<html>
<head>
<base href="https://bugs.freedesktop.org/" />
</head>
<body>
<p>
<div>
<b><a class="bz_bug_link
bz_status_NEW "
title="NEW - [HSW, regression, bisected] i965/pixel_read: Use meta_pbo_GetTexSubImage for PBO ReadPixels (ef0499af255ecd)"
href="https://bugs.freedesktop.org/show_bug.cgi?id=89314#c18">Comment # 18</a>
on <a class="bz_bug_link
bz_status_NEW "
title="NEW - [HSW, regression, bisected] i965/pixel_read: Use meta_pbo_GetTexSubImage for PBO ReadPixels (ef0499af255ecd)"
href="https://bugs.freedesktop.org/show_bug.cgi?id=89314">bug 89314</a>
from <span class="vcard"><a class="email" href="mailto:neil@linux.intel.com" title="Neil Roberts <neil@linux.intel.com>"> <span class="fn">Neil Roberts</span></a>
</span></b>
<pre>Just to be a bit more specific — the test doesn't ensure that the driver rounds
in any particular way, it only ensures that it rounds the same way regardless
of whether a PBO is used or not. Ie, it compares the results of getting the
data with and without a PBO to make sure they are the same. This is difficult
for us to implement because when a PBO is used we will let the GPU do the
rounding, otherwise we will do it on the CPU. The GPU seems to have slightly
inaccurate results when the fractional part is a little bit above 0.5. Rounding
either way is explicitly allowed by the GLES spec so the GPU is not a problem.
However I can't find any mention of whether it has to be consist about the
choice of rounding so maybe we can argue that it is allowed by omission.</pre>
</div>
</p>
<hr>
<span>You are receiving this mail because:</span>
<ul>
<li>You are the QA Contact for the bug.</li>
</ul>
</body>
</html>