RFR: 8255271: Avoid generating duplicate interpreter entries for subword types

[Claes Redestad]

On Fri, 23 Oct 2020 13:08:27 GMT, Coleen Phillimore <coleenp at openjdk.org> wrote:

>> Top-of-stack optimizations are generally missing for byte, char, short and boolean, and for several types of entry points we already avoid generating entries and instead redirect to the int variant.
>> 
>> This patch removes generation of "specialized" variants for byte, char, short and boolean more thoroughly after verifying they all generate the same code as their int specialization. This slightly reduces overhead of generating the interpreter and the size thereof.
>
> src/hotspot/share/interpreter/templateInterpreter.cpp line 112:
> 
>> 110:   _entry[ztos] = ientry;
>> 111:   _entry[ctos] = ientry;
>> 112:   _entry[stos] = ientry;
> 
> next patch is to get rid of these entry points since I think nothing branches to them.

Thanks for reviewing. 

I'll file an RFE to keep at this. I didn't want to keep pulling at this thread for this RFE since it seems like quite a chunk of code will be unravelled here.

-------------

PR: https://git.openjdk.java.net/jdk/pull/816