This is the mail archive of the gcc-patches@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]

Re: [patches] Re: SSE min/max instruction support


> On Sun, Feb 25, 2001 at 12:14:14AM +0100, Jan Hubicka wrote:
> > So I've started to play with idea, that probably emit_conditional_move
> > should be pretty aggresive and should try such special cases first - min,max,
> > store flags and similar before doing the conditional move expander.
> 
> No, I think this should be done directly in ifcvt.c without
> cluttering up emit_conditional_move.
ifcvt.c is currently the only user of emit_conditional_move.
Do we want to keep it so? Perhaps it makes sense to discover some conditional
moves in the RTL generation pass directly (for instance the MIN/MAX expanders
can do).

It can be also nice to recognize -a>a?-a:a as abs. Combiner can't, ifcvt
maybe can.

Where is the right place in ifcvt to do so?
Honza
> 
> 
> r~


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]