6502.org Forum  Projects  Code  Documents  Tools  Forum
It is currently Sat Nov 23, 2024 5:50 am

All times are UTC




Post new topic Reply to topic  [ 16 posts ]  Go to page 1, 2  Next
Author Message
PostPosted: Wed Aug 05, 2020 2:59 pm 
Offline

Joined: Sat Apr 11, 2020 7:28 pm
Posts: 344
In a C64, the VIC-II controls the CPU, giving it the clock and stopping it
through RDY and AEC pins.

In a C64, the VIC-II receives two clock ticks. A "COLOR CLOCK" for the video
color generation, and a "DOT CLOCK" for its internal workings, where it gets
divided for external CPU clocking. A 8701 chip gets those clock given to the
VIC-II.

What about this?

Put a clock in a board of double the rate a 8701 needs to be fed with
(double of 14.31818 or 17.734472 Mhz).

Make the clock send: one tick to the 8701, and the other one to a 65c02/816
chip, so both the chips would receive its intended tick one each following
the other, alternately.

The VIC-II would behave as always, and the 65c02/816 would too. Both of them
would be able to access the RAM never at the same time, because their clocks
would never overlap.

But maybe that wouldn't stop the VIC-II to need being able to stop the CPU
to gets its reads during badlines, sprites and so.

Also, the VIC-II wouldn't be able to stand access for read or write to its
registers at such a high speed from the CPU (although the datasheet does not
list a "minimal" time for that - or I can't read a dataheet).

What do you think?


Top
 Profile  
Reply with quote  
PostPosted: Fri Aug 07, 2020 3:11 am 
Offline

Joined: Tue May 05, 2009 2:49 pm
Posts: 113
If I understand you correctly, you'd do this:

Image

This assumes the VIC-II can handle a cycle width half as wide as it normally sees, but for sake of argument...

As you can see, PHI0 is still 1MHz, and a bad cycle still consumes 100% of it, meaning the entire cycle is not available for the CPU.

(For reference, this timing diagram is from WaveDrom - https://wavedrom.com/editor.html and the code to create it is:

Code:
{signal: [
             ["Original",
                 {name: 'DOTCLOCK',wave: 'hl.h.l.h.l.h.l.h.l.h.l.h.l.h.l.h.l'},
              {name: 'PHI0',    wave: 'hl...............h...............l'},
              {name: 'VICCYC',  wave: 'x1...............x................'},
              {name: 'BADCYC',  wave: 'x1...............................x'},
              {name: 'CPURAM',  wave: 'hx...............1...............x'},
              ],
              {},
             ["Revised",
                 {name: 'DBLCLK',  wave: 'hlhlhlhlhlhlhlhlhlhlhlhlhlhlhlhlhl'},
                 {name: 'DOTCLOCK',wave: 'hl..hl..hl..hl..hl..hl..hl..hl..hl'},
                 {name: 'CPUCLOCK',wave: 'hlhl..hl..hl..hl..hl..hl..hl..hl..'},
              {name: 'PHI0',    wave: 'hl...............h...............l'},
              {name: 'VICCYC',  wave: 'x1...............x................'},
              {name: 'BADCYC',  wave: 'x1...............................x'},
                 {name: 'CPURAM',  wave: '1x1x..1x..1x..1x..1x..1x..1x..1x..'},
           ]
             
          ],             
 config: { hscale: 1 },
 head:     { text:'C6C02/816/VIC-II Timing' },
}


An alternative would be to send 8 double speed clocks to the VIC, and then 8 periods of no clock, and let the CPU use that second set of 8 clocks to do work, but I doubt VIC would operate with such an irregular clock, so I think that is a non starter.

The best bet is to throw each address/data line set into a mux with some really fast RAM (if you're planning to run the 65C02 at 8MHz, that means RAM needs to be <125nS, but you need to access it only on the high half cycle, so 62nS or better) and a latch. In the front half of each half cycle, you allow VIC to access, latching the data into a register and then presenting it to the VIC as DATA for the remainder of that half cycle. On the back half of the cycle, let the CPU have the RAM directly. Most of the VIC accesses would go unused, but each device would have total access to RAM (some provision needs to be made for when the CPU needs to talk to the VIC, but to do that, it needs to slow down to 1MHz, and then you're waiting on the VIC anyway:

Image

Code:
{signal: [
             ["Revised",
                 {name: 'DBLCLK',  wave: 'hlhlhlhlhlhlhlhlhlhlhlhlhlhlhlhlhl'},
                 {name: 'DOTCLOCK',wave: 'hl.h.l.h.l.h.l.h.l.h.l.h.l.h.l.h.l'},
                 {name: 'CPUCLOCK',wave: 'hl.h.l.h.l.h.l.h.l.h.l.h.l.h.l.h.l'},
              {name: 'PHI0',    wave: 'hl...............h...............l'},
              {name: 'RAMCLOCK',wave: 'hl.h.l.h.l.h.l.h.l.h.l.h.l.h.l.h.l'},
              {name: 'VICRAM',  wave: 'x1.x.............1.x..............'},
                 {name: 'CPURAM',  wave: '1x.1.x.1.x.1.x.1.x.1.x.1.x.1.x.1.x'},
           ]
             
          ],             
 config: { hscale: 1 },
 head:     { text:'C6C02/816/VIC-II Timing #2' },
}


Realistically, the VIC address lines won't settle down that fast, so you'd need to put the VIC data access about midway through the 1MHz half cycle (but, I didn't feel like updating my diagram, so just imagine it). In any case, that's the best way to use a 1MHz VIC-II and an 8MHz '02/816 and not have to deal with bad cycles.

Jim


Top
 Profile  
Reply with quote  
PostPosted: Fri Aug 07, 2020 5:39 pm 
Offline

Joined: Sat Apr 11, 2020 7:28 pm
Posts: 344
Another way could be giving VIC-II its own RAM (32K, 64K...) where it would read everything from, and make the CPU access that RAM only to store graphic data, and access a different physical RAM for running code, calculations and so.

The VIC-II would run thinking it is actually halting the CPU, but the CPU wouldn't be being halted, as AEC wouldn't be tied to RDY in the CPU.

But the CPU should be aware of when the VIC-II is reading the VRAM, so it wouldn't try to access it at the same time as it. Could be AEC line of the VIC-II used for such an awareness, instead of halting the CPU?


Top
 Profile  
Reply with quote  
PostPosted: Fri Aug 07, 2020 6:07 pm 
Offline

Joined: Tue May 05, 2009 2:49 pm
Posts: 113
If the CPU tries to access the graphics data when VIC is accessing, it'd have to be halted for the duration of the VIC access time (could be up to 65uS). So, your idea is more prone to stalling. If you mirror the RAM, then reads can be done from other RAM, while writes would still stall.

Jim


Top
 Profile  
Reply with quote  
PostPosted: Fri Aug 07, 2020 8:22 pm 
Offline

Joined: Sat Apr 11, 2020 7:28 pm
Posts: 344
brain wrote:
If the CPU tries to access the graphics data when VIC is accessing, it'd have to be halted for the duration of the VIC access time (could be up to 65uS). So, your idea is more prone to stalling. If you mirror the RAM, then reads can be done from other RAM, while writes would still stall.

Jim


That's why I thought using the AEC pin not for halting the CPU but for signaling: the code that would update the graphics data in VRAM would check throught AEC if access should be allowed.

Every time a read or write to a memory location inside the VRAM would be tried, the first thing to do would be "is AEC active?", and if "Yes", then do a NOP, and ask for AEC status again. When AEC wouln't be active that would mean that the VIC is not accessing the VRAM area, and the access would be allowed. But again, all of this by software, instead of relying on the CPU internal RDY signal.

Would it be safely assumed that a memory access by the VIC lasts half, or a full cycle of PHI0?


Top
 Profile  
Reply with quote  
PostPosted: Fri Aug 07, 2020 10:03 pm 
Offline

Joined: Tue May 05, 2009 2:49 pm
Posts: 113
I understand, and you can do NOP instead of halting the CPU, but the effect is the same (62-65 uS of the CPU waiting). A badline is 62-65 cycles of consecutive double VIC-II accesses, with no RAM time available for the CPU.

Jim


Top
 Profile  
Reply with quote  
PostPosted: Fri Aug 07, 2020 11:22 pm 
Offline

Joined: Sat Apr 11, 2020 7:28 pm
Posts: 344
brain wrote:
I understand, and you can do NOP instead of halting the CPU, but the effect is the same (62-65 uS of the CPU waiting). A badline is 62-65 cycles of consecutive double VIC-II accesses, with no RAM time available for the CPU.

Jim


Well... you are right. Even if the VIC-II have no access to the rest of the RAM, the bus would be shared amongst all the chips in this setup, so the CPU would be waiting for access any RAM, so...

What about a dual bus? I mean... Electronics experts would tell better, but there must be a way to have the VIC-II to have its own RAM, physically separated from the rest of the system, and the CPU his. But when certain addresses are touched by the CPU, and the VIC-II is not accessing those same locations, the magic would make the CPU access that video-bus, read/write there, and leave out. And I mean, physically leaving out, because the magic would keep the system-bus and the video-bus apart until needed again, so if the CPU is doing anything else like calculations or reading or writing data from or to ports, or whatever, it would be free to do it while the VIC-II is only keeping busy its video-bus.

It would be like... the computer having a video card with its own memory.

feasible? doable?


Top
 Profile  
Reply with quote  
PostPosted: Fri Aug 07, 2020 11:37 pm 
Offline

Joined: Tue May 05, 2009 2:49 pm
Posts: 113
Sure:

https://www.digikey.com/product-detail/ ... D/11523492

Jim


Top
 Profile  
Reply with quote  
PostPosted: Sat Aug 08, 2020 2:47 pm 
Offline

Joined: Wed Aug 16, 2017 5:35 am
Posts: 10
tokafondo, what do You think about asynchronous dual-port SRAM?
Try use it like this.
Here you can find them at reasonable prices, for example.


Top
 Profile  
Reply with quote  
PostPosted: Sun Aug 09, 2020 2:43 pm 
Offline

Joined: Sat Apr 11, 2020 7:28 pm
Posts: 344
Well, it seems dual port ram could do it... But it seems to be that an address decoder would be still needed, and some sort of PLA-like chip, because of the way the VIC-II work.

Anyway: is a dual port the only solution? It's like, the DPRAM would be the chip that would be managing a dual bus thing... but now I think that there should be a way for the CPU to manage the VIC-II registers, and having this DPRAM solution wouldn't allow the CPU doing that.

There must be something else... Something that allows isolating the VIC-II from the rest of the system when needed, so the CPU have free access to the RAM. Beamracer seems to be doing exactly that.


Top
 Profile  
Reply with quote  
PostPosted: Sun Aug 09, 2020 3:33 pm 
Offline
User avatar

Joined: Thu Dec 11, 2008 1:28 pm
Posts: 10986
Location: England
In a related discussion over on cbm-hackers, user 'smf' posted a link to this project, which runs a 4MHz 6502 in a C64:
http://wiki.projekt64.filety.pl/doku.ph ... kt64/turbo
via https://csdb.dk/release/?id=102360


Top
 Profile  
Reply with quote  
PostPosted: Sun Aug 09, 2020 3:51 pm 
Offline
User avatar

Joined: Thu May 14, 2015 9:20 pm
Posts: 155
Location: UK
The answer is, how much custom or glue logic do you want to throw at the problem.

There are other systems where various methods have been used to either stop the microprocessor only when it’s accessing the RAM where the video data is located. Or to allow the main microprocessor to continue to execute a program or read/write data during the time that video system is using the RAM where the display data is located.

For the former, having different physical memory chips with tristate buffers to allow the microprocessor to access the RAM chips which hold the video data is possible. When not accessing video RAM, the uP runs at full speed. When it tries to access the video RAM, it’s paused until the video system has a time slot where it does not need access.

Or the video RAM can be completely separate from the microprocessor memory map, with tristate latches used to allow writes to the video RAM only during the time that the video system does not need access. The microprocessor loads up the required address and the data then that is used to access the separate bus. This can even be made to be a FIFO buffer that automatically does the writes without the microprocessor having to sit waiting for access.

But if you are going that far, maybe a better solution would be to produce a new video controller using a FPGA.

Mark


Top
 Profile  
Reply with quote  
PostPosted: Sun Aug 09, 2020 6:50 pm 
Offline

Joined: Sat Apr 11, 2020 7:28 pm
Posts: 344
Thank you. Yes, the FPGA would be the easy way... And I say easy because it could be designed as desired.

But the VIC-II is not a bad graphics chip. It does what it does nicely, and it has been more than proven the things it can do when programmed properly.

So why not boost it with the capabilities of a faster processor? And that's the talk here. Thanks.


Top
 Profile  
Reply with quote  
PostPosted: Sun Aug 09, 2020 6:53 pm 
Offline

Joined: Tue May 05, 2009 2:49 pm
Posts: 113
Beamracer implements a dual port functionality in the CPLD/FPGA, just like I described initially.

Yes dual port RAM would prevent accessing the VIC registers. You'd have to have a "bypass" that allows the CPU to talk directly to the VIC-II

And, 1024MAK is right. Anything is possible, but it requires lots of glue logic to implement.

Jim


Top
 Profile  
Reply with quote  
PostPosted: Mon Aug 10, 2020 5:11 am 
Offline
User avatar

Joined: Fri Nov 09, 2012 5:54 pm
Posts: 1431
Don't know if this is helpful, but I now remembered that there had been a project for running the C64 at 2MHz,
by resorting to a 6502A and SRAM (instead of DRAM).


Top
 Profile  
Reply with quote  
Display posts from previous:  Sort by  
Post new topic Reply to topic  [ 16 posts ]  Go to page 1, 2  Next

All times are UTC


Who is online

Users browsing this forum: No registered users and 25 guests


You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot post attachments in this forum

Search for:
Jump to: