XSEDE Startup allocation

[phpisciuneri]

• https://portal.xsede.org/allocations-overview#types-startup

[phpisciuneri]

@labrinid

[phpisciuneri]

Process from XSEDE web page:

Startup and Education allocations are easy to request with fast approval time. The request process requires a CV and short abstract; PIs may apply at any time. Startup allocations are appropriate for code development, experimentation with XSEDE platforms, or for application testing. Most first-time XSEDE users begin with a Startup allocation. PIs may request a single computational or visualization resource or a combination of resources. Startup requests may total up to 200,000 CPU-hours (or Service Units, SUs), though limits for individual resources may vary. SPs providing allocatable storage resources consider storage startups in combination with compute and visualization resources and some SPs allow storage-only requests. For resource-specific startup limits view the table above.

The PI should assess the amount of time that will be necessary to accomplish the computational project. If you are just getting started and not certain of your computational needs, then apply for a Startup allocation request. For a project that requires a significant amount of time or data storage, apply for a Research allocation.

[phpisciuneri]

Attached is my up-to-date CV: cv.pdf

[phpisciuneri]

Draft of proposal (for startup they just request an abstract): proposal.pdf

[phpisciuneri]

@labrinid @angenzheng our startup proposal is submitted for XSEDE resources. I will let you know when the status is updated.

Request Title: Load Balancing, Graph Repartitioning, and Data Migration for the Simulation Turbulent Reacting Flows
Request Number: ASC160085
PI Name: Patrick  Pisciuneri
Request Type: New
Allocation Type: Startup
Uploaded Documents:PI CV: cv-pisciuneri.pdf

[AngenZheng]

@phpisciuneri @labrinid can we also use the same proposal for the MPI cluster resource allocation? I was told that they will start to charge the usage in the next few months.

Thanks, Angen

[phpisciuneri]

@AngenZheng yes I have been meaning to work on a proposal for Pitt resources as well. Maybe we can talk more about this tomorrow.

At any rate the startup allocation successfully went through today. We have 50,000 SUs on stampede. I am able to log in and it looks like my old files are still there. So I will get to work on getting the last version of the tg code and paragon running on stampede.

@AngenZheng It seems that they changed the way you log in. You now use a TACC cell phone app to generate a token. If you log into the TACC portal it forces you to set this up: https://portal.tacc.utexas.edu/home

@labrinid Do you have an XSEDE account? I wasn't able to find you. If not you can create one at the XSEDE user portal site and then I can add you to the allocation. If you don't ever intend on logging into the system then you really don't need it. Please let me know.

[AngenZheng]

Thank you! Patrick.

We have submitted one proposal to SAM for continuing using the cluster last week. Hopefully, it will be approved.

Also, I am working on a computer science journal paper and would like to include some experiments using the tg benchmark if possible. But one of the modules is not linking while compiling tg code. Do you know what may be the reason for this?

[ 95%] Linking CXX executable tg libiplmcfdlib.a(scalar_average.cpp.o): In function iplmcfd::Iplmc::scalar_average()': scalar_average.cpp:(.text+0x1e6): undefined reference toCKHML' ext/chemkinpp/libchemkinpp.a(thermomixture.cpp.o): In function chemkinpp::detail::cklib_ckwyp(double const&, double const*, double const*, int const*, double const*, double*)': thermomixture.cpp:(.text+0x1): undefined reference toCKWYP' ext/chemkinpp/libchemkinpp.a(thermomixture.cpp.o): In function boost::detail::sp_counted_impl_p<mytools::fortran::file::impl_>::dispose()': thermomixture.cpp:(.text._ZN5boost6detail17sp_counted_impl_pIN7mytools7fortran4file5impl_EE7disposeEv[_ZN5boost6detail17sp_counted_impl_pIN7mytools7fortran4file5impl_EE7disposeEv]+0x14): undefined reference toMYTOOLS_FORTRAN_CLOSEUNIT'

Best! Angen

On Tue, Jan 24, 2017 at 9:19 PM, phpisciuneri notifications@github.com wrote:

@AngenZheng https://github.com/AngenZheng yes I have been meaning to work on a proposal for Pitt resources as well. Maybe we can talk more about this tomorrow.

At any rate the startup allocation successfully went through today. We have 50,000 SUs on stampede. I am able to log in and it looks like my old files are still there. So I will get to work on getting the last version of the tg code and paragon running on stampede.

@AngenZheng https://github.com/AngenZheng It seems that they changed the way you log in. You now use a TACC cell phone app to generate a token. If you log into the TACC portal it forces you to set this up: https://portal.tacc.utexas.edu/home

@labrinid https://github.com/labrinid Do you have an XSEDE account? I wasn't able to find you. If not you can create one at the XSEDE user portal site and then I can add you to the allocation. If you don't ever intend on logging into the system then you really don't need it. Please let me know.

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/phpisciuneri/tg/issues/5#issuecomment-275001182, or mute the thread https://github.com/notifications/unsubscribe-auth/AF1QVyCL1S6b916pHqGNYwlYBz7zZQ-Mks5rVrE1gaJpZM4KghGq .

-- Sincerely yours, Angen Zheng, PhD Student Department of Computer Science University of Pittsburgh Pittsburgh, PA, USA Office: 6805 Sennott Square Homepage: http://people.cs.pitt.edu/~anz28/

[phpisciuneri]

Looks like linking errors with chemkin library. Where are you trying to compile?

[AngenZheng]

It's on the MPI cluster.

Angen

On Tue, Jan 24, 2017 at 9:49 PM, phpisciuneri notifications@github.com wrote:

Looks like linking errors with chemkin library. Where are you trying to compile?

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/phpisciuneri/tg/issues/5#issuecomment-275005165, or mute the thread https://github.com/notifications/unsubscribe-auth/AF1QV1SI_33rfJ1tB5nL5KshUuZti4LIks5rVrhHgaJpZM4KghGq .

-- Sincerely yours, Angen Zheng, PhD Student Department of Computer Science University of Pittsburgh Pittsburgh, PA, USA Office: 6805 Sennott Square Homepage: http://people.cs.pitt.edu/~anz28/

[phpisciuneri]

Here is the script I am using to build on the mpi machine. See how this differs from what you are using/doing:

# source this script to build
# $ . build-mpi.sh

# set environment
module purge
export SYS_TYPE=haswell
source /opt/sam/spack/share/spack/setup-env.sh
module load intel
spack load boost
spack load hwloc%intel
spack load openmpi@1.8.6
module load trilinos@12.0.1%intel@15.0.3-736pgl4

spack load cmake@3.3.0

# create build directory
mkdir -p build

cd build

# configure and generate makefile
CC=icc CXX=icpc FC=ifort cmake \
    -DCMAKE_BUILD_TYPE=release \
    -DCHEMKINPP_IMPLICIT_BLAS=off \
    ..

[AngenZheng]

For some reason, the problem is still there even if I use the same script as you post. It is not urge problem though, I am still working on optimizing the repartitioning code.

Thanks, Angen

On Tue, Jan 24, 2017 at 9:53 PM, phpisciuneri notifications@github.com wrote:

Here is the script I am using to build on the mpi machine. See how this differs from what you are using/doing:

source this script to build# $ . build-mpi.sh

set environment

module purgeexport SYS_TYPE=haswellsource /opt/sam/spack/share/spack/setup-env.sh module load intel spack load boost spack load hwloc%intel spack load openmpi@1.8.6 module load trilinos@12.0.1%intel@15.0.3-736pgl4

spack load cmake@3.3.0

create build directory

mkdir -p build cd build

configure and generate makefile

CC=icc CXX=icpc FC=ifort cmake \ -DCMAKE_BUILD_TYPE=release \ -DCHEMKINPP_IMPLICIT_BLAS=off \ ..

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/phpisciuneri/tg/issues/5#issuecomment-275005594, or mute the thread https://github.com/notifications/unsubscribe-auth/AF1QV09UISTfFjnnofQiQZFcndI4pdy5ks5rVrkUgaJpZM4KghGq .

-- Sincerely yours, Angen Zheng, PhD Student Department of Computer Science University of Pittsburgh Pittsburgh, PA, USA Office: 6805 Sennott Square Homepage: http://people.cs.pitt.edu/~anz28/