Author: Jeff Larkin [email protected] Some examples also from Keita Teranishi
These examples were put together for a Winter 2012 workshop at ORNL introducing the Titan Cray XK6 system. They are designed as only simple examples of how to build using various programming models and compilers available on that system using the Cray build environment. Because the examples were devised during a time of rapid development of the respective packages, these examples, expecially the Makefiles, need to be made current from time to time. If you have issues, please email me at the address above for support.
-
Example 1 demonstrates building a simple, single-node test case for CUDA for C and CUDA Fortran.
-
Example 2 extends example 1 for building with MPI as well. Running the resulting executable with more than 1 process per node will give errors.
-
Example 3 builds a simple OpenCL example. The Makefile must change certain flaps for support under different compilers. The following warning is expected when building using the PGI compiler
cc -ta=nvidia:4.1 -c -o openclExample.o openclExample.c PGC-W-0267-#warning -- Need to implement some method to align data here (/opt/nvidia/cudatoolkit/5.0.33.103/include/CL/cl_platform.h: 408) PGC/x86-64 Linux 12.8-0: compilation completed with warnings
-
Example 4 demonstrates calling libsci_acc from a C codeand requires the Cray Compiler Environment (CCE)
-
Example 5 demonstrates calling libsci_acc from CUDA Fortran (CURRENTLY BROKEN)
-
Example 6 demonstrates calling CUDA from OpenACC. It currently only supports the CCE.
-
Example 7 demonstrates calling libsci_acc from OpenACC. (WORKING BUT OUTDATED)
-
Example 7 demonstrates calling libsci_acc from PGIACC. (CURRENTLY BROKEN)