BUG: Computer crash with numpy 1.23.4 and eigvals / eigvalsh on large matrices #22516
Description
Describe the issue:
My entire computer crashes when computing eigenvalues on the pip precompiled numpy but not the source/compiled version.
The version of LAPACK/BLAS changes and the pip version was multithreaded but the locally compiled version was not.
The matrices were not large enough to cause memory pressure on this machine. I have used the same numpy version on different machines without any issue so I am including CPU information as well:
broken on: Intel(R) Core(TM) i9-7900X CPU @ 3.30GHz
It worked on a Ryzen 5900x however.
Reproduce the code example:
import numpy as np
np.linalg.eigvals(np.random.random((1000,1000)))Error message:
N/A - computer crashes and no logs are available afaik.
Help getting relevant logs appreciated.NumPy/Python version information:
Broken
>>> import sys, numpy; print(numpy.__version__, sys.version)
1.23.4 3.10.6 (main, Aug 10 2022, 11:40:04) [GCC 11.3.0]
>>> numpy.show_config()
openblas64__info:
libraries = ['openblas64_', 'openblas64_']
library_dirs = ['/usr/local/lib']
language = c
define_macros = [('HAVE_CBLAS', None), ('BLAS_SYMBOL_SUFFIX', '64_'), ('HAVE_BLAS_ILP64', None)]
runtime_library_dirs = ['/usr/local/lib']
blas_ilp64_opt_info:
libraries = ['openblas64_', 'openblas64_']
library_dirs = ['/usr/local/lib']
language = c
define_macros = [('HAVE_CBLAS', None), ('BLAS_SYMBOL_SUFFIX', '64_'), ('HAVE_BLAS_ILP64', None)]
runtime_library_dirs = ['/usr/local/lib']
openblas64__lapack_info:
libraries = ['openblas64_', 'openblas64_']
library_dirs = ['/usr/local/lib']
language = c
define_macros = [('HAVE_CBLAS', None), ('BLAS_SYMBOL_SUFFIX', '64_'), ('HAVE_BLAS_ILP64', None), ('HAVE_LAPACKE', None)]
runtime_library_dirs = ['/usr/local/lib']
lapack_ilp64_opt_info:
libraries = ['openblas64_', 'openblas64_']
library_dirs = ['/usr/local/lib']
language = c
define_macros = [('HAVE_CBLAS', None), ('BLAS_SYMBOL_SUFFIX', '64_'), ('HAVE_BLAS_ILP64', None), ('HAVE_LAPACKE', None)]
runtime_library_dirs = ['/usr/local/lib']
Supported SIMD extensions in this NumPy install:
baseline = SSE,SSE2,SSE3
found = SSSE3,SSE41,POPCNT,SSE42,AVX,F16C,FMA3,AVX2,AVX512F,AVX512CD,AVX512_SKX
not found = AVX512_KNL,AVX512_KNM,AVX512_CLX,AVX512_CNL,AVX512_ICL
Working version on same machine
>>> import sys, numpy; print(numpy.__version__, sys.version)
1.23.4 3.10.6 (main, Aug 10 2022, 11:40:04) [GCC 11.3.0]
>>> numpy.show_config()
blas_armpl_info:
NOT AVAILABLE
blas_mkl_info:
NOT AVAILABLE
blis_info:
NOT AVAILABLE
openblas_info:
NOT AVAILABLE
accelerate_info:
NOT AVAILABLE
atlas_3_10_blas_threads_info:
NOT AVAILABLE
atlas_3_10_blas_info:
NOT AVAILABLE
atlas_blas_threads_info:
NOT AVAILABLE
atlas_blas_info:
language = c
define_macros = [('HAVE_CBLAS', None), ('NO_ATLAS_INFO', -1)]
libraries = ['f77blas', 'cblas', 'atlas', 'f77blas', 'cblas']
library_dirs = ['/usr/lib/x86_64-linux-gnu']
blas_opt_info:
language = c
define_macros = [('HAVE_CBLAS', None), ('NO_ATLAS_INFO', -1)]
libraries = ['f77blas', 'cblas', 'atlas', 'f77blas', 'cblas']
library_dirs = ['/usr/lib/x86_64-linux-gnu']
lapack_armpl_info:
NOT AVAILABLE
lapack_mkl_info:
NOT AVAILABLE
openblas_lapack_info:
NOT AVAILABLE
openblas_clapack_info:
NOT AVAILABLE
flame_info:
NOT AVAILABLE
atlas_3_10_threads_info:
NOT AVAILABLE
atlas_3_10_info:
NOT AVAILABLE
atlas_threads_info:
NOT AVAILABLE
atlas_info:
language = f77
libraries = ['lapack', 'f77blas', 'cblas', 'atlas', 'f77blas', 'cblas']
library_dirs = ['/usr/lib/x86_64-linux-gnu']
define_macros = [('NO_ATLAS_INFO', -1)]
lapack_opt_info:
language = f77
libraries = ['lapack', 'f77blas', 'cblas', 'atlas', 'f77blas', 'cblas']
library_dirs = ['/usr/lib/x86_64-linux-gnu']
define_macros = [('NO_ATLAS_INFO', -1)]
Supported SIMD extensions in this NumPy install:
baseline = SSE,SSE2,SSE3
found = SSSE3,SSE41,POPCNT,SSE42,AVX,F16C,FMA3,AVX2,AVX512F,AVX512CD,AVX512_SKX
not found = AVX512_KNL,AVX512_KNM,AVX512_CLX,AVX512_CNL,AVX512_ICL
Context for the issue:
A lot of my work involves getting eigenvalues of matrices, but I do have a fallback which works.