| |
Frantic
Registered: Mar 2003
Posts: 1648 |
C64 Codebase
Hello!
It was a while ago now that C64 Codebase Wiki opened its doors. I would just like to encourage good coders sympathetic with this project to actually add some code there. This is important in order to keep the quality of the site. I feel that the quality is somehow fading a little with the stuff that has been added during the last 6 months, to generalize a little.
If you are a decent coder: Just think for a moment about all those sources that you have lying around on your C64 disks and PC harddrives. Codebase needs you!
http://codebase64.org/doku.php
If someone feel like donating some cool stuff, I could use that as prices in some kind of add-good-stuff-to-codebase-competition or so.
A good codebase is a good base for the future C64 scene! ;)
|
|
... 127 posts hidden. Click here to view all posts....
|
| |
Ninja
Registered: Jan 2002
Posts: 411 |
Stoff: Hey, you succeeded, good news :) Can you add the mag in which my original article can be found to the codebase entry? I forgot which it was.
In general: After the initial 2x2-routine, Oswald and I also played with 4x4-versions, but the gain in speed is _much_ less noticable, of course. We considered that those 8 pages can usually be used for something different which will please the demo-watcher more. So, in practice, we hardly use it. |
| |
St0fF
Registered: Oct 2002
Posts: 40 |
@Ninja: Sorry, I downloaded it from the-dreams.de, like you suggested. I don't know where you published it in the first place.
But I have to object just a little: those 8 pages are not wasted at all. It's less than $20 bytes for the nmi-routines, so stuffing it with other data ain't no problem... and compared to an auxiliary timer routine, which fires the irq in cycle 0 - we have a gain of around 44 cycles each interrupt. That's more than 1000cycles more each frame. |
| |
Oswald
Registered: Apr 2002
Posts: 5094 |
Quote: Stoff: Hey, you succeeded, good news :) Can you add the mag in which my original article can be found to the codebase entry? I forgot which it was.
In general: After the initial 2x2-routine, Oswald and I also played with 4x4-versions, but the gain in speed is _much_ less noticable, of course. We considered that those 8 pages can usually be used for something different which will please the demo-watcher more. So, in practice, we hardly use it.
all credits goes to Ninja ofcourse, I'm just an end user of these routines :) |
| |
Ninja
Registered: Jan 2002
Posts: 411 |
Stoff: Hehe, forgot about the online-version. Could you link to that?
After all, it's your choice if you want the 5% extra. My experience is that the memory can also be spent on some additional GFX or such (check Salute) which might add more to the part than the speedup. If you have those pages free anyhow, sure go for it. For me, though, with most effects having unrolled code and huge tables, 8 pages is quite a lot. YMMV. |
| |
St0fF
Registered: Oct 2002
Posts: 40 |
... done - I just linked the articles-page. btw. it was Vandalism News #43 (stated on that articles-page).
About the 8 pages - what I mean is: you normally need some data-tables with your unrolled loops. Most of them do not need to be "a full page", but they are, just to not loose cycles for page boundary crossings. That's where some embedding of the routines might take place ;) Just stay tuned, I'm coding. It's been some time since the last Neoplasia-release of any kind ...
Another question just came to my mind while re-reading my article. My calculations about starting the timers - as a trained eye might already have noticed - add/subtract one cycle. This value just came from testing/trial and crash. But now I have an idea in mind: may it just be that an Interrupt is the only operation that happens in a 6510, which is not able to prefetch the next opcode? So this "one cycle earlier" starting of the NMI-timer and the "one additional cycle more delay" to the jitter-timer might be just that: the NMI actually needs another cycle for "1st command prefetch"? |
| |
Frantic
Registered: Mar 2003
Posts: 1648 |
Thanks again for your efforts! Also thanks to HCL for adding some info as well! All in all a good initiative that more people should follow! |
| |
bepp
Registered: Jun 2010
Posts: 265 |
On my wishlist is: more pictures... A simple screenshot next to an effect/routine would be very informative for somebody who doesn't speak the lingo but who likes to get a better understanding :) |
| |
JackAsser
Registered: Jun 2002
Posts: 2014 |
Quote: Thanks for all the flowers. It indeed helps a lot if you code stuff you want to explain yourself. I mean, I had Ninjas source, but I restarted from scratch - it was also my first own "half variance delay" routine in there (3 A4-pages of sick and mostly wrong calculations ...).
I guess one has to turn 30 to feel the fun again, just the way it was with 15 years ...
@Jackasser: I've tried a lot to loose another cycle - but failed. How would you get 8 cycles of jitter?
What I mean is that with the one CIA you'd get 0..6 cycles of jitter and with the other CIA you'd get 1..7. I.e. treat the different CIA models as just one more cycle of jitter to take care of.
|
| |
Frantic
Registered: Mar 2003
Posts: 1648 |
@Bepp: Good point! I added hints and notes about adding pictures here and there in order to encourage people to add more (pedagogically motivated) pictures. |
| |
MagerValp
Registered: Dec 2001
Posts: 1078 |
Quoting Frantic(pedagogically motivated) pictures.
Goatse and lolcats in 3... 2... |
| Previous - 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12 | 13 | 14 - Next |
