RFR: 8222852: Reduce String concat combinator tree shapes by folding constants into prependers

[Remi Forax]

That's nice !

In StringConcatFactory, i will change
  prefixConstant + constantValue (resp suffixConstant + constantValue)
to 
  prefixConstant.concat(constantValue)

to not depend on any string concat implementations of javac.

regards,
Rémi

----- Mail original -----
> De: "Claes Redestad" <claes.redestad at oracle.com>
> À: "Peter Levart" <peter.levart at gmail.com>, "Aleksey Shipilev" <shade at redhat.com>, "core-libs-dev"
> <core-libs-dev at openjdk.java.net>
> Envoyé: Lundi 29 Avril 2019 11:34:11
> Objet: Re: RFR: 8222852: Reduce String concat combinator tree shapes by folding constants into prependers

> Hi Peter,
> 
> On 2019-04-28 23:34, Claes Redestad wrote:
>>>>
>>>> What do you think?
>>>
>>> This is what I meant by "simplified wiring logic":
>>>
>>> http://cr.openjdk.java.net/~plevart/jdk-dev/8222852_StringConcatOpt/webrev.02p/
>>>
>>>
>>> I haven't tried this though, but if it works correctly and if JIT
>>> inlines the strategy for each call site, it should also eliminate the
>>> dead code paths that are detected from null/non-null bound constants
>>> and the end result should be equivalent machine code. But that's just
>>> in theory...
>> 
>> thanks for picking this up and experimenting further with it!
> 
> having run the numbers on your patch, I'm warming to this approach:
> 
> - It seems the branches testing for null do get eliminated (no
> regressions in neither existing nor new micro trying to provoke
> pollution between LFs)
> - Startup numbers generally a bit better, and the amortized/repeated
> bootstrap cost of always binding in prefix and suffix appears to be
> insignificant
> - The logic for handling constants is nicer and less spread out; bonus
> points for folding successive suffix constants!
> 
> http://cr.openjdk.java.net/~redestad/8222852/open.03/
> 
> I merged some of the cleanups and comments and think this is good to
> go rather than splitting it up into a follow-up RFE.
> 
> Thanks!
> 
> /Claes