[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date: Fri, 06 Jun 2014 14:42:29 -0700
From: "H. Peter Anvin" <hpa@...or.com>
To: Andy Lutomirski <luto@...capital.net>
CC: "H. Peter Anvin" <hpa@...ux.intel.com>,
Linus Torvalds <torvalds@...ux-foundation.org>,
Andrew Morton <akpm@...ux-foundation.org>,
Cyrill Gorcunov <gorcunov@...nvz.org>,
Linux Kernel Mailing List <linux-kernel@...r.kernel.org>,
Ingo Molnar <mingo@...nel.org>,
Sasha Levin <sasha.levin@...cle.com>,
Stefani Seibold <stefani@...bold.net>,
Thomas Gleixner <tglx@...utronix.de>,
Pavel Emelyanov <xemul@...allels.com>
Subject: Re: [GIT PULL] x86/vdso changes for v3.16
On 06/06/2014 02:39 PM, Andy Lutomirski wrote:
>>
>> For accessing memory members doing it via a pointer is pretty much TRT,
>> but for things that might be in register it is undesirable to force it
>> out to memory.
>
> Do you also believe in the folklore that GCC can optimize code
> sequences like the things in that header? Because I'm pretty sure
> that no clang or gcc version I've ever seen can do it.
>
I have seen gcc do some pretty sophisticated memory elision lately.
Don't know if that includes byte swaps.
> On the other hand, even a factor of ten in the time it takes to run
> vdso2c is completely irrelevant.
Yep, as I noted in the patch I sent (which is broken - updated one
included here.)
-hpa
View attachment "0001-x86-vdso-Use-tools-le_byteshift.h-for-littleendian-a.patch" of type "text/x-patch" (6875 bytes)
Powered by blists - more mailing lists