topic Re: imx6 opencl zero copy in i.MX Processors

imx6 opencl zero copy

liujkian — Sun, 27 Jan 2013 12:49:45 GMT

Hi, i have a question about the opencl map(clEnqueueMapBuffer ) mechanism. if a opencl host program allocs a block of memory by itself (such as malloc) not clCreateBuffer,then it calls clEnqueueMapBuffer to map the memory to gpu.can gpu use the mapped memory directly (or DMA )without any data copy (or DMA transfer )to vivante opencl runtime lib or to gpu global memory. Any Tips?thanks

Re: imx6 opencl zero copy

liujkian — Mon, 28 Jan 2013 15:14:17 GMT

anyone can help me?thanks

Re: imx6 opencl zero copy

andre_silva — Fri, 08 Mar 2013 19:28:20 GMT

I´m checking this information and will be back soon with the answer.

Re: imx6 opencl zero copy

arthassin — Mon, 11 Mar 2013 10:17:26 GMT

I've tried this on imx6. Seems like the buffer is still copied. No any performance improvement was achieved.

Re: imx6 opencl zero copy

andre_silva — Mon, 11 Mar 2013 19:20:54 GMT

I just checked and cl_mem must be used to allocate memory clEnqueueMapBuffer, there isn´t another way to do fill cpu buffers to be used in gpu, so, zero copies in this case is not possible.

regards,

Andre

Re: imx6 opencl zero copy

korabelnikov — Tue, 22 Jun 2021 15:26:28 GMT

Hi,

Is any updates or tips on it? How to make host<->opencl memory access faster?

data transfers take a half of computation time on imx6qp