[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [f-cpu] Execution unit port



Yann Guidon a écrit :
> 
> hello,
> 
> nicO wrote:
> > Hello,
> >
> > I'm looking for the "usual" definition of the port of a unit :
> >
> > 3 data in,
> > 2 data out,
> "as much as needed"
> (ASU could have 4 output ports but this is not yet completely determined).
> 
> > SIMD flag,
> > clk,
> > enable or start,
> > reset,
> > something for the exception ?
> except for the SR unit, no EU needs to report exceptions
> (protection violation, but the pipeline is stalled at issue
> when the SR EU is used).
> 
> FP units can report exceptions if the "IEEE flag" is set
> (again : the pipeline is stopped until we are sure that the
> result is ok).
> 
> concerning the "necessary flags" for example, ROP2 requires the 2 mode
> bits and 4 function bits. Some transation and buffering of the 4 function bits
> is performed during the Xbar cycle.
> 
> > If there is something to say if the outed data is correct or not ?
> no. Data are discarded "implicitely" by the Xbar write stage,
> by choosing a proper source for the proper register and with the right write mask.
> 
> > Does i miss a signal ?
> no, except for very special cases, it does the trick.
> there's no magic :-)

Yep, but we need something more precise. In that case, it will be much
more easy to add or remove unit. Special case could became very boring
to handel...

nicO

> 
> > nicO
> WHYGEE
> ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
> *************************************************************
> To unsubscribe, send an e-mail to majordomo@seul.org with
> unsubscribe f-cpu       in the body. http://f-cpu.seul.org/
*************************************************************
To unsubscribe, send an e-mail to majordomo@seul.org with
unsubscribe f-cpu       in the body. http://f-cpu.seul.org/