Message-ID: <a610aeb2-b1cb-0d57-8078-97d1051481f9@redhat.com>
Date:   Sun, 10 Feb 2019 21:08:40 -0500
From:   Waiman Long <longman@...hat.com>
To:     Peter Zijlstra <peterz@...radead.org>,
        Ingo Molnar <mingo@...hat.com>,
        Will Deacon <will.deacon@....com>,
        Thomas Gleixner <tglx@...utronix.de>
Cc:     linux-kernel@...r.kernel.org, linux-alpha@...r.kernel.org,
        linux-arm-kernel@...ts.infradead.org,
        linux-hexagon@...r.kernel.org, linux-ia64@...r.kernel.org,
        linuxppc-dev@...ts.ozlabs.org, linux-sh@...r.kernel.org,
        sparclinux@...r.kernel.org, linux-xtensa@...ux-xtensa.org,
        linux-arch@...r.kernel.org, x86@...nel.org,
        Arnd Bergmann <arnd@...db.de>, Borislav Petkov <bp@...en8.de>,
        "H. Peter Anvin" <hpa@...or.com>,
        Davidlohr Bueso <dave@...olabs.net>,
        Linus Torvalds <torvalds@...ux-foundation.org>,
        Andrew Morton <akpm@...ux-foundation.org>,
        Tim Chen <tim.c.chen@...ux.intel.com>
Subject: Re: [PATCH] locking/rwsem: Remove arch specific rwsem files

On 02/10/2019 09:00 PM, Waiman Long wrote:
> As the generic rwsem-xadd code is using the appropriate acquire and
> release versions of the atomic operations, the arch specific rwsem.h
> files will not be that much faster than the generic code as long as the
> atomic functions are properly implemented. So we can remove those arch
> specific rwsem.h and stop building asm/rwsem.h to reduce maintenance
> effort.
>
> Currently, only x86, alpha and ia64 have implemented architecture
> specific fast paths. I don't have access to alpha and ia64 systems for
> testing, but they are legacy systems that are not likely to be updated
> to the latest kernel anyway.
>
> By using a rwsem microbenchmark, the total locking rates on a 4-socket
> 56-core 112-thread x86-64 system before and after the patch were as
> follows (mixed means equal # of read and write locks):
>
>                       Before Patch              After Patch
>    # of Threads  wlock   rlock   mixed     wlock   rlock   mixed
>    ------------  -----   -----   -----     -----   -----   -----
>         1        27,373  29,409  28,170    28,773  30,164  29,276
>         2         7,697  14,922   1,703     7,435  15,167   1,729
>         4         6,987  14,285   1,490     7,181  14,438   1,330
>         8         6,650  13,652     761     6,918  13,796     718
>        16         6,434  15,729     713     6,554  16,030     625
>        32         5,590  15,312     552     6,124  15,344     471
>        64         5,980  15,478      61     5,668  15,509      58
>
> There were some run-to-run variations for the multi-thread tests. For
> x86-64, using the generic C code fast path seems to be a liitle bit
> faster than the assembly version especially for read-lock and when lock
> contention is low.  Looking at the assembly version of the fast paths,
> there are assembly to/from C code wrappers that save and restore all
> the callee-clobbered registers (7 registers on x86-64). The assembly
> generated from the generic C code doesn't need to do that. That may
> explain the slight performance gain here.
>
> The generic asm rwsem.h can also be merged into kernel/locking/rwsem.h
> as no other code other than those under kernel/locking needs to access
> the internal rwsem macros and functions.
>
> Signed-off-by: Waiman Long <longman@...hat.com>

I have decided to break the rwsem patchset that I sent out on last
Thursday into 3 parts. This patch is part 0 as it touches a number of
arch specific files and so have the widest distribution. I would like to
get it merged first. Part 1 will be patches 1-10 (except 4) of my
original rwsem patchset. This part moves things around, adds more
debugging capability and lays the ground work for the next part. Part 2
will contains the remaining patches which are the real beef of the whole
patchset.

Cheers,
Longman

Powered by blists - more mailing lists