[foreign-jextract] RFR: MemorySegmentPool + Allocator [v6]

[Maurizio Cimadamore]

On Tue, 20 Apr 2021 15:36:47 GMT, Radoslaw Smogura <github.com+7535718+rsmogura at openjdk.org> wrote:

>> (Preview)
>> 
>> The MemorySegmentPool is a pool maintaining memory segments, optionally can expose allocator which can be bound to other scope, and which will return allocated segments back to pool.
>> 
>> However the best results has been achieved by using getSegmentEntry & putSegmentEntry methods.
>> 
>> The pool is intended to be used by long running applications (i.e. like global shared pool), where fast allocation and de-allocation of segments is critical (was designed during implementation of I/O subsystem with Panama, as a pool for temporary buffers between system I/O methods and Java byte arrays from InputStreams).
>> 
>> The pool uses hand-made SpinLockQueue as the Deque from JDK offers too much functionality and overhead.
>
> Radoslaw Smogura has updated the pull request incrementally with one additional commit since the last revision:
> 
>   Use VH insted of volatile
>   
>   Replacee
>   //    while ((int) LOCK.compareAndExchange(this, 0, 1) != 1) { }
>       while (!LOCK.compareAndSet(this, 0, 1)) { }
>   
>   Performane around 34 ns / oop

test/micro/org/openjdk/bench/jdk/incubator/foreign/StrLenTest.java line 158:

> 156:         try(ResourceScope scope = ResourceScope.newConfinedScope()) {
> 157:             final var allocator = memorySegmentPool.allocatorForScope(scope);
> 158:             return (int)STRLEN.invokeExact(CLinker.toCString(str, allocator).address());

This comparison is not 100% identical - here we're calling CLinker.toCString, while the low level benchmark just does a bulk copy. For some reason the latter seems to work much better, and I get better numbers if I do that in this benchmark too.

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

PR: https://git.openjdk.java.net/panama-foreign/pull/509