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.
psblas3/test/comm/spmv
Stack-1 87eca22895 [WIP] RMA swapdata optimization, nest builder updates, GPU/scorep test tooling
- 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 [WIP] RMA swapdata optimization, nest builder updates, GPU/scorep test tooling 2 weeks ago
README.md [UPDATE] Updated communicationtest to read also external matrices 3 months ago
extract_scorep.sh [WIP] RMA swapdata optimization, nest builder updates, GPU/scorep test tooling 2 weeks ago
gpu_spmv.sh [WIP] RMA swapdata optimization, nest builder updates, GPU/scorep test tooling 2 weeks ago
psb_c_spmv_test.f90 [UPDATE] Added comm tests for all precisions and fixed spmm test in computational_routines 1 month ago
psb_d_spmv_test.f90 [UPDATE] Added comm tests for all precisions and fixed spmm test in computational_routines 1 month ago
psb_s_spmv_test.f90 [UPDATE] Added comm tests for all precisions and fixed spmm test in computational_routines 1 month ago
psb_spmv_test.f90 [WIP] RMA swapdata optimization, nest builder updates, GPU/scorep test tooling 2 weeks ago
psb_z_spmv_test.f90 [UPDATE] Added comm tests for all precisions and fixed spmm test in computational_routines 1 month ago
strong_spmv.sh [WIP] RMA swapdata optimization, nest builder updates, GPU/scorep test tooling 2 weeks ago

README.md

spmv overlap communication test

This test was added after introducing different communication schemes in PSBLAS.

It exercises the overlapped SpMV communication path inside psb_spmm.

Communication pattern:

  • split exchange/computation flow (start + local compute + wait)
  • halo/overlap update through internal swap routines used by SpMV kernels
  • same matrix/vector workload repeated across schemes for timing comparison

Communication schemes compared:

  • psb_comm_isend_irecv_
  • psb_comm_ineighbor_alltoallv_
  • psb_comm_persistent_ineighbor_alltoallv_

Unlike swapdata/, which checks direct halo exchange, this test covers the overlapped SpMV workflow.

Run options

  • Default PDE-generated matrix: ./runs/psb_spmv_kernel [--gpu=TRUE|FALSE] [--nooverlap]
  • External matrix: ./runs/psb_spmv_kernel [--gpu=TRUE|FALSE] --matrix=<path> [--fmt=MM|HB] [--nooverlap]

The overlap path is enabled by default; pass --nooverlap to force the non-overlapped halo-update path.

When --matrix is provided, the benchmark reads and distributes that matrix instead of generating the 3D PDE test matrix.