Skip to content

Describe numpy/MKL issues as comment in dpnp_eig_c #1006

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 2 commits into from
Oct 18, 2021
Merged

Describe numpy/MKL issues as comment in dpnp_eig_c #1006

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
0f1288a
Desribe numpy/MKL issues as comment in dpnp_eig_c
densmirn Oct 14, 2021
e215eb3
Merge branch 'master' into bug/import_order2
Alexander-Makaryev Oct 18, 2021
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
9 changes: 9 additions & 0 deletions dpnp/backend/kernels/dpnp_krnl_common.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -401,6 +401,15 @@ void dpnp_eig_c(const void* array_in, void* result1, void* result2, size_t size)
const std::int64_t scratchpad_size = mkl_lapack::syevd_scratchpad_size<double>(
DPNP_QUEUE, oneapi::mkl::job::vec, oneapi::mkl::uplo::upper, size, lda);

// https://github.com/IntelPython/dpnp/issues/1005
// Test tests/test_linalg.py::test_eig_arange raises 2 issues in dpnp_eig_c on CPU
// 1. Call of mkl_lapack::syevd_scratchpad_size<double> returns wrong value that causes out of memory issue.
// 2. Call of the function oneapi::mkl::lapack::syevd causes segfault.
// Example of the command to reproduce the issues:
// SYCL_DEVICE_FILTER=cpu pytest tests/test_linalg.py::test_eig_arange[2-float64]
// High-level reason of the issues is numpy is imported before dpnp in third party tests.
// Low-level reason of the issues could be related to MKL runtime library loaded during numpy import.

double* scratchpad = reinterpret_cast<double*>(dpnp_memory_alloc_c(scratchpad_size * sizeof(double)));

event = mkl_lapack::syevd(DPNP_QUEUE, // queue
Expand Down