OpenCL force buffer to stay on the GPU -


is there way force opencl keep data in global memory buffer on chip (i.e. never gets swapped out system memory)? want reserve portion of gpu's memory own needs, , want data put there remain on gpu regardless of whether other applications start saturating gpu's memory.

thanks!

gpu's aren't linked cpus memory management unit, don't page faults cpu. block device peripherals controlled device drivers.

for opencl when allocate memory on gpu creating cl_mem object , enqueueing (writing) gpu stick around until explicitly release clreleasememobject. reuse buffer need not release , keep track of cl_mem object.


Comments

Popular posts from this blog

jasper reports - Fixed header in Excel using JasperReports -

media player - Android: mediaplayer went away with unhandled events -

python - ('The SQL contains 0 parameter markers, but 50 parameters were supplied', 'HY000') or TypeError: 'tuple' object is not callable -