From jgsled@bottom.bic.mni.mcgill.ca Fri Mar 9 15:55:23 2001 >From jgsled Fri Mar 9 15:55:23 2001 Return-Path: Received: (from jgsled@localhost) by bottom.bic.mni.mcgill.ca (8.9.3/8.9.3) id PAA07097 for stever@bic.mni.mcgill.ca; Fri, 9 Mar 2001 15:55:23 -0500 (EST) Date: Fri, 9 Mar 2001 15:55:22 -0500 From: John SLED To: Steve ROBBINS Subject: Re: Questions about your minc software from MPI Cologne N3 package Message-ID: <20010309155522.C24475@bottom.bic.mni.mcgill.ca> References: <200103011621.RAA22152@minto.mpg.de> <20010305105913.D13705@bottom.bic.mni.mcgill.ca> Mime-Version: 1.0 Content-Type: multipart/mixed; boundary="gKMricLos+KVdGMg" Content-Disposition: inline User-Agent: Mutt/1.3.9i In-Reply-To: <20010305105913.D13705@bottom.bic.mni.mcgill.ca>; from stever@bic.mni.mcgill.ca on Mon, Mar 05, 2001 at 10:59:13AM -0500 Status: RO Content-Length: 5822 Lines: 173 --gKMricLos+KVdGMg Content-Type: text/plain; charset=us-ascii Content-Disposition: inline Dear Helmut, My C++ code dates from a time when C++ compilers were more lax. I've attached to this message a set of patches which should clean up the difficulties you mentioned. If N3 is a subdirectory of the current directory then patch -p0 < mni_n3_1.04_patch_20010309 should make some changes. In addition, a colleague had forwarded me some changes that are specific to solaris which I have included at the end of this message. with regards, John Sled McConnell Brain Imaging Centre Tel: (514) 398-8547 Montreal Neurological Institute Fax: (514) 398-2975 3801 University Street, Room WB208 Email: jgsled@bic.mni.mcgill.ca Montreal, Quebec H3A 2B4 Canada --------------------------------------- PREAMBLE: N3 built on Sun-Ultra60 running Solaris 7, gcc 2.95.2 and gnu-make 3.78. All the other required packages (PERL, NETCDF, MINC, MNI-PERLLIB) built as per instructions with no problems. All file paths will be relative to the N3 source directory. BEFORE RUNNING CONFIGURE Add double-quotes (") around $CPP_COMPILER on line 1071 of N3/configure AFTER RUNNING CONFIGURE In N3/Makefile.include add -Wno-non-template-friend to the CCFLAGS line In N3/extern/src/CLAPACK/make.inc Change cc to gcc (twice). In N3/extern/src/CLAPACK/F2CLIBS/libF77/makefile Change cc to gcc. Remove the -x option from the ld line (obsoleted by Solaris). In N3/src/Base/templates/Matrix.c Line 253 change "0" to "(Type)0" Line 1684 change "factor" to "(double) factor" Line 2987, 3006 change "m" to "(double)m" In N3/src/Base/templates/Matrix.h Line 565, 573 change "nElements()" to "(double)nElements()" In N3/src/Base/templates/SimpleArray.c Line 28 insert "#include " (for the finite() function). In N3/src/SharpenHist/sharpen_hist.c Line 44 insert "#include " (for the finite() function). THEN make build make test make install Also an unadvertised feature is that gzip/gunzip should be in the path when running the software. N3 then passes the test-suite. --gKMricLos+KVdGMg Content-Type: text/plain; charset=us-ascii Content-Disposition: attachment; filename="mni_n3_1.04_patch_20010309" diff -r -Nau --exclude-from=exclude_from_diff N3.sav/src/Base/include/Spline.h N3/src/Base/include/Spline.h --- N3.sav/src/Base/include/Spline.h Fri Aug 23 15:36:39 1996 +++ N3/src/Base/include/Spline.h Tue Mar 6 10:08:26 2001 @@ -101,7 +101,7 @@ return fit(coord2FlMat(coord), values); } #endif - operator ! () const { return !_fitted; } + int operator ! () const { return !_fitted; } operator double () const { return operator () (_tempPoint); } virtual double operator () (const float *point = 0) const = 0; diff -r -Nau --exclude-from=exclude_from_diff N3.sav/src/Base/include/backProp.h N3/src/Base/include/backProp.h --- N3.sav/src/Base/include/backProp.h Tue Feb 18 16:38:18 1997 +++ N3/src/Base/include/backProp.h Tue Mar 6 10:12:14 2001 @@ -60,9 +60,9 @@ Boolean _softmaxAtOutputLayer; // Consts and default values - static const _LUT_LENGTH; + static const int _LUT_LENGTH; static const long _SEED; - static const _N_CYCLES; + static const int _N_CYCLES; static const float _LEARNING_RATE, _MOMENTUM, _TEMPERATURE; static const float _MAX_ERROR, _MAX_D_ERROR; static const unsigned _SHUFFLE_INTERVAL; diff -r -Nau --exclude-from=exclude_from_diff N3.sav/src/Base/src/TrainingSet.c N3/src/Base/src/TrainingSet.c --- N3.sav/src/Base/src/TrainingSet.c Sat Oct 4 17:25:07 1997 +++ N3/src/Base/src/TrainingSet.c Tue Mar 6 10:11:36 2001 @@ -22,7 +22,7 @@ #include #include -const DEFAULT_TRAINING_SET_SIZE = 250; +const int DEFAULT_TRAINING_SET_SIZE = 250; /*************** * Example class diff -r -Nau --exclude-from=exclude_from_diff N3.sav/src/Base/src/dcomplex.c N3/src/Base/src/dcomplex.c --- N3.sav/src/Base/src/dcomplex.c Fri Aug 29 13:44:49 1997 +++ N3/src/Base/src/dcomplex.c Tue Mar 6 10:13:03 2001 @@ -23,7 +23,7 @@ // A few functions that define (bogus) math ops for complex -static _errorCount_dcomplex = 100; +static int _errorCount_dcomplex = 100; int operator < (const dcomplex&, const dcomplex&) { if (_errorCount_dcomplex) { diff -r -Nau --exclude-from=exclude_from_diff N3.sav/src/Base/src/fcomplex.c N3/src/Base/src/fcomplex.c --- N3.sav/src/Base/src/fcomplex.c Thu Aug 28 17:43:33 1997 +++ N3/src/Base/src/fcomplex.c Tue Mar 6 10:12:42 2001 @@ -21,7 +21,7 @@ // A few functions that define (bogus) math ops for complex -static _errorCount_fcomplex = 100; +static int _errorCount_fcomplex = 100; int operator < (const fcomplex&, const fcomplex&) { if (_errorCount_fcomplex) { diff -r -Nau --exclude-from=exclude_from_diff N3.sav/src/SharpenHist/sharpen_hist.c N3/src/SharpenHist/sharpen_hist.c --- N3.sav/src/SharpenHist/sharpen_hist.c Tue Aug 10 14:51:13 1999 +++ N3/src/SharpenHist/sharpen_hist.c Tue Mar 6 11:40:51 2001 @@ -104,7 +104,7 @@ if(args.blur_flag == FALSE) { // compute filtered distribution f - f = real(ifft(pmultEquals(fft(X_padded,0,1), filter),0,1)); + f = real(ifft(pmultEquals(asCompMat(X_padded).fft(0,1), filter),0,1)); #ifdef HAVE_MATLAB if(args.debug_flag == TRUE) f.saveMatlab("distribution.mat", "f"); @@ -127,8 +127,8 @@ // compute mapping Mat Y_padded = - pdivEquals(real(ifft(pmultEquals(fft(moment,0,1), blur),0,1)), - real(ifft(pmultEquals(fft(f,0,1),blur),0,1))); + pdiv(real(ifft(pmultEquals(asCompMat(moment).fft(0,1), blur),0,1)), + real(ifft(pmultEquals(asCompMat(f).fft(0,1),blur),0,1))); Y = Y_padded(offset,offset+X.getrows()-1,0,0); // remove any NANs of INFs from Y --gKMricLos+KVdGMg--