diff options
Diffstat (limited to 'numpy/core')
| -rw-r--r-- | numpy/core/src/umath/dispatching.c | 3 | ||||
| -rw-r--r-- | numpy/core/src/umath/ufunc_object.c | 30 |
2 files changed, 23 insertions, 10 deletions
diff --git a/numpy/core/src/umath/dispatching.c b/numpy/core/src/umath/dispatching.c index 878ab97d7..8638c4e2e 100644 --- a/numpy/core/src/umath/dispatching.c +++ b/numpy/core/src/umath/dispatching.c @@ -205,7 +205,8 @@ resolve_implementation_info(PyUFuncObject *ufunc, if (!matches) { continue; } - + // TODO: Right now, there should only be a single match possible, so + // this can go into the NEXT pr. /* The resolver matches, but we have to check if it is better */ if (best_dtypes != NULL) { int current_best = -1; /* -1 neither, 0 current best, 1 new */ diff --git a/numpy/core/src/umath/ufunc_object.c b/numpy/core/src/umath/ufunc_object.c index 1e1d7baad..c50f69c09 100644 --- a/numpy/core/src/umath/ufunc_object.c +++ b/numpy/core/src/umath/ufunc_object.c @@ -958,14 +958,21 @@ convert_ufunc_arguments(PyUFuncObject *ufunc, if (!out_op_DTypes[i]->legacy) { all_legacy = NPY_FALSE; } - if (PyArray_NDIM(out_op[i]) != 0) { - all_scalar = NPY_FALSE; - continue; + if (PyArray_NDIM(out_op[i]) == 0) { + any_scalar = NPY_TRUE; } else { - any_scalar = NPY_TRUE; + all_scalar = NPY_FALSE; + continue; } - /* Special case if it was a Python scalar, to allow "weak" promotion */ + // TODO: Refactor this into a helper function! + // We probably need to move the special object array-conversion till + // After the promotion :(. If this is NOT a forced-legacy promotion?! + // We may need to convert "twice" (but not more than we currently do) + /* + * Special case if it was a Python scalar, to allow "weak" promotion, + * so replace the arrays DType with the Python scalar DType. + */ if ((PyObject *)out_op[i] != obj) { PyArray_DTypeMeta *scalar_DType = NULL; if (PyLong_CheckExact(obj)) { @@ -1049,7 +1056,7 @@ check_for_trivial_loop(PyArrayMethodObject *ufuncimpl, PyArrayObject **op, PyArray_Descr **dtypes, NPY_CASTING casting, npy_intp buffersize) { - npy_intp i, nin = ufuncimpl->nin, nop = nin + ufuncimpl->nout; + int i, nin = ufuncimpl->nin, nop = nin + ufuncimpl->nout; for (i = 0; i < nop; ++i) { /* @@ -1567,6 +1574,8 @@ execute_ufunc_loop(PyArrayMethod_Context *context, int masked, NPY_END_THREADS; NPY_AUXDATA_FREE(auxdata); + // TODO: use the same pattern as we changed to in the trivial-loop + // have to dealloc the iterator, but can do that first. if (res == 0 && !(flags & NPY_METH_NO_FLOATINGPOINT_ERRORS)) { /* NOTE: We could check float errors even when `res < 0` */ const char *name = ufunc_get_name_cstr((PyUFuncObject *)context->caller); @@ -4313,8 +4322,8 @@ _get_dtype(PyObject *dtype_obj) { /* - * Finish conversion parsing of the signature. NumPy always only honored - * the type number for passed in descriptors/dtypes. + * Finish conversion parsing of the DType signature. NumPy always only + * honored the type number for passed in descriptors/dtypes. * The `dtype` argument is interpreted as the first output DType (not * descriptor). * Unlike the dtype of an `out` array, it influences loop selection! @@ -4530,7 +4539,10 @@ resolve_descriptors(int nop, retval = 0; } else { - /* Rare fall-back to the legacy resolver to pass `operands` */ + /* + * Fall-back to legacy resolver using `operands`, used exclusively + * for datetime64/timedelta64 and custom ufuncs (in pyerfa/astropy). + */ retval = ufunc->type_resolver(ufunc, casting, operands, NULL, dtypes); } |