Dear all,
I'm using amber12 pmemd.cuda_SPFP to run MD simulation. But there is a fatal error. The running process is always terminated sooner or later(I changed the random number and retried many times), with nothing error message in the md.out file. When I check the log file, I found the error posted by some others before. But I'm not able to solve this odd problem by the ways they proposed. My amber12 is compiled with gnu and the cuda driver version is 310.19, and after been compiled, all of the cuda test have passed.
My machine information:
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
Linux distribution: CentOS 6.3
GPU: 4*M2090
CPU: 2*Intel(R) Xeon(R) CPU E5-2620 0 . 2.00GHz
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
My input file and log file is as below:
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
md script
&cntrl
!irest=1, ntx=5, ! restart MD
imin=0, ! run MD
ntb=2, ntp=1, ! periodic box
ntpr=5000, ntwr=10000, ntwx=5000,
ntt=3, gamma_ln=2.0, tempi = 300.0, temp0 = 300.0,
pres0=1.0, taup=2.0, ig=4052, ! rand number
ntf=2, ntc=2, ! SHAKE
nstlim=100000000, dt=0.002, ! steptime
cut=8.0, ntr=1,
ioutfm=1,
/
RESTRAINT
100.0
RES 1 599
END
END
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
cudaMemcpy GpuBuffer::Download failed unspecified launch failure
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
Any help will be appreciated. Thanks a lot!
Yours,
Jun
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Wed Dec 12 2012 - 01:00:02 PST
