| Age | Commit message (Collapse) | Author |
|---|
|
Special notes:
- _nvcc_arch_flags is always called with exact arguments, no need for
default values
- min_driver_version has its args annotation loosened because it has to
fit the constraints of the module interface?
|
|
The min_driver_version function has an extensive, informative custom
error message, so leave that in place.
The other two functions didn't have much information there, and it's
fairly evident that the cuda compiler itself is the best thing to have
here. Moreover, there was some fairly gnarly code to validate the
allowed values, which we can greatly simplify by uplifting the
typechecking parts to the dedicated decorators that are both really good
at it, and have nicely formatted error messages complete with reference
to the problematic functions.
|
|
It is pretty trivial and more confusing when standalone, especially the
use of a sentinel "unknown" string as a standin for "this isn't one of
the allowed object types". Much easier to directly raise an error in the
fallthrough/else.
|
|
It is only ever invoked once. It has a statically known signature based
on how it's used, so there's no good reason to allow it to accept any
defaults -- and defaulting to None messes with type safety.
Saturate is always given, so don't claim it is only sometimes.
|
|
This officially only ever accepted string or array of strings.
|
|
They're hardly used, and can be gotten directly from the Environment
instead
|
|
This replaces all of the Apache blurbs at the start of each file with an
`# SPDX-License-Identifier: Apache-2.0` string. It also fixes existing
uses to be consistent in capitalization, and to be placed above any
copyright notices.
This removes nearly 3000 lines of boilerplate from the project (only
python files), which no developer cares to look at.
SPDX is in common use, particularly in the Linux kernel, and is the
recommended format for Meson's own `project(license: )` field
|
|
|
|
|
|
In particular, CUDA 12 removes support for Kepler (3.x) entirely.
Unusually, however, it does not introduce any new architectures,
or even compute capabilities.
|
|
|
|
Agrees with Release Notes, Section 1.1, Table 3.
https://docs.nvidia.com/cuda/cuda-toolkit-release-notes/index.html
|
|
The upper bound on the CUDA Compute Capability is not always known.
Avoid checking cuda_hi_limit_gpu_architecture if it remains undefined.
|
|
Co-authored-by: David Seifert <soap@gentoo.org>
|
|
It turns out we don't generally need to proxy every compiler ever
through the top-level package. The number of times we directly poke at
one is negligible and direct imports are pretty clean.
|
|
|
|
Thanks to `ModuleInfo`, all modules are just named `foo.py` instead of
`unstable_foo.py`, which simplifies the import method a bit. This also
allows for accurate FeatureNew/FeatureDeprecated use, as we know when
the module was added and if/when it was stabilized.
|
