[Bug 37028] Amnesia/HPL2 Demo: Strange graphical bugs on r600g

[bugzilla-daemon at freedesktop.org]

https://bugs.freedesktop.org/show_bug.cgi?id=37028

--- Comment #12 from Maggioni Marcello <hayarms at gmail.com> 2011-06-07 09:15:59 PDT ---
Created an attachment (id=47666)
 --> (https://bugs.freedesktop.org/attachment.cgi?id=47666)
Further tests

Hi, your test application crashes on my system with and without the patch.

I determined that the result that the query should return is 68698 samples each
execution.

So I modified your test application to see how many times that value is
returned instead of some other random values (the modded app is attached) and
the output is this:

[hades at artemis ~]$ ./query 
Correct values: 17
Not correct values: 33
Correct values: 50
Not correct values: 0
Correct values: 50
Not correct values: 0
Correct values: 50
Not correct values: 0
Correct values: 50
Not correct values: 0
Correct values: 50

.....

As you can see at the beginning the probability of wrong queries is high and
then it corrects itself out. All the subsequent queries seem to be correct.
This happens both with your patch and without it.

I tried to step inside the gallium code (r600_query_result) to see what values
are returned.
When the value is correct I get values like:

(gdb) n
1641                            query->result += end - start;
2: /x end = 0x80000001824291f0
1: /x start = 0x8000000182424e0c
...

when I get an error I get values for start and end like :
(gdb) n
1641                            query->result += end - start;
2: /x end = 0xffffffffffff0000
1: /x start = 0xffff0000ffffffff
(gdb) n
1636            for (i = 0; i < size; i += 4) {
2: /x end = 0xffffffffffff0000
1: /x start = 0xffff0000ffffffff
(gdb) 


That are obviously wrong.

The very first r600_query_result execution gives correct results, then from the
second r600_query_result execution on the results are incorrect and then ,
after some runs, the values become correct again.

I don't know how to interpret this. Maybe a drm bug? I use kernel 2.6.39

-- 
Configure bugmail: https://bugs.freedesktop.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.