<html>
<head>
<base href="https://bugs.freedesktop.org/" />
</head>
<body><span class="vcard"><a class="email" href="mailto:daniel.sebald@ieee.org" title="Dan Sebald <daniel.sebald@ieee.org>"> <span class="fn">Dan Sebald</span></a>
</span> changed
<a class="bz_bug_link
bz_status_NEW "
title="NEW - Drivers/DRI/swrast"
href="https://bugs.freedesktop.org/show_bug.cgi?id=89586">bug 89586</a>
<br>
<table border="1" cellspacing="0" cellpadding="8">
<tr>
<th>What</th>
<th>Removed</th>
<th>Added</th>
</tr>
<tr>
<td style="text-align:right;">Attachment #114342 is obsolete</td>
<td>
</td>
<td>1
</td>
</tr></table>
<p>
<div>
<b><a class="bz_bug_link
bz_status_NEW "
title="NEW - Drivers/DRI/swrast"
href="https://bugs.freedesktop.org/show_bug.cgi?id=89586#c40">Comment # 40</a>
on <a class="bz_bug_link
bz_status_NEW "
title="NEW - Drivers/DRI/swrast"
href="https://bugs.freedesktop.org/show_bug.cgi?id=89586">bug 89586</a>
from <span class="vcard"><a class="email" href="mailto:daniel.sebald@ieee.org" title="Dan Sebald <daniel.sebald@ieee.org>"> <span class="fn">Dan Sebald</span></a>
</span></b>
<pre>Created <span class=""><a href="attachment.cgi?id=114710" name="attach_114710" title="Changeset to fix vertical lines and fine tune positive_unzoom_x() and negative_unzoom_x()">attachment 114710</a> <a href="attachment.cgi?id=114710&action=edit" title="Changeset to fix vertical lines and fine tune positive_unzoom_x() and negative_unzoom_x()">[details]</a></span> <a href='page.cgi?id=splinter.html&bug=89586&attachment=114710'>[review]</a>
Changeset to fix vertical lines and fine tune positive_unzoom_x() and
negative_unzoom_x()
Attached is an update to the SWRAST legacy changeset. With this change, the
driver passes all tests in Piglit gl-1.0-pixelzoom check.
The main addition to the changeset over the last changeset is the inclusion of
a tolerance for the ceil() and floor() functions. The issue is that with
single precision float division and multiplication the formula
(xz - xr) / xfactor
can be off by a fair amount, on the order of 10e-5. I printed out some numbers
the driver was using in cases there the gl-1.0-pixelzoom alternating-line test
was failing. The numbers agree exactly with this example result:
octave:3> single(53)/single(400) * single(400)
ans = 52.9999961853027
I put in a tolerance of 0.00004 on the rounding functions. By my very rough
estimate, I think it is possible to scale input images with dimension of about
100,000 down to typical screen sizes without the added tolerance causing its
own sort of artifact.</pre>
</div>
</p>
<hr>
<span>You are receiving this mail because:</span>
<ul>
<li>You are the QA Contact for the bug.</li>
<li>You are the assignee for the bug.</li>
</ul>
</body>
</html>