February 2021 - petsc-dev - mailman3.cels.anl.gov

Re: [petsc-dev] error with flags PETSc uses for determining AVX
by Jed Brown 13 Feb '21

13 Feb '21

Use -march=native or similar. The default target is basic x86_64, which has only SSE2. Barry Smith <bsmith(a)petsc.dev> writes: > PETSc source has code like defined(__AVX2__) in the source but it does not seem to be able to find any of these macros (icc or gcc) on the petsc-02 system > > Are these macros supposed to be defined? How does on get them to be defined? Why are they not define? What am I doing wrong? > > Keep reading > > $ lscpu > Architecture: x86_64 > CPU op-mode(s): 32-bit, 64-bit > Byte Order: Little Endian > CPU(s): 64 > On-line CPU(s) list: 0-63 > Thread(s) per core: 2 > Core(s) per socket: 16 > Socket(s): 2 > NUMA node(s): 2 > Vendor ID: GenuineIntel > CPU family: 6 > Model: 85 > Model name: Intel(R) Xeon(R) Gold 5218 CPU @ 2.30GHz > Stepping: 7 > CPU MHz: 1000.603 > CPU max MHz: 2301.0000 > CPU min MHz: 1000.0000 > BogoMIPS: 4600.00 > Virtualization: VT-x > L1d cache: 32K > L1i cache: 32K > L2 cache: 1024K > L3 cache: 22528K > NUMA node0 CPU(s): 0-15,32-47 > NUMA node1 CPU(s): 16-31,48-63 > Flags: fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss ht tm pbe syscall nx pdpe1gb rdtscp lm constant_tsc art arch_perfmon pebs bts rep_good nopl xtopology nonstop_tsc cpuid aperfmperf pni pclmulqdq dtes64 monitor ds_cpl vmx smx est tm2 ssse3 sdbg fma cx16 xtpr pdcm pcid dca sse4_1 sse4_2 x2apic movbe popcnt tsc_deadline_timer aes xsave avx f16c rdrand lahf_lm abm 3dnowprefetch cpuid_fault epb cat_l3 cdp_l3 invpcid_single intel_ppin ssbd mba ibrs ibpb stibp ibrs_enhanced tpr_shadow vnmi flexpriority ept vpid fsgsbase tsc_adjust bmi1 avx2 smep bmi2 erms invpcid cqm mpx rdt_a avx512f avx512dq rdseed adx smap clflushopt clwb intel_pt avx512cd avx512bw avx512vl xsaveopt xsavec xgetbv1 xsaves cqm_llc cqm_occup_llc cqm_mbm_total cqm_mbm_local dtherm ida arat pln pts pku ospke avx512_vnni md_clear flush_l1d arch_capabilities > > Test program > > #if defined(__FMA__) > #error FMA > #endif > > #if defined(__AVX512F__) > #error AVX512F > #endif > > #if defined(__AVX2__) > #error AVX2 > #endif > > > icc mytest.c > /usr/lib/gcc/x86_64-linux-gnu/7/../../../x86_64-linux-gnu/Scrt1.o: In function `_start': > (.text+0x20): undefined reference to `main'

1 0

error with flags PETSc uses for determining AVX
by Barry Smith 13 Feb '21

13 Feb '21

PETSc source has code like defined(__AVX2__) in the source but it does not seem to be able to find any of these macros (icc or gcc) on the petsc-02 system Are these macros supposed to be defined? How does on get them to be defined? Why are they not define? What am I doing wrong? Keep reading $ lscpu Architecture: x86_64 CPU op-mode(s): 32-bit, 64-bit Byte Order: Little Endian CPU(s): 64 On-line CPU(s) list: 0-63 Thread(s) per core: 2 Core(s) per socket: 16 Socket(s): 2 NUMA node(s): 2 Vendor ID: GenuineIntel CPU family: 6 Model: 85 Model name: Intel(R) Xeon(R) Gold 5218 CPU @ 2.30GHz Stepping: 7 CPU MHz: 1000.603 CPU max MHz: 2301.0000 CPU min MHz: 1000.0000 BogoMIPS: 4600.00 Virtualization: VT-x L1d cache: 32K L1i cache: 32K L2 cache: 1024K L3 cache: 22528K NUMA node0 CPU(s): 0-15,32-47 NUMA node1 CPU(s): 16-31,48-63 Flags: fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss ht tm pbe syscall nx pdpe1gb rdtscp lm constant_tsc art arch_perfmon pebs bts rep_good nopl xtopology nonstop_tsc cpuid aperfmperf pni pclmulqdq dtes64 monitor ds_cpl vmx smx est tm2 ssse3 sdbg fma cx16 xtpr pdcm pcid dca sse4_1 sse4_2 x2apic movbe popcnt tsc_deadline_timer aes xsave avx f16c rdrand lahf_lm abm 3dnowprefetch cpuid_fault epb cat_l3 cdp_l3 invpcid_single intel_ppin ssbd mba ibrs ibpb stibp ibrs_enhanced tpr_shadow vnmi flexpriority ept vpid fsgsbase tsc_adjust bmi1 avx2 smep bmi2 erms invpcid cqm mpx rdt_a avx512f avx512dq rdseed adx smap clflushopt clwb intel_pt avx512cd avx512bw avx512vl xsaveopt xsavec xgetbv1 xsaves cqm_llc cqm_occup_llc cqm_mbm_total cqm_mbm_local dtherm ida arat pln pts pku ospke avx512_vnni md_clear flush_l1d arch_capabilities Test program #if defined(__FMA__) #error FMA #endif #if defined(__AVX512F__) #error AVX512F #endif #if defined(__AVX2__) #error AVX2 #endif icc mytest.c /usr/lib/gcc/x86_64-linux-gnu/7/../../../x86_64-linux-gnu/Scrt1.o: In function `_start': (.text+0x20): undefined reference to `main'

1 0

Fortran macOS Linkage Warning
by Jacob Faibussowitsch 13 Feb '21

13 Feb '21

Hello All, I’ve been getting these warnings for weeks, at first I figured it was due to gfortran brew package not being updated for Big Sur, but since it has persisted through a few gfortran brew updates I’m thinking it may be petsc related. This is on clean build on master. ld: warning: object file (master-arch-darwin-c-debug/obj/sys/classes/bag/f2003-src/fsrc/bagenum.o) was built for newer macOS version (11.2) than being linked (11.0) ld: warning: object file (master-arch-darwin-c-debug/obj/sys/f90-src/fsrc/f90_fwrap.o) was built for newer macOS version (11.2) than being linked (11.0) ld: warning: object file (master-arch-darwin-c-debug/obj/sys/objects/f2003-src/fsrc/optionenum.o) was built for newer macOS version (11.2) than being linked (11.0) ld: warning: object file (master-arch-darwin-c-debug/obj/sys/fsrc/somefort.o) was built for newer macOS version (11.2) than being linked (11.0) ld: warning: object file (master-arch-darwin-c-debug/obj/sys/f90-mod/petscsysmod.o) was built for newer macOS version (11.2) than being linked (11.0) Best regards, Jacob Faibussowitsch (Jacob Fai - booss - oh - vitch) Cell: (312) 694-3391

1 0

Re: [petsc-dev] PETSC MPI error
by Barry Smith 09 Feb '21

09 Feb '21

Ok, they use the OpenMPI version naming in their include files but changed the name in MPI_Get_library_version() to make it look like they wrote their own MPI :-). I guess we are too picky in our testing, in pinit.c you can change to /* check for OpenMPI version, it is not part of the MPI ABI initiative (is it part of another initiative that needs to be handled?) */ #elif defined(OMPI_MAJOR_VERSION) { char *ver,bs[32],*bsf; ierr = PetscStrstr(mpilibraryversion,"Open MPI",&ver);if (ierr) return ierr; if (ver) { PetscSNPrintf(bs,32,"v%d.%d",OMPI_MAJOR_VERSION,OMPI_MINOR_VERSION); ierr = PetscStrstr(ver,bs,&bsf);if (ierr) return ierr; if (!bsf) { fprintf(stderr,"PETSc Error --- Open MPI library version \n%s does not match what PETSc was compiled with %d.%d, aborting\n",mpilibraryversion,OMPI_MAJOR_VERSION,OMPI_MINOR_VERSION); return PETSC_ERR_MPI_LIB_INCOMP; } } #endif > On Feb 9, 2021, at 2:49 PM, Mark Adams <mfadams(a)lbl.gov> wrote: > > I see this message when I run at problem on Fugaku: > > 05:45 master= ~/petsc/src/snes/tests/data$ more hello.5067177.err.1.9 > PETSc Error --- Open MPI library version > FUJITSU MPI Library 4.0.0 (4.0.1fj4.0.0) does not match what PETSc was compiled with 4.0, aborting > > This is on master. Any ideas? > Thanks, > Mark > <make.log><configure.log>

1 0

PETSC MPI error
by Mark Adams 09 Feb '21

09 Feb '21

I see this message when I run at problem on Fugaku: 05:45 master= ~/petsc/src/snes/tests/data$ more hello.5067177.err.1.9 PETSc Error --- Open MPI library version FUJITSU MPI Library 4.0.0 (4.0.1fj4.0.0) does not match what PETSc was compiled with 4.0, aborting This is on master. Any ideas? Thanks, Mark

1 0

Re: [petsc-dev] About parallel of ILU
by Mark Adams 09 Feb '21

09 Feb '21

On Tue, Feb 9, 2021 at 4:31 AM Chen Gang <569615491(a)qq.com> wrote: > Dear Professor, > > I have a question about the bjacobi+ilu, I'm in the place > > src/ksp/ksp/tutorials > > and running the ex2 with > > mpirun -np 40 -bind-to core ./ex2 -m 2000 -n 2000 -ksp_max_it 1000000 > -ksp_rtol 1e-10 -ksp_type gmres -pc_type bjacobi -sub_pc_type ilu > -sub_pc_factor_levels 10 -ksp_monitor_short > > on the Intel machine with 40 cores, the petsc is complied with MKL and > ICC/MPIICC with O3. > > I found the numerical results are not the same from time to time, although > they all right, the iterations are not exactly the same from time to time. > Where have I done wrong? > Humm, iterations should rarely change. You can see floating point scale differences. Try jacobi instead of ilu and/or try not using MKL. > > > > > > I'm looking forward to hear you! > > Best, > > Gang Chen > > Sichuan Univertsity, > School of Mathematics, > Chengdu, China >

1 0

Re: [petsc-dev] About parallel of ILU
by Matthew Knepley 09 Feb '21

09 Feb '21

On Tue, Feb 9, 2021 at 4:30 AM Chen Gang <569615491(a)qq.com> wrote: > Dear Professor, > > I have a question about the bjacobi+ilu, I'm in the place > > src/ksp/ksp/tutorials > > and running the ex2 with > > mpirun -np 40 -bind-to core ./ex2 -m 2000 -n 2000 -ksp_max_it 1000000 > -ksp_rtol 1e-10 -ksp_type gmres -pc_type bjacobi -sub_pc_type ilu > -sub_pc_factor_levels 10 -ksp_monitor_short > > on the Intel machine with 40 cores, the petsc is complied with MKL and > ICC/MPIICC with O3. > > I found the numerical results are not the same from time to time, although > they all right, the iterations are not exactly the same from time to time. > Where have I done wrong? > MPI does not guarantee the order that reductions are calculated, so that you can get slightly different answers in parallel. Thanks, Matt > I'm looking forward to hear you! > > Best, > > Gang Chen > > Sichuan Univertsity, > School of Mathematics, > Chengdu, China > -- What most experimenters take for granted before they begin their experiments is infinitely more interesting than any results to which their experiments lead. -- Norbert Wiener https://www.cse.buffalo.edu/~knepley/ <http://www.cse.buffalo.edu/~knepley/>

1 0

About parallel of ILU
by Chen Gang 09 Feb '21

09 Feb '21

Dear Professor, I have a question about the bjacobi+ilu, I'm in the place   src/ksp/ksp/tutorials  and running the ex2 with mpirun -np 40 -bind-to core ./ex2 -m 2000 -n 2000  -ksp_max_it 1000000 -ksp_rtol 1e-10 -ksp_type gmres -pc_type bjacobi -sub_pc_type ilu  -sub_pc_factor_levels 10 -ksp_monitor_short on the Intel machine with 40 cores, the petsc is complied with MKL and ICC/MPIICC with O3. I found the numerical results are not the same from time to time, although they all right, the iterations are not exactly the same from time to time. Where have I done wrong? I'm looking forward to hear you! Best, Gang Chen Sichuan Univertsity, School of Mathematics, Chengdu, China

1 0

MatSeqAIJCUSPARSEBuildILUUpperTriMatrix question
by Mark Adams 06 Feb '21

06 Feb '21

I am adding GPU LU factorizations to cuSparse and Kokkos matrices and I see that the cuSparse solvers have ILU in the name. It looks to me like it should work for an LU ... but I wanted to check and see if there are any thoughts. Thanks, Mark

1 0

petsc-3.14.4 now available
by Satish Balay 03 Feb '21

03 Feb '21

Dear PETSc users, The patch release petsc-3.14.4 is now available for download. http://www.mcs.anl.gov/petsc/download/index.html Satish

1 0