> With ADCs in the design, I can use the adcX_clk, while I want to use a clock 
> a little higher than adcX_clk to process with the data. How can I deal with 
> the second clock?
The short answer is that you can't.  Simulink can create a fake 2x clock for 
you by adding enable/clock gating lines to all your other logic and clocking 
the entire physical device at twice the speed but this rarely works unless 
you're wanting to run everything very slowly. In all cases, the Simulink domain 
is always clocked by one clock domain. 

This multi-clock region requirement came through quite strongly at the last 
CASPER workshop and is something we're thinking about, especially in light of 
overlapping PFB requirements. It's a tricky thing though. FWIW, Matlab's HDL 
coder can apparently support such things, and looks like a promising 
alternative to XSG in many other ways too. It's also something that's being 
built into the new MyHDL toolflow from the start. But these new 'flows are all 
a long way away if they ever get off the ground.

The current CASPER toolflow is designed so that all the DSP operations happen 
synchronously with the input data. This locally-syncronous model is actually 
rather efficient and is in general a good one to follow. I'd suggest you try'n 
tweak your processing logic to support synchronous data.

In terms of 10GbE cores and other peripherals requiring additional clock 
domains (such as the bus infrastructure), this is handled automatically in the 
toolflow by the base system, coded in HDL (see the XPS infrastructure/pcores 
directory for the hairy details). In general, you don't have to worry about 
this stuff.

Jason

Reply via email to