May 2013 - devel - mailman3.cels.anl.gov

Re: [mpich-devel] "create new ticket" text
by Ken Raffenetti 30 May '13

30 May '13

Mike, I have updated the text on this page to reflect the new names/locations. Ken ----- Original Message ----- > From: "Michael Blocksome" <blocksom(a)us.ibm.com> > To: devel(a)mpich.org > Sent: Wednesday, May 29, 2013 2:01:38 PM > Subject: [mpich-devel] "create new ticket" text > > FYI, the text on the trac page ... > > https://trac.mpich.org/projects/mpich/newticket > > ... talks about "MPICH2" and "mpich-discuss(a)mcs.anl.org". I think > these terms have migrated away by now. > > Michael Blocksome > Blue Gene Messaging > blocksom(a)us.ibm.com >

1 0

Re: [mpich-devel] MPIR_ERRTEST_ARGNULL and count=0 (or type_size=0)
by Jeff Hammond 30 May '13

30 May '13

The check for NO_OP in *get_accumulate was most certainly present. That was the first thing I checked when MPI_Get_accumulate(MPI_REPLACE) failed to behave like MPI_Get w.r.t. ARMCI-MPI test codes. I'm going to supplement the RMA test suite so that this issue and ones like it would have been caught by MPICH rather than waiting until the very last ARMCI-MPI test to be revealed. Jeff On Thu, May 30, 2013 at 4:53 AM, Jim Dinan <james.dinan(a)gmail.com> wrote: > Ugh, deep shame upon me and whoever reviewed my patch! Lisandro is > correct. :) > > For accumulate ops, the op can also cause a buffer to be ignored (e.g. > GACC/FOP with NO_OP). I think there is code to handle this, but keep an > eye out for it. > > ~Jim. > > On 05/29/2013 03:08 PM, Lisandro Dalcin wrote: >> On 29 May 2013 22:49, Jeff Hammond <jhammond(a)alcf.anl.gov> wrote: >>> It seems that I didn't understand MPI_BOTTOM completely, so the error >>> checks needed to be different than described below, but the high-level >>> issues are the same. >>> >>> Interested parties may consult >>> http://trac.mpich.org/projects/mpich/ticket/1863 for details and >>> further discussion. >>> >> >> Please take a look at the macro MPIR_ERRTEST_USERBUFFER . >> >> >> >> -- >> Lisandro Dalcin >> --------------- >> CIMEC (INTEC/CONICET-UNL) >> Predio CONICET-Santa Fe >> Colectora RN 168 Km 472, Paraje El Pozo >> 3000 Santa Fe, Argentina >> Tel: +54-342-4511594 (ext 1011) >> Tel/Fax: +54-342-4511169 >> > -- Jeff Hammond Argonne Leadership Computing Facility University of Chicago Computation Institute jhammond(a)alcf.anl.gov / (630) 252-5381 http://www.linkedin.com/in/jeffhammond https://wiki.alcf.anl.gov/parts/index.php/User:Jhammond ALCF docs: http://www.alcf.anl.gov/user-guides

1 0

Re: [mpich-devel] MPIR_ERRTEST_ARGNULL and count=0 (or type_size=0)
by Jim Dinan 30 May '13

30 May '13

Ugh, deep shame upon me and whoever reviewed my patch! Lisandro is correct. :) For accumulate ops, the op can also cause a buffer to be ignored (e.g. GACC/FOP with NO_OP). I think there is code to handle this, but keep an eye out for it. ~Jim. On 05/29/2013 03:08 PM, Lisandro Dalcin wrote: > On 29 May 2013 22:49, Jeff Hammond <jhammond(a)alcf.anl.gov> wrote: >> It seems that I didn't understand MPI_BOTTOM completely, so the error >> checks needed to be different than described below, but the high-level >> issues are the same. >> >> Interested parties may consult >> http://trac.mpich.org/projects/mpich/ticket/1863 for details and >> further discussion. >> > > Please take a look at the macro MPIR_ERRTEST_USERBUFFER . > > > > -- > Lisandro Dalcin > --------------- > CIMEC (INTEC/CONICET-UNL) > Predio CONICET-Santa Fe > Colectora RN 168 Km 472, Paraje El Pozo > 3000 Santa Fe, Argentina > Tel: +54-342-4511594 (ext 1011) > Tel/Fax: +54-342-4511169 >

1 0

Re: [mpich-devel] MPIR_ERRTEST_ARGNULL and count=0 (or type_size=0)
by Jeff Hammond 29 May '13

29 May '13

I don't know if you read http://trac.mpich.org/projects/mpich/ticket/1863, but the RMA code in the repo is using MPIR_ERRTEST_ARGNULL in a few places, which caused me problems in ARMCI-MPI. Most of the RMA functions, in particular the ones that existed in MPI-2, were missing any buffer checks. The buffer checks that were present were added to the RMW routines. My patch is relative to remotes/origin/HEAD. "I don't always build MPI from source, but when I do, I use the latest development version of MPICH" </dos equis mem> Best, Jeff On Wed, May 29, 2013 at 4:14 PM, Lisandro Dalcin <dalcinl(a)gmail.com> wrote: > On 29 May 2013 23:55, Jeff Hammond <jhammond(a)alcf.anl.gov> wrote: >> Well that's no fun. I was hoping to make it look like I did some >> actual work here :-) >> > > The actual work is figuring out the issue!!! I've spent enough time > with these bugs in other MPI implementations. As I know by testing > myself that MPICH does the right thing for MPI 2.x calls, I guesses > you were missing a more easy way to do it. > > BTW, check this: > > $ grep -r MPIR_ERRTEST_USERBUFFER src/mpi/rma > > I get no output, I think other RMA calls also need better error > checking of buffer addresses... > > > -- > Lisandro Dalcin > --------------- > CIMEC (INTEC/CONICET-UNL) > Predio CONICET-Santa Fe > Colectora RN 168 Km 472, Paraje El Pozo > 3000 Santa Fe, Argentina > Tel: +54-342-4511594 (ext 1011) > Tel/Fax: +54-342-4511169 -- Jeff Hammond Argonne Leadership Computing Facility University of Chicago Computation Institute jhammond(a)alcf.anl.gov / (630) 252-5381 http://www.linkedin.com/in/jeffhammond https://wiki.alcf.anl.gov/parts/index.php/User:Jhammond ALCF docs: http://www.alcf.anl.gov/user-guides

1 0

Re: [mpich-devel] MPIR_ERRTEST_ARGNULL and count=0 (or type_size=0)
by Lisandro Dalcin 29 May '13

29 May '13

On 29 May 2013 23:55, Jeff Hammond <jhammond(a)alcf.anl.gov> wrote: > Well that's no fun. I was hoping to make it look like I did some > actual work here :-) > The actual work is figuring out the issue!!! I've spent enough time with these bugs in other MPI implementations. As I know by testing myself that MPICH does the right thing for MPI 2.x calls, I guesses you were missing a more easy way to do it. BTW, check this: $ grep -r MPIR_ERRTEST_USERBUFFER src/mpi/rma I get no output, I think other RMA calls also need better error checking of buffer addresses... -- Lisandro Dalcin --------------- CIMEC (INTEC/CONICET-UNL) Predio CONICET-Santa Fe Colectora RN 168 Km 472, Paraje El Pozo 3000 Santa Fe, Argentina Tel: +54-342-4511594 (ext 1011) Tel/Fax: +54-342-4511169

1 0

Re: [mpich-devel] MPIR_ERRTEST_ARGNULL and count=0 (or type_size=0)
by Jeff Hammond 29 May '13

29 May '13

Well that's no fun. I was hoping to make it look like I did some actual work here :-) Thanks, Jeff On Wed, May 29, 2013 at 3:08 PM, Lisandro Dalcin <dalcinl(a)gmail.com> wrote: > On 29 May 2013 22:49, Jeff Hammond <jhammond(a)alcf.anl.gov> wrote: >> It seems that I didn't understand MPI_BOTTOM completely, so the error >> checks needed to be different than described below, but the high-level >> issues are the same. >> >> Interested parties may consult >> http://trac.mpich.org/projects/mpich/ticket/1863 for details and >> further discussion. >> > > Please take a look at the macro MPIR_ERRTEST_USERBUFFER . > > > > -- > Lisandro Dalcin > --------------- > CIMEC (INTEC/CONICET-UNL) > Predio CONICET-Santa Fe > Colectora RN 168 Km 472, Paraje El Pozo > 3000 Santa Fe, Argentina > Tel: +54-342-4511594 (ext 1011) > Tel/Fax: +54-342-4511169 -- Jeff Hammond Argonne Leadership Computing Facility University of Chicago Computation Institute jhammond(a)alcf.anl.gov / (630) 252-5381 http://www.linkedin.com/in/jeffhammond https://wiki.alcf.anl.gov/parts/index.php/User:Jhammond ALCF docs: http://www.alcf.anl.gov/user-guides

1 0

Re: [mpich-devel] MPIR_ERRTEST_ARGNULL and count=0 (or type_size=0)
by Lisandro Dalcin 29 May '13

29 May '13

On 29 May 2013 22:49, Jeff Hammond <jhammond(a)alcf.anl.gov> wrote: > It seems that I didn't understand MPI_BOTTOM completely, so the error > checks needed to be different than described below, but the high-level > issues are the same. > > Interested parties may consult > http://trac.mpich.org/projects/mpich/ticket/1863 for details and > further discussion. > Please take a look at the macro MPIR_ERRTEST_USERBUFFER . -- Lisandro Dalcin --------------- CIMEC (INTEC/CONICET-UNL) Predio CONICET-Santa Fe Colectora RN 168 Km 472, Paraje El Pozo 3000 Santa Fe, Argentina Tel: +54-342-4511594 (ext 1011) Tel/Fax: +54-342-4511169

1 0

Re: [mpich-devel] MPIR_ERRTEST_ARGNULL and count=0 (or type_size=0)
by Jeff Hammond 29 May '13

29 May '13

It seems that I didn't understand MPI_BOTTOM completely, so the error checks needed to be different than described below, but the high-level issues are the same. Interested parties may consult http://trac.mpich.org/projects/mpich/ticket/1863 for details and further discussion. Jeff On Wed, May 29, 2013 at 7:41 AM, Jeff Hammond <jhammond(a)alcf.anl.gov> wrote: > I see Jim used MPIR_ERRTEST_ARGNULL in (r)get_accumulate, which makes > sense (I am working on verifying that they should be in the others as > well). However, these calls are not guarded by "if (count)", which > means that one presumably (I have not written a test for it but it > seems obvious) cannot use addr=NULL with count=0. It seems to me that > 0-byte transfers are find with NULL buffers (this was recently > verified for MPI_Ssend because it's a common way to implement p2p > synchronization). > > It seems prudent to change something like this: > > MPIR_ERRTEST_COUNT(origin_count, mpi_errno); > MPIR_ERRTEST_DATATYPE(origin_datatype, "origin_datatype", > mpi_errno); > MPIR_ERRTEST_ARGNULL(origin_addr, "origin_addr", mpi_errno); > > To this: > > MPIR_ERRTEST_COUNT(origin_count, mpi_errno); > MPIR_ERRTEST_DATATYPE(origin_datatype, "origin_datatype", > mpi_errno); > if (origin_count) MPIR_ERRTEST_ARGNULL(origin_addr, > "origin_addr", mpi_errno); > > Then I got to thinking, what if someone - almost certainly not > intentionally but because of a general implementation - uses > MPI_Type_vector to create a zero-length vector type. I don't see why > one can't do a count>0 transfer with with a NULL buffer if the > datatype is empty. > > I wrote the following test of a 0-length vector and see that it works > just fine with bcast. > > #include <stdio.h> > #include <stdlib.h> > > #include <mpi.h> > > int main(int argc, char ** argv) { > int rank, nproc; > MPI_Datatype emptyvec; > MPI_Aint lb, ext; > > MPI_Init(&argc, &argv); > > MPI_Comm_rank(MPI_COMM_WORLD, &rank); > MPI_Comm_size(MPI_COMM_WORLD, &nproc); > > if (rank == 0) printf("Starting MPI datatype test with %d > processes\n", nproc); > > MPI_Type_vector(0,0,0,MPI_DOUBLE,&emptyvec); > MPI_Type_commit(&emptyvec); > MPI_Type_get_extent(emptyvec, &lb, &ext); > MPI_Type_size(emptyvec, &size); > if (rank == 0) printf("lb = %zu ext = %zu size = %d \n", lb, ext, size); > > MPI_Bcast(NULL, 0, emptyvec, 0, MPI_COMM_WORLD); > MPI_Bcast(NULL, 1, emptyvec, 0, MPI_COMM_WORLD); > > MPI_Type_free(&emptyvec); > > if (rank == 0) printf("Test complete: PASS.\n"); > > MPI_Finalize(); > > return 0; > } > > This suggests that it should be valid to use addr=NULL and count>0 as > long as the extent of the type is 0, hence the aforementioned code > should really be: > > int origin_datatype_size; > ... > MPIR_ERRTEST_COUNT(origin_count, mpi_errno); > MPIR_ERRTEST_DATATYPE(origin_datatype, "origin_datatype", > mpi_errno); > MPID_Datatype_get_size_macro(origin_datatype, origin_datatype_size); > if (origin_count && origin_datatype_size) > MPIR_ERRTEST_ARGNULL(origin_addr, "origin_addr", mpi_errno); > > Do people agree with this reasoning? > > Thanks, > > Jeff > > -- > Jeff Hammond > Argonne Leadership Computing Facility > University of Chicago Computation Institute > jhammond(a)alcf.anl.gov / (630) 252-5381 > http://www.linkedin.com/in/jeffhammond > https://wiki.alcf.anl.gov/parts/index.php/User:Jhammond > ALCF docs: http://www.alcf.anl.gov/user-guides -- Jeff Hammond Argonne Leadership Computing Facility University of Chicago Computation Institute jhammond(a)alcf.anl.gov / (630) 252-5381 http://www.linkedin.com/in/jeffhammond https://wiki.alcf.anl.gov/parts/index.php/User:Jhammond ALCF docs: http://www.alcf.anl.gov/user-guides

1 0

"create new ticket" text
by Michael Blocksome 29 May '13

29 May '13

FYI, the text on the trac page ... https://trac.mpich.org/projects/mpich/newticket ... talks about "MPICH2" and "mpich-discuss(a)mcs.anl.org". I think these terms have migrated away by now. Michael Blocksome Blue Gene Messaging blocksom(a)us.ibm.com

1 0

MPIR_ERRTEST_ARGNULL and count=0 (or type_size=0)
by Jeff Hammond 29 May '13

29 May '13

I see Jim used MPIR_ERRTEST_ARGNULL in (r)get_accumulate, which makes sense (I am working on verifying that they should be in the others as well). However, these calls are not guarded by "if (count)", which means that one presumably (I have not written a test for it but it seems obvious) cannot use addr=NULL with count=0. It seems to me that 0-byte transfers are find with NULL buffers (this was recently verified for MPI_Ssend because it's a common way to implement p2p synchronization). It seems prudent to change something like this: MPIR_ERRTEST_COUNT(origin_count, mpi_errno); MPIR_ERRTEST_DATATYPE(origin_datatype, "origin_datatype", mpi_errno); MPIR_ERRTEST_ARGNULL(origin_addr, "origin_addr", mpi_errno); To this: MPIR_ERRTEST_COUNT(origin_count, mpi_errno); MPIR_ERRTEST_DATATYPE(origin_datatype, "origin_datatype", mpi_errno); if (origin_count) MPIR_ERRTEST_ARGNULL(origin_addr, "origin_addr", mpi_errno); Then I got to thinking, what if someone - almost certainly not intentionally but because of a general implementation - uses MPI_Type_vector to create a zero-length vector type. I don't see why one can't do a count>0 transfer with with a NULL buffer if the datatype is empty. I wrote the following test of a 0-length vector and see that it works just fine with bcast. #include <stdio.h> #include <stdlib.h> #include <mpi.h> int main(int argc, char ** argv) { int rank, nproc; MPI_Datatype emptyvec; MPI_Aint lb, ext; MPI_Init(&argc, &argv); MPI_Comm_rank(MPI_COMM_WORLD, &rank); MPI_Comm_size(MPI_COMM_WORLD, &nproc); if (rank == 0) printf("Starting MPI datatype test with %d processes\n", nproc); MPI_Type_vector(0,0,0,MPI_DOUBLE,&emptyvec); MPI_Type_commit(&emptyvec); MPI_Type_get_extent(emptyvec, &lb, &ext); MPI_Type_size(emptyvec, &size); if (rank == 0) printf("lb = %zu ext = %zu size = %d \n", lb, ext, size); MPI_Bcast(NULL, 0, emptyvec, 0, MPI_COMM_WORLD); MPI_Bcast(NULL, 1, emptyvec, 0, MPI_COMM_WORLD); MPI_Type_free(&emptyvec); if (rank == 0) printf("Test complete: PASS.\n"); MPI_Finalize(); return 0; } This suggests that it should be valid to use addr=NULL and count>0 as long as the extent of the type is 0, hence the aforementioned code should really be: int origin_datatype_size; ... MPIR_ERRTEST_COUNT(origin_count, mpi_errno); MPIR_ERRTEST_DATATYPE(origin_datatype, "origin_datatype", mpi_errno); MPID_Datatype_get_size_macro(origin_datatype, origin_datatype_size); if (origin_count && origin_datatype_size) MPIR_ERRTEST_ARGNULL(origin_addr, "origin_addr", mpi_errno); Do people agree with this reasoning? Thanks, Jeff -- Jeff Hammond Argonne Leadership Computing Facility University of Chicago Computation Institute jhammond(a)alcf.anl.gov / (630) 252-5381 http://www.linkedin.com/in/jeffhammond https://wiki.alcf.anl.gov/parts/index.php/User:Jhammond ALCF docs: http://www.alcf.anl.gov/user-guides

1 0