[libcxx] r275114 - Don't compute modulus of hash if it is smaller than the bucket count.

Duncan P. N. Exon Smith via cfe-commits cfe-commits at lists.llvm.org
Mon Jul 18 10:43:21 PDT 2016


> On 2016-Jul-17, at 13:11, Eric Fiselier <eric at efcs.ca> wrote:
> 
> Hi Duncan,
> 
> It's possibly expected. It depends on what operation it's performing. I expected a bit of a performance drop in some cases but I have a plan to fix those.
> Do you have a link to LNT?

LNT is documented here:
  http://llvm.org/docs/TestingGuide.html
  http://llvm.org/docs/lnt/

We have a public bot that submits results here:
  http://llvm.org/perf/

However I think Chris tracked this down using some regression bisection tools on our internal instance.  I'm looking now to see if the public instance caught a regression on -O3 -flto:
  http://llvm.org/perf/db_default/v4/nts/machine/42

> 
> On Wed, Jul 13, 2016 at 6:41 PM, Duncan P. N. Exon Smith <dexonsmith at apple.com> wrote:
> Hmm.  I implied there were other regressions, but I just finished scanning them.  Shootout-C++/hash2 is the only major one.  The others were small, and only at -O0.
> 
> > On 2016-Jul-13, at 17:38, Duncan P. N. Exon Smith via cfe-commits <cfe-commits at lists.llvm.org> wrote:
> >
> > We saw mixed results from this on LNT, including some major regressions.  For example, on x86_64, SingleSource/Benchmarks/Shootout-C++/hash2 regressed 18.5% at -O3 and over 20% at -Os.
> >
> > Is this expected?
> 
> ^ Still interested in an answer, though ;).
> 
> >
> >> On 2016-Jul-11, at 15:02, Eric Fiselier via cfe-commits <cfe-commits at lists.llvm.org> wrote:
> >>
> >> Author: ericwf
> >> Date: Mon Jul 11 17:02:02 2016
> >> New Revision: 275114
> >>
> >> URL: http://llvm.org/viewvc/llvm-project?rev=275114&view=rev
> >> Log:
> >> Don't compute modulus of hash if it is smaller than the bucket count.
> >>
> >> This cleans up a previous optimization attempt in hash, and results in
> >> additional performance improvements over that previous attempt. Additionally
> >> this new optimization does not hinder the power of 2 bucket count optimization.
> >>
> >> Modified:
> >>   libcxx/trunk/include/__hash_table
> >>
> >> Modified: libcxx/trunk/include/__hash_table
> >> URL: http://llvm.org/viewvc/llvm-project/libcxx/trunk/include/__hash_table?rev=275114&r1=275113&r2=275114&view=diff
> >> ==============================================================================
> >> --- libcxx/trunk/include/__hash_table (original)
> >> +++ libcxx/trunk/include/__hash_table Mon Jul 11 17:02:02 2016
> >> @@ -90,7 +90,8 @@ inline _LIBCPP_INLINE_VISIBILITY
> >> size_t
> >> __constrain_hash(size_t __h, size_t __bc)
> >> {
> >> -    return !(__bc & (__bc - 1)) ? __h & (__bc - 1) : __h % __bc;
> >> +    return !(__bc & (__bc - 1)) ? __h & (__bc - 1) :
> >> +        (__h < __bc ? __h : __h % __bc);
> >> }
> >>
> >> inline _LIBCPP_INLINE_VISIBILITY
> >> @@ -2201,8 +2202,7 @@ __hash_table<_Tp, _Hash, _Equal, _Alloc>
> >>        if (__nd != nullptr)
> >>        {
> >>            for (__nd = __nd->__next_; __nd != nullptr &&
> >> -                (__hash == __nd->__hash_
> >> -                    || __constrain_hash(__nd->__hash_, __bc) == __chash);
> >> +                __constrain_hash(__nd->__hash_, __bc) == __chash;
> >>                                                           __nd = __nd->__next_)
> >>            {
> >>                if ((__nd->__hash_ == __hash) && key_eq()(__nd->__value_, __k))
> >> @@ -2231,8 +2231,7 @@ __hash_table<_Tp, _Hash, _Equal, _Alloc>
> >>        if (__nd != nullptr)
> >>        {
> >>            for (__nd = __nd->__next_; __nd != nullptr &&
> >> -                (__hash == __nd->__hash_
> >> -                    || __constrain_hash(__nd->__hash_, __bc) == __chash);
> >> +                  __constrain_hash(__nd->__hash_, __bc) == __chash;
> >>                                                           __nd = __nd->__next_)
> >>            {
> >>                if ((__nd->__hash_ == __hash) && key_eq()(__nd->__value_, __k))
> >>
> >>
> >> _______________________________________________
> >> cfe-commits mailing list
> >> cfe-commits at lists.llvm.org
> >> http://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
> >
> > _______________________________________________
> > cfe-commits mailing list
> > cfe-commits at lists.llvm.org
> > http://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
> 
> 



More information about the cfe-commits mailing list