[Bug 89314] [HSW, regression, bisected] i965/pixel_read: Use meta_pbo_GetTexSubImage for PBO ReadPixels (ef0499af255ecd)

[bugzilla-daemon at freedesktop.org]

https://bugs.freedesktop.org/show_bug.cgi?id=89314

Jason Ekstrand <jason at jlekstrand.net> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |idr at freedesktop.org,
                   |                            |neil at linux.intel.com

--- Comment #8 from Jason Ekstrand <jason at jlekstrand.net> ---
Ok, I did a little more digging.  One of the offending values is the 10-bit
value 680.  According to GDB,

(680 / (float)0x3ff) * 0xff == 169.50145

The software path rounds this up to 170 as it should be.  The hardware,
howevver, rounds it down to 169 which is clearly wrong.  Since the dEQP test
requires bit-accurate results, the test fails.

I'm not sure what we want to do here.  One option would be to simply disable
the PBO path for RGB10_A2 since the hardware doesn't round correctly.  Given
how thurough the dEQP test is, it's actually kind of encouraging that RGB10_A2
is the only one that failed.

I'm also CC'ing Ian and Neil since they may want to chip in here.

-- 
You are receiving this mail because:
You are the QA Contact for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.freedesktop.org/archives/intel-3d-bugs/attachments/20150227/c1fbefbf/attachment.html>