Prevent CUDAFuture from using uninitialized device index

[mrshenli]

Stack from ghstack:

• Let child CUDAFuture wait for parent CUDAFuture's CUDAEvents #51506 Let child CUDAFuture wait for parent CUDAFuture's CUDAEvents
• Prevent CUDAFuture from using uninitialized device index #51505 Prevent CUDAFuture using uninitialized device index

Differential Revision: D26187380

[facebook-github-bot]

💊 CI failures summary and remediations

As of commit dacfdbf (more details on the Dr. CI page):

---

• 1/1 failures possibly* introduced in this PR
  • 1/1 non-CircleCI failure(s)

---

This comment was automatically generated by Dr. CI (expand for details).

Follow this link to opt-out of these comments for your Pull Requests.

Please report bugs/suggestions to the (internal) Dr. CI Users group.

[codecov]

Codecov Report

Merging #51505 (dacfdbf) into gh/mrshenli/290/base (718e4b1) will decrease coverage by 0.00%.
The diff coverage is n/a.

@@                   Coverage Diff                    @@
##           gh/mrshenli/290/base   #51505      +/-   ##
========================================================
- Coverage                 80.86%   80.86%   -0.01%     
========================================================
  Files                      1942     1942              
  Lines                    211376   211376              
========================================================
- Hits                     170926   170925       -1     
- Misses                    40450    40451       +1     

[lw]

This works for me. An alternative to possibly consider (up to you) is to store the current device as an optional<...>, and thus then use at::cuda::OptionalCUDAGuard, in order to leave the current device entirely untouched in case of error.

[facebook-github-bot]

@mrshenli merged this pull request in a3f2fe0.