Initial commit

initial v1.1 which was obtained from MVS ivm account with gitignore added

.gitignore

+#This is a gitignore file for parlib
+
+#object files
+*.o
+#linked libraries
+*.a
+#copied manual and headers
+man/
+include/

ParLib.src/Makefile

+.SUFFIXES:      .c .o .a
+
+AR          = ar crl
+RANLIB      = ranlib
+
+INSTALLDIR	= /gpfs/NETHOME/ivm6/CM/ParLib.v1.1
+#INSTALLDIR	= $(HOME)
+
+
+default: all
+all: setvars libparlib.a libparlibf.a
+
+#
+# For INM's HP-cluster
+#
+   CC 			= mpicc
+   LIBS			= -lmpich
+#   LIBPATH		= -L/opt/mpich-gm/lib
+#   INCPATH		= -I/opt/mpich-gm/include
+   DEFINES		= -DFORTRANUNDERSCORE
+
+setvars:
+
+libparlib.a: bexchange.o transpose.o
+	$(AR) $@ $?
+	$(RANLIB) $@
+
+libparlibf.a: parlibf.o bexchangef.o transposef.o
+	$(AR) $@ $?
+	$(RANLIB) $@
+
+.c.o:
+		$(CC) $(INCPATH) -c $(DEFINES) $<
+
+clean:
+		rm -f *.o *.a
+
+rebuild: clean all
+
+install: libparlib.a libparlibf.a
+	@if !(test -d $(INSTALLDIR)) ; then \
+	   echo Creating directory $(INSTALLDIR)...; \
+	   mkdir $(INSTALLDIR) ; \
+	fi
+	@if !(test -d $(INSTALLDIR)/lib) ; then \
+	   echo Creating directory $(INSTALLDIR)/lib...; \
+	   mkdir $(INSTALLDIR)/lib ; \
+	fi
+	cp $? $(INSTALLDIR)/lib
+	@if !(test -d $(INSTALLDIR)/include) ; then \
+	   echo Creating directory $(INSTALLDIR)/include...; \
+	   mkdir $(INSTALLDIR)/include ; \
+	fi
+	cp -r parlib.h parlibf.h $(INSTALLDIR)/include
+	@if !(test -d $(INSTALLDIR)/man) ; then \
+	   echo Creating directory $(INSTALLDIR)/man...; \
+	   mkdir $(INSTALLDIR)/man ; \
+	fi
+	@if !(test -d $(INSTALLDIR)/man/man3) ; then \
+	   echo Creating directory $(INSTALLDIR)/man/man3...; \
+	   mkdir $(INSTALLDIR)/man/man3 ; \
+	fi
+	cp -r man/* $(INSTALLDIR)/man/man3
+	ln -s $(INSTALLDIR)/man/man3/P_BExchange_init.3 \
+      $(INSTALLDIR)/man/man3/P_BExchange_start.3
+	ln -s $(INSTALLDIR)/man/man3/P_BExchange_init.3 \
+      $(INSTALLDIR)/man/man3/P_BExchange_end.3
+	ln -s $(INSTALLDIR)/man/man3/P_BExchange_init.3 \
+      $(INSTALLDIR)/man/man3/P_BExchange_free.3
+	ln -s $(INSTALLDIR)/man/man3/P_Transpose_init.3 \
+      $(INSTALLDIR)/man/man3/P_Transpose_start.3
+	ln -s $(INSTALLDIR)/man/man3/P_Transpose_init.3 \
+      $(INSTALLDIR)/man/man3/P_Transpose_end.3
+	ln -s $(INSTALLDIR)/man/man3/P_Transpose_init.3 \
+      $(INSTALLDIR)/man/man3/P_Transpose_free.3
+
+parlibf.o: parlib.h
+bexchange.o: parlib.h
+bexchangef.o: parlib.h
+transpose.o: parlib.h
+transposef.o: parlib.h

ParLib.src/README

+
+                              INM ParLib v1.0
+-------------------------------------------------------------------------------
+author: Val Gloukhov
+         Institute of Numerical Mathematics of the Russian Academy 
+         of Sciences
+         (gluhoff@inm.ras.ru)
+
+latest update: 11/30/2001
+-------------------------------------------------------------------------------
+
+INM ParLib is a high level library facilitating the coding of finite 
+difference and spectral approximation weather and climate prediction 
+models on distributed memory parallel computers.

ParLib.src/bexchange.c

+#include "parlib.h"
+
+/*  
+ *  Error codes:
+ *     0 - success
+ *     1 - nonpositive number of dimensions
+ *     2 - wrong communicated dimension
+ *     3 - negative boundary width
+ *     4 - nonpositive dimension
+ *     5 - boundary width exceeds the array block length
+ */
+int P_BExchange_init ( ndims, stride, blklen, bdim, overlap, datatype, 
+		comm, period, bexchange )
+   int				ndims, *stride, *blklen, bdim, overlap[2], period;
+   MPI_Datatype	datatype;
+   MPI_Comm			comm;
+	BExchange		*bexchange;
+{
+	int				nproc, iproc, direct, idim, sendproc[2], recvproc[2];
+	int				count, strd, sbind[2], rbind[2], send[2], recv[2];
+   MPI_Aint			fsize;
+   MPI_Datatype	oldtype, btype[2];
+   MPI_Request		sreq[2], rreq[2]; 
+
+/*
+ * Check input parameters
+ */
+   if ( ndims < 1 ) {return 1;}
+   if ( bdim < 1 || bdim > ndims ) {return 2;}
+   if ( overlap[0] == 0 && overlap[1] == 0 ) {return 0;} /* success */
+   for ( idim = 0; idim < ndims; idim++ ) {
+      if ( stride[idim] <= 0 ) {return 4;}
+   }
+   for ( direct = 0; direct < 2; direct++ ) {
+      if ( overlap[direct] < 0 ) {return 3;}
+      if ( overlap[direct] > blklen[bdim-1] ) {return 5;}
+   }
+/*
+ * Define the number of processors in the group and the rank
+ */
+   MPI_Comm_size ( comm, &nproc );
+   if ( nproc == 0 ) {return 0;} /* success */
+   MPI_Comm_rank ( comm, &iproc );
+   if ( iproc == MPI_UNDEFINED ) {return 0;} /* the process does not belong to the group */
+   sendproc[0] = ( iproc == 0 ? nproc-1 : iproc-1 );
+   recvproc[0] = ( iproc == nproc-1 ? 0 : iproc+1 );
+	sendproc[1] = recvproc[0];
+	recvproc[1] = sendproc[0];
+   send[0] = iproc > 0 || period; 
+   recv[0] = iproc < nproc-1 || period;
+   send[1] = recv[0];
+   recv[1] = send[0];
+   MPI_Type_extent ( datatype, &fsize );
+/*
+ * Define data types for the boundaries
+ */
+    for ( direct = 0; direct < 2; direct++ ) {
+      if ( overlap[direct] > 0 ) {
+         oldtype = datatype;
+         strd = 1;
+         for ( idim = 0; idim < ndims; idim++ ) {
+            if ( idim+1 == bdim ) {
+               count = overlap[direct];
+            } else {
+               count = blklen[idim];
+            }
+            MPI_Type_hvector ( count, 1, strd * fsize, oldtype, 
+					&btype[direct] );
+            if ( idim > 0 ) {
+               MPI_Type_free ( &oldtype );
+            }
+            oldtype = btype[direct];
+            strd = strd * stride[idim];
+         }
+         MPI_Type_commit ( &btype[direct] );
+      }
+   }
+/*
+ * Determine the begining of boundaries
+ */
+   strd = 1;
+   for ( idim = 0; idim < bdim - 1; idim++ ) {
+      strd = strd * stride[idim];
+   }
+   sbind[0] = 0;
+   rbind[0] = blklen[bdim-1]*strd;
+   sbind[1] = (blklen[bdim-1]-overlap[1])*strd;
+   rbind[1] = -overlap[1]*strd;
+
+   for ( direct = 0; direct < 2; direct++ ) {
+      bexchange->overlap[direct] = overlap[direct];
+      bexchange->send[direct] = send[direct];
+      bexchange->recv[direct] = recv[direct];
+      bexchange->btype[direct] = btype[direct];
+      bexchange->sendproc[direct] = sendproc[direct];
+      bexchange->recvproc[direct] = recvproc[direct];
+      bexchange->sbind[direct] = sbind[direct];
+      bexchange->rbind[direct] = rbind[direct];
+	}
+	bexchange->comm = comm;
+	bexchange->fsize = fsize;
+	return 0;
+}
+
+int P_BExchange_start ( a, bexchange )
+   void				*a; 
+	BExchange		*bexchange;
+{
+	int				direct, overlap[2], send[2], recv[2], btype[2]; 
+	int				sendproc[2], recvproc[2], sbind[2], rbind[2];
+	MPI_Comm			comm;
+   MPI_Request		sreq[2], rreq[2]; 
+   MPI_Aint			fsize;
+	char				*ach = (char *) a;
+
+   for ( direct = 0; direct < 2; direct++ ) {
+      overlap[direct] = bexchange->overlap[direct];
+      send[direct] = bexchange->send[direct];
+      recv[direct] = bexchange->recv[direct];
+      btype[direct] = bexchange->btype[direct];
+      sendproc[direct] = bexchange->sendproc[direct];
+      recvproc[direct] = bexchange->recvproc[direct];
+      sbind[direct] = bexchange->sbind[direct];
+      rbind[direct] = bexchange->rbind[direct];
+   }
+	comm = bexchange->comm;
+	fsize = bexchange->fsize;
+
+   for ( direct = 0; direct < 2; direct++ ) {
+      if ( overlap[direct] > 0 ) {
+         if ( send[direct] ) {
+            MPI_Isend ( ach+sbind[direct]*fsize, 1, btype[direct], 
+					sendproc[direct], 0, comm, &sreq[direct] );
+         }
+         if ( recv[direct] ) {
+            MPI_Irecv ( ach+rbind[direct]*fsize, 1, btype[direct], 
+					recvproc[direct], 0, comm, &rreq[direct] );
+         }
+      }
+   }
+   for ( direct = 0; direct < 2; direct++ ) {
+      bexchange->sreq[direct]=sreq[direct];
+      bexchange->rreq[direct]=rreq[direct];
+   }
+	return 0;
+}
+
+int P_BExchange_end ( bexchange ) 
+	BExchange    *bexchange;
+{
+   MPI_Status		status;
+   int 				direct, overlap[2], send[2], recv[2];
+   MPI_Request		sreq[2], rreq[2]; 
+
+   for ( direct = 0; direct < 2; direct++ ) {
+      overlap[direct] = bexchange->overlap[direct];
+      send[direct] = bexchange->send[direct];
+      recv[direct] = bexchange->recv[direct];
+      sreq[direct] = bexchange->sreq[direct];
+      rreq[direct] = bexchange->rreq[direct];
+   }
+
+   for ( direct = 0; direct < 2; direct++ ) {
+      if ( overlap[direct] > 0 ) {
+         if ( send[direct] ) {
+            MPI_Wait ( &sreq[direct], &status );
+         }
+         if ( recv[direct] ) {
+            MPI_Wait ( &rreq[direct], &status );
+         }
+      }
+   }
+	return 0;
+}
+
+int P_BExchange_free ( bexchange ) 
+	BExchange    *bexchange;
+{
+   int direct, overlap[2];
+   MPI_Datatype btype[2];
+   for ( direct = 0; direct < 2; direct++ ) {
+      overlap[direct] = bexchange->overlap[direct];
+      btype[direct] = bexchange->btype[direct];
+	}
+   for ( direct = 0; direct < 2; direct++ ) {
+      if ( overlap[direct] > 0 ) {
+         MPI_Type_free ( &btype[direct] );
+      }
+   }
+	return 0;
+}
+
+int P_BExchange ( a, ndims, stride, blklen, bdim, overlap, datatype,
+		comm, period )
+   void *a; 
+   MPI_Datatype datatype;
+   int ndims, *stride, *blklen, bdim, overlap[2]; 
+   int period;
+   MPI_Comm comm;
+{
+	BExchange bexchange;
+	int ierr;
+   if ( ierr = P_BExchange_init ( ndims, stride, blklen, bdim, overlap, 
+		datatype, comm, period, &bexchange ) != 0 ) { return ierr; }
+   P_BExchange_start ( a, &bexchange );
+   P_BExchange_end ( &bexchange );
+   P_BExchange_free ( &bexchange );
+   return 0;
+}

ParLib.src/bexchangef.c

+#include <stdlib.h>
+#include "parlib.h"
+
+#ifdef FORTRANUNDERSCORE
+void p_bexchange_init_ ( ndims, stride, blklen, bdim, overlap, datatype, 
+		comm, period, bexchange, ierr )
+#elif defined(FORTRANDOUBLEUNDERSCORE)
+void p_bexchange_init__ ( ndims, stride, blklen, bdim, overlap, datatype,
+		comm, period, bexchange, ierr )
+#else
+void p_bexchange_init ( ndims, stride, blklen, bdim, overlap, datatype, 
+		comm, period, bexchange, ierr )
+#endif
+
+   MPI_Fint *datatype, *ndims, *stride, *blklen, *bdim, *overlap, 
+		*period, *ierr, *comm; 
+	BExchange **bexchange;
+{
+   *bexchange = (BExchange *) malloc ( sizeof (BExchange) );
+   *ierr = P_BExchange_init ( (int)*ndims, (int *)stride, (int *)blklen,
+	(int)*bdim, (int *)overlap, MPI_Type_f2c(*datatype), 
+	MPI_Comm_f2c(*comm), (int)*period, *bexchange );
+}
+
+#ifdef FORTRANUNDERSCORE
+void p_bexchange_start_ ( a, bexchange, ierr )
+#elif defined(FORTRANDOUBLEUNDERSCORE)
+void p_bexchange_start__ ( a, bexchange, ierr )
+#else
+void p_bexchange_start ( a, bexchange, ierr )
+#endif
+
+   void *a;
+   BExchange **bexchange;
+	MPI_Fint *ierr;
+{
+   *ierr = P_BExchange_start ( a, *bexchange );
+}
+
+#ifdef FORTRANUNDERSCORE
+void p_bexchange_end_ ( bexchange, ierr )
+#elif defined(FORTRANDOUBLEUNDERSCORE)
+void p_bexchange_end__ ( bexchange, ierr )
+#else
+void p_bexchange_end ( bexchange, ierr )
+#endif
+
+   BExchange **bexchange;
+   MPI_Fint *ierr;
+{
+   *ierr = P_BExchange_end ( *bexchange );
+}
+
+#ifdef FORTRANUNDERSCORE
+void p_bexchange_free_ ( bexchange, ierr )
+#elif defined(FORTRANDOUBLEUNDERSCORE)
+void p_bexchange_free__ ( bexchange, ierr )
+#else
+void p_bexchange_free ( bexchange, ierr )
+#endif
+   BExchange **bexchange;
+   MPI_Fint *ierr;
+{
+   P_BExchange_free ( *bexchange );
+   free ( *bexchange );
+}
+
+#ifdef FORTRANUNDERSCORE
+void p_bexchange_ ( a, ndims, stride, blklen, bdim, overlap, datatype,
+		comm, period, ierr )
+#elif defined(FORTRANDOUBLEUNDERSCORE)
+void p_bexchange__ ( a, ndims, stride, blklen, bdim, overlap, datatype, 
+		comm, period, ierr )
+#else
+void p_bexchange ( a, ndims, stride, blklen, bdim, overlap, datatype,
+		comm, period, ierr )
+#endif
+	void *a;
+   int *ndims, *stride, *blklen, *bdim, *overlap, *period, *ierr;
+   MPI_Datatype *datatype;
+   MPI_Comm *comm;
+{
+   *ierr = P_BExchange ( a, (int)*ndims, (int *)stride, (int *)blklen, 
+		(int)*bdim, (int *)overlap, MPI_Type_f2c(*datatype), 
+		MPI_Comm_f2c(*comm), (int)*period );
+}

ParLib.src/doc/manual.aux

+\relax 
+\catcode`"\active
+\select@language{russian}
+\@writefile{toc}{\select@language{russian}}
+\@writefile{lof}{\select@language{russian}}
+\@writefile{lot}{\select@language{russian}}
+\citation{T126}
+\citation{AGCM}
+\citation{Ilghiz}
+\citation{Swarztrauber}
+\citation{Barros}
+\@writefile{toc}{\contentsline {section}{\numberline {1}\IeC {\CYRV }\IeC {\cyrv }\IeC {\cyre }\IeC {\cyrd }\IeC {\cyre }\IeC {\cyrn }\IeC {\cyri }\IeC {\cyre }}{1}}
+\@writefile{toc}{\contentsline {section}{\numberline {2}\IeC {\CYRI }\IeC {\cyrs }\IeC {\cyrp }\IeC {\cyro }\IeC {\cyrl }\IeC {\cyrsftsn }\IeC {\cyrz }\IeC {\cyro }\IeC {\cyrv }\IeC {\cyra }\IeC {\cyrn }\IeC {\cyri }\IeC {\cyre } \IeC {\cyrb }\IeC {\cyri }\IeC {\cyrb }\IeC {\cyrl }\IeC {\cyri }\IeC {\cyro }\IeC {\cyrt }\IeC {\cyre }\IeC {\cyrk }\IeC {\cyri }}{1}}
+\@writefile{toc}{\contentsline {subsection}{\numberline {2.1}\IeC {\CYRK }\IeC {\cyro }\IeC {\cyrm }\IeC {\cyrp }\IeC {\cyri }\IeC {\cyrl }\IeC {\cyrya }\IeC {\cyrc }\IeC {\cyri }\IeC {\cyrya } \IeC {\cyri } \IeC {\cyrl }\IeC {\cyri }\IeC {\cyrn }\IeC {\cyrk }\IeC {\cyro }\IeC {\cyrv }\IeC {\cyrk }\IeC {\cyra } \IeC {\cyrp }\IeC {\cyrr }\IeC {\cyro }\IeC {\cyrg }\IeC {\cyrr }\IeC {\cyra }\IeC {\cyrm }\IeC {\cyrm }}{2}}
+\@writefile{toc}{\contentsline {subsection}{\numberline {2.2}\IeC {\CYRV }\IeC {\cyrery }\IeC {\cyrz }\IeC {\cyro }\IeC {\cyrv } \IeC {\cyrb }\IeC {\cyri }\IeC {\cyrb }\IeC {\cyrl }\IeC {\cyri }\IeC {\cyro }\IeC {\cyrt }\IeC {\cyre }\IeC {\cyrch }\IeC {\cyrn }\IeC {\cyrery }\IeC {\cyrh } \IeC {\cyrp }\IeC {\cyrr }\IeC {\cyro }\IeC {\cyrc }\IeC {\cyre }\IeC {\cyrd }\IeC {\cyru }\IeC {\cyrr }}{2}}
+\@writefile{toc}{\contentsline {subsubsection}{\numberline {2.2.1}\IeC {\CYRO }\IeC {\cyrb }\IeC {\cyrm }\IeC {\cyre }\IeC {\cyrn } \IeC {\cyrg }\IeC {\cyrr }\IeC {\cyra }\IeC {\cyrn }\IeC {\cyri }\IeC {\cyrch }\IeC {\cyrn }\IeC {\cyrery }\IeC {\cyrm }\IeC {\cyri } \IeC {\cyrz }\IeC {\cyrn }\IeC {\cyra }\IeC {\cyrch }\IeC {\cyre }\IeC {\cyrn }\IeC {\cyri }\IeC {\cyrya }\IeC {\cyrm }\IeC {\cyri }}{2}}
+\@writefile{toc}{\contentsline {subsubsection}{\numberline {2.2.2}\IeC {\CYRT }\IeC {\cyrr }\IeC {\cyra }\IeC {\cyrn }\IeC {\cyrs }\IeC {\cyrp }\IeC {\cyro }\IeC {\cyrn }\IeC {\cyri }\IeC {\cyrr }\IeC {\cyro }\IeC {\cyrv }\IeC {\cyra }\IeC {\cyrn }\IeC {\cyri }\IeC {\cyre }}{3}}
+\bibcite{T126}{1}
+\bibcite{AGCM}{2}
+\bibcite{Ilghiz}{3}
+\bibcite{Swarztrauber}{4}
+\bibcite{Barros}{5}
+\bibcite{Rodriguez}{6}
+\bibcite{Henderson}{7}

ParLib.src/doc/manual.log

+This is TeX, Version 3.14159 (Web2C 7.3.1) (format=latex 2001.1.2)  2 NOV 2001 14:34
+**manual.tex
+(manual.tex
+LaTeX2e <1999/12/01> patch level 1
+Babel <v3.6Z> and hyphenation patterns for american, russian, nohyphenation, lo
+aded.
+(/usr/share/texmf/tex/latex/base/article.cls
+Document Class: article 1999/09/10 v1.4a Standard LaTeX document class
+(/usr/share/texmf/tex/latex/base/size12.clo
+File: size12.clo 1999/09/10 v1.4a Standard LaTeX file (size option)
+)
+\c@part=\count79
+\c@section=\count80
+\c@subsection=\count81
+\c@subsubsection=\count82
+\c@paragraph=\count83
+\c@subparagraph=\count84
+\c@figure=\count85
+\c@table=\count86
+\abovecaptionskip=\skip41
+\belowcaptionskip=\skip42
+\bibindent=\dimen102
+)
+(/usr/share/texmf/tex/latex/base/inputenc.sty
+Package: inputenc 1999/09/17 v0.992 Input encoding file 
+
+(/usr/share/texmf/tex/latex/cyrillic/koi8-r.def
+File: koi8-r.def 1999/11/16 v1.0d Input encoding file
+))
+(/usr/share/texmf/tex/generic/babel/babel.sty
+Package: babel 1999/09/09 v3.6Z The Babel package
+
+(/usr/share/texmf/tex/generic/babel/russianb.ldf
+Language: russianb 1999/08/27 v1.1l Russian support from the babel system
+
+(/usr/share/texmf/tex/generic/babel/babel.def
+File: babel.def 1999/09/09 v3.6Z Babel common definitions
+\babel@savecnt=\count87
+\U@D=\dimen103
+)
+(/usr/share/texmf/tex/latex/cyrillic/t2aenc.def
+File: t2aenc.def 1999/11/29 v1.0c Cyrillic encoding definition file
+)
+LaTeX Info: Redefining \latintext on input line 142.
+LaTeX Info: Redefining \textlatin on input line 147.
+Package babel Info: Making " an active character on input line 210.
+)) (manual.aux
+LaTeX Font Info:    Try loading font information for T2A+cmr on input line 3.
+
+(/usr/share/texmf/tex/latex/cyrillic/t2acmr.fd
+File: t2acmr.fd 1999/01/07 v1.0 Computer Modern Cyrillic font definitions
+))
+\openout1 = `manual.aux'.
+
+LaTeX Font Info:    Checking defaults for OML/cmm/m/it on input line 5.
+LaTeX Font Info:    ... okay on input line 5.
+LaTeX Font Info:    Checking defaults for T1/cmr/m/n on input line 5.
+LaTeX Font Info:    ... okay on input line 5.
+LaTeX Font Info:    Checking defaults for OT1/cmr/m/n on input line 5.
+LaTeX Font Info:    ... okay on input line 5.
+LaTeX Font Info:    Checking defaults for OMS/cmsy/m/n on input line 5.
+LaTeX Font Info:    ... okay on input line 5.
+LaTeX Font Info:    Checking defaults for OMX/cmex/m/n on input line 5.
+LaTeX Font Info:    ... okay on input line 5.
+LaTeX Font Info:    Checking defaults for U/cmr/m/n on input line 5.
+LaTeX Font Info:    ... okay on input line 5.
+LaTeX Font Info:    Checking defaults for T2A/cmr/m/n on input line 5.
+LaTeX Font Info:    ... okay on input line 5.
+ [1
+
+]
+Overfull \hbox (0.26997pt too wide) in paragraph at lines 38--39
+[]\T2A/cmr/m/n/12  ----,  - --  
+- -
+ []
+
+LaTeX Font Info:    Try loading font information for T2A+cmtt on input line 45.
+
+(/usr/share/texmf/tex/latex/cyrillic/t2acmtt.fd
+File: t2acmtt.fd 1999/01/07 v1.0 Computer Modern Cyrillic font definitions
+)
+Overfull \hbox (1.98051pt too wide) in paragraph at lines 44--46
+\T2A/cmr/m/n/12 --- --- - - \T2A/cmtt/m/n/12 -I
+\T2A/cmr/m/it/12 __--\T2A/cmtt/m/n/12 /include\T2A/cmr/m/n/12 ,
+ []
+
+[1]
+LaTeX Font Info:    External font `cmex10' loaded for size
+(Font)              <12> on input line 60.
+LaTeX Font Info:    External font `cmex10' loaded for size
+(Font)              <8> on input line 60.
+LaTeX Font Info:    External font `cmex10' loaded for size
+(Font)              <6> on input line 60.
+
+Overfull \hbox (1.32059pt too wide) in paragraph at lines 75--82
+\T2A/cmr/m/n/12 -, -- - -- \T2A/cmtt/m/n/12 MPI_REA
+L \T2A/cmr/m/n/12  \T2A/cmtt/m/n/12 MPI_DOUBLE_PRECISION\T2A/cmr/m/n/12 . 
+ []
+
+[2] [3] [4] [5] [6] (manual.aux) ) 
+Here is how much of TeX's memory you used:
+ 975 strings out of 20896
+ 10539 string characters out of 197516
+ 59080 words of memory out of 350001
+ 3925 multiletter control sequences out of 10000+15000
+ 12345 words of font info for 33 fonts, out of 400000 for 1000
+ 24 hyphenation exceptions out of 1000
+ 24i,8n,24p,457b,303s stack positions out of 1500i,100n,500p,50000b,4000s
+
+Output written on manual.dvi (7 pages, 19380 bytes).

ParLib.src/doc/manual.tex

+% LaTeX document
+\documentclass[a4paper,12pt]{article}
+\usepackage[koi8-r]{inputenc}
+\usepackage[russian]{babel}
+\begin{document}
+\begin{titlepage}
+\begin{center}
+
+\Huge{\textbf{
+  \\
+   \\
+INM~ParLib ( 1.0)}}
+
+\vspace{0.1\textheight}
+\Large{~.~.} \\
+
+\vspace{0.2\textheight}
+\Large{  ~ \\
+, \today}
+
+\end{center}
+\end{titlepage}
+
+\section{}
+         INM ParLib,        ~\cite{T126}  ~\cite{AGCM}       .
+
+             ,        ,        ,           .
+
+          . 
+               .
+         .
+
+  ,              ,        .
+       ,        ,    .
+       ,    .
+  ~\cite{Ilghiz}, \cite{Swarztrauber}          ,         ~\cite{Barros}.
+
+ ,        ,              (          ,  ).
+
+\section{ }
+    ,    ,    .
+
+\subsection{   }
+      MPI            ,   MPI.
+ ,      {\tt -I{\it \_\_\-\-}/include},  ~-- {\tt -L{\it \_\_}/lib -lparlib}, a          {\tt -lparlibf}.
+
+\subsection{  }
+      
+\begin{verbatim}
+      INCLUDE 'parlibf.h'
+\end{verbatim}
+ ,   
+\begin{verbatim}
+#include 'parlib.h'
+\end{verbatim}
+ .
+
+\subsubsection{  }
+
+  $\cal A$        .
+          ().
+,      .
+       
+\begin{verbatim}
+      DIMENSION A(STRIDE(1), ..., STRIDE(NDIMS))
+\end{verbatim}
+ {\tt NDIMS}~--   , {\tt STRIDE}~--  .
+ ,  {\tt BDIM}~--   ,    $\cal A$         ,         ,      .
+
+        \\{\tt P\_BExchange}  :
+\begin{verbatim}
+      CALL  P_BEXCAHNGE (ARR, NDIMS, STRIDE, BLKLEN, BDIM, 
+     1 OVERLAP, DATATYPE, COMM, PERIOD, IERROR)
+\end{verbatim}
+ {\tt ARR}~--      {\tt A};  {\tt BLKLEN}~--      {\tt A},    ,  ,        {\tt ARR(1:BLKLEN(1), ..., 1:BLKLEN(NDIMS))}.
+     {\tt OVERLAP},          {\tt OVERLAP(1)}   {\tt BDIM},     ~--   {\tt OVERLAP(2)}. 
+  {\tt PERIOD}       . 
+  ,              .
+    MPI {\tt COMM},     {\tt A}   MPI {\tt DATATYPE}.
+,    {\tt MPI\_REAL}  {\tt MPI\_DOUBLE\_PRECISION}.
+%   ,           
+
+         . 
+       :
+\begin{verbatim}
+      CALL P_BEXCAHNGE_INIT (NDIMS, STRIDE, BLKLEN, BDIM, 
+     1 OVERLAP, DATATYPE, COMM, PERIO, BEXCHANGE, IERROR)
+      CALL P_BEXCAHNGE_START (ARR, BEXCHANGE, IERROR)
+      CALL P_BEXCAHNGE_END (BEXCHANGE, IERROR)
+      CALL P_BEXCAHNGE_FREE (BEXCHANGE, IERROR)
+\end{verbatim}
+      ,     ,    ,    {\tt BEXCHANGE},   ,   ,    , , ,      . 
+ ,   {\tt P\_BExcahnge\_start}  {\tt P\_BExchange\_end}       ,    .
+
+\subsubsection{}
+,  $\cal A$        {\tt DIM\_SOURCE}   {\tt LBLKS\_SOURCE(IPROC)},  {\tt IPROC}~--  ,   {\tt DIM\_DEST}    .
+    $\cal A$   ,          {\tt LBLKS\_DEST(IPROC)}   {\tt DIM\_DEST}      {\tt DIM\_SOURCE}.
+ {\tt ARR\_SOURCE}~--      $\cal A$  ,  {\tt ARR\_DEST}~--  ,        :
+\begin{verbatim}
+      CALL P_TRANSPOSE (NDIMS, ARR_SOURCE, DIM_SOURCE, 
+     1 LBLKS_SOURCE, ARR_DEST, DIM_DEST, LBLKS_DEST, 
+     2 STRIDE, BLKLEN, OVERLAP, DATATYPE, COMM, PERIOD, 
+     3 DIAG, IERROR)
+\end{verbatim}
+ {\tt NDIMS}~--    $\cal A$, {\tt DATATYPE}~--  MPI,    , {\tt COMM}~--  MPI,   .
+,    $\cal A$      ,         . 
+ ,      ,  {\tt DIAG}    {\tt .FALSE.}   .
+
+    $\cal A$      {\tt A\_SOURCE},   ~--  {\tt A\_DEST}.
+      
+\begin{verbatim}
+      DIMENSION A_SOURCE(DIM_SOURCE(1), ..., DIM_SOURCE(NDIMS))
+      DIMENSION A_DEST(DIM_DEST(1), ..., DIM_DEST(NDIMS))
+\end{verbatim}
+  {\tt DIM\_SOURCE}    :
+$$
+   \texttt{DIM\_SOURSE(IDIM)}=\left\{
+   \begin{array}{l}
+      \texttt{BLKLEN(DIM\_SOURCE)} \textrm{,  } \texttt{IDIM=DIM\_SOURCE}\\
+      \texttt{STRIDE(IDIM)} \textrm{,   }
+   \end{array}\right.
+$$
+  {\tt DIM\_DEST}:
+$$
+   \texttt{DIM\_DEST(IDIM)}=\left\{
+   \begin{array}{l}
+      \texttt{BLKLEN(DIM\_DEST)} \textrm{,  } \texttt{IDIM=DIM\_DEST}\\
+      \texttt{STRIDE(IDIM)} \textrm{,   }
+   \end{array}\right.
+$$
+
+    {\tt A\_SOURCE}  {\tt A\_DEST}       ,   :
+\begin{verbatim}
+      ARR_SOURCE(1:BLK_SOURCE(1), ..., 1:BLK_SOURCE(NDIMS)
+      ARR_DEST(1:BLK_DEST(1), ..., 1:BLK_DEST(NDIMS)
+\end{verbatim}
+    :
+$$
+   \texttt{BLK\_SOURCE(IDIM)}=\left\{
+   \begin{array}{l}
+      \texttt{LBLKS\_SOURCE(IPROC)} \textrm{,  } \texttt{IDIM=DIM\_SOURCE} \\
+      \sum\texttt{LBLKS\_DEST} \textrm{,  } \texttt{IDIM=DIM\_DEST} \\
+      \texttt{BLKLEN(IDIM)} \textrm{,   }
+   \end{array}\right.
+$$
+$$
+   \texttt{BLK\_DEST(IDIM)}=\left\{
+   \begin{array}{l}
+      \texttt{LBLKS\_DEST(IPROC)} \textrm{,  } \texttt{IDIM=DIM\_DEST} \\
+      \sum\texttt{LBLKS\_SOURE} \textrm{,  } \texttt{IDIM=DIM\_SOURCE} \\
+      \texttt{BLKLEN(IDIM)} \textrm{,   }
+   \end{array}\right.
+$$
+
+       :
+\begin{verbatim}
+      CALL P_TRANSPOSE_INIT (NDIMS, DIM_SOURCE, LBLKS_SOURCE, 
+     1 DIM_DEST, LBLKS_DEST, STRIDE, BLKLEN, OVERLAP, DATATYPE, 
+     2 COMM, PERIOD, DIAG, TRANSP, IERROR)
+      CALL P_TRANSPOSE_START (ARR_SOURCE, ARR_DEST, TRANSP, 
+     1 IERROR)
+      CALL P_TRANSPOSE_END (TRANSP, IERROR)
+      CALL P_TRANSPOSE_FREE (TRANSP, IERROR)
+\end{verbatim}
+ {\tt TRANSP}~--  .
+
+\begin{thebibliography}{}
+\bibitem{T126} ~.~.        T126       . Optimization of Finite Element Approximations, Splines and Wavelets (OFEA'2001). Abstracts of International conference (June~25--29, 2001, St.-Petersburg, Russia), 2001, 184~p.
+
+\bibitem{AGCM} .~A.~Tolstykh, V.~N.~Gloukhov. Implementation of global atmospheric models on parallel computers. {\it  .}    ~, .
+
+\bibitem{Ilghiz} ~.~.       .    .  , , 1999.
+
+\bibitem{Swarztrauber} P.~N.~Swarztrauber, S.~W.~Hammond. A comparison of optimal FFTs on torus and hypercube multicomputers. {\it Parallel computing},~27 (2001), pp.~847--859.
+
+\bibitem{Barros} Barros,~S.~R.~M., Kauranne,~T. On the parallelization of global spectral weather models. {\it Parallel Computing},~20 (1994), pp.~1335--1356.
+
+\bibitem{Rodriguez} B.~Rodriguez, L.~Hart, T.~Henderson. Performance and portability in parallel computing: a weather forecasr view. {\it High Performance Computing in the Geosciences}, 1995, Kluwer Academic Publishers, Netherlands, pp.~1--23.
+
+\bibitem{Henderson} T.~Henderson, D.~Shaffer, M.~Govett, L.~Hart. SMS User's Guide. Advanced Computing Branch, Aviation Division, NOAA Forecast system laboratory, Boulder, 2001.
+
+\end{thebibliography}
+
+\end{document}

ParLib.src/parlib.h

+#ifndef _MPI_INCLUDE
+#include <mpi.h>
+#endif
+
+typedef struct BExchange {
+	int				overlap[2], send[2], recv[2], btype[2]; 
+	int				sendproc[2], recvproc[2], sbind[2], rbind[2];
+	MPI_Comm			comm;
+   MPI_Request		sreq[2], rreq[2]; 
+   MPI_Aint			fsize;
+} BExchange;
+
+typedef struct Transposition {
+	MPI_Datatype *stype, *rtype;
+	int *sbeg, *rbeg;
+	MPI_Comm comm;
+	int nproc, iproc;
+	MPI_Request *sreq, *rreq;
+	MPI_Aint fsize;
+} Transposition;
+
+int P_BExchange_init ( int, int*, int*, int, int*, MPI_Datatype, 
+	MPI_Comm, int, BExchange* );
+int P_BExchange_start ( void*, BExchange* );
+int P_BExchange_end ( BExchange* );
+int P_BExchange_free ( BExchange* );
+int P_BExchange ( void*, int, int*, int*, int, int*, MPI_Datatype,
+	MPI_Comm, int );
+
+int P_Transpose_init ( int , int, int*, int, int*, int*, int*, int*, 
+	MPI_Datatype, MPI_Comm, int, Transposition* );
+int P_Transpose_start ( void*, void*, Transposition* );
+int P_Transpose_end ( Transposition* );
+int P_Transpose_free ( Transposition* );
+int P_Transpose ( int, void*, int, int*, void*, int, int*, int*, int*, 
+	int*, MPI_Datatype, MPI_Comm, int );

ParLib.src/parlibf.c

+#include "parlib.h"
+
+// Additionall calls to make this version 1.1 compliant with version 2.1
+// -------------------------------------------------------------------------- //
+#ifdef FORTRANUNDERSCORE
+void parlib_init_()
+#elif defined(FORTRANDOUBLEUNDERSCORE)
+void parlib_init__()
+#else
+void parlib_init()
+#endif
+{
+}
+
+#ifdef FORTRANUNDERSCORE
+void parlib_deinit_()
+#elif defined(FORTRANDOUBLEUNDERSCORE)
+void parlib_deinit__()
+#else
+void parlib_deinit()
+#endif
+{
+}
+// -------------------------------------------------------------------------- //

ParLib.src/parlibf.h

+      INTEGER HANDLE_SIZE 
+      PARAMETER (HANDLE_SIZE = 2) 

ParLib.src/transpose.c

+#include <stdlib.h>
+#include "parlib.h"
+
+int P_Transpose_init ( ndims, dim_source, lblks_source, dim_dest, 
+		lblks_dest, stride, blklen, overlap, datatype, comm, period,
+		transp )
+	int ndims, dim_source, *lblks_source, dim_dest, *lblks_dest, *stride;
+   int *blklen, *overlap;
+	MPI_Datatype datatype;
+   MPI_Comm comm;
+	int period;
+   Transposition *transp;
+{
+	int idim, nproc, iproc, ip, strd, count;
+	int wblka, wblkb, begb;
+	int ifsta, ifstb, idir, suma, sumb;
+	MPI_Aint fsize;
+	MPI_Datatype oldtype, *stype, *rtype;
+	int *sbeg, *rbeg;
+/*
+ * Check input parameters
+ */
+	if ( ndims < 2) { return 1; }
+	if ( dim_source < 1 || dim_source > ndims ) { return 2; }
+	if ( dim_dest < 1 || dim_dest > ndims ) { return 3; }
+   if ( dim_source == dim_dest ) { return 4; }
+   for ( idim = 0; idim < ndims; idim++ ) {
+		if ( stride[idim] <= 0) { return 5; }
+	}
+   for (idir = 0; idir < 2; idir++ ) {
+		if ( overlap[idir] < 0 ) { return 6; }
+	}
+/*
+ * Define the number of processors in the group and the rank
+ */
+	MPI_Comm_size ( comm, &nproc );
+	if ( nproc == 0 ) { return 0; } 
+	MPI_Comm_rank ( comm, &iproc );
+	if ( iproc == MPI_UNDEFINED ) { return 0; }
+
+	suma = sumb = 0;
+	for ( ip = 0; ip < nproc; ip++ ) {
+		suma += lblks_source[ip];
+		sumb += lblks_dest[ip];
+		if ( lblks_source[ip] <= 0 ) { return 14; }
+		if ( lblks_dest[ip] <= 0 ) { return 15; }
+	}
+
+	if ( lblks_source[iproc] > blklen[dim_source-1] ) { return 8; }
+   if ( lblks_dest[iproc] > blklen[dim_dest-1] ) { return 9; }
+	if ( suma > stride[dim_source-1] ) { return 10; }
+   if ( sumb > stride[dim_dest-1] ) { return 11; }
+	for ( idim = 0; idim < ndims; idim++ ) {
+		if ( idim != dim_source-1 && idim != dim_dest-1 ) {
+			if ( blklen[idim] > stride[idim] ) { return 7; }
+		}
+	}
+	if ( overlap[0] > lblks_dest[0] ) { return 12; }
+	if ( overlap[1] > lblks_dest[nproc-1] ) { return 13; }
+
+	MPI_Type_extent ( datatype, &fsize );
+/*
+ * Allocate memory
+ */
+	stype = transp->stype =
+		(MPI_Datatype *) malloc ( sizeof(MPI_Datatype)*nproc ); 
+	rtype = transp->rtype =
+		(MPI_Datatype *) malloc ( sizeof(MPI_Datatype)*nproc ); 
+	sbeg = transp->sbeg = (int *) malloc ( sizeof(int)*nproc ); 
+	rbeg = transp->rbeg = (int *) malloc ( sizeof(int)*nproc ); 
+/*
+ * Define data types for the blocks and the beginings of the blocks
+ */
+   ifsta = ifstb = 1;
+   for ( ip = 0; ip < nproc; ip++ ) {
+      wblka = lblks_source[iproc];
+      wblkb = lblks_dest[ip];
+      if ( ip > 0 || period ) wblkb += overlap[0];
+      if ( ip < nproc-1 || period ) wblkb += overlap[1];
+
+      oldtype = datatype;
+      strd = 1;
+      for ( idim = 0; idim < ndims; idim++ ) {
+         if ( idim == dim_source-1 ) {
+            count = wblka;
+         } else if ( idim == dim_dest-1 ) {
+            count = wblkb;
+         } else {
+            count = blklen[idim];
+         }
+         MPI_Type_hvector ( count, 1, strd*fsize, oldtype, stype+ip );
+			if ( idim > 0 ) { MPI_Type_free ( &oldtype ); }
+         oldtype = stype[ip];
+         if ( idim == dim_source-1 ) {
+            strd *= blklen[idim];
+         } else {
+            strd *= stride[idim];
+         }
+      }
+      MPI_Type_commit ( stype+ip );
+
+      wblka = lblks_source[ip];
+      wblkb = lblks_dest[iproc];
+      if ( iproc > 0 || period ) wblkb += overlap[0];
+      if ( iproc < nproc-1 || period ) wblkb += overlap[1];
+      oldtype = datatype;
+      strd = 1;
+      for ( idim = 0; idim < ndims; idim++ ) {
+         if ( idim == dim_source-1 ) {
+            count = wblka;
+         } else if ( idim == dim_dest-1 ) {
+            count = wblkb;
+         } else {
+            count = blklen[idim];
+         }
+         MPI_Type_hvector ( count, 1, strd*fsize, oldtype, rtype+ip );
+			if ( idim > 0 ) { MPI_Type_free ( &oldtype ); }
+         oldtype = rtype[ip];
+         if ( idim == dim_dest-1 ) {
+            strd *= blklen[idim];
+         } else {
+            strd *= stride[idim];
+         }
+      }
+      MPI_Type_commit ( rtype+ip );
+
+      begb = ifstb;
+      if ( ip > 0 || period ) begb -= overlap[0];
+      strd = 1;
+      for ( idim = 0; idim < dim_dest-1; idim++ ) {
+         if ( idim == dim_source-1 ) {
+            strd *= blklen[idim];
+         } else {
+            strd *= stride[idim];
+         }
+      }
+      sbeg[ip] = strd*(begb-1);
+
+      rbeg[ip] = 0;
+      if ( iproc > 0 || period ) rbeg[ip] -= overlap[0]*strd;
+
+      strd = 1;
+      for ( idim = 0; idim < dim_source-1; idim++ ) {
+         if ( idim == dim_dest-1 ) {
+            strd *= blklen[idim];
+         } else {
+            strd *= stride[idim];
+         }
+      }
+      rbeg[ip] += strd*(ifsta-1);
+      ifsta += lblks_source[ip];
+      ifstb += lblks_dest[ip];
+	}
+	transp->nproc = nproc;
+	transp->iproc = iproc;
+	transp->comm = comm;
+	transp->fsize = fsize;
+	return 0;
+}
+
+int P_Transpose_start ( arr_source, arr_dest, transp )
+	void *arr_source, *arr_dest;
+	Transposition *transp;
+{
+	char *arr_source_ch = (char *) arr_source;
+	char *arr_dest_ch = (char *) arr_dest;
+	int nproc = transp->nproc;
+	int iproc = transp->iproc;
+	MPI_Aint fsize = transp->fsize;
+	int *sbeg = transp->sbeg;
+	int *rbeg = transp->rbeg;
+	MPI_Datatype *stype = transp->stype;
+	MPI_Datatype *rtype = transp->rtype;
+	MPI_Comm comm = transp->comm;
+	MPI_Request *sreq, *rreq;
+	int ip;
+	char *src, *dest;
+
+   if ( nproc == 0 ) { return 0; }
+   if ( iproc == MPI_UNDEFINED ) { return 0; }
+/*
+ * Allocate memory
+ */
+	sreq = transp->sreq = 
+		(MPI_Request *) malloc ( sizeof(MPI_Request)*nproc );
+	rreq = transp->rreq = 
+		(MPI_Request *) malloc ( sizeof(MPI_Request)*nproc );
+/*
+ * Start the communication
+ */
+   for ( ip = 0; ip < nproc; ip++ ) {
+	   dest = arr_dest_ch+rbeg[ip]*fsize;
+		src = arr_source_ch+sbeg[ip]*fsize;
+		if ( dest == src && iproc == ip ) {
+			rreq[ip] = sreq[ip] = MPI_REQUEST_NULL;
+      } else {
+         MPI_Irecv ( dest, 1, rtype[ip], ip, 0, comm, rreq+ip );
+         MPI_Isend ( src, 1, stype[ip], ip, 0, comm, sreq+ip );
+		}
+	}
+	return 0;
+}
+
+int P_Transpose_end ( transp )
+	Transposition *transp;
+{
+	int nproc = transp->nproc;
+	int iproc = transp->iproc;
+	MPI_Request *sreq = transp->sreq; 
+	MPI_Request *rreq = transp->rreq; 
+	int ip;
+	MPI_Status status;
+
+   if ( nproc == 0 ) { return 0; }
+   if ( iproc == MPI_UNDEFINED ) { return 0; }
+
+   for ( ip = 0; ip < nproc; ip++ ) {
+      MPI_Wait ( rreq+ip, &status );
+      MPI_Wait ( sreq+ip, &status );
+   }
+	return 0;
+}
+
+int P_Transpose_free ( transp )
+	Transposition *transp;
+{
+	int nproc = transp->nproc;
+	int iproc = transp->iproc;
+	int *sbeg = transp->sbeg;
+	int *rbeg = transp->rbeg;
+	MPI_Datatype *stype = transp->stype;
+	MPI_Datatype *rtype = transp->rtype;
+	MPI_Request *sreq = transp->sreq; 
+	MPI_Request *rreq = transp->rreq; 
+	int ip;
+
+   if ( nproc == 0 ) { return 0; }
+   if ( iproc == MPI_UNDEFINED ) { return 0; }
+
+   for ( ip = 0; ip < nproc; ip++ ) {
+      MPI_Type_free ( rtype+ip );
+      MPI_Type_free ( stype+ip );
+   }
+	free ( stype );
+	free ( rtype );
+	free ( sbeg );
+   free ( rbeg );
+	free ( sreq );
+   free ( rreq );
+	return 0;
+}
+
+int P_Transpose ( ndims, arr_source, dim_source, lblks_source, arr_dest, 
+		dim_dest, lblks_dest, stride, blklen, overlap, datatype, comm, 
+		period )
+	void *arr_source, *arr_dest;
+	int ndims, dim_source, *lblks_source, dim_dest, *lblks_dest, *stride; 
+	int *blklen, *overlap;
+	MPI_Datatype datatype;
+   MPI_Comm comm;
+	int period;
+{
+   Transposition transp;
+	int ierr;
+	if ( ierr = P_Transpose_init ( ndims, dim_source, lblks_source, 
+		dim_dest, lblks_dest, stride, blklen, overlap, datatype, comm, 
+		period, &transp ) != 0 ) 
+	{
+		return ierr;
+	}
+	P_Transpose_start ( arr_source, arr_dest, &transp );
+	P_Transpose_end ( &transp );
+	P_Transpose_free ( &transp );
+	return 0;
+}

ParLib.src/transposef.c

+#include <stdlib.h>
+#include "parlib.h"
+
+#ifdef FORTRANUNDERSCORE
+void p_transpose_init_ ( ndims, dim_source, lblks_source, dim_dest, 
+		lblks_dest, stride, blklen, overlap, datatype, comm, period,
+		transp, ierr )
+#elif defined(FORTRANDOUBLEUNDERSCORE)
+void p_transpose_init__ ( ndims, dim_source, lblks_source, dim_dest, 
+		lblks_dest, stride, blklen, overlap, datatype, comm, period,
+		transp, ierr )
+#else
+void p_transpose_init ( ndims, dim_source, lblks_source, dim_dest, 
+		lblks_dest, stride, blklen, overlap, datatype, comm, period,
+		transp, ierr )
+#endif
+	MPI_Fint *ndims, *dim_source, *lblks_source, *dim_dest, *lblks_dest; 
+	MPI_Fint *stride, *blklen, *overlap, *datatype, *comm, *period;
+   MPI_Fint **transp, *ierr;
+{
+	*transp = (MPI_Fint *) malloc ( sizeof(Transposition) );
+	*ierr = P_Transpose_init ( *ndims, *dim_source, lblks_source, 
+		*dim_dest, lblks_dest, stride, blklen, overlap, 
+		MPI_Type_f2c(*datatype), MPI_Comm_f2c(*comm), *period,
+		(Transposition *) *transp );
+}
+
+#ifdef FORTRANUNDERSCORE
+void p_transpose_start_ ( arr_source, arr_dest, transp, ierr )
+#elif defined(FORTRANDOUBLEUNDERSCORE)
+void p_transpose_start__ ( arr_source, arr_dest, transp, ierr )
+#else
+void p_transpose_start ( arr_source, arr_dest, transp, ierr )
+#endif
+	void *arr_source, *arr_dest;
+	MPI_Fint **transp, *ierr;
+{
+   *ierr = P_Transpose_start ( arr_source, arr_dest, 
+		(Transposition *) *transp );
+}
+
+#ifdef FORTRANUNDERSCORE
+void p_transpose_end_ ( transp, ierr )
+#elif defined(FORTRANDOUBLEUNDERSCORE)
+void p_transpose_end__ ( transp, ierr )
+#else
+void p_transpose_end ( transp, ierr )
+#endif
+	MPI_Fint **transp, *ierr;
+{
+	*ierr = P_Transpose_end ( (Transposition *) *transp );
+}
+
+#ifdef FORTRANUNDERSCORE
+void p_transpose_free_ ( transp, ierr )
+#elif defined(FORTRANDOUBLEUNDERSCORE)
+void p_transpose_free__ ( transp, ierr )
+#else
+void p_transpose_free ( transp, ierr )
+#endif
+	MPI_Fint **transp, *ierr;
+{
+	*ierr = P_Transpose_free ( (Transposition *) *transp );
+	free ( *transp ); 
+}
+
+#ifdef FORTRANUNDERSCORE
+void p_transpose_ ( ndims, arr_source, dim_source, lblks_source, arr_dest,
+		dim_dest, lblks_dest, stride, blklen, overlap, datatype, comm, 
+		period, ierr )
+#elif defined(FORTRANDOUBLEUNDERSCORE)
+void p_transpose__ ( ndims, arr_source, dim_source, lblks_source, arr_dest,
+		dim_dest, lblks_dest, stride, blklen, overlap, datatype, comm, 
+		period, ierr )
+#else
+void p_transpose ( ndims, arr_source, dim_source, lblks_source, arr_dest, 
+		dim_dest, lblks_dest, stride, blklen, overlap, datatype, comm, 
+		period, ierr )
+#endif
+	void *arr_source, *arr_dest;
+	MPI_Fint *ndims, *dim_source, *lblks_source, *dim_dest, *lblks_dest; 
+	MPI_Fint *stride, *blklen, *overlap, *datatype, *comm, *period;
+	MPI_Fint *ierr;
+{
+	*ierr = P_Transpose ( *ndims, arr_source, *dim_source, lblks_source, 
+		arr_dest, *dim_dest, lblks_dest, stride, blklen, overlap, 
+		MPI_Type_f2c(*datatype), MPI_Comm_f2c(*comm), *period );
+}