Unable to run tutorial amg
Opened this issue · 1 comments
I tried "amg" in the tutorial but there were problems. The software worked for "CG solver" and "CG solver with AMG preconditioner, ONEPASS COARSENING, DIRECT INTERPOLATION" but failed for "CG with AMG preconditioner, AG COARSENING (PMIS), AG INTERPOLATION". The output is below. Any ideas?
Peter Yim
./amg ../testdata/mat65k.mtx
Device Info
Name: Unknown CPU
Vendor: Intel(R) Corporation
Type: CPU
Available: 1
Max Compute Units: 8
Max Work Group Size: 8192
Global Mem Size: 32728408064
Local Mem Size: 32768
Local Mem Type: 2
Host Unified Memory: 1
Reading matrix...
Reading matrix completed.
-- CG solver (no preconditioner, warmup) --
Solver time: 1.58202
Relative residual: 9.60749e-09
Iterations: 453
Relative deviation from result: 1.13681e-08
-- CG with AMG preconditioner, ONEPASS COARSENING, DIRECT INTERPOLATION --
- Setup phase (ViennaCL types)...
Setup time: 0.767442
- CG solver (ViennaCL types)...
Solver time: 0.294805
Relative residual: 9.2538e-09
Iterations: 12
Relative deviation from result: 1.27865e-08
-- CG with AMG preconditioner, AG COARSENING (PMIS), AG INTERPOLATION --
amg: /home/ec2-user/viennacl-dev/viennacl/ocl/handle.hpp:195: const viennacl::ocl::context& viennacl::ocl::handle<OCL_TYPE>::context() const [with OCL_TYPE = cl_mem*]: Assertion `p_context != __null && bool("Logic error: Accessing dangling context from handle.")' failed.
Aborted
Able to get a crash from the GPU also
Device Info
Name: Hainan
Vendor: Advanced Micro Devices, Inc.
Type: GPU
Available: 1
Max Compute Units: 5
Max Work Group Size: 256
Global Mem Size: 2146349056
Local Mem Size: 32768
Local Mem Type: 1
Host Unified Memory: 0
Reading matrix...
Reading matrix completed.
-- CG solver (no preconditioner, warmup) --
Solver time: 65.5498
Relative residual: 9.54852e-09
Iterations: 1206
Relative deviation from result: 4.3179e-07
-- CG with AMG preconditioner, ONEPASS COARSENING, DIRECT INTERPOLATION --
- Setup phase (ViennaCL types)...
terminate called after throwing an instance of 'viennacl::linalg::amg_coarse_problem_too_large_exception'
what(): No further coarsening possible (437 coarse points). Consider changing the strong connection threshold or increasing the coarsening cutoff.
Aborted (core dumped)