https://software.intel.com/content/www/us/en/develop/documentation/mpi-developer-guide-linux/top/additional-supported-features/asynchronous-progress-control.html

It states "and a partial support for non-blocking collectives ( MPI_Ibcas t, MPI_Ireduce , and MPI_Iallreduce )."  I do not know what partial support means but you can try setting the variables and see if that helps.

On Jan 22, 2021, at 11:20 AM, Viet H.Q.H. <hqhviet@tohoku.ac.jp> wrote:

Dear Victor and Berry,

Thank you so much for your answers.

I fixed the code with the bug in the PetscCommSplitReductionBegin function as commented by Brave.

     ierr = PetscCommSplitReductionBegin (PetscObjectComm ((PetscObject) u));

It was also a mistake to set the vector size too small.
I just set a vector size of 100000000 and ran the code on 4 nodes with 2 processors per node. The result is as follows

The time used for the asynchronous calculation: 0.022043
+ | u | = 10000.
The time used for the synchronous calculation: 0.016188
+ | b | = 10000.

Asynchronous computation still takes a longer time.

I also confirmed that PETSC_HAVE_MPI_IALLREDUCE is defined in the file $PETSC_DIR/include/petscconf.h

I built Petsc by using the following script

#!/usr/bin/bash
set -e
DATE="21.01.18"
MPIIT_DIR="/work/A/intel/2018_update2/compilers_and_libraries_2018.2.199/linux/mpi/intel64"
MKL_DIR="/work/A/intel/2018_update2/compilers_and_libraries_2018.2.199/linux/mkl"
INSTL_DIR="${HOME}/local/petsc-3.14.3"
BUILD_DIR="${HOME}/tmp/petsc/build_${DATE}"
PETSC_DIR="${HOME}/tmp/petsc"

cd ${PETSC_DIR}
./configure --force --prefix=${INSTL_DIR} --with-mpi-dir=${MPIIT_DIR}  --with-fortran-bindings=0 --with-mpiexe=${MPIIT_DIR}/bin/mpiexec --with-valgrind-dir=${HOME}/local/valgrind --with-blaslapack-dir=${MKL_DIR} --download-make --with-debugging=0 COPTFLAGS='-O3 -march=native -mtune=native' CXXOPTFLAGS='-O3 -march=native -mtune=native' FOPTFLAGS='-O3 -march=native -mtune=native'

make PETSC_DIR=${HOME}/tmp/petsc PETSC_ARCH=arch-linux2-c-opt all
make PETSC_DIR=${HOME}/tmp/petsc PETSC_ARCH=arch-linux2-c-opt install 

Intel 2018 also complies with the MPI-3 standard.

Are there specific settings for Intel MPI to obtain the performance of the MPI_IALLREDUCE function?

Sincerely,

Viet.

On Fri, Jan 22, 2021 at 11:20 AM Barry Smith <bsmith@petsc.dev> wrote:

  ierr = VecNormBegin(u,NORM_2,&norm1);
    ierr = PetscCommSplitReductionBegin(PetscObjectComm((PetscObject)Ax)); 

How come you call this on Ax and not on u? For clarity, if nothing else, I think you should call it on u.

comb.c has 

/*

      Split phase global vector reductions with support for combining the

   communication portion of several operations. Using MPI-1.1 support only

      The idea for this and much of the initial code is contributed by

   Victor Eijkhout.

       Usage:

             VecDotBegin(Vec,Vec,PetscScalar *);

             VecNormBegin(Vec,NormType,PetscReal *);

             ....

             VecDotEnd(Vec,Vec,PetscScalar *);

             VecNormEnd(Vec,NormType,PetscReal *);

       Limitations:

         - The order of the xxxEnd() functions MUST be in the same order

           as the xxxBegin(). There is extensive error checking to try to

           insure that the user calls the routines in the correct order

*/

#include <petsc/private/vecimpl.h>    /*I   "petscvec.h"    I*/

static PetscErrorCode MPIPetsc_Iallreduce(void *sendbuf,void *recvbuf,PetscMPIInt count,MPI_Datatype datatype,MPI_Op op,MPI_Comm comm,MPI_Request *request)

{

  PETSC_UNUSED PetscErrorCode ierr;

  PetscFunctionBegin;

#if defined(PETSC_HAVE_MPI_IALLREDUCE)

  ierr = MPI_Iallreduce(sendbuf,recvbuf,count,datatype,op,comm,request);CHKERRMPI(ierr);

#elif defined(PETSC_HAVE_MPIX_IALLREDUCE)

  ierr = MPIX_Iallreduce(sendbuf,recvbuf,count,datatype,op,comm,request);CHKERRQ(ierr);

#else

  ierr = MPIU_Allreduce(sendbuf,recvbuf,count,datatype,op,comm);CHKERRQ(ierr);

  *request = MPI_REQUEST_NULL;

#endif

  PetscFunctionReturn(0);

}

So first check if $PETSC_DIR/include/petscconf.h has 

PETSC_HAVE_MPI_IALLREDUCE

if it does not then the standard MPI reduce is called. 

If this is set then any improvement depends on the implementation of iallreduce inside the MPI you are using. 

Barry

On Jan 21, 2021, at 6:52 AM, Viet H.Q.H. <hqhviet@tohoku.ac.jp> wrote:

Hello Petsc developers and supporters,

I would like to confirm the performance of asynchronous computations of inner product computation overlapping with matrix-vector multiplication computation by the below code.

 PetscLogDouble tt1,tt2;
    KSP ksp;
    //ierr = VecSet(c,one);
    ierr = VecSet(c,one);
    ierr = VecSet(u,one);
    ierr = VecSet(b,one);

    ierr = KSPCreate(PETSC_COMM_WORLD,&ksp); CHKERRQ(ierr);
    ierr = KSP_MatMult(ksp,A,x,Ax); CHKERRQ(ierr);


    ierr = PetscTime(&tt1);CHKERRQ(ierr);
    ierr = VecNormBegin(u,NORM_2,&norm1);
    ierr = PetscCommSplitReductionBegin(PetscObjectComm((PetscObject)Ax));
    ierr = KSP_MatMult(ksp,A,c,Ac);
    ierr = VecNormEnd(u,NORM_2,&norm1);
    ierr = PetscTime(&tt2);CHKERRQ(ierr);

    ierr = PetscPrintf(PETSC_COMM_WORLD, "The time used for the asynchronous calculation: %f\n",tt2-tt1); CHKERRQ(ierr);
    ierr = PetscPrintf(PETSC_COMM_WORLD,"+ |u| =  %g\n",(double) norm1); CHKERRQ(ierr);


    ierr = PetscTime(&tt1);CHKERRQ(ierr);
    ierr = VecNorm(b,NORM_2,&norm2); CHKERRQ(ierr);
    ierr = KSP_MatMult(ksp,A,c,Ac);
    ierr = PetscTime(&tt2);CHKERRQ(ierr);

    ierr = PetscPrintf(PETSC_COMM_WORLD, "The time used for the synchronous calculation: %f\n",tt2-tt1); CHKERRQ(ierr);
    ierr = PetscPrintf(PETSC_COMM_WORLD,"+ |b| =  %g\n",(double) norm2); CHKERRQ(ierr);

On a cluster with two or four nodes, the asynchronous computation is always much slower than synchronous computation.

The time used for the asynchronous calculation: 0.000203
+ |u| =  100.
The time used for the synchronous calculation: 0.000006
+ |b| =  100.

Are there any necessary settings on MPI or Petsc to gain performance of asynchronous computation?

Thank you very much for anything you can provide.
Sincerely,
Viet.