On 22 December 2011 16:54, Eric Anholt <span dir="ltr"><<a href="mailto:eric@anholt.net">eric@anholt.net</a>></span> wrote:<br><div class="gmail_quote"><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
Here's today's patch series for gen7 transform feedback. It runs on<br>
top of a kernel patch at people.freedesktop.org:~anholt/linux on the<br>
gen7-reset-sol branch. I expected it to be easy, but not this easy.<br></blockquote><div><br>This is fantastic, Eric. I'm really pleased how quickly this is coming together.<br><br>Other than a few minor comments that I've already sent out, the series is:<br>
Reviewed-by: Paul Berry <<a href="mailto:stereotype441@gmail.com">stereotype441@gmail.com</a>><br> </div><blockquote class="gmail_quote" style="margin:0pt 0pt 0pt 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">
<br>
Remaining test failures:<br>
<br>
tessellation polygon flat_last warn<br>
tessellation quad_strip flat_last warn<br>
tessellation quads flat_last warn<br>
tessellation triangle_fan flat_first fail<br></blockquote><div><br>I'm sorry to hear that triangle_fan flat_first fails--both AMD and nVidia pass that test, so it's a bit embarrassing for Intel to fail it. But it's an obscure case (who flatshades trifans anyhow, especially when using transform feedback?) and I can't see any way of fixing it on Gen7 without firing up the GS, which seems like *way* overkill. I'm far less bothered by the 3 warnings, because nVidia gets those exact same warnings too.<br>
<br>So IMHO, it's ok to leave these 4 tests failing.<br><br>Incidentally, Gen6 gets the exact same 4 failures, plus 4 additional failures for "tessellation triangle_strip". I'll try to fix "tessellation triangle_strip" tomorrow.<br>
</div><blockquote class="gmail_quote" style="margin:0pt 0pt 0pt 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">
<br>
Also, it looks like none of the current piglit tests test transform<br>
feedback across batchbuffers. I don't expect major problems there,<br>
given that we have a userland count of verts emitted.<br></blockquote><div><br>I have 3 more work days left in the year, so I'll try to implement those tests before you need them.<br></div></div>