Linux PARISC architecture development
 help / color / mirror / Atom feed
* [parisc-linux] [mingo@elte.hu: [patch] spinlock consolidation, v2]
@ 2005-06-03 17:31 Matthew Wilcox
  2005-06-04 11:55 ` Joel Soete
  2005-06-06  6:05 ` [parisc-linux] Re: [patch] spinlock consolidation, v2 Grant Grundler
  0 siblings, 2 replies; 16+ messages in thread
From: Matthew Wilcox @ 2005-06-03 17:31 UTC (permalink / raw)
  To: parisc-linux


Anyone have time to test this out?

----- Forwarded message from Ingo Molnar <mingo@elte.hu> -----

Date:	Fri, 3 Jun 2005 17:40:29 +0200
From:	Ingo Molnar <mingo@elte.hu>
Subject: [patch] spinlock consolidation, v2

the latest version of the spinlock consolidation patch can be found at:

  http://redhat.com/~mingo/spinlock-patches/consolidate-spinlocks.patch

the patch is now complete in the sense that it does everything i wanted 
it to do. If you have any other suggestions (or i have missed to 
incorporate an earlier suggestion of yours), please yell.

Changes:

 - all architectures have been converted to the new spinlock code.
   arm, i386, ia64, ppc, ppc64, s390/s390x, x64 was build-tested via
   crosscompilers. Alpha, m32r, mips, parisc, sh, sparc, sparc64 has
   not been tested yet, but should be mostly fine. x86 and x64 was
   boot-tested in all relevant .config variations. It all brought a nice 
   reduction in source code size:

     62 files changed, 1455 insertions(+), 3025 deletions(-)

   Al, would you be interested in checking this patch on your build 
   farm? It should build on all architectures, UP and SMP alike.

   (NOTE: i've switched sparc32, sparc64, alpha, ppc, parisc to use the 
    generic spinlock debugging code. I believe the generic debugging 
    code is now capable enough to be a replacement - but especially the 
    Sparc ones are pretty advanced; so if i've missed some important
    feature please let me know and i'll implement it in the generic 
    code.)

 - linux/spinlock_types.h: new, pure header file that can be used
   by other headers to define spinlock fields - without having to
   pull in all the other include files that are needed on the
   implementational side. (Roman Zippel)

 - lib/spinlock_debug.c: got rid of the __FILE__/__LINE__ debug output 
   (suggested by Ingo Oeser), and streamlined the debug output. 
   Implemented 'lockup detection' which is a must for architectures that 
   dont have the equivalent of an NMI watchdog. (but is useful on other 
   architectures as well.) Both spinlocks and rwlocks are now fully 
   debugged.

 - linux/spinlock.h: got rid of the ATOMIC_DEC_AND_LOCK cruft. This is 
   achieved by not doing the UP-nonpreempt-nondebug specific 
   optimization but letting it pick the generic _atomic_dec_and_lock 
   function. The assembly looks sane on x86 (no locked ops, etc.) so 
   there's no performance problem. Other architectures should work fine 
   too, those which implement _atomic_dec_and_lock unconditionally might 
   want to review whether they want to use the CONFIG_HAVE_DEC_LOCK 
   mechanism to get the optimized (generic) version of the function on 
   UP.

 - asm-generic/spinlock_types_up.h: further simplifications (suggested
   by Arjan van de Ven), typo fixed

 - linux/spinlock_up.h: since this an UP-nondebug branch now, the macros 
   were simplified and streamlined significantly.

 - asm-generic/spinlock_up.h: further cleanups, reordering of op
   definitions into 'natural' op order.

 - asm-i386/spinlock.h and asm-x86_64/spinlock.h: reordering of ops, 
   cleanups

 - include/linux/spinlock.h: more cleanups, reordering

 - linux/spinlock_smp.h: cleanups, reordering of prototypes

 - kernel/spinlock.c: fixed bug in generic_raw_read_trylock and renamed 
   it to generic__raw_read_trylock to ease conversion.

 - lib/kernel_lock.c: simplification

 - (lots of small details i forgot)

	Ingo

----- End forwarded message -----

-- 
"Next the statesmen will invent cheap lies, putting the blame upon 
the nation that is attacked, and every man will be glad of those
conscience-soothing falsities, and will diligently study them, and refuse
to examine any refutations of them; and thus he will by and by convince 
himself that the war is just, and will thank God for the better sleep 
he enjoys after this process of grotesque self-deception." -- Mark Twain
_______________________________________________
parisc-linux mailing list
parisc-linux@lists.parisc-linux.org
http://lists.parisc-linux.org/mailman/listinfo/parisc-linux

^ permalink raw reply	[flat|nested] 16+ messages in thread
end of thread, other threads:[~2005-06-21 11:22 UTC | newest]

Thread overview: 16+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2005-06-03 17:31 [parisc-linux] [mingo@elte.hu: [patch] spinlock consolidation, v2] Matthew Wilcox
2005-06-04 11:55 ` Joel Soete
2005-06-04 18:03   ` Grant Grundler
2005-06-05 19:13     ` Joel Soete
2005-06-06  6:05 ` [parisc-linux] Re: [patch] spinlock consolidation, v2 Grant Grundler
2005-06-06  7:36   ` Ingo Molnar
2005-06-06 15:12     ` Grant Grundler
2005-06-06 16:31       ` Ingo Molnar
2005-06-06 17:55         ` Grant Grundler
     [not found]     ` <20050606175029.GC24437@colo.lackof.org>
2005-06-12  6:49       ` Ingo Molnar
2005-06-12  7:25         ` Grant Grundler
2005-06-12  7:34           ` Ingo Molnar
2005-06-13  6:29             ` Grant Grundler
2005-06-13  7:44               ` Ingo Molnar
2005-06-13 18:39                 ` Grant Grundler
2005-06-21 11:22                   ` Ingo Molnar

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox