VMD-L Mailing List
From: Josh Vermaas (vermaas2_at_illinois.edu)
Date: Sun Mar 31 2013 - 14:14:42 CDT
- Next message: Win Liu: "Re: One same issue happened as before in fftk"
- Previous message: JC Gumbart: "Re: FFTK bond optimization +idlepoll request"
- In reply to: Francesco Pietra: "FFTK bond optimization +idlepoll request"
- Next in thread: Francesco Pietra: "Re: FFTK bond optimization +idlepoll request"
- Reply: Francesco Pietra: "Re: FFTK bond optimization +idlepoll request"
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ] [ attachment ]
Hi Francesco,
The easy solution is just to get a second namd binary that is CPU only,
and point to that one when using FFTK. For typical molecule
parameterizations where the number of atoms is less than 100, the merits
of using a GPU-accelerated version of NAMD (or even using more than 1
CPU) are questionable. The speedup simply isn't there (and may actually
be slower), as the number of interactions that need to be computed is
low relative to the overhead of splitting the work up. So don't worry
about it, and pick up a simpler binary.
-Josh Vermaas
On 03/31/2013 01:14 PM, Francesco Pietra wrote:
> Hello:
> With VMD for LINUXAMD64, version 1.9.2a20 (March 12, 2013), FFTK
> plugin, from OptGeom, for a diatomic molecule, skipping charge
> optimization, I tried CalcBonded OptBonded.
>
> At RunOptimization, error
>
> Charm++: standalone mode (not using charmrun)
> Converse/Charm++ Commit ID: v6.4.0-beta1-0-g5776d21
> FATAL ERROR: Please add +idlepoll to command line for proper performance.
> [0] Stack Traceback:
> [0:0] CmiAbort+0x70 [0xa6302d]
> [0:1] _Z8NAMD_diePKc+0x4a [0x55e03a]
> [0:2] _ZN7BackEnd4initEiPPc+0xd4 [0x565d84]
> [0:3] main+0x10 [0x529190]
> [0:4] __libc_start_main+0xfd [0x7fa6e82afead]
> [0:5] [0x52bead]
> [0] Stack Traceback:
> [0:0] [0xa62598]
> [0:1] CmiAbort+0xa5 [0xa63062]
> [0:2] _Z8NAMD_diePKc+0x4a [0x55e03a]
> [0:3] _ZN7BackEnd4initEiPPc+0xd4 [0x565d84]
> [0:4] main+0x10 [0x529190]
> [0:5] __libc_start_main+0xfd [0x7fa6e82afead]
> [0:6] [0x52bead]
> ------------- Processor 0 Exiting: Called CmiAbort ------------
> Reason: FATAL ERROR: Please add +idlepoll to command line for proper
> performance.
>
> Charm++ fatal error:
> FATAL ERROR: Please add +idlepoll to command line for proper performance.
>
> Charm++: standalone mode (not using charmrun)
> Converse/Charm++ Commit ID: v6.4.0-beta1-0-g5776d21
> FATAL ERROR: Please add +idlepoll to command line for proper performance.
> [0] Stack Traceback:
> [0:0] CmiAbort+0x70 [0xa6302d]
> [0:1] _Z8NAMD_diePKc+0x4a [0x55e03a]
> [0:2] _ZN7BackEnd4initEiPPc+0xd4 [0x565d84]
> [0:3] main+0x10 [0x529190]
> [0:4] __libc_start_main+0xfd [0x7fa6e82afead]
> [0:5] [0x52bead]
> [0] Stack Traceback:
> [0:0] [0xa62598]
> [0:1] CmiAbort+0xa5 [0xa63062]
> [0:2] _Z8NAMD_diePKc+0x4a [0x55e03a]
> [0:3] _ZN7BackEnd4initEiPPc+0xd4 [0x565d84]
> [0:4] main+0x10 [0x529190]
> [0:5] __libc_start_main+0xfd [0x7fa6e82afead]
> [0:6] [0x52bead]
> ------------- Processor 0 Exiting: Called CmiAbort ------------
> Reason: FATAL ERROR: Please add +idlepoll to command line for proper
> performance.
>
> Charm++ fatal error:
> FATAL ERROR: Please add +idlepoll to command line for proper performance.
>
> while executing
> "::exec /opt/namd2.9-cuda4.0-plumed-1.3_2013-03-12/bin/namd2
> min-bondangles.conf"
> ("eval" body line 1)
> invoked from within
> "eval ::exec [list $exec_path] [lrange $args 1 end]"
> (procedure "::ExecTool::exec" line 14)
> invoked from within
> "::ExecTool::exec $namdbin $minName.conf"
> (procedure "::ForceFieldToolKit::BondAngleOpt::optBondsAngles"
> line 68)
> invoked from within
> "$function $p0list"
> (procedure "construct_initial_simplex" line 22)
> invoked from within
> "construct_initial_simplex [lindex $args 0] $scale"
> (procedure "::Optimize::Opt0::handle" line 91)
> invoked from within
> "$opt initsimplex $baInitial $scale"
> (procedure "::ForceFieldToolKit::BondAngleOpt::optimize" line 320)
> invoked from within
> "::ForceFieldToolKit::BondAngleOpt::optimize"
> (procedure "::ForceFieldToolKit::gui::baoptRunOpt" line 40)
> invoked from within
> "::ForceFieldToolKit::gui::baoptRunOpt "
> invoked from within
> ".fftk_gui.hlf.nb.bondangleopt.runOpt invoke "
> invoked from within
> ".fftk_gui.hlf.nb.bondangleopt.runOpt instate {pressed !disabled} {
> .fftk_gui.hlf.nb.bondangleopt.runOpt state !pressed;
> .fftk_gui.hlf.nb.bondangleopt...."
> (command bound to event)
>
> This is a CPU/GPU system (and the GPUs where activated before runopt)
> where the command line is
>
> charmrun $NAMD_HOME/bin/namd2 filename.conf +p(numver of CPUs) +idlepoll
>
> With FFTK I simply added the full path to namd2 executable
>
> Where to add the requested "+idlepoll"? If added after "namd" of the
> FFTK line, the error is "namd2 binary not found"
>
>
> I am following old directions for FFTK. Where to find a manual for
> current version?
>
> Thanks
> francesco pietra
>
>
- Next message: Win Liu: "Re: One same issue happened as before in fftk"
- Previous message: JC Gumbart: "Re: FFTK bond optimization +idlepoll request"
- In reply to: Francesco Pietra: "FFTK bond optimization +idlepoll request"
- Next in thread: Francesco Pietra: "Re: FFTK bond optimization +idlepoll request"
- Reply: Francesco Pietra: "Re: FFTK bond optimization +idlepoll request"
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ] [ attachment ]