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] Fixed family and model detection for AMD CPU's


Please find the attached recreated patch to this mail. Bugzilla ID included in Glibc Changelog and git commit log.

Problem: In GLIBC Family and model values detection are incorrect. 
Why this fix? 
Based on arch ISA flags string and memory routines are selected. In future patches, will be defining ISA flags based on family 
and model values which will help in selecting the correct string or memory routines.

Is it OK to apply?

Thanks,
Amit Pawar


-----Original Message-----
From: Pawar, Amit 
Sent: Friday, November 20, 2015 11:45 AM
To: 'Joseph Myers'
Cc: 'Florian Weimer'; 'libc-alpha@sourceware.org'
Subject: RE: [PATCH] Fixed family and model detection for AMD CPU's

Ping!

-----Original Message-----
From: Pawar, Amit 
Sent: Friday, November 06, 2015 10:32 AM
To: 'Joseph Myers'
Cc: Florian Weimer; libc-alpha@sourceware.org
Subject: RE: [PATCH] Fixed family and model detection for AMD CPU's

I have filed a bug with document reference. Please do check the link https://sourceware.org/bugzilla/show_bug.cgi?id=19214
This is only a correctness patch and can help in future for right selection of string and memory routines.

Amit Pawar

-----Original Message-----
From: Joseph Myers [mailto:joseph@codesourcery.com] 
Sent: Thursday, November 05, 2015 6:20 PM
To: Pawar, Amit
Cc: Florian Weimer; libc-alpha@sourceware.org
Subject: RE: [PATCH] Fixed family and model detection for AMD CPU's

On Thu, 5 Nov 2015, Pawar, Amit wrote:

> Hi Florian,
> 
> This is a correctness patch where family and model detection was wrong.

That doesn't answer the question.  What do you mean by "wrong"?  What is the background documentation that shows that the proposed change makes the logic correct?

A proper explanation might be along the lines of "processor X reports values A, B, C from CPUID; the existing code follows paths P and Q that wrongly identify it as processor Y, but, as explained in document R (URL), CPUID values on AMD processors should be interpreted as way S, meaning that the logic should change in way T".  And with a bug filed in Bugzilla if this was user-visible (e.g. inappropriate function implementations being selected in glibc because of the wrong choices).

--
Joseph S. Myers
joseph@codesourcery.com

Attachment: 0001-BZ-19214-Fix-for-family-and-model-detection-for-AMD-.patch
Description: 0001-BZ-19214-Fix-for-family-and-model-detection-for-AMD-.patch


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