numpy 1.23.x

[mkoeppe]

(split out from #34081)

​https://numpy.org/doc/stable/release/1.23.0-notes.html# (released)

The upgrade looks harmless.

1.23.x requires GCC >= 6.5. (This has been documented in numpy/numpy#22155 - to be included in 1.23.3.)

We set #34266 as a dependency, which increases our requirement from GCC >= 6.3 to GCC >= 8)

Previous update:

#32423 Update numpy to 1.22.x, scipy 1.8.x, networkx 2.8.x

Depends on #34266

CC: @sagetrac-tmonteil

Component: packages: standard

Author: Matthias Koeppe

Branch/Commit: 03fad7d

Reviewer: Dima Pasechnik, John Palmieri

Issue created by migration from https://trac.sagemath.org/ticket/34110

[mkoeppe]

Branch: u/mkoeppe/numpy_1_23

[mkoeppe]

Commit: d87212c

[mkoeppe]

Last 10 new commits:

e1998da	Merge #32423
507fc2e	build/pkgs/numpy: Update to 1.23.0
58d62f4	build/bin/write-dockerfile.sh: ADD src/VERSION.txt
99f3617	Merge #34017
3fab415	build/pkgs/numpy: Switch to using .tar.gz
443b3d3	build/pkgs/numpy/spkg-install.in: Fix 32bit build on x86_64
9cd5ea6	Merge #34017
ee79c2e	build/pkgs/numpy/spkg-install.in: Fix 32bit build on x86_64 (fixup)
2534f5a	build/pkgs/numpy/patches/21891.patch: New
d87212c	Merge #32423

[mkoeppe]

Dependencies: #32423

[mkoeppe]

Reviewer: https://github.com/mkoeppe/sage/actions/runs/2611393906, https://github.com/mkoeppe/sage/actions/runs/2611393910

[mkoeppe]

comment:4

On debian-stretch-standard (https://github.com/mkoeppe/sage/runs/7184259105?check_suite_focus=true):

In file included from numpy/core/src/common/simd/avx512/avx512.h:75:0,
                   from numpy/core/src/common/simd/simd.h:50,
                   from /sage/local/var/lib/sage/venv-python3.10.3/var/tmp/sage/build/numpy-1.23.0/src/numpy/core/src/npysort/x86-qsort.dispatch.cpp:17,
                   from build/src.linux-x86_64-3.10/numpy/core/src/npysort/x86-qsort.dispatch.avx512_skx.cpp:21:
  numpy/core/src/common/simd/avx512/arithmetic.h: In function 'npy_uint32 npyv_reducemax_u32(npyv_u32)':
  numpy/core/src/common/simd/avx512/arithmetic.h:387:89: error: invalid conversion from 'int' to '_MM_PERM_ENUM' [-fpermissive]
           npyv_u32 a3 = _mm512_max_epu32(a2, _mm512_shuffle_epi32(a2, (1<<6 | 0<<4 | 3<<2 | 2)));
                                                                       ~~~~~~~~~~~~~~~~~~~~^~~~
...
 In file included from build/src.linux-x86_64-3.10/numpy/core/src/npysort/x86-qsort.dispatch.avx512_skx.cpp:21:0:
  /sage/local/var/lib/sage/venv-python3.10.3/var/tmp/sage/build/numpy-1.23.0/src/numpy/core/src/npysort/x86-qsort.dispatch.cpp: In function 'npy_int partition_vec(type_t*, npy_intp, npy_intp, zmm_t, zmm_t, zmm_t*, zmm_t*)':
  /sage/local/var/lib/sage/venv-python3.10.3/var/tmp/sage/build/numpy-1.23.0/src/numpy/core/src/npysort/x86-qsort.dispatch.cpp:654:64: error: there are no arguments to '_knot_mask16' that depend on a template parameter, so a declaration of '_knot_mask16' must be available [-fpermissive]
       vtype::mask_compressstoreu(arr + left, _knot_mask16(gt_mask), curr_vec);
                                                                  ^

(regression from numpy 1.22.x). Reported at numpy/numpy#21920

[sagetrac-git]

Changed commit from d87212c to 8f5860a

[sagetrac-git]

Branch pushed to git repo; I updated commit sha1. New commits:

4cce3ef	Merge tag '9.7.beta4' into t/34110/numpy_1_23
8f5860a	build/pkgs/numpy: Update to 1.23.1

[mkoeppe]

comment:7

ubuntu-trusty-toolchain-gcc_9 (https://github.com/mkoeppe/sage/runs/7532140899?check_suite_focus=true) and other ubuntu releases

sage -t --random-seed=291926986539572225544641036716487640141 src/sage/matrix/matrix_double_dense.pyx
**********************************************************************
File "src/sage/matrix/matrix_double_dense.pyx", line 3642, in sage.matrix.matrix_double_dense.Matrix_double_dense.exp
Failed example:
    A.exp()  # tol 1e-15
Expected:
    [51.968956198705044  74.73656456700327]
    [112.10484685050491 164.07380304920997]
Got:
    [ 51.96895619870499  74.73656456700321]
    [112.10484685050481 164.07380304920983]
Tolerance exceeded in 1 of 4:
    51.968956198705044 vs 51.96895619870499, tolerance 2e-15 > 1e-15
**********************************************************************
File "src/sage/matrix/matrix_double_dense.pyx", line 3655, in sage.matrix.matrix_double_dense.Matrix_double_dense.exp
Failed example:
    A.exp()   # tol 1e-15
Expected:
    [51.968956198705044  74.73656456700327]
    [112.10484685050491 164.07380304920997]
Got:
    [ 51.96895619870499  74.73656456700321]
    [112.10484685050481 164.07380304920983]
Tolerance exceeded in 1 of 4:
    51.968956198705044 vs 51.96895619870499, tolerance 2e-15 > 1e-15
**********************************************************************

sage -t --random-seed=291926986539572225544641036716487640141 src/sage/symbolic/constants_c_impl.pxi
**********************************************************************
File "src/sage/symbolic/constants_c_impl.pxi", line 164, in sage.symbolic.constants_c_impl.E.__pow__
Failed example:
    e^A  # rel tol 1e-15
Expected:
    [51.968956198705044  74.73656456700327]
    [112.10484685050491 164.07380304920997]
Got:
    [ 51.96895619870499  74.73656456700321]
    [112.10484685050481 164.07380304920983]
Tolerance exceeded in 1 of 4:
    51.968956198705044 vs 51.96895619870499, tolerance 2e-15 > 1e-15

[mkoeppe]

Changed reviewer from https://github.com/mkoeppe/sage/actions/runs/2611393906, https://github.com/mkoeppe/sage/actions/runs/2611393910 to https://github.com/mkoeppe/sage/actions/runs/2748565810, https://github.com/mkoeppe/sage/actions/runs/2743631299

[sagetrac-git]

Branch pushed to git repo; I updated commit sha1. New commits:

7aec01a

src/sage/matrix/matrix_double_dense.pyx, src/sage/symbolic/constants_c_impl.pxi: Increase tolerances for RDF matrix exp

[sagetrac-git]

Changed commit from 8f5860a to 7aec01a

[sagetrac-git]

Branch pushed to git repo; I updated commit sha1. This was a forced push. New commits:

[sagetrac-git]

Changed commit from 7aec01a to 8f5860a

[mkoeppe]

Changed dependencies from #32423 to #34266