- petsc-dev - mailman3.cels.anl.gov

Re: Bitkeeper going commercial-only
by Satish Balay 12 Apr '05

12 Apr '05

On Tue, 12 Apr 2005, Matthew Knepley wrote: > They have promised us free licenses. Thats good to know. Is this something new? > In fact, its pretty easy for any academic, open source project to > get a free license. It turns out Victor already had one for 2 years > now. Victor uses the 'free' version [bk-3.2.3] with 'academic licence'. From what I understand all free versions will stop working. Only 'Bk Pro' will work after June 1st. So I would think that any new free academic/research licenses would be off Bk Pro'. Satish

1 0

Re: Bitkeeper going commercial-only
by Matthew Knepley 12 Apr '05

12 Apr '05

Aron Ahmadia <aron.ahmadia(a)gmail.com> writes: > So this is really crappy news, I really liked BK. How is this going to > affect PETSc and what are your future short-term and long-term plans? > > I have to figure out what I'm doing with COW, and I was wondering how you > guys were dealing with this. They have promised us free licenses. In fact, its pretty easy for any academic, open source project to get a free license. It turns out Victor already had one for 2 years now. My guess is once all the Linux kernel assholes stop caring about BK, it will go back to being completely free. <Commence beating numbskull, ignorant kernel programmers> Matt -- "Failure has a thousand explanations. Success doesn't need one" -- Sir Alec Guiness

1 0

Bitkeeper going commercial-only
by Aron Ahmadia 12 Apr '05

12 Apr '05

So this is really crappy news, I really liked BK. How is this going to affect PETSc and what are your future short-term and long-term plans? I have to figure out what I'm doing with COW, and I was wondering how you guys were dealing with this. ~Aron

1 0

Re: Adding Harwell-Boeing support in PETSc?
by Hong Zhang 07 Apr '05

07 Apr '05

Victor, > > I don't know if Roldan ever updated that. Obviously, he hasn't update until this morning. I've put your mmio.c in petsc repository to be used with mm2petsc.c Thanks, Hong

1 0

Re: Adding Harwell-Boeing support in PETSc?
by Victor Eijkhout 07 Apr '05

07 Apr '05

At 9:40 AM -0500 4/7/05, Hong Zhang wrote: >It seems the bug is on when to raise error flag? Yes. Here's my email exchange with Roldan Pozo: At 4:03 PM -0400 5/7/03, pozo(a)nist.gov wrote: >Actually, the ANSI standard says it returns the number of characters >"transmitted" (Sec. 4.9.6.1), which may or may not be the same as the >number of characters printed. In any case, you are right: the code >as it appears will return a pessimistic error code. I guess I was >thinking of fscanf, which returns the number of items succesfully >assigned. (So much for symmetry in function interface design...) > >The offending line should be replaced with "if (ret_code < 0 )". >Thanks for pointing this out, Victor. > >Regards, >Roldan > >---------------------------------------------------------------- >Quoting Victor Eijkhout <eijkhout(a)cs.utk.edu>: > >> >> I don't see how this could ever work: >> >> int mm_write_banner(FILE *f, MM_typecode matcode) >> { >> char *str = mm_typecode_to_str(matcode); >> int ret_code; >> >> ret_code = fprintf(f, "%s %s\n", MatrixMarketBanner, str); >> free(str); >> if (ret_code !=2 ) >> return MM_COULD_NOT_WRITE_FILE; >> else >> return 0; >> } >> >> >> According to my copy of the C standard (ok, the reviled >> Schildt-annotated edition) fprintf returns the number of characters > > written. How can that be 2? and later At 11:26 PM -0400 5/7/03, pozo(a)nist.gov wrote: > > >The offending line should be replaced with "if (ret_code < 0 )". >> >> And the same goes for other write functions. I had to replace this in >> at least one other location. > >Right. (I mean, "write.") > >I'll update the files on the web site. I don't know if Roldan ever updated that. -- Victor Eijkhout <eijkhout(a)cs.utk.edu>, 329 Claxton, Comp Sci, UT, Knoxville TN 37996. tel: 865 974 9308 (W), 865 673 6998 (H), 865 974 8296 (F) http://www.cs.utk.edu/~eijkhout/

1 0

Re: Adding Harwell-Boeing support in PETSc?
by Hong Zhang 07 Apr '05

07 Apr '05

Victor, The mmio.c I sent to you was obtained few years ago. Your mmio.c diff from the current mmio.c on diff Victor_mmio.c mmio.c 10d9 < #include <stdlib.h> 12a12 > #include <malloc.h> 251c251 < if (fprintf(f, "%d %d\n", M, N) < 0) --- > if (fprintf(f, "%d %d\n", M, N) != 2) 393c393 < if (ret_code < 0 ) --- > if (ret_code !=2 ) 441c441,453 < --- > > > /** > * Create a new copy of a string s. strdup() is a common routine, but > * not part of ANSI C, so it is included here. Used by mm_typecode_to_str(). > * > */ > char *strdup(const char *s) > { > int len = strlen(s); > char *s2 = (char *) malloc((len+1)*sizeof(char)); > return strcpy(s2, s); > } 446a459 > char *strdup(const char *); It seems the bug is on when to raise error flag? I have been using several matrices converted from matrix market using their mmio.c. I just want make sure the bug has not affected the correctness of my converted matrices. I'll put your mmio.c in petsc repository. Thanks, Hong On Thu, 7 Apr 2005, Victor Eijkhout wrote: > At 1:41 PM -0500 4/5/05, Hong Zhang wrote: > >The files mm2petsc.c, mmio.c and mmio.h are attached. > > Years ago I found a couple of bugs in mmio.c and reported them to > Roldan Pozo. Apparently they haven't been fixed. The good boy is too > busy with other things. > -- > Victor Eijkhout <eijkhout(a)cs.utk.edu>, 329 Claxton, Comp Sci, UT, > Knoxville TN 37996. > tel: 865 974 9308 (W), 865 673 6998 (H), 865 974 8296 (F) > http://www.cs.utk.edu/~eijkhout/

1 0

Re: Adding Harwell-Boeing support in PETSc?
by Victor Eijkhout 06 Apr '05

06 Apr '05

At 1:41 PM -0500 4/5/05, Hong Zhang wrote: >The files mm2petsc.c, mmio.c and mmio.h are attached. Years ago I found a couple of bugs in mmio.c and reported them to Roldan Pozo. Apparently they haven't been fixed. The good boy is too busy with other things. -- Victor Eijkhout <eijkhout(a)cs.utk.edu>, 329 Claxton, Comp Sci, UT, Knoxville TN 37996. tel: 865 974 9308 (W), 865 673 6998 (H), 865 974 8296 (F) http://www.cs.utk.edu/~eijkhout/

1 0

Re: Adding Harwell-Boeing support in PETSc?
by Barry Smith 05 Apr '05

05 Apr '05

Rich, You could certainly write a "converter" in Fortran if that is easier that then just does the usual MatView() to save it in PETSc binary format for later loading by a PETSc solver code like src/ksp/examples/tutorials. If you do we could add it to src/mat/examples/tests Barry Writing a parallel loader just seems like a huge waste of time, over just reading the ascii in sequentially and then saving as a Petsc matrix for later reading in in parallel. On Tue, 5 Apr 2005, Richard Tran Mills wrote: > Barry Smith wrote: > > My understanding is that MatrixMarket format is different from > > Harwell-Boeing? > > Yes, they are different. MatrixMarket format is a coordinate format that is > considerably easier to parse than Harwell-Boeing, which was really designed > with Fortran in mind. The matrices available at Matrix Market are usually > available in both formats. > > > I could not find a Harwell-Boeing loader in src/mat/examples/tests > > (though there is a writer). Does anyone have C (PETSc?) code to read > > in Harwell-Boeing? If so we could merge it into the MatrixMarket (coordiate > > storage) reader code and then (with a command line switch) read in either > > format and dump in PETSc format? > > There are some freely distributable and modifiable C routines for working with > HB files available at > > http://math.nist.gov/~KRemington/harwell_io/harwell_io.html > > Incorporating some of these routines into a PETSc viewer seems like the best > bet. I don't know of anything else in C. I have usually used Saad's SPARSKIT > Fortran routines for working with Harwell-Boeing matrices. > > Cheers, > Richard > >

1 0

Re: Adding Harwell-Boeing support in PETSc?
by Hong Zhang 05 Apr '05

05 Apr '05

Barry, I get Harwell-Boeing matrices from Matrix Market repository, where each HB matrix is provided by both Matrix Market and HB format. HB format takes smaller space though. I never used HB format. I wrote the code that convert MM format into petsc binary, which is stored under ~/datafiles/matrices/MtxMarket. Hong On Tue, 5 Apr 2005, Barry Smith wrote: > > My understanding is that MatrixMarket format is different from > Harwell-Boeing? > > I could not find a Harwell-Boeing loader in src/mat/examples/tests > (though there is a writer). Does anyone have C (PETSc?) code to read > in Harwell-Boeing? If so we could merge it into the MatrixMarket > (coordiate storage) reader code and then (with a command line switch) read in > either format and dump in PETSc format? > > Barry > > > On Tue, 5 Apr 2005, Victor Eijkhout wrote: > > > At 10:32 AM -0500 4/5/05, Hong Zhang wrote: > > > We have code that read MatrixMarket format, and write > > > into a petsc binary file. Then you can use petsc MatLoad() > > > for viewing or computation. > > > > I'm interested in this too. I wrote my own version. Maybe a good idea to add > > this to the petsc distribution? > > > >

1 0

Re: Adding Harwell-Boeing support in PETSc?
by Hong Zhang 05 Apr '05

05 Apr '05

Victor and Richard, The files mm2petsc.c, mmio.c and mmio.h are attached. I just update it with petsc-dev (petsc-2.2.1 should be fine). Let me know if you see problem/bug. If you further improve it, please give me your improvement as well. Thanks, Hong On Tue, 5 Apr 2005, Victor Eijkhout wrote: > At 10:32 AM -0500 4/5/05, Hong Zhang wrote: > >We have code that read MatrixMarket format, and write > >into a petsc binary file. Then you can use petsc MatLoad() > >for viewing or computation. > > I'm interested in this too. I wrote my own version. Maybe a good idea > to add this to the petsc distribution? > -- > Victor Eijkhout <eijkhout(a)cs.utk.edu>, 329 Claxton, Comp Sci, UT, > Knoxville TN 37996. > tel: 865 974 9308 (W), 865 673 6998 (H), 865 974 8296 (F) > http://www.cs.utk.edu/~eijkhout/ > >

1 0