I wrote a test program in fortran and used gfortran 4.9.2(tdm64-gcc 4.9.2) with openmp to make me familiar with parallel processing. Theprogram does matrix inversion.
I tested for different floating point precisions. For single, double and quadruple floating point precisioneverything seems ok. Check of the results reveals that some off-diagonalelements of the identity matrix resulting from multiplication of the originalmatrix with its inverse are <> 0 with the expected magnitude. However for extended floating point precision off-diagonalelements of the identity matrix for thread 0 are <> 0 with the expectedmagnitude, for threads 1 to 3 they are a factor 1000 (!) larger and moreoveridentical to those of the identity matrix in the double floating pointprecision calculation. See the attached files with results of calculations fordifferent floating point precisions. The calculations were performed on a machine with a 4-corei5 cpu and windows 7 64 bit. The CodeBlock IDE was used. Where is the error? With regards Pieter Bedijn - retired scientist and software developer pbed...@aol.com
test1omp_rkind04.lis
Description: Binary data
test1omp_rkind08.lis
Description: Binary data
test1omp_rkind10.lis
Description: Binary data
test1omp_rkind16.lis
Description: Binary data
------------------------------------------------------------------------------ New Year. New Location. New Benefits. New Data Center in Ashburn, VA. GigeNET is offering a free month of service with a new server in Ashburn. Choose from 2 high performing configs, both with 100TB of bandwidth. Higher redundancy.Lower latency.Increased capacity.Completely compliant. http://p.sf.net/sfu/gigenet
_______________________________________________ Mingw-w64-public mailing list Mingw-w64-public@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/mingw-w64-public
