microblaze question

If you need a real small CPU, take a look at

ftp://137.193.64.130/pub/mproz/mproz3_e.pdf

Only a few dozen flip-flops and a few hundred gates.

i have a microblaze in a spartan 3e 250 and it takes about 50%

yes. you can run from block ram. you can't include a bunch of libraries (for example, printf won't fit).

alan nishioka snipped-for-privacy@nishioka.com

Mmm, and only to be programmed in assembler, I assume. Picoblaze is even smaller... Right now I'm running an opencore 8051 which works fine and can be programmed in C, but for the project I'm doing it could be that this runs too slow and cannot handle to dataflow entering a FIFO. microblaze would certainly solve it unless I have to add hardware. Thanks anyway. Taco

Thanks a lot, that's exactly what I needed to know. taco

What is the complexity of your code?

If you want to minimize the code to fit inside block RAMs the 8051 might not be your best choice. You can find some comparison data on my proc4 web page:

As a note, a MCU based on proc4 core fits in a spartan3 50 (i.e. the smallest) using only the internal RAM, making use of [almost] all pins (as I/Os).

-- mmihai

I took a quick look at your webpage [1] and have a few questions:

Do you have a C-compiler for your processor? (You only say that it is programmable in a high-level language on your webpage.)

Do you know if commercial compilers for Z80 and 8051 will produce better results than sdcc?

I also note that the 8051 results were done with 6 clocks per machine cycle. AFAIK there are much faster 8051 cores available.

/Andreas

Yeah, same question. I think this is not the case and if it's not a standard high level language, moving to a different processor is not very easy.

The answer is yes. The Keil compiler produces better optimized code for

Mmm. to get the graphs in some better direction? Taco

No, it's not C ...

I don't have access to comercial compilers for Z80/8051.

Yes, I know. The 8051 simulator I've used ( 8051sim) computes the cycles based on the original timing. I saw 2X cores which are just twice faster. The one called "single cycle" do not execute all the instruction in a single cycle (i.e. timing is dependent on the instruction, I think I saw 1-4) so you really need a core simulator.

-- mmihai