qemu-devel
[Top][All Lists]
Advanced

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

Re: [Qemu-devel] [PATCH 28/32] ppc: Avoid double translation for lvx/lvx


From: David Gibson
Subject: Re: [Qemu-devel] [PATCH 28/32] ppc: Avoid double translation for lvx/lvxl/stvx/stvxl
Date: Fri, 29 Jul 2016 13:34:56 +1000
User-agent: Mutt/1.6.2 (2016-07-01)

On Fri, Jul 29, 2016 at 12:13:01PM +1000, Benjamin Herrenschmidt wrote:
> On Fri, 2016-07-29 at 06:19 +0530, Richard Henderson wrote:
> > This, I'm not so keen on.
> > 
> > (1) The helper, since it writes to registers controlled by tcg, must be 
> > described to clobber all registers.  Which will noticeably increase memory 
> > traffic to ENV.  For instance, you won't be able to hold the guest register 
> > holding the address in a host register across the call.
> 
> Ah I wasn't aware of this. How do you describe such a clobber ? Can I describe
> specifically which one is clobbered ? I didn't think TCG kept track of the 
> vector
> halves but I must admit I'm still a bit new with TCG in general.
> 
> I noticed other constructs doing that (passing a register number to an 
> opcode),
> what do I do to ensure the right clobbers are there ?
> 
> > > (2) We're going to have to teach tcg about 16-byte data types soon 
> > > anyway, for 
> > the proper emulation of 16-byte atomic operations.
> 
> Is anybody working on this already ? I thought about that approach as
> it would definitely make things easier for that and a couple of other
> cases such as lq/stq.

What should I do with this in the short term?  Leave it in
ppc-for-2.8, or remove it for now pending possible changes?

-- 
David Gibson                    | I'll have my music baroque, and my code
david AT gibson.dropbear.id.au  | minimalist, thank you.  NOT _the_ _other_
                                | _way_ _around_!
http://www.ozlabs.org/~dgibson

Attachment: signature.asc
Description: PGP signature


reply via email to

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