These are chat archives for bjz/gfx-rs

13th
Jul 2014
Dzmitry Malyshau
@kvark
Jul 13 2014 00:40
@bjz please have a look at #95 as well
Brendan Zabarauskas
@brendanzab
Jul 13 2014 00:48
@kvark interesting to note that you have gone back to call/cast!
Dzmitry Malyshau
@kvark
Jul 13 2014 00:49
@bjz indeed, but there is a 3rd group there - implicitly set, because it only contains SwapBuffers. This is not a Cast since it's executed by the device common dispatcher, not the backend
Brendan Zabarauskas
@brendanzab
Jul 13 2014 00:50
If ping/pong are not used, could you either remove them or document them? It is extra work for the reader to understand what they do.
Dzmitry Malyshau
@kvark
Jul 13 2014 00:50
sure, I'll remove them then
Brendan Zabarauskas
@brendanzab
Jul 13 2014 00:51
Will that remove Token too?
I was confused about that
Dzmitry Malyshau
@kvark
Jul 13 2014 00:51
no, token is used for every call...
Brendan Zabarauskas
@brendanzab
Jul 13 2014 00:52
ohh, I see
Dzmitry Malyshau
@kvark
Jul 13 2014 00:52
it should have different types for each call, but I can't put that many generic parameters in... (see Problems section in the PR description)
Brendan Zabarauskas
@brendanzab
Jul 13 2014 00:53
@kvark btw, small patch: bjz/gfx-rs#96
Dzmitry Malyshau
@kvark
Jul 13 2014 01:00
@bjz I don't think we should give even a slightest hint that compatibiltiy GL features are supported.
Corey Richardson
@cmr
Jul 13 2014 01:01
+1, we don't need it, want it, or use it.
Brendan Zabarauskas
@brendanzab
Jul 13 2014 01:02
fixed
Dzmitry Malyshau
@kvark
Jul 13 2014 01:22
now I'm confused: GL-4 help reference on glGetError also has these stack variants
Brendan Zabarauskas
@brendanzab
Jul 13 2014 01:23
Yup
but it doesn't get generated
hence the commenting out
Dzmitry Malyshau
@kvark
Jul 13 2014 01:23
ok, good
Brendan Zabarauskas
@brendanzab
Jul 13 2014 01:23
:P
Dzmitry Malyshau
@kvark
Jul 13 2014 01:23
@bjz thanks!
Brendan Zabarauskas
@brendanzab
Jul 13 2014 01:24
better than getting a random number you have to google for :P
Dzmitry Malyshau
@kvark
Jul 13 2014 02:05
updated #95
@bjz notice how the Future is gone :P
Brendan Zabarauskas
@brendanzab
Jul 13 2014 02:11
Oooh, nice!
:clap:
Dzmitry Malyshau
@kvark
Jul 13 2014 02:37
@bjz @cmr I've moved resource stuff into a separate module, and that concludes my TODO list for the resource cache. I'm kinda ready to merge, unless you have any ideas on the current problems I mentioned in the PR description (or unless you think we should approach the whole thing differently).
Brendan Zabarauskas
@brendanzab
Jul 13 2014 02:43
No I'm thinking it looks pretty good.
Dzmitry Malyshau
@kvark
Jul 13 2014 02:45
@bjz Awesome! I really dislike all the need to call demand but it looks to be the only way around, at least until the borrow checker improves... Besides, this is well hidden from the user.
Brendan Zabarauskas
@brendanzab
Jul 13 2014 02:45
So is demand only because of the borrow checker?
Dzmitry Malyshau
@kvark
Jul 13 2014 02:47
mostly. I'm not entirely sure we'll get rid of it when BC is fixed, but most cases are gonna be trivialized.
Brendan Zabarauskas
@brendanzab
Jul 13 2014 02:48
Could you add a // comment explaining your intensions?
sorry
Dzmitry Malyshau
@kvark
Jul 13 2014 02:49
sure. will do
Do you mean to comment on the demand or on everything? :)
Brendan Zabarauskas
@brendanzab
Jul 13 2014 02:53
just on the definition of demand :P
Dzmitry Malyshau
@kvark
Jul 13 2014 02:57
@bjz Does my answer about the Token make sense?
Dzmitry Malyshau
@kvark
Jul 13 2014 03:02
@bjz comments are in
Brendan Zabarauskas
@brendanzab
Jul 13 2014 03:13
I think so
@cmr cool to merge?
Corey Richardson
@cmr
Jul 13 2014 03:18
Yes.
Brendan Zabarauskas
@brendanzab
Jul 13 2014 03:20
lets do it!
Corey Richardson
@cmr
Jul 13 2014 04:15
@kvark you said you are doing texturing this weekend?
Marvin Löbel
@Kimundi
Jul 13 2014 10:34
Interesting read, though I didn't understand most of the terminology there :P
Dzmitry Malyshau
@kvark
Jul 13 2014 12:05
@cmr yeah, I was going to (unless someone else wants?). Weekend turns out to be pretty busy with home tasks, so texturing is gonna stretch over the next week.
Dzmitry Malyshau
@kvark
Jul 13 2014 12:53
This message was deleted
Brendan Zabarauskas
@brendanzab
Jul 13 2014 19:00
@cmr Do you think we can get rid of Ack by using a sync_channel instead?
/me is so confused with Ack
ack
lol
Corey Richardson
@cmr
Jul 13 2014 19:02
@bjz no, you don't want to be blocked always, you just want to wait until the frame is finished until starting the next one.
Brendan Zabarauskas
@brendanzab
Jul 13 2014 19:02
ah ok
Corey Richardson
@cmr
Jul 13 2014 19:02
An alternative I mentioned a while back is sending each frame as a single message
Brendan Zabarauskas
@brendanzab
Jul 13 2014 19:02
is there anyway we can make this easier to grok?
Corey Richardson
@cmr
Jul 13 2014 19:03
This might not always be good though, if the device can make useful progress while the application is still sending commands.
Brendan Zabarauskas
@brendanzab
Jul 13 2014 19:03
maybe by making a new channel thing in our comm lib?
Corey Richardson
@cmr
Jul 13 2014 19:03
But it has the benefit of not leaning so heavily on our channels.
Rust channels are pretty fast, but you only get ~500k channel ops per frame, which is potentially limiting.
Brendan Zabarauskas
@brendanzab
Jul 13 2014 19:04
It just looks pretty error-prone having this Ack channel hanging around exposed
mmk
Corey Richardson
@cmr
Jul 13 2014 19:05
well the swap_ack isn't public
you need to call end_frame
Brendan Zabarauskas
@brendanzab
Jul 13 2014 19:05
yeah, but from the point of view of us
and new devs on gfx-rs
Corey Richardson
@cmr
Jul 13 2014 19:07
yeah, it is getting complex.
Brendan Zabarauskas
@brendanzab
Jul 13 2014 19:08
That is why I made a new thing for the close request (although I think I could make it easier to understand)
brendanzab @bjz is currently documenting it
Brendan Zabarauskas
@brendanzab
Jul 13 2014 19:09
even though these APIs are internal, it's important for maintainability for us to remember what these things do
Brendan Zabarauskas
@brendanzab
Jul 13 2014 19:23
@cmr #97?
Corey Richardson
@cmr
Jul 13 2014 19:35
While I'm waiting for the clang devs to get back to me I might try setting up a headless X on travis.
So we can actually run our tests.