Page 3 of 5

Re: 65C02 in verilog - extended version of Arlet's core

Posted: Fri Sep 15, 2017 10:34 am
by Windfall
BigEd wrote:
So in some handwavey and incorrect sense that's 500GHz of 6502 power for $15k...
Isn't it a bargain :D

Re: 65C02 in verilog - extended version of Arlet's core

Posted: Fri Sep 15, 2017 12:52 pm
by hoglet
BigEd wrote:
So in some handwavey and incorrect sense that's 500GHz of 6502 power for $15k...
== $30 / GHz.

Or with PiTubeDirect on a Pi Zero you get one core of 275 MHz for ~$8.

== $29 / GHz

So surprisingly similar!

Dave

Re: 65C02 in verilog - extended version of Arlet's core

Posted: Fri Sep 15, 2017 1:44 pm
by Windfall
hoglet wrote:
BigEd wrote:
So in some handwavey and incorrect sense that's 500GHz of 6502 power for $15k...
== $30 / GHz.

Or with PiTubeDirect on a Pi Zero you get one core of 275 MHz for ~$8.

== $29 / GHz

So surprisingly similar!

Dave
Heh :D Yes it seems that way, doesn't it. Even if you could find something to keep 2000 cores busy.

But I only paid $1500 and $8 is a somewhat optimistic figure. :wink:

Re: 65C02 in verilog - extended version of Arlet's core

Posted: Fri Sep 22, 2017 1:18 am
by Windfall
Even the ginormously complicated 32016 (with FPU), in the form of the 32016 Second Processor (http://www.zeridajh.org/hardware/soft32 ... /index.htm), will run at 170 MHz on this board ...

:shock: :shock: :shock:

Re: 65C02 in verilog - extended version of Arlet's core

Posted: Sun Sep 24, 2017 7:01 am
by kakemoms
Just to continue the wavy bistromatics, my unpublished cutdown 8-bit n6502 gets to 100MHz with all instructions 1 cycle (efficiently making it 300+ times faster than 1MHz 6502), takes 250 LUTS and has 1KB address space. Thus one can fit 37 cores into a MachXO3-9400 that costs $13.71..

That gives the equivalent of 11.1GHz or $1.23USD/GHz... :D

Re: 65C02 in verilog - extended version of Arlet's core

Posted: Sun Sep 24, 2017 7:04 am
by BigEd
Sounds like something worth publishing kakmoms! Do you intend to release it? If you do, please don't get blocked by the idea of waiting until it is finished - sharing works in progress is also helpful.

Re: 65C02 in verilog - extended version of Arlet's core

Posted: Sun Sep 24, 2017 12:24 pm
by Windfall
kakemoms wrote:
Just to continue the wavy bistromatics, my unpublished cutdown 8-bit n6502 gets to 100MHz
Kind of useless if you don't mention on what hardware.
kakemoms wrote:
with all instructions 1 cycle (efficiently making it 300+ times faster than 1MHz 6502), takes 250 LUTS and has 1KB address space.
And somewhere a nose was growing rapidly ... :)

Re: 65C02 in verilog - extended version of Arlet's core

Posted: Sun Sep 24, 2017 12:53 pm
by BigEd
Is that hostility I sense John? There's no need for that.

Re: 65C02 in verilog - extended version of Arlet's core

Posted: Sun Sep 24, 2017 1:05 pm
by Windfall
BigEd wrote:
Is that hostility I sense John? There's no need for that.
Uh .. No ? See the smiley ? :D

Re: 65C02 in verilog - extended version of Arlet's core

Posted: Sun Sep 24, 2017 1:13 pm
by BigEd
There are people who park on double yellow lines and put their hazard lights on. It doesn't entirely change what they've done...

But like you, I'd be interested to see more of kakemom's creation.

Re: 65C02 in verilog - extended version of Arlet's core

Posted: Sun Sep 24, 2017 1:18 pm
by Windfall
BigEd wrote:
There are people who park on double yellow lines and put their hazard lights on. It doesn't entirely change what they've done...

But like you, I'd be interested to see more of kakemom's creation.
I kinda think an all 1-cycle 250-LUT 6502 is a complete (but nice) fantasy. But hey, we shall see (or actually probably not if I'm right). :D

Re: 65C02 in verilog - extended version of Arlet's core

Posted: Sun Sep 24, 2017 2:10 pm
by Dr Jefyll
The project is described as a cutdown 6502, not a complete implementation. It's fair to ask how drastic the cutting down was -- and it seems it must've been drastic indeed! -- but the project is unpublished. If it remains so then I'm OK with that. But it'd be interesting to hear some further detail.

Re: 65C02 in verilog - extended version of Arlet's core

Posted: Sun Sep 24, 2017 3:10 pm
by Windfall
Dr Jefyll wrote:
The project is described as a cutdown 6502, not a complete implementation.
Indeed. But, in my view, it stops being a 6502 if it doesn't execute the expected instruction set. Better call it '6502 inspired' or '6502 challenged', then, if it doesn't. :D I'd expect 'cutdown' to refer to some external property (like the 1 KB memory limitation).

Re: 65C02 in verilog - extended version of Arlet's core

Posted: Sun Sep 24, 2017 9:19 pm
by kakemoms
Windfall wrote:
Dr Jefyll wrote:
The project is described as a cutdown 6502, not a complete implementation.
Indeed. But, in my view, it stops being a 6502 if it doesn't execute the expected instruction set. Better call it '6502 inspired' or '6502 challenged', then, if it doesn't. :D I'd expect 'cutdown' to refer to some external property (like the 1 KB memory limitation).
Well, you are a purist, which is fine. Ultimately 6502 code does or doesn't run on other versions of the 6502, its all about compability level. Try running commodore 64 code on a 2MHz 6502, and it will probably fail due to some
timing-critical routine.

So this is a 6502-like cpu with fewer instructions and some special considerations, but its code will compile on any 6502 assembler.

I may publish it once its finalized.

Re: 65C02 in verilog - extended version of Arlet's core

Posted: Sun Sep 24, 2017 10:12 pm
by Windfall
kakemoms wrote:
Windfall wrote:
Dr Jefyll wrote:
The project is described as a cutdown 6502, not a complete implementation.
Indeed. But, in my view, it stops being a 6502 if it doesn't execute the expected instruction set. Better call it '6502 inspired' or '6502 challenged', then, if it doesn't. :D I'd expect 'cutdown' to refer to some external property (like the 1 KB memory limitation).
Well, you are a purist, which is fine. Ultimately 6502 code does or doesn't run on other versions of the 6502, its all about compability level. Try running commodore 64 code on a 2MHz 6502, and it will probably fail due to some
timing-critical routine.

So this is a 6502-like cpu with fewer instructions and some special considerations, but its code will compile on any 6502 assembler.
Okay. Let's just say that your 6502 is not for existing 6502 code. Only new code, adhering to specific and as yet unseen restrictions.