Samuel Thibault
58330fd908
Add stencil example
|
14 years ago | |
|---|---|---|
| .. | ||
| 0.5.out | 14 years ago | |
| 0.out | 14 years ago | |
| 1.out | 14 years ago | |
| 2.out | 14 years ago | |
| 3.out | 14 years ago | |
| 4.out | 14 years ago | |
| 6.out | 14 years ago | |
| Makefile.am | 14 years ago | |
| README | 14 years ago | |
| life.c | 14 years ago | |
| life_cuda.cu | 14 years ago | |
| mpi.out | 14 years ago | |
| results | 14 years ago | |
| run | 14 years ago | |
| shadow.cu | 14 years ago | |
| shadow.h | 14 years ago | |
| stencil-blocks.c | 14 years ago | |
| stencil-kernels.c | 14 years ago | |
| stencil-tasks.c | 14 years ago | |
| stencil.c | 14 years ago | |
| stencil.h | 14 years ago | |
README
This is a sample 3D stencil application (here just using the game of life rules
for simplicity), split on the z axis.
This is a suggest order of read:
life.c
life.cu: Heart of the stencil computation: compute a new state from an old one.
shadow.cu
shadow.h: Perform replication of data on X and Y edges, to fold the domain on
itself through mere replication of the source state.
stencil.h: Declarations
stencil-kernels.c: Computation Kernels
stencil-blocks.c: Manage block and tags allocation
stencil-tasks.c: Schedule tasks for updates and saves
stencil.c: Main application
*.out: various results according to beta value (communication vs computation
penalty ratio), run make pics or make view to get pictures.
mpi.out: results on MPI.
results: a few results
for simplicity), split on the z axis.
This is a suggest order of read:
life.c
life.cu: Heart of the stencil computation: compute a new state from an old one.
shadow.cu
shadow.h: Perform replication of data on X and Y edges, to fold the domain on
itself through mere replication of the source state.
stencil.h: Declarations
stencil-kernels.c: Computation Kernels
stencil-blocks.c: Manage block and tags allocation
stencil-tasks.c: Schedule tasks for updates and saves
stencil.c: Main application
*.out: various results according to beta value (communication vs computation
penalty ratio), run make pics or make view to get pictures.
mpi.out: results on MPI.
results: a few results