strange error from NAMD with CUDA

From: t s (tsigalko2003_at_gmail.com)
Date: Thu Nov 05 2009 - 13:55:14 CST

Hi Everybody,

I met one strange problem with CUDA and NAMD, anyone could give me any
advice?

I compiled the NAMD (CVS, latest version) on my Mac Pro laptop. I installed
the CUDA 2.2 (including driver and sdk and so on).
I first compiled the charmrun (version 6.1.3) I chose the most simple one (
single cpu without mpi) . It is quite smooth.
Then I used the fftw/tcl lib from the namd website to compile the namd2. It
is also quite good.
But when I executed the namd2 code, using
chamrun ++local namd2 +idlepoll input.conf > output.log&

the program can run severl second.
Everything is good untill the namd reached the last step before real
calculation of energies. I copied the output information as follwing
I am totally lost about the infomation of "absolute/relative imprecision "
anyone can help me?
Thanks in advance.

Bests,
Tao

output information
"Info: Startup phase 5 took 0.00714087 s, 646.957 MB of memory in use
LDB: Measuring processor speeds ... Done.
Info: Startup phase 6 took 0.00641608 s, 646.957 MB of memory in use
Info: CREATING 1288 COMPUTE OBJECTS
Info: useSync: 1 useProxySync: 0
Info: NONBONDED TABLE R-SQUARED SPACING: 0.0625
Info: NONBONDED TABLE SIZE: 769 POINTS
Info: ABSOLUTE IMPRECISION IN FAST TABLE FORCE: 2.64698e-23 AT 11.9138
Info: RELATIVE IMPRECISION IN FAST TABLE FORCE: 1.40436e-16 AT 11.9138
Info: ABSOLUTE IMPRECISION IN SCOR TABLE FORCE: 4.10282e-22 AT 11.9138
Info: RELATIVE IMPRECISION IN SCOR TABLE FORCE: 3.56912e-15 AT 11.9138
Info: ABSOLUTE IMPRECISION IN VDWA TABLE ENERGY: 9.86076e-32 AT 11.8295
Info: RELATIVE IMPRECISION IN VDWA TABLE ENERGY: 6.75473e-14 AT 11.9138
Info: ABSOLUTE IMPRECISION IN VDWA TABLE FORCE: 4.62223e-32 AT 11.9138
Info: RELATIVE IMPRECISION IN VDWA TABLE FORCE: 2.12237e-15 AT 11.9138
Info: ABSOLUTE IMPRECISION IN VDWB TABLE ENERGY: 4.1359e-25 AT 11.8295
Info: RELATIVE IMPRECISION IN VDWB TABLE ENERGY: 7.56853e-15 AT 11.9138
Info: ABSOLUTE IMPRECISION IN VDWB TABLE FORCE: 3.87741e-26 AT 11.9138
Info: RELATIVE IMPRECISION IN VDWB TABLE FORCE: 5.97409e-16 AT 11.9138
Updated CUDA force table with 8192 elements.
create ComputeNonbondedCUDA
Pe 0 found 390 unique exclusion lists needing 21408 bytes
FATAL ERROR: CUDA error memcpy to exclusions: invalid argument
------------- Processor 0 Exiting: Called CmiAbort ------------
Reason: FATAL ERROR: CUDA error memcpy to exclusions: invalid argument

[0] Stack Traceback:
Fatal error on PE 0> FATAL ERROR: CUDA error memcpy to exclusions: invalid
argument

  [0] 0 namd2 0x00413958 CmiAbort + 104
  [1] 1 namd2 0x00002d23 _Z8NAMD_diePKc + 99
"

This archive was generated by hypermail 2.1.6 : Wed Feb 29 2012 - 15:53:27 CST