Hi, I used a matrix multiplication and take the time of a execution in both of softcores and executing the algorithm on MOR1K the performance is twice faster than on OR1200, someone knows why the MOR1K is much better than OR1200?

I'm happy to hear that. The mor1kx was rewritten from scratch to be better, smaller and faster. It seems like the hard work has paid off :)

But it's twice faster than OR1200 because of the six stage pipeline? (Sorry,I forgot to say that I'm using the mor1k4.1 cappuccino).

Hi suelin,

Did you run the same romcode(instructions) in mor1kx and or1200 tests?

I run rtl simulation to observe the behavior.
In or1200(cache disabled) environment, it's 3 wb_clk cycles per instruction execution.
In or1200(cache enabled) environment, it's 4 wb_clk cycles per instruction execution.
In mor1kx(cappuccino) environment, it's 5~6 wb_clk cycles per instruction execution.

Nick

Hi Nick!

So, the ROM is the same on OR1200 and MOR1KX.
Other observation that I forgot to do is that I'm using the board Atlys.

Suelin

It's hard to say exactly why there is such a large difference. mor1kx is faster, but it could also be different cache sizes, more efficient memory accesses, serial vs parallel multipliers and plenty of other things.

Anyhow, or1200 isn't actively developed anymore, so we would recommend you to move to mor1kx anyway

//Olof