On Sat, 2 Apr 2011 03:04:51 +0530, Kunal Puri <[email protected]> wrote: > Hi, I'm new to OpenCL and was playing arounf with PyOpenCL's interactive > features. My question is regarding the memory size output from > get_device_info: > > device.global_mem_size = 1073741824 (1024 MB) > device.max_mem_alloc_size = 536870912 (512MB) > device.local_mem_size = 32768 (32 KB) > > The device being queried is the CPU on my laptop (there is no GPU). So I am > limited by allocating at maximum 512 MB of data on the device, this through > cl.Buffer commands in the host code. In addition to this, the size of the > variables I declare with the qualifier '__local' in the kernel code cannot > exceed 32KB. So what happens to the other 512 MB (global_mem - > max_mem_alloc)? > > Also, when using PyOpenCL, do we do away with kernel arguments having the > qualifier '__local'?
Whose OpenCL implementation are you using? If AMD CPU, those just seem to be hard limits at the moment... Here's what I get: GLOBAL_MEM_SIZE: 1073741824 (I have 1.5 GiB physical) MAX_MEM_ALLOC_SIZE: 536870912 Also see here: http://developer.amd.com/support/KnowledgeBase/Lists/KnowledgeBase/DispForm.aspx?ID=123 Andreas
pgpHNqf8RuUfC.pgp
Description: PGP signature
_______________________________________________ PyOpenCL mailing list [email protected] http://lists.tiker.net/listinfo/pyopencl
