add back the iteration based seeding for Array

[KristofferC]

These were removed in #472 and while I was a bit uncomfortable with it I did not protest. However, now after doing some benchmarks I see that this has a runtime cost that is very much nonsignificant.

Below is a profile and the big peaks are from seed!.

If I use the iteration-based seed function they completely disappear. In order to not break anything, I added these back but with a restriction to Array so that e.g. GPU arrays will keep using the broadcast based seeding.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 87.60%. Comparing base (1592fe9) to head (bafdb3b).
⚠️ Report is 48 commits behind head on master.

Additional details and impacted files

@@            Coverage Diff             @@
##           master     #634      +/-   ##
==========================================
+ Coverage   87.28%   87.60%   +0.32%     
==========================================
  Files          10       10              
  Lines         912      936      +24     
==========================================
+ Hits          796      820      +24     
  Misses        116      116              

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[devmotion]

I wonder if one could use map! instead of broadcasting? I would assume that this would still be GPU-compatible but possibly there is (almost) no runtime cost for Array?

[KristofferC]

Should check if this is still useful on later julia versions.

[devmotion]

This was already fixed by #739, as far as I can tell. That PR reverted the seeding to an iteration-based approach for all arrays (which broke the - untested - GPU support...).

[KristofferC]

Thanks for the reference, I had missed that one.