Alan, Thanks for the testing. I have briefly looked at ndiff in the past. Useful for this kind of thing!
I seem to recall that the python issues are partly a 32-bit / 64-bit issue. I think that on my old 32 bit system (no longer in use) I used to see some non-significant differences with the python examples. In particular the issue in the standard error with example 23 is due to this. Python integers as signed 32 bit integers on a 32-bit system. The FCI variables are 32-bit unsigned, so have to be stored in a long integer on a 32-bit python system, hence the L tagged on the end. I failed to find an easy way to get rid of this, but that doesn't mean there isn't one. No problem on 64-bit systems. Andrew On Sun, Oct 21, 2012 at 12:21:36AM -0700, Alan Irwin wrote: > I used ndiff (a really useful fuzzy diff) to analyze the differences > between the C and Python results. > > If you are interested in using that application yourself, you > can download, verify, and signature check it > using > > wget ftp://ftp.math.utah.edu/pub/misc/ndiff-2.00.tar.gz > wget ftp://ftp.math.utah.edu/pub/misc/ndiff-2.00.tar.gz.sig > gpg --verify ndiff-2.00.tar.gz.sig > > Furthermore, after unpacking copy the two attached files (constituting > a new CMake-based build system for ndiff I just implemented in the > last hour) into the top of the ndiff-2.00 source tree then build it > using > > mkdir build_dir > cmake ../ndiff-2.0.0 > make > > This creates the newly compiled ndiff executable in the separate build > tree which is really all you need. This build system for ndiff should > work on Unix without issues, but I will leave it to Arjen to try it > out in the Windows case, if he is interested in a fuzzy diff which > ignores small relative or absolute numerical differences between the > two files. > > The command > > for INDEX in 03 04 05 06 08 09 11 12 14a 15 16 17 18 19 20 21 22 23 25 \ > 26 27 29 33; do echo "index = $INDEX"; \ > /home/software/ndiff/build_dir/ndiff -relerr 1.e-03 x${INDEX}c.psc \ > x${INDEX}p.psc ; done |less > > showed relatively few (just examples 17, 19, 20, 22, 25, and 29) of > those C-Python PostScript differences were numerically significant. > Furthermore, when I checked > those visually, the only really large difference was the 3rd page > of example 19 which was completely messed up in the Python version > for some reason. The example 23 stdout difference was > > 1c1 > < For example 23 prior to page 12 the FCI is 0x80000000 > --- > >For example 23 prior to page 12 the FCI is 0x80000000L > > which is a trivial although it would be nice to get > rid of this difference. > > In sum, aside from the 3rd page of example 19, python on Wine seems to > be working well. Arjen, do you get similar Python results (small > relative differences but no large ones other than page 3 of > example 19) on Windows? > > Alan > __________________________ > Alan W. Irwin > > Astronomical research affiliation with Department of Physics and Astronomy, > University of Victoria (astrowww.phys.uvic.ca). > > Programming affiliations with the FreeEOS equation-of-state > implementation for stellar interiors (freeeos.sf.net); the Time > Ephemerides project (timeephem.sf.net); PLplot scientific plotting > software package (plplot.sf.net); the libLASi project > (unifont.org/lasi); the Loads of Linux Links project (loll.sf.net); > and the Linux Brochure Project (lbproject.sf.net). > __________________________ > > Linux-powered Science > __________________________ > # The autoconf-based build of ndiff is beginning to break down so this > # is the start of the implementation of an equivalent build system > # using CMake. This is just a first proof-of-concept attempt to build > # ndiff and not even bother to test or install it. But at least the > # build appears to work and so does the resulting ndiff executable. > > # Top-level CMakeLists.txt boilerplate. > project(ndiff C) > cmake_minimum_required(VERSION 2.8.2 FATAL_ERROR) > > # Configure everything in config.h which must have > # specific values of 1 or 0 > > set(HAVE_LONG_DOUBLE 0) > find_file(HAVE_CTYPE_H ctype.h) > if(HAVE_CTYPE_H) > set(HAVE_CTYPE_H 1) > else(HAVE_CTYPE_H) > set(HAVE_CTYPE_H 0) > endif(HAVE_CTYPE_H) > > find_file(HAVE_LIMITS_H limits.h) > if(HAVE_LIMITS_H) > set(HAVE_LIMITS_H 1) > else(HAVE_LIMITS_H) > set(HAVE_LIMITS_H 0) > endif(HAVE_LIMITS_H) > > find_file(HAVE_STDDEF_H stddef.h) > if(HAVE_STDDEF_H) > set(HAVE_STDDEF_H 1) > else(HAVE_STDDEF_H) > set(HAVE_STDDEF_H 0) > endif(HAVE_STDDEF_H) > > find_file(HAVE_STDIO_H stdio.h) > if(HAVE_STDIO_H) > set(HAVE_STDIO_H 1) > else(HAVE_STDIO_H) > set(HAVE_STDIO_H 0) > endif(HAVE_STDIO_H) > > find_file(HAVE_STDLIB_H stdlib.h) > if(HAVE_STDLIB_H) > set(HAVE_STDLIB_H 1) > else(HAVE_STDLIB_H) > set(HAVE_STDLIB_H 0) > endif(HAVE_STDLIB_H) > > find_file(HAVE_STRING_H string.h) > if(HAVE_STRING_H) > set(HAVE_STRING_H 1) > else(HAVE_STRING_H) > set(HAVE_STRING_H 0) > endif(HAVE_STRING_H) > > # Code uses HAVE_GMP rather than the expected HAVE_GMP_H > find_file(HAVE_GMP gmp.h) > if(HAVE_GMP) > set(HAVE_GMP 1) > else(HAVE_GMP) > set(HAVE_GMP 0) > endif(HAVE_GMP) > > configure_file( > ${CMAKE_CURRENT_SOURCE_DIR}/config.h.cmake > ${CMAKE_CURRENT_BINARY_DIR}/config.h > @ONLY > ) > > # Build the ndiff executable: > include_directories( > ${CMAKE_CURRENT_SOURCE_DIR} > ${CMAKE_CURRENT_BINARY_DIR} > ) > > set(ndiff_SRC > ndiff.c > store.c > awklib.c > regexp/regexp.c > regexp/regerror.c > ) > > add_executable(ndiff ${ndiff_SRC}) > > # That wasn't so hard, was it? > // Note that ndiff code uses old-fashioned #defines with values > // which must be either true or false, rather than the #ifdef style. > > // Define if the `long double' type works. > #define HAVE_LONG_DOUBLE @HAVE_LONG_DOUBLE@ > > // Define to `unsigned' if <sys/types.h> doesn't define. > // #define size_t > > // Define if you have the ANSI C header files. > //#define STDC_HEADERS @STDC_HEADERS@ > > // Define if you have the <ctype.h> header file. > #define HAVE_CTYPE_H @HAVE_CTYPE_H@ > > // Define if you have the <limits.h> header file. > #define HAVE_LIMITS_H @HAVE_LIMITS_H@ > > // Define if you have the <stddef.h> header file. > #define HAVE_STDDEF_H @HAVE_STDDEF_H@ > > // Define if you have the <stdio.h> header file. > #define HAVE_STDIO_H @HAVE_STDIO_H@ > > // Define if you have the <stdlib.h> header file. > #define HAVE_STDLIB_H @HAVE_STDLIB_H@ > > // Define if you have the <string.h> header file. > #define HAVE_STRING_H @HAVE_STRING_H@ > > // Define if have the <gmp.h> header file. > #define HAVE_GMP @HAVE_GMP@ > ------------------------------------------------------------------------------ > Everyone hates slow websites. So do we. > Make your web apps faster with AppDynamics > Download AppDynamics Lite for free today: > http://p.sf.net/sfu/appdyn_sfd2d_oct > _______________________________________________ > Plplot-devel mailing list > Plplot-devel@lists.sourceforge.net > https://lists.sourceforge.net/lists/listinfo/plplot-devel ------------------------------------------------------------------------------ Everyone hates slow websites. So do we. Make your web apps faster with AppDynamics Download AppDynamics Lite for free today: http://p.sf.net/sfu/appdyn_sfd2d_oct _______________________________________________ Plplot-devel mailing list Plplot-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/plplot-devel
