TEST: 1.19.x + blas variants #196

h-vetinari · 2020-06-26T17:51:03Z

Based on #195; doing a similar investigation as in conda-forge/scipy-feedstock#130.

conda-forge-linter · 2020-06-26T17:51:11Z

Hi! This is the friendly automated conda-forge-linting service.

I just wanted to let you know that I linted all conda-recipes in your PR (recipe) and found it was in an excellent condition.

h-vetinari · 2020-06-26T19:40:26Z

This looks quite a bit better than on the scipy-side. Still it did catch some errors that might not show up anywhere else in CI.

I mentioned in conda/conda-build#3947 that IMO, foundational packages like numpy/scipy should be tested across all available blas variants. This would - effectively - provide some further CI on the various blas builds, but also make sure that no broken variants are shipped.

Note: The cells of the matrix have the same behaviour for all python versions (cpython 3.6-3.8, pypy 3.6) unless otherwise noted - hence the simplified matrix

variant	blis	mkl	netlib	openblas
linux / x86	✔️	✔️	✔️ (cpython) / ❌ (pypy)	✔️ (cpython) / ❌ (pypy)
linux / aarch	➖	➖	✔️	✔️
linux / ppc64le	➖	➖	✔️	❌ (non-uniformly)
osx	✔️	✔️	✔️	✔️
win	✔️	✔️	✔️	✔️

There were 4 additional failures not shown in the matrix above:

the failure for osx+mkl on python 3.8 is a spurious http error
aarch+netlib+pypy runs into 60min drone timeout at ~54% of the test suite
ppc64le+netlib had two build script errors on travis, but I don't have rights to restart the job

linux + x86 + netlib/openblas + pypy: segfault, test suite aborted

linalg/tests/test_linalg.py::TestCond::test_sq_cases PASSED              [ 62%]
linalg/tests/test_linalg.py::TestCond::test_empty_sq_cases PASSED        [ 62%]
Fatal Python error: Segmentation fault

Stack (most recent call first, approximate line numbers):
  File "[...snip...]/lib/python3.6/site-packages/numpy/linalg/tests/test_linalg.py", line 81 in check [jitted]
  File "[...snip...]/lib/python3.6/site-packages/numpy/linalg/tests/test_linalg.py", line 337 in check_cases [jitted]
  File "[...snip...]/lib/python3.6/site-packages/numpy/linalg/tests/test_linalg.py", line 337 in check_cases
  File "[...snip...]/lib/python3.6/site-packages/numpy/linalg/tests/test_linalg.py", line 392 in test_generalized_sq_cases
  File "[...snip...]/site-packages/_pytest/python.py", line 175 in pytest_pyfunc_call
  File "[...snip...]/site-packages/pluggy/callers.py", line 157 in _multicall [jit inlined]
  File "[...snip...]/site-packages/pluggy/manager.py", line 84 in <lambda> [jit inlined]
  File "[...snip...]/site-packages/pluggy/manager.py", line 90 in _hookexec [jitted]
  File "[...snip...]/site-packages/pluggy/hooks.py", line 272 in __call__ [jitted]
  File "[...snip...]/site-packages/pluggy/hooks.py", line 272 in __call__
  File "[...snip...]/site-packages/_pytest/python.py", line 1475 in runtest
  File "[...snip...]/site-packages/_pytest/runner.py", line 126 in pytest_runtest_call
  File "[...snip...]/site-packages/pluggy/callers.py", line 157 in _multicall [jit inlined]
  File "[...snip...]/site-packages/pluggy/manager.py", line 84 in <lambda> [jit inlined]
Tests failed for numpy-1.19.0-py36h04dfcbd_0.tar.bz2 - moving package to /home/conda/feedstock_root/build_artifacts/broken
WARNING:conda_build.build:Tests failed for numpy-1.19.0-py36h04dfcbd_0.tar.bz2 - moving package to /home/conda/feedstock_root/build_artifacts/broken
WARNING conda_build.build:tests_failed(2751): Tests failed for numpy-1.19.0-py36h04dfcbd_0.tar.bz2 - moving package to /home/conda/feedstock_root/build_artifacts/broken
TESTS FAILED: numpy-1.19.0-py36h04dfcbd_0.tar.bz2

linux + ppc64le + openblas + cpython 3.6: 1 test failure (TestCholesky.test_basic_property)

=================================== FAILURES ===================================
_______________________ TestCholesky.test_basic_property _______________________

self = <numpy.linalg.tests.test_linalg.TestCholesky object at 0x706e90d77b38>

    def test_basic_property(self):
        # Check A = L L^H
        shapes = [(1, 1), (2, 2), (3, 3), (50, 50), (3, 10, 10)]
        dtypes = (np.float32, np.float64, np.complex64, np.complex128)
    
        for shape, dtype in itertools.product(shapes, dtypes):
            np.random.seed(1)
            a = np.random.randn(*shape)
            if np.issubdtype(dtype, np.complexfloating):
                a = a + 1j*np.random.randn(*shape)
    
            t = list(range(len(shape)))
            t[-2:] = -1, -2
    
            a = np.matmul(a.transpose(t).conj(), a)
            a = np.asarray(a, dtype=dtype)
    
            c = np.linalg.cholesky(a)
    
            b = np.matmul(c, c.transpose(t).conj())
            assert_allclose(b, a,
                            err_msg="{} {}\n{}\n{}".format(shape, dtype, a, c),
>                           atol=500 * a.shape[0] * np.finfo(dtype).eps)
E           AssertionError: 
E           Not equal to tolerance rtol=1e-07, atol=5.55112e-12
E           (50, 50) <class 'numpy.complex128'>
E           [[ 86.08558859 +0.j           3.44218047+15.88127244j
E              -0.45363563 -0.44661413j ...   2.64035992 -9.58169853j
E              -3.23064882 +6.99917231j   3.72228944 -3.50497908j]
E            [  3.44218047-15.88127244j  97.31130762 +0.j
E               4.95610003 +3.51207641j ... -24.44048428 -0.53073368j
E             -21.33047678 +4.24795683j  -2.7632562  +5.87612999j]
E            [ -0.45363563 +0.44661413j   4.95610003 -3.51207641j
E             108.58323952 +0.j         ...   6.24573557 -3.07029841j
E               7.36481721 -5.16506877j  -1.1160803  -0.23457131j]
E            ...
E            [  2.64035992 +9.58169853j -24.44048428 +0.53073368j
E               6.24573557 +3.07029841j ...  97.33841401 +0.j
E              13.64351655 +0.62184078j   3.54722947 +1.06034335j]
E            [ -3.23064882 -6.99917231j -21.33047678 -4.24795683j
E               7.36481721 +5.16506877j ...  13.64351655 -0.62184078j
E              99.49995802 +0.j         -15.73440912 -2.65289272j]
E            [  3.72228944 +3.50497908j  -2.7632562  -5.87612999j
E              -1.1160803  +0.23457131j ...   3.54722947 -1.06034335j
E             -15.73440912 +2.65289272j  89.70910493 +0.j        ]]
E           [[ 9.27823198+0.j          0.        +0.j          0.        +0.j
E             ...  0.        +0.j          0.        +0.j
E              0.        +0.j        ]
E            [ 0.3709953 -1.71167012j  9.70792746+0.j          0.        +0.j
E             ...  0.        +0.j          0.        +0.j
E              0.        +0.j        ]
E            [-0.04889246+0.04813569j  0.52087651-0.35499305j 10.40101916+0.j
E             ...  0.        +0.j          0.        +0.j
E              0.        +0.j        ]
E            ...
E            [ 0.28457576+1.03270737j -2.3463717 -0.03497098j  0.71336226+0.38319797j
E             ...  2.96242819+0.j          0.        +0.j
E              0.        +0.j        ]
E            [-0.3481966 -0.75436488j -2.31692309-0.3473546j   0.81411515+0.58790833j
E             ...  0.99318458+0.51943463j  1.12603697+0.j
E              0.        +0.j        ]
E            [ 0.40118521+0.37776368j -0.23336466-0.69046409j -0.11904651+0.06872805j
E             ...  0.0339795 -0.51402601j  0.96743148-0.46502072j
E              1.24939718+0.j        ]]
E           Mismatched elements: 436 / 2500 (17.4%)
E           Max absolute difference: 94.66308979
E           Max relative difference: 16.00607492
E            x: array([[ 86.085589 +0.j      ,   3.44218 +15.881272j,
E                    -0.453636 -0.446614j, ...,   2.64036  -9.581699j,
E                    -3.230649 +6.999172j,   3.722289 -3.504979j],...
E            y: array([[ 86.085589 +0.j      ,   3.44218 +15.881272j,
E                    -0.453636 -0.446614j, ...,   2.64036  -9.581699j,
E                    -3.230649 +6.999172j,   3.722289 -3.504979j],...

../_test_env_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_/lib/python3.6/site-packages/numpy/linalg/tests/test_linalg.py:1736: AssertionError
=============================== warnings summary ===============================

linux + ppc64le + openblas + cpython 3.7: 4 test failures

=========================== short test summary info ============================
FAILED core/tests/test_multiarray.py::TestDot::test_dot_3args - AssertionError: 
FAILED linalg/tests/test_linalg.py::TestCond::test_sq_cases - AssertionError:...
FAILED linalg/tests/test_linalg.py::TestMatrixPower::test_power_is_minus_one[dt3]
FAILED linalg/tests/test_linalg.py::TestCholesky::test_basic_property - numpy...
= 4 failed, 10394 passed, 568 skipped, 26 deselected, 19 xfailed, 5 xpassed, 233 warnings in 424.23s (0:07:04) =

linux + ppc64le + openblas + cpython 3.8: 3 test failures

=========================== short test summary info ============================
FAILED core/tests/test_multiarray.py::TestDot::test_dot_3args - AssertionError: 
FAILED linalg/tests/test_linalg.py::TestCond::test_sq_cases - AssertionError:...
FAILED linalg/tests/test_linalg.py::TestCholesky::test_basic_property - numpy...
= 3 failed, 10398 passed, 567 skipped, 26 deselected, 19 xfailed, 5 xpassed, 233 warnings in 401.93s (0:06:41) =

h-vetinari · 2020-06-26T19:43:44Z

Maybe worth pinging @conda-forge/help-pypy on this. The segfaults on linux+pypy for numpy 1.19 affect both this PR and #195.

mattip · 2020-06-27T17:59:48Z

What version of OpenBLAS are you testing with? 0.3.10 was released recently and fixes an interaction that would cause a segfault on PyPy on azure machines.

h-vetinari · 2020-06-27T22:40:10Z

@mattip
It was running with libopenblas 0.3.9 before, but even after bumping to 0.3.10 (see conda-forge/blas-feedstock#45), the segfault with openblas persists. Interestingly (and I have no good reason why), the segfault is now gone for the netlib variant (perhaps it's not fully deterministic).

There's more stuff that I don't understand, actually. Bumping blis from 0.6.1 to 0.7.0 has introduced a segfault on cpython 3.6, 12 test failures on 3.7, but runs through fine on 3.8...? 🤔

h-vetinari · 2020-06-27T23:46:14Z

After bumping blis to 0.7.0 and openblas to 0.3.10 (see conda-forge/blas-feedstock#45), the picture changed a bit, in some ways unexpectedly.

In short:

the linux+x86+pypy segfault remains, but seems non-deterministic (in a run before it disappeared only from the netlib build, not the openblas one, now it did appeared for the first time in the mkl build, while disappearing for the others)
linux+ppc64le+openblas failures remained
both pypy builds on aarch time out
win+blis is now failing
TestCholesky.test_basic_property seems to be the most common failure across variants
the failure of win+mkl+py37 is likely spurious

variant	blis	mkl	netlib	openblas
linux / x86	✔️	✔️ (cpython) / ❌ (pypy)	✔️	✔️
linux / aarch	➖	➖	✔️ (cpython) / ❌ (pypy)	✔️ (cpython) / ❌ (pypy)
linux / ppc64le	➖	➖	✔️	❌ (non-uniformly)
osx	✔️	✔️	✔️	✔️
win	❌	✔️	✔️	✔️

linux + x86 + mkl + pypy: segfault, test suite aborted

linalg/tests/test_linalg.py::TestCond::test_sq_cases PASSED              [ 62%]
linalg/tests/test_linalg.py::TestCond::test_empty_sq_cases PASSED        [ 62%]
Fatal Python error: Segmentation fault

Stack (most recent call first, approximate line numbers):
  File "[...snip...]/lib/python3.6/site-packages/numpy/linalg/tests/test_linalg.py", line 81 in check [jitted]
  File "[...snip...]/lib/python3.6/site-packages/numpy/linalg/tests/test_linalg.py", line 337 in check_cases [jitted]
  File "[...snip...]/lib/python3.6/site-packages/numpy/linalg/tests/test_linalg.py", line 337 in check_cases
  File "[...snip...]/lib/python3.6/site-packages/numpy/linalg/tests/test_linalg.py", line 392 in test_generalized_sq_cases
  File "[...snip...]/site-packages/_pytest/python.py", line 175 in pytest_pyfunc_call
  File "[...snip...]/site-packages/pluggy/callers.py", line 157 in _multicall [jit inlined]
  File "[...snip...]/site-packages/pluggy/manager.py", line 84 in <lambda> [jit inlined]
  File "[...snip...]/site-packages/pluggy/manager.py", line 90 in _hookexec [jitted]
  File "[...snip...]/site-packages/pluggy/hooks.py", line 272 in __call__ [jitted]
  File "[...snip...]/site-packages/pluggy/hooks.py", line 272 in __call__
  File "[...snip...]/site-packages/_pytest/python.py", line 1475 in runtest
  File "[...snip...]/site-packages/_pytest/runner.py", line 126 in pytest_runtest_call
  File "[...snip...]/site-packages/pluggy/callers.py", line 157 in _multicall [jit inlined]
  File "[...snip...]/site-packages/pluggy/manager.py", line 84 in <lambda> [jit inlined]
Tests failed for numpy-1.19.0-py36h04dfcbd_0.tar.bz2 - moving package to /home/conda/feedstock_root/build_artifacts/broken
WARNING:conda_build.build:Tests failed for numpy-1.19.0-py36h04dfcbd_0.tar.bz2 - moving package to /home/conda/feedstock_root/build_artifacts/broken
WARNING conda_build.build:tests_failed(2751): Tests failed for numpy-1.19.0-py36h04dfcbd_0.tar.bz2 - moving package to /home/conda/feedstock_root/build_artifacts/broken
TESTS FAILED: numpy-1.19.0-py36h04dfcbd_0.tar.bz2

linux + ppc64le + openblas + cpython: 3 test failures

=========================== short test summary info ============================
FAILED linalg/tests/test_linalg.py::TestCond::test_sq_cases - AssertionError:...
FAILED linalg/tests/test_linalg.py::TestCholesky::test_basic_property - numpy...
FAILED linalg/tests/test_linalg.py::TestTensorinv::test_tensorinv_result - As...
= 3 failed, 10395 passed, 568 skipped, 26 deselected, 19 xfailed, 5 xpassed, 233 warnings in 507.77s (0:08:27) =

linux + ppc64le + openblas + pypy: 4 test failures

=========================== short test summary info ============================
FAILED core/tests/test_multiarray.py::TestDot::test_dot_3args - AssertionError: 
FAILED linalg/tests/test_linalg.py::TestInv::test_sq_cases - AssertionError: ...
FAILED linalg/tests/test_linalg.py::TestCond::test_sq_cases - AssertionError:...
FAILED linalg/tests/test_linalg.py::TestCholesky::test_basic_property - numpy...
= 4 failed, 10337 passed, 612 skipped, 26 deselected, 21 xfailed, 5 xpassed, 233 warnings in 993.28s (0:16:33) =

windows + blis + cpython: 11-12 failed tests

This one seems to be a bit fickle as well - in the previous run, it appeared for py36 & py37 but not py38, not it appeared for py37 & py38 but not py36.

=========================== short test summary info ===========================
FAILED core/tests/test_multiarray.py::TestMatmul::test_matmul_object - Assert...
FAILED linalg/tests/test_linalg.py::TestSolve::test_sq_cases - AssertionError...
FAILED linalg/tests/test_linalg.py::TestSolve::test_generalized_sq_cases - As...
FAILED linalg/tests/test_linalg.py::TestInv::test_sq_cases - AssertionError: ...
FAILED linalg/tests/test_linalg.py::TestInv::test_generalized_sq_cases - Asse...
FAILED linalg/tests/test_linalg.py::TestPinv::test_nonsq_cases - AssertionErr...
FAILED linalg/tests/test_linalg.py::TestPinv::test_generalized_sq_cases - Ass...
FAILED linalg/tests/test_linalg.py::TestPinv::test_generalized_nonsq_cases - ...
FAILED linalg/tests/test_linalg.py::TestDet::test_sq_cases - AssertionError: ...
FAILED linalg/tests/test_linalg.py::TestDet::test_generalized_sq_cases - Asse...
FAILED linalg/tests/test_linalg.py::TestMatrixPower::test_power_is_minus_one[dt13]
FAILED linalg/tests/test_linalg.py::TestCholesky::test_basic_property - Asser...
= 12 failed, 10401 passed, 493 skipped, 19 xfailed, 1 xpassed, 228 warnings in 545.60s (0:09:05) =

mattip · 2020-06-28T08:50:11Z

Thanks for the summary. Numpy regularly builds on travisCI machines with PyPy, both in the numpy/numpy CI and on the MacPython/num-wheels CI. I wonder what is different on the Azure builds?

mattip · 2020-06-28T09:03:46Z

In the build log for pypy I am seeing libcblas: 3.8.0-16_mkl. Am I looking at the wrong logs? I expected to see 3.10.

h-vetinari · 2020-06-28T09:23:11Z

@mattip: In the build log for pypy I am seeing libcblas: 3.8.0-16_mkl. Am I looking at the wrong logs? I expected to see 3.10.

You're looking at the right logs, but I think you're mixing up the reference-lapack version numbers with openblas (note: 3.9.0 isn't in conda-forge yet because MKL isn't compatible yet). libblas / libcblas / liblapack / liblapacke get the lapack version number, and a build string suffix indicating which blas flavour (blis / mkl / netlib = reference-lapack / openblas) - you were looking at a run with mkl.

As I noted in the last comment, the segfault on pypy seems non-deterministic. It appeared for openblas in 2 out of 3 runs, for netlib 1/3 times and for mkl 1/3 times. One of the runs has been deleted by azure, but here's one where the segfault occurred with openblas. I haven't investigated which test follows linalg/tests/test_linalg.py::TestCond::test_empty_sq_cases PASSED [ 62%], but it always seems to happen there.

mattip · 2020-06-28T09:46:27Z

Ahh, right sorry. But in the linux_blas_implopenblasnumpy1.18python3.6.____73_pypytarget_platformlinux-64.yaml log, I see libopenblas: 0.3.9-h5ec1e0e_0. Shouldn't that be 0.3.10?

In NumPy we use add a check like this (from tools/openblas.py:

def test_version(expected_version, ilp64=get_ilp64()):
    """
    Assert that expected OpenBLAS version is
    actually available via NumPy
    """
    import numpy
    import ctypes

    dll = ctypes.CDLL(numpy.core._multiarray_umath.__file__)
    if ilp64 == "64_":
        get_config = dll.openblas_get_config64_
    else:
        get_config = dll.openblas_get_config
    get_config.restype=ctypes.c_char_p
    res = get_config()
    print('OpenBLAS get_config returned', str(res))
    if not expected_version:
        expected_version = OPENBLAS_V
    check_str = b'OpenBLAS %s' % expected_version.encode()
    print(check_str)
    assert check_str in res, '%s not found in %s' %(expected_version, res)
    if ilp64:
        assert b"USE64BITINT" in res
    else:
        assert b"USE64BITINT" not in res

h-vetinari · 2020-06-28T11:13:47Z

@mattip: Shouldn't that be 0.3.10?

You're right, sorry. The run I linked was before conda-forge/blas-feedstock#45 was merged. The first run after that had the segfault, but got deleted by azure since I restarted before everything was through. Let's try that again.

mattip · 2020-06-29T17:43:09Z

this is a tough PR. The starts must align just perfectly for everything to pass in the same run.

h-vetinari · 2020-06-29T17:50:47Z

this is a tough PR. The starts must align just perfectly for everything to pass in the same run.

It seems to me the best answer would be to figure out what causes the segfault and fix it 😉
So far, the pypy build has always failed for at least one of the blas variants.

The blis+win builds are also failing consistently, as are the linux+ppc builds. So we'll probably need more than lucky stars to pull this into the green 😅

mattip · 2020-07-14T20:28:23Z

I wonder if there is a path to release a 1.19 NumPy for at least the most common platforms.

mattip · 2020-07-14T21:00:25Z

I see that OpenBLAS is being built with DYNAMIC_ARCH. Is a base `TARGET being set as well? See for instance MacPython/openblas-libs#34

In [this comment](https://github.com/OpenMathLib/OpenBLAS/issues/2388#issuecomment-582267407), and elsewhere, they recommends using `TARGET` as a minimal platform when `DYNAMIC_ARCH` is used which will be the fallback when no better dynamic loop can be found. The default is the build platform, which may be problematic if the build is done on a more advanced machine than the test.

h-vetinari · 2021-05-08T18:35:37Z

No more 1.19 patch releases expected, time to close this PR

…nda-forge-pinning 2022.04.12.21.56.28

This reverts commit 6dc144f.

h-vetinari · 2022-04-13T00:57:20Z

Revival (new PyPy builds and BLAS updates): all green except PPC

Due to rebuilding 1.19 for pypy3.8/3.9, much less several relevant BLAS (& infrastructure) changes, I thought I'd revive this PR for one last update.

From 9 failures out of 92 runs, we're now at 12 failures (PPC-only) out of 108 runs.

Notable

Added accelerate BLAS flavour on osx
Testing against PyPy 3.8 and 3.9 added everywhere but for osx-arm
Big bumps for openblas, blis & MKL
In the meantime, the previous blis errors have been tracked down and fixed
The broken pipe errors on windows should be gone thanks to the pagefile work
Fixed the error reporting of the test suite (used to be broken due to missing sys.exit wrapper)
Switched to running the full test suite; emulation keeps running only label='fast' tests.

Details

variant	before	after
linux + aarch + pypy3.6 + openblas	one test failure (probably flaky as it's "only" from hypothesis)	gone 🥳
linux + ppc + pypy + netlib	segfault	gone 🥳
linux + ppc	fine otherwise (on travis)	test failures due to emulation problems (on azure)
win + blis	11-12 failures across all python versions	fixed 🥳
win	3 failures due to `The process tried to write to a nonexistent pipe.`	gone 🥳
win + pypy	...	two spurious failures in `test_may_share_memory_easy_fuzz`, resolved by restart

lib	before	after	updated version	updated build
`numpy`	`1.19.5`	`1.19.5`
`libblas`	`3.9.0-7`	`3.9.0-14`		X
`blis`	`0.8.0-1`	`0.9.0-0`	X
`openblas`	`0.3.12-pthreads-1`	`0.3.20-pthreads-0`	X
`mkl`	`2020.4-304`	`2022.0.1-803`	X
`netlib`	`3.9.0-3`	`3.9.0-5`		X
`pypy`	`7.3.3-3`	`7.3.9-1` (pypy38/39) `7.3.7-3` (pypy37)	X
`qemu-user-static`	?	`6.1.0-8`

variant	accelerate	blis	mkl	netlib	openblas	sum*
linux / x86	➖	✔️	✔️	✔️	✔️	-
linux / aarch	➖	➖	➖	✔️	✔️	-
linux / ppc64le	➖	➖	➖	✖️	✖️	12F
osx / arm	✔️	➖	➖	✔️	✔️	-
osx / x86	✔️	✔️	✔️	✔️	✔️	-
win / x86	➖	✔️	✔️	✔️	✔️	-
sum*	-	-	-	6F	6F	12F

* sum of Failures (out of a total of 108 CI combinations being tested)

Build logs:
Azure

h-vetinari requested review from jakirkham, msarahan, ocefpaf, pelson and rgommers as code owners June 26, 2020 17:51

h-vetinari marked this pull request as draft June 26, 2020 17:51

h-vetinari mentioned this pull request Jun 27, 2020

numpy v1.19.0 #195

Merged

3 tasks

h-vetinari closed this Jun 27, 2020

h-vetinari reopened this Jun 27, 2020

h-vetinari closed this Jun 27, 2020

h-vetinari reopened this Jun 27, 2020

h-vetinari closed this Jun 28, 2020

h-vetinari reopened this Jun 28, 2020

isuruf closed this Jul 16, 2020

isuruf reopened this Jul 16, 2020

h-vetinari changed the title ~~WIP: 1.19.5 + blas variants~~ TEST: 1.19.5 + blas variants Feb 10, 2021

h-vetinari closed this May 8, 2021

h-vetinari deleted the 1.19_blas_var branch May 8, 2021 18:35

h-vetinari mentioned this pull request Jun 29, 2021

TEST: 1.21.x + blas variants #237

Closed

h-vetinari mentioned this pull request Jan 3, 2022

TEST: 1.22.x + blas variants #252

Closed

h-vetinari changed the title ~~TEST: 1.19.5 + blas variants~~ TEST: 1.19.x + blas variants Jan 3, 2022

h-vetinari restored the 1.19_blas_var branch April 6, 2022 21:43

h-vetinari deleted the 1.19_blas_var branch April 6, 2022 21:56

h-vetinari mentioned this pull request Apr 7, 2022

moving from master to main mega-issue conda-forge/conda-forge.github.io#1162

Closed

40 tasks

h-vetinari restored the 1.19_blas_var branch April 12, 2022 22:23

h-vetinari reopened this Apr 12, 2022

h-vetinari changed the base branch from master to numpy119 April 12, 2022 22:23

h-vetinari added 3 commits April 13, 2022 09:27

test for all blas variants

11da5a6

MNT: Re-rendered with conda-build 3.21.8, conda-smithy 3.19.0, and co…

661e9f1

…nda-forge-pinning 2022.04.12.21.56.28

Revert "skip test suite on PPC due to QEMU bugs"

f5d4c8b

This reverts commit 6dc144f.

h-vetinari force-pushed the 1.19_blas_var branch from b1d98ba to f5d4c8b Compare April 12, 2022 22:32

conda-forge deleted a comment from conda-forge-linter Apr 12, 2022

only run fast tests on travis

a92855e

h-vetinari closed this Apr 19, 2022

h-vetinari deleted the 1.19_blas_var branch April 21, 2022 11:08

h-vetinari mentioned this pull request Jun 23, 2022

TEST: 1.23.x + blas variants #273

Closed

h-vetinari mentioned this pull request Jan 27, 2023

TEST: 1.24.x + blas variants #288

Closed

This was referenced Jun 18, 2023

TEST: 1.25.x + blas variants #293

Closed

Update following the NumPy 1.25.0 release. scipy/oldest-supported-numpy#69

Merged

h-vetinari mentioned this pull request Jan 9, 2024

TEST: 1.26.x + blas variants #307

Closed

h-vetinari mentioned this pull request Jun 17, 2024

TEST: 2.0.x + blas variants #317

Draft

h-vetinari mentioned this pull request Aug 19, 2024

TEST: 2.1.x + blas variants #326

Draft

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

TEST: 1.19.x + blas variants #196

TEST: 1.19.x + blas variants #196

h-vetinari commented Jun 26, 2020

conda-forge-linter commented Jun 26, 2020

h-vetinari commented Jun 26, 2020 •

edited

Loading

h-vetinari commented Jun 26, 2020

mattip commented Jun 27, 2020

h-vetinari commented Jun 27, 2020

h-vetinari commented Jun 27, 2020

mattip commented Jun 28, 2020

mattip commented Jun 28, 2020

h-vetinari commented Jun 28, 2020 •

edited

Loading

mattip commented Jun 28, 2020

h-vetinari commented Jun 28, 2020

mattip commented Jun 29, 2020

h-vetinari commented Jun 29, 2020

mattip commented Jul 14, 2020

mattip commented Jul 14, 2020

h-vetinari commented May 8, 2021

h-vetinari commented Apr 13, 2022

TEST: 1.19.x + blas variants #196

TEST: 1.19.x + blas variants #196

Conversation

h-vetinari commented Jun 26, 2020

conda-forge-linter commented Jun 26, 2020

h-vetinari commented Jun 26, 2020 • edited Loading

h-vetinari commented Jun 26, 2020

mattip commented Jun 27, 2020

h-vetinari commented Jun 27, 2020

h-vetinari commented Jun 27, 2020

mattip commented Jun 28, 2020

mattip commented Jun 28, 2020

h-vetinari commented Jun 28, 2020 • edited Loading

mattip commented Jun 28, 2020

h-vetinari commented Jun 28, 2020

mattip commented Jun 29, 2020

h-vetinari commented Jun 29, 2020

mattip commented Jul 14, 2020

mattip commented Jul 14, 2020

h-vetinari commented May 8, 2021

h-vetinari commented Apr 13, 2022

Revival (new PyPy builds and BLAS updates): all green except PPC

Notable

Details

h-vetinari commented Jun 26, 2020 •

edited

Loading

h-vetinari commented Jun 28, 2020 •

edited

Loading