2). If i copy both, the data in the local memory is broken at a certain offset. There is no other local memory used in the kernel and i have no idea what i am doing wrong.
__private event_t events; __local float3 p_edges_1[16*48]; __local float3 p_edges_2[16*48]; events = async_work_group_copy(p_edges_1 , pM_edge1 + global_offset_p1,(size_t)(48*16),0); events = async_work_group_copy(p_edges_2 , pM_edge2 + global_offset_p2,(size_t)(48*16),0); wait_group_events(2, events); [...] //acces p_edges_1[X] ... p_edges_2[Y]
ok .. sorry for my posting .. it seems to be buggy somehow.
When i put the local memory into the kernel signature it works.
....,__local float3* p_edges_1,__local float3* p_edges_2)