parse_json regression in String allocations

[vtjnash]

This commit (#48996) caused a regression in parse_json. Maybe related to #49241? @oscardssmith

[oscardssmith]

I don't see how it would be related to #49241 since #48996 doesn't affect 1.9...

[gbaraldi]

It's not a 1.9 regression, it's a master regression 1.9 is fine.

[oscardssmith]

is there a known mechanism by which improving effects could regress performance?

[vtjnash]

#49241 is the only mechanism I know of currently, which is why I linked to that. Depends on what was changed in that commit though

[gbaraldi]

We went from

to

So we are boxing a lot of ints and tuples that we didn't before. The values themselves don't matter because they have a different rate.

[oscardssmith]

Is this fixed by #50444?

[gbaraldi]

No

[oscardssmith]

Why not? was this issue not caused by #49241? If not, do we have another potential reason why improving effects would regress performance?

[gbaraldi]

This is a separate issue that was just triggered by that change. But @aviatesk looked into it and I think it's something due to results not being inlineable and fun.

[JeffBezanson]

duplicate of #50458.