Here are the results of timing the Csound command rendering examples/xanadu.csd:

COMPILER                              VERSION   RENDERING TIME (seconds)   BENCHMARK

Intel C++ Compiler                    10.1      25.824                      76.2
Microsoft Visual C++ Express Edition  15.0      30.169                      89.0
MinGW                                 4.2.1     33.909                     100.O

In all cases, Csound was compiled to run fast on all x86 but tuned for Core 2 Duo, without getting really finicky about individual options.

This is without even attempting to use Intel's high-performance Math Kernel Library or Intel Performance Primitives to implement the vector-vector and vector-scalar arithmetic in aops.c. I think that might provide a further similar increment in performance. Intel also has a drop-in replacement for FFTW that is faster than FFTW.

The Intel compiler had trouble with the big C++ file generated by SWIG for CsoundAC, so I did not compile that.

On Windows, note that the Intel compiler MUST be used alongside an installation of the Microsoft compiler, but the free Express Edition of the Microsoft compiler does the job.

Also note, and this is the most important thing, some high-end optimizations are missing from the Microsoft compiler's Express Edition. It is quite likely that the gap between Microsoft and Intel would narrow, or even vanish, if I had used the Professional Edition of the Microsoft compiler. Even so, Microsoft does not have anything like the Intel performance libraries.

Regards,
Mike



-------------------------------------------------------------------------
This SF.net email is sponsored by: Microsoft
Defy all challenges. Microsoft(R) Visual Studio 2008.
http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/
_______________________________________________
Csound-devel mailing list
Csound-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/csound-devel