[openrisc] Re: Floating point (yes/no)?

["Damjan Lampret" <lampret@opencores.org>]

Hi Chris,

No, OR1200 is w/o floating point.

RTEMS uses heavy find first bit instruction? How?

I know that instruction is useful for implementing fast floating-point
emulation (for normalization) but what RTEMS scheduler has to do with this
instruction?

Anyway I was planning to add it however right now I have such troubles with
OR1200 forcing it to run at 100MHz that I'd rather kick out shift
instructions than add fidn first bit instruction ;-)

Anyway do you want to add this instruction to the architecture for use in
future implementations?

regards,
Damjan

PS Lets move this conversation to openrisc mailing list to make list alive
again.

----- Original Message -----
From: "Chris Ziomkowski" <chris@asics.ws>
To: <lampret@opencores.org>
Sent: Friday, May 11, 2001 9:41 AM
Subject: Floating point (yes/no)?


> Damjan,
>
> Did you say you were or were not planning on implementing
> the floating point module for Or1200? I know you mentioned
> it but I can't remember what you said.
>
> The reason I asked is that the RTEMS scheduler makes heavy
> use of a find first bit instruction, which we don't have.
> I can of course, write this in software, but I can also
> make a hack using the integer to floating point instruction
> if it is available.
>
> This is strictly an efficiency thing. I'm not suggesting
> changing the architecture (maybe another possibility for
> Or2k), and I'm not suggesting floating point for Or1200
> if you weren't including it.
>
> I'm just asking. Can I assume floating point is available
> or should I implement it in software?
>
> Thanks,
>
> Chris
> chris@asics.ws
>
>
>
>
>
>
> _______________________________________________________
> www.asics.ws - Solutions for your ASIC needs
>