[Bug 91153] [ivb] bright pixels appear when mousing over LibreOffice Impress

bugzilla-daemon at freedesktop.org bugzilla-daemon at freedesktop.org
Mon Jul 6 12:42:57 PDT 2015


https://bugs.freedesktop.org/show_bug.cgi?id=91153

--- Comment #21 from Joe Peterson <joe at wildlava.com> ---
Ok, according to my bisection (and I want to test the "good" version a little
more before I declare it good), the following commit appears to "fix" it:

   e47eb0c sna: Don't unroll BLT points

Looking at the diff for that commit, I do see something potentially suspect in
the old code in the second unrolled block:

-                       if (n_this_time & 4) {
-                               *((uint64_t *)b + 0) = pt_add(cmd, p+0, dx,
dy);
-                               *((uint64_t *)b + 1) = pt_add(cmd, p+1, dx,
dy);
-                               *((uint64_t *)b + 2) = pt_add(cmd, p+2, dx,
dy);
-                               *((uint64_t *)b + 3) = pt_add(cmd, p+3, dx,
dy);
-                               b += 8;
-                               p += 8;

Note that both b and p are incremented by 8. In the previous block, which looks
more correct they are:

-                       if (n_this_time & 4) {
-                               *((uint64_t *)b + 0) = pt_add(cmd, p+0, 0, 0);
-                               *((uint64_t *)b + 1) = pt_add(cmd, p+1, 0, 0);
-                               *((uint64_t *)b + 2) = pt_add(cmd, p+2, 0, 0);
-                               *((uint64_t *)b + 3) = pt_add(cmd, p+3, 0, 0);
-                               b += 8;
-                               p += 4;

I didn't analyze the unrolled code completely yet, but I wonder if that is
where the bug was before the code was "rolled". I'll look at this a bit more
later, but I wanted to get where I was so far to you now...

-- 
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-gfx-bugs/attachments/20150706/05120f4c/attachment.html>


More information about the intel-gfx-bugs mailing list