FIXED_486_STRING ?

FIXED_486_STRING ?

[Denis Zaitsev]

What is the state of the FIXED_486_STRING macro?  It is used once thru
all the kernel tree - in include/asm-i386/string.h - and it seems that
its role is to disable a usage of string-486.h completely...  Am I
right?

Re: FIXED_486_STRING ?

[Keith Owens]

On Sat, 13 Apr 2002 22:47:43 +0600, 
Denis Zaitsev <zzz@cd-club.ru> wrote:
>What is the state of the FIXED_486_STRING macro?  It is used once thru
>all the kernel tree - in include/asm-i386/string.h - and it seems that
>its role is to disable a usage of string-486.h completely...  Am I
>right?

Dead code, it has been dead since at least 2.0.21.  Unless somebody
wants to fix the string-486 code, delete FIXED_486_STRING,
CONFIG_X86_USE_STRING_486 and include/asm-i386/string-486.h.

Re: FIXED_486_STRING ?

[Dave Jones]

On Sun, Apr 14, 2002 at 10:27:18AM +1000, Keith Owens wrote:

 > Dead code, it has been dead since at least 2.0.21.  Unless somebody
 > wants to fix the string-486 code, delete FIXED_486_STRING,
 > CONFIG_X86_USE_STRING_486 and include/asm-i386/string-486.h.

I proposed doing this a few months back, then someone stepped forward
who had worked on these routines recently and fixed up whatever problems
they originally exhibited.  Deleting the dead code from 2.2 / 2.4 probably
makes sense, but it'd be nice to have the 2.5 ones fixed up.

    Dave.

-- 
| Dave Jones.        http://www.codemonkey.org.uk
| SuSE Labs

Re: FIXED_486_STRING ?

[Denis Zaitsev]

On Sun, Apr 14, 2002 at 02:44:06AM +0200, Dave Jones wrote:
> On Sun, Apr 14, 2002 at 10:27:18AM +1000, Keith Owens wrote:
> 
>  > Dead code, it has been dead since at least 2.0.21.  Unless somebody
>  > wants to fix the string-486 code, delete FIXED_486_STRING,
>  > CONFIG_X86_USE_STRING_486 and include/asm-i386/string-486.h.
> 
> I proposed doing this a few months back, then someone stepped forward
> who had worked on these routines recently and fixed up whatever problems
> they originally exhibited.  Deleting the dead code from 2.2 / 2.4 probably
> makes sense, but it'd be nice to have the 2.5 ones fixed up.
> 
>     Dave.
> 
Yes, the special code for Pentium is important for it, as that code is
faster and shorter.  The last signature in string-486.h is of
2000/05/09, so, if this is true, the problems have not been fixed.
So, what is the real situation?  I would like to fix the code, and it
seems that I'm not alone :)

Re: FIXED_486_STRING ?

[Dave Jones]

On Mon, Apr 15, 2002 at 06:38:26AM +0600, Denis Zaitsev wrote:
 > Yes, the special code for Pentium is important for it, as that code is
 > faster and shorter.  The last signature in string-486.h is of
 > 2000/05/09, so, if this is true, the problems have not been fixed.
 > So, what is the real situation?  I would like to fix the code, and it
 > seems that I'm not alone :)

Petko Manolov <lz5mj@yahoo.com> did some work on them circa 2.4.0test.
His patch is at http://www.dce.bg/~petkan/linux/string-486.diff
Aparently there are still 1-2 problems with these routines which is
why he hasn't pushed for inclusion I guess.

-- 
| Dave Jones.        http://www.codemonkey.org.uk
| SuSE Labs

Re: FIXED_486_STRING ?

[Denis Zaitsev]

On Mon, Apr 15, 2002 at 03:03:55AM +0200, Dave Jones wrote:
> Petko Manolov <lz5mj@yahoo.com> did some work on them circa 2.4.0test.
> His patch is at http://www.dce.bg/~petkan/linux/string-486.diff
> Aparently there are still 1-2 problems with these routines which is
> why he hasn't pushed for inclusion I guess.
> 
These patches are included...  But the string-486.h itself is turned
off by FIXED_486_STRING.  BTW, what are the problems?

Re: FIXED_486_STRING ?

[Dave Jones]

On Mon, Apr 15, 2002 at 07:25:56AM +0600, Denis Zaitsev wrote:
 > On Mon, Apr 15, 2002 at 03:03:55AM +0200, Dave Jones wrote:
 > > Petko Manolov <lz5mj@yahoo.com> did some work on them circa 2.4.0test.
 > > His patch is at http://www.dce.bg/~petkan/linux/string-486.diff
 > These patches are included... 

Look again.

(davej@noodles:linux-2.4.19-pre6)$ cat ../string-486.diff | patch -p1 -F1 --dry-run
patching file include/asm-i386/string-486.h
Hunk #13 FAILED at 365.
Hunk #14 succeeded at 388 (offset -5 lines).
Hunk #15 succeeded at 409 (offset -5 lines).
Hunk #16 succeeded at 452 (offset -5 lines).
Hunk #17 succeeded at 516 (offset -5 lines).
1 out of 17 hunks FAILED -- saving rejects to file include/asm-i386/string-486.h.rej 

Almost still applies except for one hunk.
 
 > But the string-486.h itself is turned
 > off by FIXED_486_STRING.  BTW, what are the problems?

Not sure off-hand. I would hazard a guess that they copied
too little/too much, but Petko would be a better person
to ask.

-- 
| Dave Jones.        http://www.codemonkey.org.uk
| SuSE Labs

Re: FIXED_486_STRING ?

[Denis Zaitsev]

On Mon, Apr 15, 2002 at 03:36:59AM +0200, Dave Jones wrote:
> On Mon, Apr 15, 2002 at 07:25:56AM +0600, Denis Zaitsev wrote:
>  > On Mon, Apr 15, 2002 at 03:03:55AM +0200, Dave Jones wrote:
>  > > Petko Manolov <lz5mj@yahoo.com> did some work on them circa 2.4.0test.
>  > > His patch is at http://www.dce.bg/~petkan/linux/string-486.diff
>  > These patches are included... 
> 
> Look again.
> 
> (davej@noodles:linux-2.4.19-pre6)$ cat ../string-486.diff | patch -p1 -F1 --dry-run
> patching file include/asm-i386/string-486.h
> Hunk #13 FAILED at 365.
> Hunk #14 succeeded at 388 (offset -5 lines).
> Hunk #15 succeeded at 409 (offset -5 lines).
> Hunk #16 succeeded at 452 (offset -5 lines).
> Hunk #17 succeeded at 516 (offset -5 lines).
> 1 out of 17 hunks FAILED -- saving rejects to file include/asm-i386/string-486.h.rej 
> 
> Almost still applies except for one hunk.

Yes, indeed.  I've missed the fact that it is the latest patch.  Thanks.

>  
>  > But the string-486.h itself is turned
>  > off by FIXED_486_STRING.  BTW, what are the problems?
> 
> Not sure off-hand. I would hazard a guess that they copied
> too little/too much, but Petko would be a better person
> to ask.
> 

I've already sent him a letter...