This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: Address side effect patches
- To: Joern Rennecke <amylaar at cygnus dot co dot uk>
- Subject: Re: Address side effect patches
- From: "Michael P. Hayes" <michaelh at ongaonga dot chch dot cri dot nz>
- Date: Sun, 31 May 1998 23:31:46 GMT
- Cc: michaelh at ongaonga dot chch dot cri dot nz (Michael P. Hayes), egcs at cygnus dot com, gcc2 at cygnus dot com
- References: <199805311159.LAA17127@ongaonga.chch.cri.nz><199805311309.OAA27594@phal.cygnus.co.uk>
Joern Rennecke writes:
> > q = p; *++q; *q++. While the algorithm I used can generate quite good
> > sequences of address side effects, it could be improved if a better
> > flow graph was available and if there was a cost mechanism for
> > addressing modes.
>
> There actually is one already: the ADDRESS_COST target macro. Look into
> tm.texi for details.
I meant to say, a better address cost mechanism. On many machines,
the address cost for an operand is dependent upon the other insn
operands, but the current mechanism doesn't provide any context.
Michael.