<html>
<head>
<base href="https://bugs.freedesktop.org/" />
</head>
<body>
<p>
<div>
<b><a class="bz_bug_link
bz_status_NEW "
title="NEW - [G965 ILK G45] Regression: 24 piglit regressions in glsl-1.10"
href="https://bugs.freedesktop.org/show_bug.cgi?id=92621#c4">Comment # 4</a>
on <a class="bz_bug_link
bz_status_NEW "
title="NEW - [G965 ILK G45] Regression: 24 piglit regressions in glsl-1.10"
href="https://bugs.freedesktop.org/show_bug.cgi?id=92621">bug 92621</a>
from <span class="vcard"><a class="email" href="mailto:apinheiro@igalia.com" title="Alejandro Piñeiro (freenode IRC: apinheiro) <apinheiro@igalia.com>"> <span class="fn">Alejandro Piñeiro (freenode IRC: apinheiro)</span></a>
</span></b>
<pre>Ok, I found the problem. Some comparisons are being wrongly removed with
unpack_flags.
When I added this function:
bool reads_flag(unsigned c)
{
if (!reads_flag())
return false;
switch (predicate) {
case BRW_PREDICATE_NONE:
return false;
case BRW_PREDICATE_ALIGN16_REPLICATE_X:
return c == 0;
case BRW_PREDICATE_ALIGN16_REPLICATE_Y:
return c == 1;
case BRW_PREDICATE_ALIGN16_REPLICATE_Z:
return c == 2;
case BRW_PREDICATE_ALIGN16_REPLICATE_W:
return c == 3;
default:
return true;
}
}
I was only checking the predicate, but as the one without parameter shows:
bool reads_flag()
{
return predicate || opcode == VS_OPCODE_UNPACK_FLAGS_SIMD4X2;
}
We also need to check the opcode.
Adding that checks fixes the regressions I tested manually. I will do a full
test and send a patch to the list soon.</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>