Date: Wed, 21 Nov 2012 23:53:06 +0100 From: magnum <john.magnum@...hmail.com> To: john-dev@...ts.openwall.com Subject: Re: Problems on recent AMD driver On 21 Nov, 2012, at 22:06 , magnum <john.magnum@...hmail.com> wrote: > ...but per nvidia whitepapers, the following (which is what I do, and others before me) is recommended for using "pinned" or "page-locked" memory. And it's a mix of the above: > So in this last example, we map it once and never unmap - and we are *also* using clEnqueueWriteBuffer(). If you only go per the Khronos spec, this seems to violate things. From the spec of clEnqueueMapBuffer: "The behavior of OpenCL function calls that enqueue commands that write or copy to regions of a memory object that are mapped is undefined". Maybe I should ask on the Khronos forum to get this straightened out for certain. I just did: http://www.khronos.org/message_boards/viewtopic.php?f=28&t=5523 Let's hope someone who actually has a clue will answer. magnum
Powered by blists - more mailing lists
Confused about mailing lists and their use? Read about mailing lists on Wikipedia and check out these guidelines on proper formatting of your messages.