[gmx-users] OpenCL compile error

Geert Konijnendijk geert at konijnendijk.info
Mon Feb 22 18:31:09 CET 2016


Dear GMX users,

When I try to run a simulation, I get the following error:

Selecting kernel for AMD
The OpenCL compilation log has been saved in "nbnxn_ocl_kernels.cl.FAILED"

-------------------------------------------------------
Program:     gmx mdrun, version 2016-dev-20160222-29943fe
Source file: src\gromacs\mdlib\nbnxn_ocl\nbnxn_ocl_jit_support.cpp (line
230)

Fatal error:
Failed to compile NBNXN kernels for GPU #Hawaii: /M3l

For more information and tips for troubleshooting, please check the GROMACS
website at http://www.gromacs.org/Documentation/Errors
-------------------------------------------------------

The log file contains the following: 

Compilation of source file failed! 
-- Used build options: -DWARP_SIZE_TEST=64 -D_AMD_SOURCE_
-DGMX_OCL_FASTGEN_ADD_TWINCUT -DEL_EWALD_ANA -DEELNAME=_ElecEw
-DVDWNAME=_VdwLJ -DCENTRAL=22 -DNBNXN_GPU_NCLUSTER_PER_SUPERCLUSTER=8
-DNBNXN_GPU_CLUSTER_SIZE=8 -DNBNXN_GPU_JGROUP_SIZE=4
-DNBNXN_AVOID_SING_R2_INC=1.0e-12f
-I"C:\path_to_gromacs\src\gromacs\mdlib\nbnxn_ocl" 
--------------LOG START---------------
C:\Users\Geert\AppData\Local\Temp\\OCL5788T8.cl:59:10: fatal error:
'nbnxn_ocl_kernels_fastgen_add_twincut.clh' file not found
#include FLAVOR_LEVEL_GENERATOR
         ^
C:\Users\Geert\AppData\Local\Temp\\OCL5788T8.cl:45:36: note: expanded from
macro 'FLAVOR_LEVEL_GENERATOR'
    #define FLAVOR_LEVEL_GENERATOR
"nbnxn_ocl_kernels_fastgen_add_twincut.clh"
                                   ^

error: Clang front-end compilation failed!
Frontend phase failed compilation.
Error: Compiling CL to IR
---------------LOG END----------------

When I try to compile the kernel in AMD CodeXL with the same build options
as listed in the log above, it fails with the same error for most GPU
architectures, but it succeeds for some. 

I am running on Windows 10, with the OpenCL library from the AMD App SDK
3.0. 

Is this a known problem and is there a simple solution, or could this be a
bug in the AMD OpenCL compiler or the gromacs OpenCL code?

Thanks in advance!

Geert



More information about the gromacs.org_gmx-users mailing list