* Update to CUDA 11.3 on Windows.
* Update PowerShell to 7.1.3.
* Update CUDA to 11.3 on Linux.
* "Explode" VM provisioning scripts for more consistent feedback during deploy. This resolves the deployment script often hanging with no feedback as to why.
* [libdatachannel] Fix use of deprecated result_type typedef.
Submitted upstream as https://github.com/paullouisageneau/libdatachannel/pull/413
* [libvpx] Get the libvpx outputs from the correct place.
(Perhaps VS2019 version 16.10 moved where these are placed? I've been defensive and left an attempt to find from the old location in place.)
* [chromaprint] Support implementations where lrintf is an intrinsic.
* Add provision-entire-image script.
* [cudnn] Disable download-on-the-fly due to licensing concerns.
* Add libnccl to Linux VMs.
* [wangle] Disable x64-windows due to an ICE.
* [cmake] Update cmake to 3.20.1 to avoid https://gitlab.kitware.com/cmake/cmake/-/issues/21571 race
* [libudis86] Fix passing a bogus working directory which fails on CMake 3.20.x
* [dartsim] Disable unit tests, examples, and tutorials, some of which have CMake authoring errors rejected by 3.20.1.
* Add thrust to the cuda installees.
* [tensorflow] Put .bzl in CURRENT_BUILDTREES_DIR to avoid running out of disk space in CI and to respect --clean-after-build.
* [dimcli] Skip port broken by changes in VS2019 project system.
* [upb] Disable an additional warning.
* [libhv] Fix typo DISABLE_PARALLEL => DISABLE_PARALLEL_CONFIGURE
* Update pools
* [cudnn] add other install dir on linux
* [cudnn] add other locations for linux
* [cudnn] fix env variables
* [cudnn] fix references
* [cudnn] use also cuda module as a hint
* [cudnn] fix refs
* [cudnn] auto-download only if cuda version matches
* [cudnn] auto download also for cuda 10.2
* [cudnn] update version ref
* Update ports/cuda/vcpkg_find_cuda.cmake
Co-authored-by: Robert Schumacher <roschuma@microsoft.com>
* Update ports/cuda/vcpkg_find_cuda.cmake
Co-authored-by: Robert Schumacher <roschuma@microsoft.com>
* Update ports/cudnn/portfile.cmake
Co-authored-by: Robert Schumacher <roschuma@microsoft.com>
* apply review fixes
* fix baseline refs
Co-authored-by: Robert Schumacher <roschuma@microsoft.com>
* [cudnn] add port and enable it in darknet
* [OpenCV4] fix ffmpeg feature on uwp
* [darknet] enable integration with cuda/cudnn also with opencv3
* [darknet] update to latest revision