[Libreoffice-commits] core.git: Branch 'libreoffice-4-4' - vcl/generic

[Michael Stahl]

 vcl/generic/glyphs/gcach_layout.cxx |    2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

New commits:
commit 8a58579a35bb275c985ff0fd412c92f10cd8a085
Author: Michael Stahl <mstahl at redhat.com>
Date:   Tue Feb 24 21:32:54 2015 +0100

    tdf#89666: vcl: speed up HbLayoutEngine line layout for large paragraphs
    
    When formatting a 180k char Writer paragraph, most of the time is spent
    in vcl::ScriptRun::next(), which is called twice per line from
    SwTxtGuess::Guess(), once via GetTxtBreak() and once via GetTxtSize().
    
    In the second call, from GetTxtSize(), the end position of the line is
    known, and passed to vcl, and iterating beyond that position seems
    pointless.
    
    This reduces vcl::ScriptRun::next() from 24 to 11 billion callgrind
    cycles when built with GCC 4.9.2 -m32 -Os.
    
    Change-Id: Ia23fcccaf5ef9c9ecdcb54bfc8f0f8a043c8711e
    (cherry picked from commit 7fde44c85620f8079bc4863fe3f7ea1f69a0f88c)
    Reviewed-on: https://gerrit.libreoffice.org/14646
    Reviewed-by: Caolán McNamara <caolanm at redhat.com>
    Tested-by: Caolán McNamara <caolanm at redhat.com>

diff --git a/vcl/generic/glyphs/gcach_layout.cxx b/vcl/generic/glyphs/gcach_layout.cxx
index abd04ba..f7523a1 100644
--- a/vcl/generic/glyphs/gcach_layout.cxx
+++ b/vcl/generic/glyphs/gcach_layout.cxx
@@ -372,7 +372,7 @@ bool HbLayoutEngine::layout(ServerFontLayout& rLayout, ImplLayoutArgs& rArgs)
 
     rLayout.Reserve(nGlyphCapacity);
 
-    vcl::ScriptRun aScriptRun(reinterpret_cast<const UChar *>(rArgs.mpStr), rArgs.mnLength);
+    vcl::ScriptRun aScriptRun(reinterpret_cast<const UChar *>(rArgs.mpStr), rArgs.mnEndCharPos);
 
     Point aCurrPos(0, 0);
     while (true)