<div>You are probably monopolizing the processing power of the graphics card while the code is running. When you step through the code, it runs only one cycle (not cpu cycle, one computation) and then leaves the GPU alone while waiting for you to click "next step".</div>
<div> </div>
<div>Andy<br><br></div>
<div class="gmail_quote">On Mon, Jan 24, 2011 at 3:16 PM, Aaron Birenboim <span dir="ltr"><<a href="mailto:aaron@boim.com">aaron@boim.com</a>></span> wrote:<br>
<blockquote class="gmail_quote" style="PADDING-LEFT: 1ex; MARGIN: 0px 0px 0px 0.8ex; BORDER-LEFT: #ccc 1px solid">I am working on a scientific computing problem (using the GPU -- OpenGL)<br>where when I try to do my test runs, I get into a state where any<br>
graphics response of my computer comes nearly to a halt,<br>the machine becomes sluggish, and it spends most of it's time<br>reporting a "wait" state.<br><br>I cannot reproduce this behavior while stepping through in<br>
a debugger. It only happens several seconds into<br>a longer full-speed run.<br><br>This test routine is dumping many tens (perhaps hundreds)<br>of MB of disk writes per second. Scores of files,<br>each of size 1 to 8 MB.<br>
<br>How can I find out what is causing this near lock-up?<br>How can I avoid it?<br><br>Is this kind of thing often seen with many mid<br>sized files being generated?<br>Could it be ab OpenGL thing, where something is not<br>
cleanly flushed, and the GPU is doing something odd?<br><br>(I am writing to floating point FBO's, and retrieving<br> results with glReadPixels())<br><br>Thanks for any advice:<br><br> aaron<br>_______________________________________________<br>
nmglug mailing list<br><a href="mailto:nmglug@lists.nmglug.org">nmglug@lists.nmglug.org</a><br><a href="http://lists.nmglug.org/listinfo.cgi/nmglug-nmglug.org" target="_blank">http://lists.nmglug.org/listinfo.cgi/nmglug-nmglug.org</a><br>
</blockquote></div><br>