This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: [patches] Re: Mode switching for i386 (Was: Re: [patches] Re: better fp truncation sequence on i386)
- To: "Prince, Timothy C" <timothy dot c dot prince at intel dot com>
- Subject: Re: [patches] Re: Mode switching for i386 (Was: Re: [patches] Re: better fp truncation sequence on i386)
- From: Jan Hubicka <jh at suse dot cz>
- Date: Tue, 26 Jun 2001 16:03:08 +0200
- Cc: "'Jan Hubicka'" <jh at suse dot cz>, Richard Henderson <rth at redhat dot com>, gcc-patches at gcc dot gnu dot org, patches at x86-64 dot org, gcc at gcc dot gnu dot org
- References: <9287DC1579B0D411AA2F009027F44C3F0AE96576@FMSMSX41>
>
>
> -----Original Message-----
> From: Jan Hubicka [mailto:jh@suse.cz]
> Sent: Tuesday, June 26, 2001 1:06 AM
> To: Richard Henderson; Jan Hubicka; gcc-patches@gcc.gnu.org;
> patches@x86-64.org; timothy.c.prince@intel.com; gcc@gcc.gnu.org
> Subject: Re: [patches] Re: Mode switching for i386 (Was: Re: [patches]
> Re: better fp truncation sequence on i386)
>
>
>
> > Two solutions. One, spill in XFmode, which some people have
> > wanted for various purposes anyway. Two, recognize that such
> >Spilling in the XFmode is about 5 times more expensive, so we shouldnot go
> for it by >>default.
> XFmode spills will be expensive if they produce cache line splits (i.e.
> 16-byte alignments must be maintained, thus eating more cache). My last
> copy of the Lahey/Fujitsu compiler had problems with this.
I do have plans on getting dynamic stack alignment working. Last time
it got suck on multiple entry prologues and dwarf2 output.
We've chatted about it with Jiri here today and so I hope to have
patch for this problem soonish.
Still I would love to see XFmode spills just as an option.
Think would get more tricky if we decide to mix i387 and SSE code.
Tnhen SSE code is not able to do XFmode spills.
Honza