[Author Prev][Author Next][Thread Prev][Thread Next][Author Index][Thread Index]

Re: [f-cpu] Status quo

Hi !

Le 2015-03-30 14:21, Nicolas Boulay a ÃcritÂ:
2015-03-30 13:42 GMT+02:00 <whygee@xxxxxxxxx>:
"one cycle" is not really possible with modern ultra-pipelined CPU.
I think of one cycle thoughtput not latency.
hey, please be accurate ;-)

Other "complex accelerators" might encounter problems,
for example TCP/IP.
That's not a coprossessor, that's an IO device.
I said TCP/IP accelerator, not just Ethernet MAC+PHY.

Another kind of pure HW accelerator is JPEG/MPEG block DCT "accelerator".

Some accelerated
ethernet board are also slower than pure SW with a stupid board, if a
round trip is needed, between cpu and the board.
it depends.
and other boards are fast.
The Wiznet chip I use is very satisfying :-)

Writing this reminds me that I had examined such a system, indeed !
http://ygdes.com/HSF2009/HSF2009_GPL.html [1]
Short version : a handled game console fits most of the criteria I
listed above
Â- small, so relatively inexpensive,
Â- graphics require tons of computations, we can hack accelerators
Â- low-power, real-time, embedded design
Â- has a "kernel" side and runs third-party "applications" (games)
I don't say we should start a gaming company. I mean :
this is the kind of realistic target system that could
both prove the points we make with F-CPU (freedom etc.)
AND test our designs against.

This could be fun. It could looks like a mix between opengraphics and

more or less...

But the board can't be too expensive, 100â, is a
maximum. Mylikmyst (http://m-labs.hk/m1.html [2] ) was great but too
expansive for a hobby (1000â).
MM is a "niche" device...

You should only sell a board, a case and a LCD screen will cost too
much. RasperryPi is only a single board.

Speaking about the Pi, I use some for development
and they are a convenient replacement for many things
until we have a working F-CPU. For example, I use
one Pi to re-flash a SPI memory, through a CGI program
accessed by JavaScript through local Ethernet.
I hope to bring more user-friendly, cheap and
free software of this kind to the F-CPU project.

The LCD can be cheap, depending on the model.
Existing game consoles have driven the price quite low,
PSP for example has a decent, cheap screen that can be sourced
all over the 'net. Other components like batteries,
connectors, even analog knobs can be easily found.
The expensive parts would be the FPGAs and the PCB.
Even for small runs with QFP, professional soldering can be
a significant cost, but i think BGA will be needed
so 4-layer or 6-layers PCB are necessary and the
soldering will have higher fault rates...

Then the intent is not to compete with Sony or Nintendo :
it's a development platform, so it contains the necessary I/O
and tools (ICE, power, programmers...) to develop the guest custom F-CPU. So the price-cost would be between the Pi and the MM if/when a small batch (10?) is manufactured. It's a modular system so the price and risk is shared between
the host (interface) board and the guest (application) board.

Nikolay asked that we only propose things that we can actually do.
I'd reinforce that and add that others must be able to follow and use what is done. Today, I have many of the necessary elements for the portable console and indeed, i've been working on the design of a cool development tool for a parallel project.
There is no market demand for a portable console of that type,
but we need to develop the F-CPU as well as show a practical application.
The technical design can then be easily re-purposed for
future applications that others may need/want.

I suppose I now have to create a sub-team and infrastructure for
this project, who is interested to help and how ?
I already have found the name :-D


To unsubscribe, send an e-mail to majordomo@xxxxxxxx with
unsubscribe f-cpu       in the body. http://f-cpu.seul.org/