More FPGA spam - dual Arm V9 and Cyclone V


Size is also a concern the Altera module alone would take up a good amount of space of the current Pandora PCB..
 
Last edited by a moderator:
While its a nice board and all.


It's lacking a few useful things like a gpu, you either have to make your own using the fpga or do pure software rendering.


We already have one of these here at work, we are doing some evaluation for some project.
 
Late to the party, but why not a hybrid FPGA/hardcore(not emulated)-processor? I imagine we could run java at rates not achieved by portable devices 1.25 times the size of the OP2! Not only that, but we can appeal to people like myself who are interested in hobby-based processor design. You could emulate systems like the COSMAC ELF properly or some more modern TTL machines, without software! This lets the pandora 2 not only be somewhat future-proof but also allows us to drastically increase performance and appeal to another group entirely! With the FPGA we (might) be able to emulate a second GPU so we could appeal to those who do CAD and other complex 3d-things. IMHO FPGAs and the likes of softcore processors are the future.
 
FPGA in a handheld - I'm not an expert at all, but my gut feeling says "maybe in 10 years?"
 
I'm not convinced by the "improve performance" bit.

I thought FPGAs were bulky, power-hungry and inefficient.
 
An FPGA like this is not a good choice for Java acceleration.. Java chips fell out of favor years ago, even ARM's implementation, because JITs on similar CPUs are faster. With an FPGA the situation is considerably worse than a Java chip implemented as an ASIC. To make use of an FPGA you need a task which you can extract a ton of parallelism out of or needs some really specialized functions, Java interpretation isn't either of these things.

The biggest reason why FPGAs are not a good idea is because no one is really prepared to write the HDL code to utilize them for the wide variety of applications that people are envisioning. In theory a decent FPGA could probably emulate things better than a similarly decent CPU, but in theory the DSP on Pandora could probably also help with emulation but no one uses it that way because it's a lot of work for an emulator that'll only benefit one device.

It's still a cool idea and it'd be interesting to see happen in some form, I just don't see it as likely to take off. Maybe someone should start with a lower risk platform, like a simple set-top box and let people see what they can develop on it. If something really novel comes out of it emulation-wise evaluate the possibility of making a handheld out of it. I remember there was supposed to be a handheld years ago with design by Ben Heck that was just this, an FPGA machine that played various platforms, but they only had Atari 2600 support prepared. Then they too decided to make it a console instead.. before ultimately scrapping it.
 
Last edited by a moderator:
I remember watching this video where the guy was talking about the problems you need to work around when trying to deploy this kind of config.

ARM <-> RAM <-> FPGA -> Screen.... the FPGA can be booted via the arm with new code on the fly.

Where do you want the IO to flow and how much are work is required to get a basic working setup?
 
Back
Top