Re: rwsem benchmark [was Re: [PATCH] rw_semaphores, optimisations try #3]

[Andrea Arcangeli <andrea@suse.de>]

[-- Attachment #1: Type: text/plain, Size: 2397 bytes --]

On Tue, Apr 24, 2001 at 02:19:28PM +0200, Andrea Arcangeli wrote:
> I'm starting the benchmarks of the C version and I will post a number update
> and a new patch in a few minutes.

(sorry for the below wrap around, just grow your terminal to read it stright)

				aa RW (reads) aa RW (writes) R1	      R2      RO      W1       WO
dh generic out of line try3	5842496	      3016649        13309316 5010534 3850228 13012701 1825789
dh generic out of line try3 #2	5823381	      3006773	     13311722 5023185 3845954 13021716 1802560
aa generic out of line buggy	6061713	      3129801	     14251500 4972932 4253814 13652385 1751857
aa generic out of line #2 buggy	6099046	      3148951	     14265389 4936267 4253432 13632914 1753608
aa generic out of line	        6133756       3167138	     14244991 5122637 4254504 13656896 1797627
aa generic out of line #2	6093079	      3145761	     14259141 5126506 4254532 13658447 1803505

dh x86 asm in line try3 	5789542       2989478	     16922653 5650211 4956250 15431139 813756
dh x86 asm in line try3 #2	5801777	      2995669	     16946132 5647272 4959828 15439790 816005
aa x86 asm in line buggy	5736978	      2962325	     17044842 5603085 4831655 16064773 860791
aa x86 asm in line #2 buggy	5799163	      2994404	     17053405 5601647 4833518 16037018 864103
aa generic in line	        5706875       2946931	     16943038 5644018 4837576 16085859 870833
aa generic in line #2	        5755126       2971578	     16924502 5639379 4836111 16073916 873499

I tagged my previous rows as "buggy", I left your try#3 at the start of each
version and I added at the end the new numbers with the -9 fixed revision of my
rwsem at the end. new graph is attached.

So nothing interesting is changed in the numbers as far I can tell after the
fixes and improvement of the fast path using "a" instead of "+a".

Unless you can provide a testcase that fails with my smarter and more compact
algorithm I suggest to Linus to merge my code into pre7.

Against pre6:

	ftp://ftp.kernel.org/pub/linux/kernel/people/andrea/patches/v2.4/2.4.4pre6/rwsem-9

against David's try2:

	ftp://ftp.kernel.org/pub/linux/kernel/people/andrea/patches/v2.4/2.4.4pre6/rwsem-9-against-dh-try2

against David's try3:

	ftp://ftp.kernel.org/pub/linux/kernel/people/andrea/patches/v2.4/2.4.4pre6/rwsem-9-against-dh-try3

I will keep doing regression testing in the next hours of course.

Andrea

[-- Attachment #2: rwsem-9.png --]
[-- Type: image/png, Size: 12674 bytes --]