Skip to Content.
Sympa Menu

charm - Re: [charm] converting mpi to ampi program

charm AT lists.cs.illinois.edu

Subject: Charm++ parallel programming system

List archive

Re: [charm] converting mpi to ampi program


Chronological Thread 
  • From: Nitin Bhat <nitin.bhat.k AT gmail.com>
  • To: vishwas vasudeva <vishvasu98 AT gmail.com>
  • Cc: charm AT cs.illinois.edu, ppl AT cs.uiuc.edu
  • Subject: Re: [charm] converting mpi to ampi program
  • Date: Fri, 22 Jun 2018 09:45:12 -0500
  • Authentication-results: illinois.edu; spf=softfail smtp.mailfrom=nitin.bhat.k AT gmail.com; dkim=pass header.d=gmail.com header.s=20161025; dmarc=pass header.from=gmail.com

Hi Vishwas, 

BigNetSim is an old research project that is no longer supported at Illinois. The more recent software that can perform network simulation is TraceR. 

You can download TraceR by using git clone https://charm.cs.illinois.edu/gerrit/tracer. This link is to download an earlier version. A newer version is developed by Nikhil Jain, an alumnus of PPL at LLNL. You can download the newer version from https://github.com/LLNL/TraceR

Let us know if you face any trouble running programs on these versions.

Thanks,
Nitin

On Jun 21, 2018, at 6:47 AM, vishwas vasudeva <vishvasu98 AT gmail.com> wrote:

Respected sir/madam,
In the conversion of MPI codes to AMPI ,
I followed all the instruction given in the Charm 2007  workshop by taking an MPI program of Matrix multiplication and without making any changes to the code i ran it using /bin/charmc with ' -language ampi  ' flag and inturn got 2 executable files namely charmrun and pgm (name given to the file with -o flag).
My intention is to run it on BigEmulator.

on running the command,(I am running it in my local machine with  single processor)
./charmrun  ++local ./pgm +vp8
I got this response,
------------------------------------------------------------------------------------
wizard@turingmachine:~/temp1/charm/net-linux-x86_64-bigemulator/tests/ampi/mm$ ./charmrun ++local ./pgm +vp8 
Charmrun> started all node programs in 0.006 seconds.
Converse/Charm++ Commit ID: v6.6.1-0-g74a2cc5
Charm++> scheduler running in netpoll mode.

Missing parameters for BlueGene machine size!
<tip> use command line options: +x, +y, or +z.

BG> BigSim emulator shutdown gracefully!
BG> Emulation took 0.003119 seconds!
------------------------------------------------------------------------------------------
However on specifying the +x +y +z parameters along with +bglog to get the trace files,with the following command,
./charmrun ++local ./pgm +x8 +y1 +z1 +vp8 +bglog
I got this response.
------------------------------------------------------------------------------------------------
wizard@turingmachine:~/temp1/charm/net-linux-x86_64-bigemulator/tests/ampi/mm$ ./charmrun ++local ./pgm +x8 +y1 +z1 +vp8 +bglogCharmrun> started all node programs in 0.007 seconds.
Converse/Charm++ Commit ID: v6.6.1-0-g74a2cc5
Charm++> scheduler running in netpoll mode.
BG info> Simulating 8x1x1 nodes with 1 comm + 1 work threads each.
BG info> Network type: bluegene.
alpha: 1.000000e-07 packetsize: 1024 CYCLE_TIME_FACTOR:1.000000e-03.
CYCLES_PER_HOP: 5 CYCLES_PER_CORNER: 75.
BG info> cpufactor is 1.000000.
BG info> floating point factor is 0.000000.
BG info> Using WallTimer for timing method. 
BG info> Generating timing log. 
CharmLB> Load balancer ignores processor background load.
CharmLB> Load balancer assumes all CPUs are same.
Trace: traceroot: ./pgm
Charmrun: error on request socket--
Socket closed before recv.
---------------------------------------------------------------------------------------------
I am running it in my laptop, so i don't think there is an issue with some switches.
Am I missing something. I have also attached the mpi program, please sort out the issue.

Thanking you,
Vishwas V.K. 




<mpi_mm.c>




Archive powered by MHonArc 2.6.19.

Top of Page