Regression due to removal of autoCache

[Anipik]

The Auto-cache was removed in 435a63b

I added new benchmark #1855
Before the change the time taken by this benchmark was 2.8sec (after converting to api version and making label as key instead of R4)

After the removal of the cache The time went upto 10s. But after the adding the cache check point me and @wschin was able to reduce it to 3.8s which is still regressed from 2.8s.

[danmoseley]

@glebuk do you use a particular label or github project to keep together bugs for perf issues? That might be handy.

[Ivanidzo4ka]

This is not regression, this is expected behavior see: #1604

[Zruty0]

Well, @Ivanidzo4ka , I believe that 3.8 vs 2.8 is still some sort of bug: I would expect manual caching to work the same way as the auto-caching used to work.

[codemzs]

agree with Ivan here, please provide more details but closing for now.