<div dir="ltr">Thank you all, for the tips. Using a smaller window doesn't change anything, neither does tweaking LP_NUM_THREADS (between 8/12/16). I am also not using MSAA. I'll have to dig some more regarding your other suggestions. I've shared a screenshot of what is being rendered at <a href="https://imgur.com/a/6P6OAWA">https://imgur.com/a/6P6OAWA</a> , although based on your observations I'm probably leaning towards being geometry bound.</div><br><div class="gmail_quote gmail_quote_container"><div dir="ltr" class="gmail_attr">On Thu, Jan 30, 2025 at 4:45 PM Erik Faye-Lund <<a href="mailto:erik.faye-lund@collabora.com">erik.faye-lund@collabora.com</a>> wrote:<br></div><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">On Thu, 2025-01-30 at 13:49 +0000, Jose Fonseca wrote:<br>
> At one point there was another SW renderer gallium driver by a team<br>
> at Intel that was specifically optimized for huge number of<br>
> primitives with no texturing (like the sort used by <a href="https://vtk.org/" rel="noreferrer" target="_blank">https://vtk.org/</a><br>
> ), but I can't remember its name or find it.<br>
<br>
Not super important, but for future reference: You're talking about<br>
OpenSWR, see these links for more info:<br>
<br>
<a href="https://gitlab.freedesktop.org/mesa/mesa/-/blob/amber/docs/drivers/openswr.rst" rel="noreferrer" target="_blank">https://gitlab.freedesktop.org/mesa/mesa/-/blob/amber/docs/drivers/openswr.rst</a><br>
<a href="https://www.openswr.org/" rel="noreferrer" target="_blank">https://www.openswr.org/</a><br>
<br>
It didn't really benchmark much better than Llvmpipe, though:<br>
<a href="https://www.phoronix.com/review/llvmpipe-swr-xeon" rel="noreferrer" target="_blank">https://www.phoronix.com/review/llvmpipe-swr-xeon</a><br>
<br>
In the end, I think it didn't add enough value to anyone willing to<br>
work on maintaining it.<br>
<br>
</blockquote></div>