<p dir="ltr">I'd prefer a normal pipe cap instead of per shader.</p>
<p dir="ltr">Marek </p>
<div class="gmail_quote">On Jun 9, 2016 6:03 AM, "Dave Airlie" <<a href="mailto:airlied@gmail.com">airlied@gmail.com</a>> wrote:<br type="attribution"><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">On 9 June 2016 at 10:48, Dave Airlie <<a href="mailto:airlied@gmail.com">airlied@gmail.com</a>> wrote:<br>
> I started this quite a while back, and decided to spend some<br>
> time on it again, since it's the basis for a bunch of the newer<br>
> GL extensions.<br>
><br>
> So this is my first pass at it, it's fairly complete, but it<br>
> definitely is missing some pieces before it's finished.<br>
><br>
> The patchset add core mesa, gallium, softpipe and llvmpipe support.<br>
> I've got a few tests in piglit using a generator for all the built-ins<br>
> and operations that all pass.<br>
><br>
> So the big thing here was OMG the conversions, and is also<br>
> the biggest missing thing in the gallium code, I haven't plumbed<br>
> through all the conversions into opcodes yet, I'm not really sure<br>
> if we need all of them, but it'll be a quite large opcode set,<br>
> and I wasn't feeling brave enough yet.<br>
><br>
> The other things is I've implement MOD/DIV opcodes that may or<br>
> may not be useful on real hardware, if gallium drivers are okay<br>
> with lowering this in the backends then I'm fine with how it is.<br>
> (and gallium docs do need updating).<br>
<br>
I've also posted this to my arb_gpu_shader_int64 branch, and I've<br>
added some more bits I've been finding as I run tests.<br>
<br>
Dave.<br>
_______________________________________________<br>
mesa-dev mailing list<br>
<a href="mailto:mesa-dev@lists.freedesktop.org">mesa-dev@lists.freedesktop.org</a><br>
<a href="https://lists.freedesktop.org/mailman/listinfo/mesa-dev" rel="noreferrer" target="_blank">https://lists.freedesktop.org/mailman/listinfo/mesa-dev</a><br>
</blockquote></div>