Graham Markall

Official Clockworkmod Recovery/ROM Manager support

2

The recovery has been in use for a couple of months by now and everything seems to work as expected. Points of note: - Touchscreen calibration option is added to...

Ambiguous overloads are allowed for calls to jitted functions inside other jitted functions

4

For example: ```python from numba import njit @njit(['(float32, float64)', '(float64, float32)']) def add(x, y): return x + y # (float64, float64) is an equal match for the (float32, float64) and...

Fix typos of "Generalized" in GUFunc-related code

2

This typo has annoyed me for long enough. As in https://numpy.org/doc/stable/reference/c-api/generalized-ufuncs.html, it should be *Generalized*, not *Generialized*.

CUDA: Fix Issue #8309 - atomics don't work on complex components

1

This resolves the issue by reverting the changes from #7999 that make `array.real()` and `array.imag()` overloads. Current status: this appears to work around the issue reported in #8309 - the...

CUDA: Support for multiple signatures

7

When a list of signatures is passed to the `@cuda.jit` decorator, the dispatcher is compiled for each signature. This does not result in a specialized kernel, even if the list...

CUDA: Replace use of deprecated NVVM IR features, questionable constructs

6

This change the following: - Data layouts. Since CUDA 11.6 the preferred datalayout includes 128 bit integers. Since these are not supported on prior toolkits, we need to keep to...

[WIP] Implement some CUDA intrinsics with `@overload`, `@overload_attribute`, and `@intrinsic`

2

This continues on the work from PR #7013, updated to use current keywords ('target' instead of 'hardware') and to fix it so that it actually works.

Drop CUDA 10.2

2

By the time of the next release (0.57, early 2023) CUDA 10.2 will be more than three years old. Supporting it will require extra maintenance overhead for changes planned in...

CUDA: Add trig ufunc support

1

Testing on CI for now, but hoping to bring up to scratch. Implements trig functions on arrays (`np.sin`, `np.cos`, etc.) for the CUDA target, where the output array is passed...

Use `omp_set_max_active_levels()` instead of `omp_set_nested()`

29

`omp_set_nested()` is deprecated and often results in a warning on recent installations. Following the discussion in Issue #5275, this commit changes the call to `omp_set_max_active_levels()`. Ideally we'd set it to...