<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.0//EN" "http://www.w3.org/TR/REC-html40/strict.dtd"><html><head><meta name="qrichtext" content="1" /><style type="text/css">p, li { white-space: pre-wrap; }</style></head><body style=" font-family:'DejaVu Sans'; font-size:9pt; font-weight:400; font-style:normal;">On Tuesday 31 March 2009, Bob Friesenhahn wrote:<br>
> If only trivial "exclusive use" is supported (i.e. only user owning<br>
> the "console" may use the GPU) then a huge number of potential<br>
> applications are blocked from taking advantage of the GPU and OpenCL<br>
> becomes a hinderance rather than a clear path forward.<br>
That's the main advantage of OpenCL over GLSL, they were also designed to be used on CPU(s), when you ask the library/driver to run an OpenCL kernel, you are not supposed to know if it will be executed on CPU(s) or on GPU, it's the library/driver which "decide" what would be the fastest solution (or possible). In other word, the scheduling is the job of the driver. As well as is security.<br>
<p style="-qt-paragraph-type:empty; margin-top:0px; margin-bottom:0px; margin-left:0px; margin-right:0px; -qt-block-indent:0; text-indent:0px; -qt-user-state:0;"><br></p>-- <br>
Cyrille Berger</p></body></html>