00:02
rasterman has quit [Quit: Gettin' stinky!]
00:48
<
alyssa >
robclark: reminder to self: I still need to procure a suzyq
00:49
* alyssa
files "figure out how to get a suzyq" on the todo list sometime in between "next week" and "never"
00:49
<
alyssa >
right in the "ne"'s section of the calendar
01:30
* HdkR
runs for the hills
02:07
vstehle1 has quit [Ping timeout: 480 seconds]
02:21
atler is now known as Guest6913
02:21
atler has joined #panfrost
02:23
Guest6913 has quit [Ping timeout: 480 seconds]
02:32
<
alyssa >
HdkR: i have already saturated my GPU vendor r/e count
04:07
jernej has quit [Ping timeout: 480 seconds]
04:11
livo has joined #panfrost
05:00
jernej has joined #panfrost
08:17
chewitt has quit [Read error: Connection reset by peer]
08:17
chewitt has joined #panfrost
08:32
vstehle has joined #panfrost
09:19
macc24 has joined #panfrost
10:50
chewitt has quit [Quit: Zzz..]
10:54
chewitt has joined #panfrost
11:46
rasterman has joined #panfrost
14:52
Danct12 has quit [Quit: Quitting]
14:55
Danct12 has joined #panfrost
15:00
Daanct12 has joined #panfrost
15:08
Danct12 has quit [Ping timeout: 480 seconds]
15:56
Daanct12 is now known as Danct12
16:53
leah has joined #panfrost
17:13
JoJo_ has joined #panfrost
17:55
davidlt has joined #panfrost
17:58
davidlt has quit []
18:30
<
alyssa >
this JIT allocation stuff is ridiculous
18:30
<
alyssa >
(..and makes it prohibitively difficult to understand what's happening..)
19:08
<
alyssa >
the harder I study it, the less I understand it
19:09
<
urja >
study softer? /jk
19:16
<
alyssa >
tearing into the first compute shader and figuring out exactly what's happening
20:07
macc24 has quit [Ping timeout: 480 seconds]
22:19
Net147_ has joined #panfrost
22:20
Net147 has quit [Remote host closed the connection]
22:20
<
anarsoul >
alyssa: hey, does panfrost implement zero-cost MSAA 4x that ARM is bragging about?
22:21
<
anarsoul >
or it writes out all the samples and then requires explicit resolve?
22:32
rcf has quit [Ping timeout: 480 seconds]
22:40
Net147_ has quit []
22:41
Net147 has joined #panfrost
22:48
Net147 has quit [Quit: Quit]
22:48
Net147 has joined #panfrost
23:03
<
alyssa >
anarsoul: It does, needs EXT_ms_rtt since it's not in core gles
23:03
<
alyssa >
in gallium it's just the difference of pipe_surface->samples and pipe_texture->samples
23:05
<
anarsoul >
what about EXT_multisampled_render_to_texture?
23:05
<
anarsoul >
with implicit multisample buffer and implicit resolve?
23:06
<
anarsoul >
oh, it's EXT_ms_rtt
23:09
<
anarsoul >
alyssa: but how is it handled for default framebuffer?
23:10
<
anarsoul >
buffer swap with multisampled default fb implies resolve
23:16
<
anarsoul >
alyssa: but how do I know in a gallium driver if I need to do resolve or not?
23:24
<
alyssa >
anarsoul: resolve if pipe_surface->nr_samples > pipe_surface->texture->nr_samples
23:24
<
anarsoul >
I see, thanks!
23:24
<
alyssa >
specifically pipe_surface->nr_samples=4 and pipe_surface->texture->nr_samples=1
23:25
<
alyssa >
also a gles3 driver has to worry about pipe_framebuffer->nr_samples as well, gles2 doesn't
23:25
<
anarsoul >
what about default fb, is mesa smart enough not to do explicit resolve if driver exposes PIPE_CAP_SURFACE_SAMPLE_COUNT?
23:26
nlhowell has joined #panfrost
23:26
<
anarsoul >
also I guess I need to implement resolve in lima_blit...
23:26
<
alyssa >
uh.... probably?
23:47
nlhowell has quit [Ping timeout: 480 seconds]
23:53
nlhowell has joined #panfrost