This is the mail archive of the libc-alpha@sourceware.org mailing list for the glibc project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: [PATCH] PowerPC - cache line size initialization


libc-alpha-owner@sourceware.org wrote on 12/21/2011 08:30:17 PM:

> From: Ulrich Drepper <drepper@gmail.com>
> To: Adhemerval Zanella <azanella@linux.vnet.ibm.com>
> Cc: "GNU C. Library" <libc-alpha@sourceware.org>
> Date: 12/21/2011 09:37 PM
> Subject: Re: [PATCH] PowerPC - cache line size initialization
> Sent by: libc-alpha-owner@sourceware.org
>
> On Wed, Dec 21, 2011 at 07:45, Adhemerval Zanella
> <azanella@linux.vnet.ibm.com> wrote:
> > This patch address the __cache_line_size initialization by moving
> it to sooner point,
> > prior any DSO initialization.
>
> No.  Just look what other architectures do.  Also, if the value is not
> optimized all you should get is less performance.  Otherwise the code
> isn't written as it should.
>

Adhermerval, just change to code to check for '0' __cache_line_size and if
equal execute as if the fill char is none zero (the path that does not use
DCBZ).

Assuming this is memset code is unique to A2/A2O only they will take the
hit (for allowing multiple cache line sizes).

This has been going on too long ...


Steven J. Munroe
Linux on Power Toolchain Architect


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]