http://www.digilentinc.com/Products/Det ... rod=NEXYS2

I hear from a fellow consumer of such that Linux support is good, and the XC3S1200E has greater capacity that the Xess XSA3S1000 unit, and is somewhat cheaper, AND has native USB interface for programming. Oh, AND it has more free I/O too -- the four Pmod ports combined give you 32 bits of I/O, and the FX2 connector has another 43.

Anyone here have any experience with this unit? What are your thoughts?

(Interesting that the next model up, which uses a Spartan 6-LX45 type chip, has twice as LUTs as the 3S1200, but the usable I/O options are greatly reduced. It appears that the GENESYS unit is intended more for multimedia applications than general homebrew hackery. The cost per LUT is about the same, though.)

Jeez, what are you planning on fitting onto that thing?!

Multiple 65816 cores with 16MB RAM per core, and DMA?

_________________
65Org16:https://github.com/ElEctric-EyE/verilog-6502

Actually, I was planning on a 32- or 64-bit stack-architecture CPU with MMU (or two, or three, or four), VGA controller logic, audio input and output, 65SIB support, and perhaps a high-bandwidth expansion backplane as well.

If I'm going to shell out the bucks, I'm going to get the best overall bang for them. My metric is cost per LUT, and I've budgeted $250 for this purpose. Since I spend 90% of my time sitting in front of a computer, instead of a soldering iron, it makes sense for me to purchase one, general purpose device instead of a plurality of smaller, less capable units.

Depends on what you're planning to do... for retrocomputing stuff/simpler projects the Altera DE1 is more useful out of the box... it has 12-bit VGA, SD slot, and SRAM which is easier to interface to for early work. But less capacity than the FPGA on this board...

Are you sure about it being less dense? My research shows the chip used in the DE1 has 20K LUTs, which is actually more than the 3S1200's 17K or so.

Altera's price for the unit is comparable to Digilent's price; actually a bit higher. But, what kills the DE1 for me is the complete lack of trivial-to-download PDFs on it (specifications, schematics, etc.), and what appears to be a substantially reduced availability of digital I/O pins at my disposal.

While it is true that I won't have all the fancy built-in I/O ports and whatnot, I would much rather have access to the digital I/O pins than I would to, say, dedicated audio connectors. Call me old-fashioned, but I want something I can stick a probe on and see on the o'scope!

Retrocomputing isn't the only thing I was going to use it for (although that is certainly a prominent part of it). I was also going to explore other projects, including but not limited to things involving ultra-wideband and amateur radio too.

Has anyone got direct experience of using the (free) Altera tools on linux? It would be nice to know how well that works.

A quick search led me to this article on using the DE2, which seems encouraging in that it worked, but it did involve running windows XP in a virtualised environment. Perhaps DE1 is better supported in native linux?

Hehe, I didn't even think to research that -- good point now that you bring it up. I already know from my associate's experience that WebPACK works on Linux (and youtube tutorial videos confirm this too), so I won't have to shell out an additional $200 for a copy of Windows plus a virtual machine installation.

GOT IT!

Now it's time to plan out my 256-bit stack CPU design. Why 256 bits? Because I can! Bwaaahahahahah!!!

Sweet! Are you getting into Verilog too?

_________________
65Org16:https://github.com/ElEctric-EyE/verilog-6502

I have a Verilog simulator (Icarus), but I haven't done anything with it yet.

When I'm ready to synthesize, I'll grab the WebPack software from Xilinx and the Adept programmer from Digilent.

BTW, I tested the board (it comes pre-programmed with a burn-in test), and it worked flawlessly. I'm quite happy.

Ooh, and it seems that there is an interesting application for super-wide address spaces: single-address-space operating systems. SASOSes have the advantage that interprocess communications are lightning fast (think, AmigaOS fast!) because all you do is pass pointers around. Memory protection occurs with a special MMU, and isolation occurs by widely separating content in a single virtual address space.

256 bits is a bit much though -- a single PTE for that would be on the order of 80 bytes (yes, BYTES) long -- 63 or so to hold the virtual page number, 3 or 4 to hold the physical page frame, and a few more for permissions and access tracking. Inverted page tables become a hard requirement when you're dealing with an address space that large.