I cannot reproduce this

AShte · ‎05-12-2016

I'm building a corray program with: -fast -qopt-report -g -coarray=distributed and I see this diagnostics in *optrpt: LOOP BEGIN at coback1.f90(90,5) remark #15344: loop was not vectorized: vector dependence prevents vectorization. First dependence is shown below. Use level 5 report for details remark #15346: vector dependence: assumed FLOW dependence between 117_dv_template line 91 and 117_dv_template line 91 remark #25439: unrolled with remainder by 4 LOOP END My code around lines 90-91 looks like this: 89 do j = 1, size2 90 do i = 1, width 91 pic(i,j) = 0.25 * ( oldpic(i-1,j) + oldpic(i+1,j) + & 92 oldpic(i,j-1) + oldpic(i,j+1) - edger(i,j) ) 93 end do 94 end do This is a simple Laplacian solver. What does "117_dv_template" refer to here? Thanks Anton

Xiaoping_D_Intel · ‎05-16-2016

It looks like the internal name of an array descriptor generated by the compiler for some array referenced at line 91. It shouldn't be exposed in the optimization report without being mapped back to the real array name in the code.

Is it possible for you to provide a standalone test case which can be compiled to reproduce the problem?

Thanks,

Xiaoping Duan

Intel Customer Support

AShte · ‎05-17-2016

I cannot reproduce this behaviour anymore.

However, extra white spaces indicate that some info

is missing:

remark #15346: vector dependence: assumed FLOW dependence between line 92 and line 92

Anyway, I uploaded 2 files:

http://eis.bris.ac.uk/~mexas/pgmio.f90

http://eis.bris.ac.uk/~mexas/coback1.f90

I build the program with:

ifort -c -coarray -fast -qopt-report pgmio.f90

ifort -c -coarray -fast -qopt-report coback1.f90

ifort -coarray coback1.o pgmio.o

I get the following on the terminal:

ipo: warning #11021: unresolved MPI_Barrier
        Referenced in /cm/shared/languages/Intel-Compiler-XE-16-U2/compilers_and_libraries_2016.2.181/linux/compiler/lib/intel64_lin/libicaf.so
ipo: warning #11021: unresolved MPI_Comm_dup
        Referenced in /cm/shared/languages/Intel-Compiler-XE-16-U2/compilers_and_libraries_2016.2.181/linux/compiler/lib/intel64_lin/libicaf.so
ipo: warning #11021: unresolved MPI_Put
        Referenced in /cm/shared/languages/Intel-Compiler-XE-16-U2/compilers_and_libraries_2016.2.181/linux/compiler/lib/intel64_lin/libicaf.so
ipo: warning #11021: unresolved MPI_Get
        Referenced in /cm/shared/languages/Intel-Compiler-XE-16-U2/compilers_and_libraries_2016.2.181/linux/compiler/lib/intel64_lin/libicaf.so
ipo: warning #11021: unresolved MPI_Isend
        Referenced in /cm/shared/languages/Intel-Compiler-XE-16-U2/compilers_and_libraries_2016.2.181/linux/compiler/lib/intel64_lin/libicaf.so
ipo: warning #11021: unresolved MPI_Init
        Referenced in /cm/shared/languages/Intel-Compiler-XE-16-U2/compilers_and_libraries_2016.2.181/linux/compiler/lib/intel64_lin/libicaf.so
ipo: warning #11021: unresolved MPI_Free_mem
        Referenced in /cm/shared/languages/Intel-Compiler-XE-16-U2/compilers_and_libraries_2016.2.181/linux/compiler/lib/intel64_lin/libicaf.so
ipo: warning #11021: unresolved MPI_Win_unlock
        Referenced in /cm/shared/languages/Intel-Compiler-XE-16-U2/compilers_and_libraries_2016.2.181/linux/compiler/lib/intel64_lin/libicaf.so
ipo: warning #11021: unresolved MPI_Alloc_mem
        Referenced in /cm/shared/languages/Intel-Compiler-XE-16-U2/compilers_and_libraries_2016.2.181/linux/compiler/lib/intel64_lin/libicaf.so
ipo: warning #11021: unresolved MPI_Comm_size
        Referenced in /cm/shared/languages/Intel-Compiler-XE-16-U2/compilers_and_libraries_2016.2.181/linux/compiler/lib/intel64_lin/libicaf.so
ipo: warning #11021: unresolved MPI_Abort
        Referenced in /cm/shared/languages/Intel-Compiler-XE-16-U2/compilers_and_libraries_2016.2.181/linux/compiler/lib/intel64_lin/libicaf.so
ipo: warning #11021: unresolved MPI_Wait
        Referenced in /cm/shared/languages/Intel-Compiler-XE-16-U2/compilers_and_libraries_2016.2.181/linux/compiler/lib/intel64_lin/libicaf.so
ipo: warning #11021: unresolved MPI_Recv
        Referenced in /cm/shared/languages/Intel-Compiler-XE-16-U2/compilers_and_libraries_2016.2.181/linux/compiler/lib/intel64_lin/libicaf.so
ipo: warning #11021: unresolved MPI_Finalized
        Referenced in /cm/shared/languages/Intel-Compiler-XE-16-U2/compilers_and_libraries_2016.2.181/linux/compiler/lib/intel64_lin/libicaf.so
ipo: warning #11021: unresolved MPI_Comm_rank
        Referenced in /cm/shared/languages/Intel-Compiler-XE-16-U2/compilers_and_libraries_2016.2.181/linux/compiler/lib/intel64_lin/libicaf.so
ipo: warning #11021: unresolved MPI_Win_lock
        Referenced in /cm/shared/languages/Intel-Compiler-XE-16-U2/compilers_and_libraries_2016.2.181/linux/compiler/lib/intel64_lin/libicaf.so
ipo: warning #11021: unresolved MPI_Finalize
        Referenced in /cm/shared/languages/Intel-Compiler-XE-16-U2/compilers_and_libraries_2016.2.181/linux/compiler/lib/intel64_lin/libicaf.so
ipo: warning #11021: unresolved MPI_Win_free
        Referenced in /cm/shared/languages/Intel-Compiler-XE-16-U2/compilers_and_libraries_2016.2.181/linux/compiler/lib/intel64_lin/libicaf.so
ipo: warning #11021: unresolved MPI_Win_set_errhandler
        Referenced in /cm/shared/languages/Intel-Compiler-XE-16-U2/compilers_and_libraries_2016.2.181/linux/compiler/lib/intel64_lin/libicaf.so
ipo: warning #11021: unresolved MPI_Win_fence
        Referenced in /cm/shared/languages/Intel-Compiler-XE-16-U2/compilers_and_libraries_2016.2.181/linux/compiler/lib/intel64_lin/libicaf.so
ipo: warning #11021: unresolved MPI_Win_create
        Referenced in /cm/shared/languages/Intel-Compiler-XE-16-U2/compilers_and_libraries_2016.2.181/linux/compiler/lib/intel64_lin/libicaf.so

Many thanks

Anton

Xiaoping_D_Intel · ‎05-18-2016

"-fast" option implies IPO (Inter-Procedure Optimization) enabled. Those warning message are generated in ipo linkage phase and can be safely ignored here. All the missing symbols are MPI functions defined in MPI runtime library.

I have opened a bug report to ask for removing the warnings. The bug report ID is DPD200411050.

Thanks,

Xiaoping Duan

Intel Customer Support