Pearu Peterson
Pearu Peterson
The task boils down to replacing ArrayPointer with BufferPointer.
Issue: ``` $ pytest -sv rbc/tests/ ========================================================================================= short test summary info ========================================================================================= FAILED rbc/tests/test_omnisci_device_selection.py::test_device_selection_single[lt-udf-any] - AssertionError: assert 'gpu' == 'cpu' FAILED rbc/tests/test_omnisci_device_selection.py::test_device_selection_pair[rt-lt-udf/udf-any/any] - AssertionError: assert ('any', 'gpu') == ('any', 'cpu')...
Also, do you think it is a good idea for RBC to know if the server has support for armadillo/mlpack? _Originally posted by @guilhermeleobas in https://github.com/xnd-project/rbc/issues/317#issuecomment-814626142_ Yes, in general, the...
Original description: Potential variadic input (and output?) capability by defining dataframe type that can have any number of columns. - Perhaps restrict to columns of the same type - Consider...
Slack conversation: https://quansight.slack.com/archives/CHSDG1C9W/p1615825572028500 Summary of tasks: - support loading external libraries dynamically via omniscidb command line arguments - implement C wrappers to mlpack, can be lightweight - call mlpack functions...
Task: find a way that avoids hacking with globals. > Why note use: This end up causing a circular import. _Originally posted by @guilhermeleobas in https://github.com/xnd-project/rbc/pull/286#discussion_r585815227_
For API, see https://github.com/xnd-project/rbc/blob/6cab3729b13cc9ef71d01c58845afc465c586f32/rbc/targetinfo.py#L133 Use https://github.com/xnd-project/rbc/blob/master/rbc/libcudart_ctypes.py to retrieve CUDA information. See https://github.com/omnisci/omniscidb-internal/blob/99e6039d1862985154bd1df166a6e764f6caaab0/QueryEngine/NativeCodegen.cpp#L913-L941 for a set target info keys. When done, add `host_cuda` support (`gpu=...`) to https://github.com/xnd-project/rbc/blob/6cab3729b13cc9ef71d01c58845afc465c586f32/rbc/remotejit.py#L678-L690 Aim: eliminate using omnisci...
https://github.com/BuffetCodes/Linear-Regression-using-PyTorch-CPP
As in title. Since this issue will change the UDF/UDTF registration interface, implement Column argument passing by reference (rather than by value, see #200 ) as this would be a...
The following example should be supported: ``` @omnisci('int32(Column, RowMultiplier, Column)') def array_sum(x, m, y): for i in range(x): y[i] = sum(x[i]) ```