[Libreoffice-commits] core.git: sc/source

[Takeshi Abe]

 sc/source/core/tool/interpr3.cxx |    3 +--
 1 file changed, 1 insertion(+), 2 deletions(-)

New commits:
commit d155f7af035e0fd818340c1affe41d3b6e97b502
Author: Takeshi Abe <tabe at fixedpoint.jp>
Date:   Wed Jun 6 19:05:01 2018 +0900

    sc: A putative micro optimization of MEDIAN() for even-length arrays
    
    As the previous run of std::nth_element() already collects all of
    smaller elements than the upper median into the first half of the
    vector (which may also contain an equal one to the upper median,
    that's OK), we can safely ignore the later half for finding the lower
    median.
    
    Change-Id: I64c97ee03b02ea363399077244dd99d0d10defcb
    Reviewed-on: https://gerrit.libreoffice.org/55370
    Tested-by: Jenkins <ci at libreoffice.org>
    Reviewed-by: Eike Rathke <erack at redhat.com>

diff --git a/sc/source/core/tool/interpr3.cxx b/sc/source/core/tool/interpr3.cxx
index b52a62fd3d20..a43ec7835193 100644
--- a/sc/source/core/tool/interpr3.cxx
+++ b/sc/source/core/tool/interpr3.cxx
@@ -3374,8 +3374,7 @@ double ScInterpreter::GetMedian( vector<double> & rArray )
     {
         double fUp = *iMid;
         // Lower median.
-        iMid = rArray.begin() + nMid - 1;
-        ::std::nth_element( rArray.begin(), iMid, rArray.end());
+        iMid = ::std::max_element( rArray.begin(), rArray.begin() + nMid);
         return (fUp + *iMid) / 2;
     }
 }