Implement GPU version of numpy.* functions in cupy.* namespace.

This is a tracker issue that lists the remaining numpy.* APIs (see also: comparison table). I've categorized them based on difficulty so that new contributors can pick the right task. Your contribution is highly welcomed and appreciated!

List of APIs

Very Easy

• numpy.asfarray (Add cupy.asfarray #6085)
• numpy.byte_bounds (hint: ndarray.data.ptr) (Add cupy.byte_bounds #7015)
• numpy.format_float_positional (added cupy.format_float_positional #6308)
• numpy.format_float_scientific (Add cupy.format_float_scientific API #6474)
• numpy.ndarray.searchsorted (added cupy.ndarray.searchsorted #7059)
• Alias: numpy.bool, numpy.long, numpy.ulong (Add cupy.bool, cupy.long and cupy.ulong #9253)
• Alias: numpy.isdtype ([backport] Add cupy.isdtype #9891)

Easy

• numpy.apply_over_axes (Add apply_over_axes API #8177)
• numpy.array_equiv (Add array_equiv API. #6254)
• numpy.asarray_chkfinite (Add asarray_chkfinite API. #6275)
• numpy.fabs (Add cupy.fabs API. #6282)
• numpy.float_power (added api for cupy.float_power #6371)
• numpy.heaviside (add cupy.heaviside api. #6798)
• numpy.isneginf (Add cupy.isneginf and cupy.isposinf #6089)
• numpy.isposinf (Add cupy.isneginf and cupy.isposinf #6089)
• numpy.mask_indices (Add cupy.mask_indices #6156)
• numpy.real_if_close (Add cupy.real_if_close API #6475)
• numpy.setdiff1d (Add cupy.setdiff1d api. #6433)
• numpy.setxor1d (add cupy.setxor1d api #6582)
• numpy.tril_indices & numpy.tril_indices_from (Add tril_indices and tril_indices_from API. #6305)
• numpy.triu_indices & numpy.triu_indices_from (Add triu_indices and triu_indices_from API. #6316)
• numpy.union1d (Add cupy.union1d API. #6357)
• numpy.bitwise_count (Add function bitwise_count #9390)
• numpy.cumulative_prod (Add cumulative_sum and cumulative_prod (Array API) #9864)
• numpy.cumulative_sum ([WIP] Add cumulative_sum #9205, Add cumulative_sum and cumulative_prod (Array API) #9864)
• numpy.matrix_transpose (Add cupy.linalg.matmul, cupy.linalg.matrix_transpose and cupy.matrix_transpose #9892)
• numpy.vecdot
• numpy.linalg.diagonal
• numpy.linalg.matmul (Add cupy.linalg.matmul, cupy.linalg.matrix_transpose and cupy.matrix_transpose #9892)
• numpy.linalg.matrix_norm
• numpy.linalg.matrix_transpose (Add cupy.linalg.matmul, cupy.linalg.matrix_transpose and cupy.matrix_transpose #9892)
• numpy.linalg.outer
• numpy.linalg.svdvals (Add cupy.linalg.svdvals function #9388)
• numpy.linalg.tensordot
• numpy.linalg.trace
• numpy.linalg.vecdot
• numpy.linalg.vector_norm
• numpy.lib.format.*
• numpy.lib.stride_tricks.sliding_window_view (Provide sliding_window_view #6956, Add cupy.lib.stride_tricks.sliding_window_view #7575)
• numpy.emath.* (numpy.lib.scimath.*) (add cupy.emath module #7295)

Medium

• numpy.block
• numpy.unstack
• numpy.delete (Implement delete function, add documentation #7359)
• numpy.geomspace (Implement numpy.geomspace #9082)
• numpy.insert (add cupy.insert api #6597)
• numpy.put_along_axis (add cupy.put_along_axis API #8199)
• numpy.row_stack (Add cupy.row_stack API. #6312)
• numpy.spacing
• numpy.vander (Add cupy.vander API. #6279)
• numpy.linalg.eig (Implement cupy.linalg.eig / cupy.linalg.eigvals clone of PR #8854 #8980)
• numpy.linalg.eigvals (Implement cupy.linalg.eig / cupy.linalg.eigvals clone of PR #8854 #8980)

Medium to Hard

• numpy.histogram_bin_edges
• numpy.ediff1d (Add cupy.ediff1d API. #6280)
• numpy.intersect1d (Add cupy.intersect1d API. #6402, Add cupy.intersect1d API #6407)
• numpy.nanpercentile
• numpy.nanquantile
• numpy.ndarray.ctypes (note: needs design discussion)
• numpy.ndarray.getfield
• numpy.ndarray.resize
• numpy.ndarray.setfield
• numpy.polynomial.*
• numpy.poly (Add cupy.poly #3547, Add cupy.poly #6697)
• numpy.polyder (Add cupy.polyder API. #6469)
• numpy.polydiv (Add cupy.polydiv #3780)
• numpy.polyint
• numpy.trapz (Add cupy.trapz #6107)
• numpy.linalg.cond (Implement cupy.linalg.cond #9140)
• numpy.linalg.multi_dot (Add linalg.multi_dot API. #6358)
• numpy.random.Generator.* (see the dedicated tracker issue for details: [Tracker] Add random distributions to new Generator #4557)

Low priority

Iterator functions

• numpy.ndenumerate
• numpy.nditer
• numpy.nested_iters

Help functions

• numpy.info
• numpy.lookfor
• numpy.source
• numpy.show_runtime (needs a design decision: we currently provide runtime info via cupy.show_config)

Internal functions

• numpy.deprecate
• numpy.deprecate_with_doc

Dtype APIs - need to filter types unsupported by CuPy

• numpy.maximum_sctype
• numpy.cast (undocumented API) (hint: see numpy/core/numerictypes.py)
• numpy.typecodes (undocumented API)
• numpy.sctypeDict (undocumented API)
• numpy.sctypes (undocumented API)
• numpy.nbytes (undocumented API)

Rarely used APIs

• numpy.einsum_path (Add cupy.einsum_path #6723)
• numpy.frompyfunc (maybe just call GUFunc?)

Steps to Contribute

Note: You will need a GPU environment to develop CuPy.

1. Fork and star ⭐ the CuPy repository 😉

2. Pick a function you want to work on. Check out the NumPy API Reference to understand what should be implemented. Please search for pull requests before start working on, somebody may have opened a PR already.

3. Implement a function in your branch. If you need help, join Gitter or just ask for help in this issue.

4. Implement test code.

5. Build CuPy and run tests to confirm that the function runs fine:
   pip install --no-build-isolation -e . && pytest tests/cupy_tests/PATH_TO_YOUR_TEST
   See the Contribution Guide for details.

6. Submit a pull-request to the main branch. (example: Implement cupy.linalg.cond #9140)

See also:

• [Tracker] Implement all scipy.* APIs in CuPy #6324
• [Tracker] Fix unimplemented options #6692