Commit Graph

17 Commits

Author SHA1 Message Date
Matthew Woehlke 2edf0fc6d7 Modules: Use new keyword-dispatched try_compile signature
Modify most of the modules that ship with CMake to use the new
try_compile / try_run signatures added by commit aa9220d3a0
(try_compile: Add keyword-dispatched signature, 2022-09-02). This
improves debugging by each invocation using its own directory so that
the results of multiple invocations can be retained.

This does not cover any invocations which provide an entire project, as
that flavor of try_compile has not yet been updated.
2022-09-14 07:52:16 -04:00
Henry Schreiner 92854bf29e CUDA: Update arch selection, restructure for maintainability
Fixes: #21712
2021-01-20 12:28:28 -05:00
Henry Schreiner a58a572459 CUDA: Add arch selection debug mode to print out CUDA versions 2021-01-20 12:27:54 -05:00
zasdfg bnm 5f6b0ba8b1 FindCUDA/select_compute_arch: Add CUDA 11.1 and SM86 support 2020-09-29 05:32:25 -04:00
zasdfg bnm 8707178c1c FindCUDA/select_compute_arch: generate PTX for the latest architecture 2020-07-08 13:35:12 -04:00
zasdfg bnm 295d1b9cb5 FindCUDA/select_compute_arch: Add CUDA 11 and Ampere support
Co-Author: Henry Schreiner <henryschreineriii@gmail.com>
2020-07-07 11:32:45 -04:00
Anton Chernov 0f9533e61b CUDA: Add CUDA 10 (Turing) detection 2018-12-07 08:51:57 -05:00
Soumith Chintala bfd93b73a0 FindCUDA: Filter unrelated content in compute capabilities output
Working around CUDA-level nvrm_gpu log statements to stdout on some
embedded platforms (ex. Drive PX2).

See-also: https://github.com/pytorch/pytorch/issues/11518#issue-359113249
2018-09-19 10:55:02 -04:00
Brad King 37ba34c0b0 FindCUDA/select_compute_arch: Restore two-component CUDA_VERSION
Since commit v3.12.0-rc1~332^2 (FindCUDA/select_compute_arch: Add
support for CUDA as a language, 2018-03-15) this module sets
`CUDA_VERSION` based on `CMAKE_CUDA_COMPILER_VERSION` when the language
is enabled.  Limit it to two components for consistency with the
normal `FindCUDA` documentation and behavior.

Fixes: #18231
2018-08-01 13:42:13 -04:00
Henry Schreiner 61b222f3ae CUDA/select_compute_arch: Fix detection on old CUDA with new card/driver
Fix version comparisons to handle patch components.  List and check
known archs for each version of CUDA so mismatching versions are not
suggested.

Fixes: #17921
2018-04-20 09:43:50 -04:00
Henry Fredrick Schreiner e743fc8e91 FindCUDA/select_compute_arch: Add support for CUDA as a language
Even though this is an internal module, we can still prepare it to
be used in another public-facing module outside of `FindCUDA`.

Issue: #16586
2018-03-20 14:19:25 -04:00
Boris Fomitchev cbf1393aa4 FindCUDA: Add CUDA9/Volta recognition 2017-08-10 11:10:07 -04:00
Bjoern Thiel 934eb497f2 FindCUDA: Fix PTX selection with multiple architectures 2017-02-22 10:35:44 -05:00
Pierre Moreau 645dd3ca53 FindCUDA: Fix typo in CUDA_DETECT_INSTALLED_GPUS doc string 2017-02-02 08:52:11 -05:00
Bjoern Thiel c1f3086de7 FindCUDA: Enable CUDA_SELECT_NVCC_ARCH_FLAGS Auto for MSVC
Fixes: #16525
2017-01-25 10:31:27 -05:00
Boris Fomitchev 2dc9a754b5 FindCUDA: Support 2.1(2.0) architecture notation
Also tweak some messages and comments.
2016-08-23 10:47:27 -04:00
Boris Fomitchev 7229ae728f FindCUDA: Refactor Android(Tegra) support
Reviewed-by: James Bigler <jamesbigler@gmail.com>
2016-06-09 09:25:55 -04:00