Skip site navigation (1)Skip section navigation (2)
Date:      Mon, 09 Feb 2015 18:08:02 -0500
From:      John Baldwin <jhb@freebsd.org>
To:        Jung-uk Kim <jkim@freebsd.org>
Cc:        svn-src-head@freebsd.org, svn-src-all@freebsd.org, src-committers@freebsd.org
Subject:   Re: svn commit: r278474 - head/sys/sys
Message-ID:  <5703806.fFzzT9N1y2@ralph.baldwin.cx>
In-Reply-To: <54D93461.8010109@FreeBSD.org>
References:  <201502092103.t19L3OAn013792@svn.freebsd.org> <3501684.yzo0NMkRZd@ralph.baldwin.cx> <54D93461.8010109@FreeBSD.org>

next in thread | previous in thread | raw e-mail | index | archive | help
On Monday, February 09, 2015 05:27:45 PM Jung-uk Kim wrote:
> On 02/09/2015 17:12, John Baldwin wrote:
> > On Monday, February 09, 2015 04:55:52 PM Jung-uk Kim wrote:
> >> On 02/09/2015 16:08, John Baldwin wrote:
> >>> On Monday, February 09, 2015 09:03:24 PM John Baldwin wrote:
> >>>> Author: jhb Date: Mon Feb  9 21:03:23 2015 New Revision:
> >>>> 278474 URL: https://svnweb.freebsd.org/changeset/base/278474
> >>>> 
> >>>> Log: Use __builtin_popcnt() to implement a BIT_COUNT()
> >>>> operation for bitsets and use this to implement CPU_COUNT()
> >>>> to count the number of CPUs in a cpuset.
> >>>> 
> >>>> MFC after:	2 weeks
> >>> 
> >>> Yes, __builtin_popcnt() works with GCC 4.2.  It should also
> >>> allow the compiler to DTRT in userland uses of this if -msse4.2
> >>> is enabled.
> >> 
> >> Back in 2012, when I submitted a similar patch, bde noted
> >> __builtin_popcount*() cannot be used with GCC 4.2 for *kernel*
> >> because it emits a library call.
> >> 
> >> http://docs.freebsd.org/cgi/mid.cgi?20121116171923.L1135
> >> 
> >> FYI...
> > 
> > Weird, I though I built a kernel with this in a tree that uses it
> > in the igb(4) driver.  We need a CPU_COUNT() no matter what, but if
> > this emits a library call under GCC I will need to add the call.
> > We could also adopt your bitcount header, though I think it is more
> > consistent to keep the loop in BIT_COUNT() and use something that
> > emulates popcountl() rather than directly using bitcount() in
> > BIT_COUNT() (primarily because the rest of sys/bitset.h is
> > structured that way: explicit loops in sys/bitset.h itself).
> 
> I think you should back it out for now and move the discussion to arch
> or hackers.  I gave it up at the time but you may have better luck. :-)
> 
> FYI, the following was the last version of my patch at the time.
> 
> https://people.freebsd.org/~jkim/bitcount5.diff

I could also just make it userland only for now?  Rui wants to use it in
userland.  However, I can back it out if that is preferred.  To be honest,
I'm not sure how valuable it is at this point to expend a lot of effort to
support GCC older than 3.4 (i.e. the non-builtin popcount approach).

-- 
John Baldwin



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?5703806.fFzzT9N1y2>