[LLVMbugs] [Bug 10889] New: -fplugin-arg-dragonegg-enable-gcc-optzns/vector-select performance regression in capacita

[bugzilla-daemon at llvm.org]

http://llvm.org/bugs/show_bug.cgi?id=10889

           Summary: -fplugin-arg-dragonegg-enable-gcc-optzns/vector-select
                    performance regression in capacita
           Product: dragonegg
           Version: trunk
          Platform: Macintosh
        OS/Version: MacOS X
            Status: NEW
          Severity: normal
          Priority: P
         Component: New Bugs
        AssignedTo: baldrick at free.fr
        ReportedBy: howarth at nitro.med.uc.edu
                CC: llvmbugs at cs.uiuc.edu


At r139290, llvm/dragonegg svn shows a major performance regression in the
run-time of the Polyhedron 2005 capacita benchmark compiled with...

/sw/lib/gcc4.6/bin/gfortran -fplugin=/sw/lib/gcc4.6/lib/dragonegg.so -msse4
-ffast-math -funroll-loops -O3 -fplugin-arg-dragonegg-enable-gcc-optzns 
capacita.f90 -o capacita

   Benchmark   Compile  Executable   Ave Run  Number   Estim
        Name    (secs)     (bytes)    (secs) Repeats   Err %
   ---------   -------  ----------   ------- -------  ------
    capacita      0.75       45408    398.86       1100.0000


vs when compiled without -fplugin-arg-dragonegg-enable-gcc-optzns...

   Benchmark   Compile  Executable   Ave Run  Number   Estim
        Name    (secs)     (bytes)    (secs) Repeats   Err %
   ---------   -------  ----------   ------- -------  ------
    capacita      0.49       41416     44.98       1100.0000

Note these are from quick.par runs as the benchmark now runs so slowly. With
the new vector-select code enabled using
-fplugin-arg-dragonegg-enable-gcc-optzns, the  capacita benchmark shows a ~9
fold regression in execution time.

-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.