This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: powerpc & unaligned block moves with fp registers
- To: kenner at vlsi1 dot ultra dot nyu dot edu (Richard Kenner)
- Subject: Re: powerpc & unaligned block moves with fp registers
- From: David Edelsohn <dje at watson dot ibm dot com>
- Date: Thu, 08 Nov 2001 16:26:06 -0500
- cc: gcc at gcc dot gnu dot org
>>>>> Richard Kenner writes:
> I guess I did not explain this clearly enough. The DImode objects
> do not need to be naturally aligned. Following the definition of
> STRICT_ALIGNMENT to 0, these instructions will work, just work more
> slowly.
Richard> You can't have it both ways! Either they are too slow, in which case
Richard> STRICT_ALIGNMENT should be set or they are *not* too slow, in which
Richard> case there is no problem.
Some architectures have gradual degradation while STRICT_ALIGNMENT
assumes a step function. This is why we use SLOW_UNALIGNED_ACCESS with
MODE and ALIGN parameters.
GCC is using STRICT_ALIGNMENT to compute a heuristic which should
be based upon SLOW_UNALIGNED_ACCESS. The dichotomy of STRICT_ALIGNMENT
does not have the flexibility to describe all architectures.
David