You cannot select more than 25 topics
Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
87eca22895
- psi_dswapdata / psb_comm_rma: RMA swapdata optimization (PSCW + Win_allocate) - psb_d_nest_builder: builder updates - cuda: Makefile + psb_d_cuda_vect_mod adjustments - test/comm: spmv/cg drivers, Makefiles, GPU + scorep/strong-scaling scripts - test/nested: halo-regime test + Makefile/CMake updates |
2 weeks ago | |
|---|---|---|
| .. | ||
| Makefile | 3 months ago | |
| README.md | 3 months ago | |
| gpu_cg.sh | 2 weeks ago | |
| psb_c_comm_cg_test.F90 | 1 month ago | |
| psb_comm_cg_test.F90 | 2 weeks ago | |
| psb_d_comm_cg_test.F90 | 1 month ago | |
| psb_s_comm_cg_test.F90 | 1 month ago | |
| psb_z_comm_cg_test.F90 | 1 month ago | |
README.md
CG no-preconditioner communication test
This test lives under test/comm/cg and builds a local executable:
- source:
psb_comm_cg_test.F90 - executable:
runs/psb_comm_cg_test
Behavior:
- generates a 3D PDE matrix using local
psb_d_gen_pde3d(double precision) - solves with
CG - uses preconditioner
NONE - runs CG three times, changing communication scheme of
xeach run
Communication pattern used in this test:
- reset solution vector
- set communication scheme on
x%v%comm_handle - run full
psb_krylov('CG', ...) - collect and compare solve time
Schemes compared:
psb_comm_isend_irecv_psb_comm_ineighbor_alltoallv_psb_comm_persistent_ineighbor_alltoallv_
How to run
From this directory:
make run(defaults:NP=4,IDIM=40)make run NP=8 IDIM=80
Default PDE-generated matrix:
./runs/psb_comm_cg_test [idim] [nrep] [nwarm] [itmax] [--gpu=TRUE|FALSE]
External matrix input:
./runs/psb_comm_cg_test [idim] [nrep] [nwarm] [itmax] --matrix=<path> [--fmt=MM|HB] [--gpu=TRUE|FALSE]
When --matrix is provided, the test reads and distributes that matrix and ignores the PDE generator.