2.5. Troubleshooting¶
This section explains error messages that may occur during \({\mathcal H}\Phi\) execution and how to resolve them.
2.5.1. Quick Reference¶
Command-line and File Errors
Error Message |
Cause |
Solution |
|---|---|---|
|
Invalid number of command-line arguments |
Run as |
|
Specified file does not exist |
Check file path and filename |
|
Cannot open input file |
Verify file existence and read permissions |
Keyword and Parameter Errors
Error Message |
Cause |
Solution |
|---|---|---|
|
Non-existent keyword specified |
Check manual for supported keywords |
|
Same keyword specified twice |
Remove duplicate keyword |
|
Required keyword not specified |
Add the required keyword |
Model and Solver Errors
Error Message |
Cause |
Solution |
|---|---|---|
|
Unsupported solver specified |
Use Lanczos, TPQ, FullDiag, CG, or TimeEvolution |
|
Unsupported model specified |
Use Hubbard, Spin, Kondo, or SpinlessFermion |
|
Unsupported model/lattice combination |
Use Expert mode or choose a supported combination |
Memory Errors
Error Message |
Cause |
Solution |
|---|---|---|
|
Memory allocation failed |
Reduce system size or use a machine with more memory |
|
Not enough memory per process |
Reduce MPI process count to increase memory per process |
MPI Errors
Error Message |
Cause |
Solution |
|---|---|---|
|
OpenMP and MPI combination is restricted |
Use either OpenMP or MPI |
|
MPI process count is not a power of 2 |
Set MPI process count to 1, 2, 4, 8, 16, etc. |
|
Insufficient MPI processes |
Increase the process count |
Hamiltonian Validation Errors (Canonical Ensemble)
The following errors occur when the combination of site count, electron count, and spin is physically impossible in canonical ensemble calculations.
Error Message |
Meaning |
|---|---|
|
Absolute value of Sz exceeds half the number of sites |
|
Electron count exceeds twice the number of sites |
|
Parity of electron count and Sz does not match |
|
Sz is too large for the electron count |
|
Sz is too large when electron count is high |
|
Sz exceeds half the number of sites in spin model |
|
Parity of site count and Sz does not match in spin model |
|
Electron count exceeds site count in Kondo model |
|
Parity of electron count and Sz does not match in Kondo model |
Solution: Check input parameters Nsite, Nelec, 2Sz and correct to physically valid combinations.
Warning Messages
Message |
Description |
|---|---|
|
Specified parameter is not used in current calculation. Comment out if unnecessary |
|
Default value is being used (not an error) |
2.5.2. Detailed Solutions for Common Errors¶
Out of Memory Error¶
Symptom: Malloc ERROR is displayed and calculation cannot start.
Cause: Hilbert space dimension is too large and required memory cannot be allocated.
Solutions:
Reduce system size
Reducing the number of sites exponentially decreases the Hilbert space dimension.
Increase MPI process count
Using more processes like
mpirun -np 16 HPhi ...distributes the Hilbert space across processes.Check available memory
Review
CHECK_Memory.datto understand required memory.
MPI Process Count Error¶
Symptom: ERROR: MPI process count is not a power of 2
Cause: \({\mathcal H}\Phi\) requires MPI process count to be a power of 2 (1, 2, 4, 8, 16, 32, …).
Solution:
# Correct examples
mpirun -np 4 HPhi -e namelist.def
mpirun -np 8 HPhi -e namelist.def
mpirun -np 16 HPhi -e namelist.def
# Wrong example (3 is not a power of 2)
mpirun -np 3 HPhi -e namelist.def # Error
File Not Found Error¶
Symptom: Error: The file (filename) does not exist.
Solutions:
Verify filenames in namelist.def
For relative paths, check position relative to execution directory
Check case sensitivity of filenames (Linux/macOS are case-sensitive)
Convergence Issues¶
Symptom: Lanczos or CG method does not converge.
Solutions:
Increase Lanczos steps
Adjust
LanczosEpsorLanczos_maxin ModPara fileChange initial vector
Try different random seeds with
initial_ivparameterDegenerate eigenvalues
Convergence may be slow when degenerate eigenvalues exist