Repository navigation
Remove the legacy backend RNG - #3163
ndgrigorian wants to merge 1 commit into
Conversation
68eef44 to
d7cce7f
Compare
315889e to
142a3d8
Compare
d7cce7f to
71b2060
Compare
142a3d8 to
160719a
Compare
|
Array API standard conformance tests for dpnp=0.22.0dev1=np2py314h8d9cdd5_29 ran successfully. |
|
Coverage is 86.23% — feature/remove-legacy-rng into feature/device-rng-random-state. No base build found for |
|
View rendered docs @ https://intelpython.github.io/dpnp/pull/3163/index.html |
71b2060 to
86d7f1b
Compare
160719a to
014a132
Compare
dpnp.random no longer uses dpnp_backend_c, so drop the dpnp_algo_random Cython module, the backend RNG kernels, DPNPRandomState, the DPNP_FN_RNG_* function pointers, the backend engines and their examples and tests, and the now-unused parts of the backend, including its GTest suite, the DPNP_BACKEND_TESTS option and call_origin. The conda recipes no longer pin onemkl-sycl-rng.
c3083f7 to
5ed41e0
Compare
edee62b to
843a343
Compare
| * Removed the unused `FourZeroOffsets_Indexer`, `FixedDimStridedIndexer` and `ThreeOffsets_FixedDimStridedIndexer` from the libtensor header `utils/offset_utils.hpp` [#3131](https://github.com/IntelPython/dpnp/pull/3131) | ||
| * Dropped support for Python 3.10 [#3173](https://github.com/IntelPython/dpnp/pull/3173) | ||
| * Removed the legacy backend implementation of `dpnp.random`, which no longer needs `onemkl-sycl-rng` [#3163](https://github.com/IntelPython/dpnp/pull/3163) | ||
| * Removed the unused parts of the legacy `dpnp_backend_c` library, including its empty GTest suite, the `DPNP_BACKEND_TESTS` CMake option and the NumPy fallback helper `call_origin` [#3163](https://github.com/IntelPython/dpnp/pull/3163) |
There was a problem hiding this comment.
call_origin was the only reader of dpnp.config.__DPNP_RAISE_EXCEPION_ON_NUMPY_FALLBACK__ (env DPNP_RAISE_EXCEPION_ON_NUMPY_FALLBACK). After this PR the setting does nothing, with no warning. Could it go into _DEPRECATED_SETTINGS in dpnp/config.py, as in #3107, with a Deprecated entry? Then the allow_fall_back_on_numpy fixture (33 uses in the CuPy random tests) can go too.
| - {{ pin_compatible('onemkl-sycl-dft', min_pin='x.x', max_pin='x') }} | ||
| - {{ pin_compatible('onemkl-sycl-lapack', min_pin='x.x', max_pin='x') }} | ||
| - {{ pin_compatible('onemkl-sycl-rng', min_pin='x.x', max_pin='x') }} | ||
| - {{ pin_compatible('onemkl-sycl-vm', min_pin='x.x', max_pin='x') }} |
There was a problem hiding this comment.
The conda-forge mkl-devel-dpcpp in host has a weak run export mkl-dpcpp, and mkl-dpcpp depends on onemkl-sycl-rng. So a package built from this recipe still installs it. rattler_recipe.yaml already lists mkl-dpcpp in ignore_run_exports; could this recipe do the same?
This PR removes the legacy RNG implementation of
dpnp.randomfromdpnp_backend_c, along with the newly-dead code in the backend, as the last part of the series splitting #3147. It is stacked on #3162Removes
dpnp_algo_randomCython module, the backend RNG kernels,DPNPRandomState, theDPNP_FN_RNG_*function pointers and the mt19937/mcg59 engines ofbackend_syclDPNP_BACKEND_TESTSCMake optioncall_originand the other helpers that only the RNG usedonemkl-sycl-rngpin in the conda recipesdpnp_backend_citself stays, sincedpnp.partitionstill uses it (#3138 removes that)