[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20150506041104.GA12080@gmail.com>
Date: Wed, 6 May 2015 06:11:04 +0200
From: Ingo Molnar <mingo2.kernel.org@...il.com>
To: Dave Hansen <dave.hansen@...ux.intel.com>
Cc: linux-kernel@...r.kernel.org,
Andy Lutomirski <luto@...capital.net>,
Borislav Petkov <bp@...en8.de>,
Fenghua Yu <fenghua.yu@...el.com>,
"H. Peter Anvin" <hpa@...or.com>,
Linus Torvalds <torvalds@...ux-foundation.org>,
Oleg Nesterov <oleg@...hat.com>,
Thomas Gleixner <tglx@...utronix.de>
Subject: Re: [PATCH 207/208] x86/fpu: Add FPU performance measurement
subsystem
* Dave Hansen <dave.hansen@...ux.intel.com> wrote:
> On 05/05/2015 10:58 AM, Ingo Molnar wrote:
> > x86/fpu: Cost of: XSAVE insn : 104 cycles
> > x86/fpu: Cost of: XRSTOR insn : 80 cycles
>
> Isn't there going to be pretty huge variability here depending on
> how much state you are xsave/xrstor'ing and if the init/modified
> optimizations are in play?
Hopefully there's such variability! :)
I thought to add measurements for that as well:
- to see the costs of this instruction family when various xstate
components are in 'init state' or not
- maybe even measure whether it can optimize based on whether things
got changed since the last save (which the SDM kind of alludes to
but which I doubt the hw does)?
This initial version only measures trivial init state save/restore
cost.
Thanks,
Ingo
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
Powered by blists - more mailing lists