[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: Re: [f-cpu] Memory convention
hi !
Ben Franchuk wrote :
>Yann Guidon wrote:
>
>> Is this really a problem ? As far as i remember, the endian flag in the
>> instructions is still present...
>> In the endian threads, there are always people wanting somethings and
>> its contrary. That's why
>> the flag is here. IIRC it's little endian by default.
>>
>> Now, anybody is free to use the endian one wants. Even though i'll be
>> the third to claim that
>> SIMD requires little endian to work properly.
>Why? Too lazy to find the lastest docs?
> But if somebody needs
>> something else,
>> the endian flag is here to fullfill his programming needs.
>I asume that the endian flag is for SIMD.
no. The load/store instructions don't care at all about the
data format, they only know the width of the register to write.
>One other problem with SIMD is that I asume it is for
>just operations with simple register indirect addressing.
>*a=*a op *b;a++;b++;
???
>If you need anything else you have to go back to
>normal processing. This limits the what you can do
>as you want to have orthagonal processing ability on
>SIMD.
i'm not sure to understand what you mean.
Can you elaborate ?
YG
*************************************************************
To unsubscribe, send an e-mail to majordomo@seul.org with
unsubscribe f-cpu in the body. http://f-cpu.seul.org/