On Wed, 22 Jun 2011, Tim Harris (RESEARCH) wrote:
Hi Nilay,
I'm not sure if this will work for your application, but I've usually
followed the existing pseudo instructions in taking inputs in fixed
registers, rather than attempting to do smarter decoding of the
instruction.
I then either (#1) run my application on the simulator, modelling the
new instruction, or (#2) run it at higher speed on real hardware, and
write a handler for the undefined-instruction fault and have the handler
provide a simple functional model of the new instruction. Being able to
do #2 has really helped me with debugging, and having the inputs in
fixed registers makes #2 much easier.
This was my fall back option was and now I have started exploring both.
Again, in this case how do we control the register size that would made
use of? Suppose I need a 32-bit value in the register. Will placing the
required value in %edi and passing %Rdi to the function called when the
instruction is encountered enough?
--
Nilay
_______________________________________________
gem5-users mailing list
[email protected]
http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users
