Hello, Éric! Please readdress your questions to the whole list.
In this case, my problem was not lammps related. I think I might have
overlooked some warnings/errors in the compilation when compiling CUDA
and the original files were for superuser only, so after a chown I could simply
recompile it.
Pablo
you have to tell LAMMPS you want to use atom style atomic/cuda,
either via the -suffix cuda command-line switch or by
using it explicitly
atom_style atomic/cuda
Steve
Dear Steve,
Thank you for your email.
Actually that’s what I do.
The point is it works with the in.melt_2.5.cuda in.melt_5.0.cuda examples but no with in.phosphate.cuda and in.rhodo.cuda.
For this two cases I get
ERROR: Invalid pair style (…/force.cpp:168
and
ERROR: Invalid atom style (…/atom.cpp:331)
respectively.
I’m using the release lammps-17Jun13.
Right now I’m trying to build lammps-26Jul13 but encountering some issue.
For the CPU version I get
…/pair_tip4p_long_omp.cpp(104): error: class “LAMMPS_NS::ThrData” has no member “timer”
thr->timer(ThrData::TIME_START);
For the GPU
*** No rule to make target `…/…/lib/cuda/Makefile.lammps’. Stop.
Do you have any idea about what’s going wrong here?
Thanks again,
Best,
Éric.
The first error is due to an oversight of mine. Please just delete those offending timer lines. They refer to code that has not yet been merged and was too careless when sending it to Steve.
Sorry,
Axel
Thank you Axel,
This “timer” issue is solved.
Éric.
Dear all,
Does someone have successfully made the GPU built of the 26Jul13 release?
I can compile the cuda library but no Makefile.lammps is created.
Thank you,
Éric.
After put the Makefile.lammps myself in lib/cuda I got this error:
…/fft3d_cuda.cpp:110: error: cannot convert âdoubleâ to âFFT_SCALARâ for argument â1â to âvoid remap_3d(FFT_SCALAR*, FFT_SCALAR*, FFT_SCALAR*, remap_plan_3d*)â
make[1]: *** [fft3d_cuda.o] Error 1
With the 17Jun13 release it worked smoothly expected when I executed it with some examples.
i’ve fixed some compilation issues in USER-CUDA and reinstated the original Makefile.lammps file (since it is adjusted automatically when compiling the cuda library for LAMMPS, unlike the other variants that need to be manually edited or copied from templates).
you can download it from here (differences between the sandia version and LAMMPS-ICMS are currently mostly cosmetic, as you can see from the list here: https://sites.google.com/site/akohlmey/software/lammps-icms#TOC-LAMMPS-ICMS-Features )
The missing file for this was added back in - will be in the next
match - so hopefully the lib/cuda build will work for you.
Steve
The missing file for this was added back in - will be in the next
match - so hopefully the lib/cuda build will work for you.
hi steve,
you also need this:
--- a/src/USER-CUDA/fix_nh_cuda.cpp<http://git.icms.temple.edu/git/?p=lammps-icms.git;a=blob;f=src/USER-CUDA/fix_nh_cuda.cpp;h=252f260885f2e19b530f23a62b676d25c45a59ef>
+++ b/src/USER-CUDA/fix_nh_cuda.cpp<http://git.icms.temple.edu/git/?p=lammps-icms.git;a=blob;f=src/USER-CUDA/fix_nh_cuda.cpp;h=c0beac218241dd17b25f2333630f50db38aea217;hb=e6f6722365f34ccbdebee4c59f4fda227f0515af>
@@ -335,7<http://git.icms.temple.edu/git/?p=lammps-icms.git;a=blob;f=src/USER-CUDA/fix_nh_cuda.cpp;h=252f260885f2e19b530f23a62b676d25c45a59ef#l335>
+335,6<http://git.icms.temple.edu/git/?p=lammps-icms.git;a=blob;f=src/USER-CUDA/fix_nh_cuda.cpp;h=c0beac218241dd17b25f2333630f50db38aea217;hb=e6f6722365f34ccbdebee4c59f4fda227f0515af#l335>@@
FixNHCuda::FixNHCuda(LAMMPS *lmp, int narg, char **arg) : Fix(lmp, narg, arg)
if (p_flag[i]) pstat_flag = 1;
if (pstat_flag) {
- box_change = 1;
if (p_flag[0] || p_flag[1] || p_flag[2]) box_change_size = 1;
if (p_flag[3] || p_flag[4] || p_flag[5]) box_change_shape = 1;
no_change_box = 1;