[PATCH 1/18] 2.6.17.9 perfmon2 patch for review: introduction

[PATCH 1/18] 2.6.17.9 perfmon2 patch for review: introduction

[Stephane Eranian]

Hello,

The following series of patches includes the generic perfmon2
subsystem and the support for i386, x86_64, and powerpc. The perfmon2
subsystem also works on MIPS and all Itanium processors. The Itanium support
is not posted because it does not easily accomodate the 100k message
limit of lkml. The powerpc support is still very preliminary.

The patches are relative to 2.6.17.9

I have already posted on the list about this subsystem. I am submitting 
today to get reviews and make progress towards getting the subsystem
merged into the mainline kernel.

The patches are split up between common and arch-specific. Each part
is further decomposed into new files and modified files. The generic
code is now split up by functionality to make reading easier.

For each new or modified files, I provide a detailed description of
the changes in each E-mail.

Thanks.

--
-Stephane

Re: [PATCH 1/18] 2.6.17.9 perfmon2 patch for review: introduction

[Christoph Hellwig]

On Wed, Aug 23, 2006 at 01:05:52AM -0700, Stephane Eranian wrote:
> Hello,
> 
> The following series of patches includes the generic perfmon2
> subsystem and the support for i386, x86_64, and powerpc. The perfmon2
> subsystem also works on MIPS and all Itanium processors. The Itanium support
> is not posted because it does not easily accomodate the 100k message
> limit of lkml. The powerpc support is still very preliminary.
> 
> The patches are relative to 2.6.17.9

pleas submit patches always against latest Linus' tree or -mm.  2.6.17.9
is already megabytes of diffs away from mainline.

Re: [PATCH 1/18] 2.6.17.9 perfmon2 patch for review: introduction

[Christoph Hellwig]

oh, and please give the patches useful subjects that descript the
patch, e.g. this one should be just:


    [PATCH 0/17] perfmon2: introduction

(yes, it's convention to number the introduction 0 and the actual patches
 1 to n)

Re: [PATCH 1/18] 2.6.17.9 perfmon2 patch for review: introduction

[Alexey Dobriyan]

On Wed, Aug 23, 2006 at 04:28:31PM +0100, Christoph Hellwig wrote:
> oh, and please give the patches useful subjects that descript the
> patch, e.g. this one should be just:
>
>
>     [PATCH 0/17] perfmon2: introduction
>
> (yes, it's convention to number the introduction 0 and the actual patches
>  1 to n)

Padding with zeros makes it even more useful:

	[PATCH 00/17]
	[PATCH 01/17]
		...
	[PATCH 17/17]

Re: [PATCH 1/18] 2.6.17.9 perfmon2 patch for review: introduction

[Christoph Hellwig]

On Wed, Aug 23, 2006 at 07:57:16PM +0400, Alexey Dobriyan wrote:
> On Wed, Aug 23, 2006 at 04:28:31PM +0100, Christoph Hellwig wrote:
> > oh, and please give the patches useful subjects that descript the
> > patch, e.g. this one should be just:
> >
> >
> >     [PATCH 0/17] perfmon2: introduction
> >
> > (yes, it's convention to number the introduction 0 and the actual patches
> >  1 to n)
> 
> Padding with zeros makes it even more useful:
> 
> 	[PATCH 00/17]
> 	[PATCH 01/17]
> 		...
> 	[PATCH 17/17]

To be honest I utterly hate that convention, and the double-padded
version [PATCH 001/17] some people use is even worse.

Re: [PATCH 1/18] 2.6.17.9 perfmon2 patch for review: introduction

[Andrew Morton]

On Wed, 23 Aug 2006 17:04:58 +0100
Christoph Hellwig <hch@infradead.org> wrote:

> > Padding with zeros makes it even more useful:
> > 
> > 	[PATCH 00/17]
> > 	[PATCH 01/17]
> > 		...
> > 	[PATCH 17/17]
> 
> To be honest I utterly hate that convention

It's so they'll correctly alphasort at the recipient's end.

I doubt if many MUAs do numeric sorting..

Re: [PATCH 1/18] 2.6.17.9 perfmon2 patch for review: introduction

[Christoph Hellwig]

On Wed, Aug 23, 2006 at 11:58:57AM -0700, Andrew Morton wrote:
> > > Padding with zeros makes it even more useful:
> > > 
> > > 	[PATCH 00/17]
> > > 	[PATCH 01/17]
> > > 		...
> > > 	[PATCH 17/17]
> > 
> > To be honest I utterly hate that convention
> 
> It's so they'll correctly alphasort at the recipient's end.

I suspect most mailers sort by date and not by subject anyway.

At least mine does :)

Re: [PATCH 1/18] 2.6.17.9 perfmon2 patch for review: introduction

[Stephane Eranian]

Andrew,
On Wed, Aug 23, 2006 at 11:58:57AM -0700, Andrew Morton wrote:
> On Wed, 23 Aug 2006 17:04:58 +0100
> Christoph Hellwig <hch@infradead.org> wrote:
> 
> > > Padding with zeros makes it even more useful:
> > > 
> > > 	[PATCH 00/17]
> > > 	[PATCH 01/17]
> > > 		...
> > > 	[PATCH 17/17]
> > 
> > To be honest I utterly hate that convention
> 
> It's so they'll correctly alphasort at the recipient's end.

That makes sense, I'll fix that in my next patch.

-- 

-Stephane

Re: [PATCH 1/18] 2.6.17.9 perfmon2 patch for review: introduction

[Andrew Morton]

On Wed, 23 Aug 2006 20:11:32 +0100
Christoph Hellwig <hch@infradead.org> wrote:

> On Wed, Aug 23, 2006 at 11:58:57AM -0700, Andrew Morton wrote:
> > > > Padding with zeros makes it even more useful:
> > > > 
> > > > 	[PATCH 00/17]
> > > > 	[PATCH 01/17]
> > > > 		...
> > > > 	[PATCH 17/17]
> > > 
> > > To be honest I utterly hate that convention
> > 
> > It's so they'll correctly alphasort at the recipient's end.
> 
> I suspect most mailers sort by date and not by subject anyway.
> 

Emails arrive out-of-order.

I used to have a `sleep 10' in my patch-bomb script to fix that, but half
an hour is too long to be sitting there hoping the DSL stays up, the kernel
stays up and that the familial hair-drier+iron+air-conditioner conspiracy
doesn't strike.  

Re: [PATCH 1/18] 2.6.17.9 perfmon2 patch for review: introduction

[Junio C Hamano]

Andrew Morton <akpm@osdl.org> writes:

> On Wed, 23 Aug 2006 17:04:58 +0100
> Christoph Hellwig <hch@infradead.org> wrote:
>
>> > Padding with zeros makes it even more useful:
>> > 
>> > 	[PATCH 00/17]
>> > 	[PATCH 01/17]
>> > 		...
>> > 	[PATCH 17/17]
>> 
>> To be honest I utterly hate that convention
>
> It's so they'll correctly alphasort at the recipient's end.
>
> I doubt if many MUAs do numeric sorting..

I wonder if 'git-format-patch --numbered' should be updated to
do the zero padding.  Right now we don't.

It should be a trivial patch to do if somebody is so inclined
(it is around ll.133 in log-tree.c).