qemu-devel
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Qemu-devel] [PATCH v2] target-mips: Use TCG registers for the FPU.


From: Aurelien Jarno
Subject: Re: [Qemu-devel] [PATCH v2] target-mips: Use TCG registers for the FPU.
Date: Fri, 5 Oct 2012 19:03:51 +0200
User-agent: Mutt/1.5.21 (2010-09-15)

On Fri, Oct 05, 2012 at 07:20:01AM -0700, Richard Henderson wrote:
> Ping.  You gave it your Reviewed-by.
> 
> r~

Yes, but I also said that with the current TCG status, it actually 
reduces the speed of the emulation instead of increasing it. I am not 
sure we really want that, even if the generated code is better.

We have to get rid of the loops on all TCG temps first, which is exactly
what the patch series I sent last week does. I have to send a new
version, but the week has been a bit crazy here and I haven't found time
to finish polishing it, nor reviewing the patches sent on the mailing
list. Let's hope the week-end will be better for that.

> On 09/18/2012 11:04 PM, Richard Henderson wrote:
> > With normal FP, this doesn't have much affect on the generated code,
> > because most of the FP operations are not CONST/PURE, and so we spill
> > registers in about the same frequency as the explicit load/stores.
> > 
> > But with Loongson multimedia instructions, which are all integral and
> > whose helpers are in fact CONST+PURE, this greatly improves the code.
> > 
> > Signed-off-by: Richard Henderson <address@hidden>
> > ---
> > 
> > As requested, only generating 64-bit fp registers now.  The generated
> > code looks quite good on i386.  It could be a tad better for x86_64,
> > but it's still better than it was.
> > 
> > 
> > r~
> > 
> > 
> >  target-mips/translate.c | 96 
> > +++++++++++++++++++++++++++----------------------
> >  1 file changed, 54 insertions(+), 42 deletions(-)
> 
> 

-- 
Aurelien Jarno                          GPG: 1024D/F1BCDB73
address@hidden                 http://www.aurel32.net



reply via email to

[Prev in Thread] Current Thread [Next in Thread]