[Mesa-dev] fbo-generatemipmap-array regression after 5a2f8b2c48a14

[Ilia Mirkin]

On Mon, Apr 21, 2014 at 12:10 PM, Brian Paul <brianp at vmware.com> wrote:
> On 04/19/2014 08:51 PM, Ilia Mirkin wrote:
>>
>> Hi Brian,
>>
>> I bisected a nv50/nvc0 regression in fbo-generatemipmap-array to
>> commit 5a2f8b2c48a14. The commit itself seems correct, but I
>> double-checked my bisection results (at least on nv50, I only assume
>> that it's the same commit causing the nvc0 regression). The
>> nv50_set_sampler_views function also seems pretty simple and correct.
>>
>> Any ideas as to what to look at?
>
>
> Off-hand, my only guess is that the nv50/nvc0's set_sampler_views() function
> isn't getting called as much as before so perhaps some driver state
> validation isn't getting done as often.

Just wanted to close the loop on this -- nothing wrong with your
commit as both of us suspected. By the time I got around to
understanding the issue, undoing your commit no longer helped. No idea
why reverting it helped in the first place when I was bisecting. The
issue ended up being due to the new u_gen_mipmap implementation which
started doing scaling blits on 1d array textures, which was broken in
nv50 and nvc0 (fixed with 8baed87212 "nv50,nvc0: fix blit 3d path for
1d array textures").

  -ilia