Polyhedron Benchmarks 2004

From: John Appleyard (spamtrap_at_polyhedron.com)
Date: 05/10/04 

Date: Mon, 10 May 2004 20:44:17 +0100

I have just revised the compiler comparison charts at www.polyhedron.com.

The execution time benchmarks have been substantially revised, with some
old ones removed, others revised, and some new ones added. In addition,
I've included the harness which we use to run the benchmarks, and a
validator for checking the correctness of the output files. We do now
encourage others to download and use the benchmarks, and to publish the
results, providing that this is done without distortion, and with due
acknowledgement to Polyhedron.

The results include the latest versions of compilers from Absoft, Intel,
Lahey, NAG, PGI and Salford, with all of these vendors showing some
significant improvements, paticularly on F90 code. All run on a Dell
Dimension 4600 with a 3.06GHz Pentium 4 and 768 MBytes RAM, dual booting
Windows XP Pro and Red Hat 9.0. We plan to publish results for these
benchmarks on a 64 bit AMD machine later this year.

The Linux compilers are

Absoft Pro Fortran 9.0
G77 3.2.2 (20030222)
Intel Fortran Compiler 8.0.046
Lahey/Fujitsu LF95 Linux 6.20a
NAG f95 5.0 (286)
PGI The Portland Group Compiler 5.1.6

On Windows, we tested

Absoft Pro Fortran 95 8.2
Compaq Visual Fortran 6.6C
Salford FTN77 4.02
Salford FTN95 4.5.0
Salford FTN95 (.NET mode) 4.5.0
G77 (Cygwin) 3.3.1
Intel Fortran Compiler 8.0.048
Lahey/Fujitsu LF95 7.10
Open Watcom 1.0

We've also updated the Diagnostics comparison charts, and were delighted
to note that a third vendor (NAG) now implements run-time checks for
uninitialised variables.

The language charts have been updated to show which compilers support
TR15581 (allocatable extensions) and OpenMP. 

-- 
John Appleyard  - (send email to john!news@.. rather than spamtrap@..)
Polyhedron Software
Programs for Programmers - QA, Compilers, Graphics, Consultancy
********* Visit our Web site on http://www.polyhedron.co.uk/ *********

Relevant Pages

  • Re: Polyhedron Benchmarks 2004
    ... > I have just revised the compiler comparison charts at www.polyhedron.com. ... > The execution time benchmarks have been substantially revised, ... > Windows XP Pro and Red Hat 9.0. ...
    (comp.lang.fortran)
  • Re: Yet Another Lisp Performance Thread
    ... commercial compilers is using them to compile a competitive compiler. ... As far as benchmarks go, there are lies, damn lies, statistics, and ... My original post was actually intended to show that Lisp can be ... work the JPEG encoder, the Lisp program can pick up even more speed. ...
    (comp.lang.lisp)
  • Re: Is There Any Reason to Even Use VC++ Anymore?
    ... I never did any benchmarks, so I was basing my statement on other reports. ... Sure the C++ compiler can ... > spend some extra time to optimize MSIL, while the C# compiler is more time ... >>> You could almost think of the syntax extensions almost independent of ...
    (microsoft.public.dotnet.languages.vc)
  • Re: Downgrading to D7
    ... There are some benchmarks on their site: ... the actual performance (call me prejudiced, but Java developpers are ... While they used Java to write the compiler, Vector Pascal is a native code ... This is why building those features in a general purpose compiler makes ...
    (borland.public.delphi.non-technical)
  • CVF, Absoft bug involving ELEMENTAL character functions
    ... with Absoft Pro Fortran 8.2. ... Compaq Visual Fortran Optimizing Compiler Version 6.6 ... Compiler internal error - please submit problem ... Absoft Pro FORTRAN ...
    (comp.lang.fortran)