RFR: 8282722: Regard mapping array in enum switches as stable for constant folding

[Paul Sandoz]

On Mon, 7 Mar 2022 07:13:20 GMT, Joshua Zhu <jzhu at openjdk.org> wrote:

> I came across a performance issue when using scatter store VectorAPI for Integer and Long simultaneously in the same application. The poor performance was caused by vector intrinsic inlining failure because of non-determined IntSpecies for a constant VectorShape of IndexMap in this scenario.
> 
> For ScatterStore operation of LongVector.SPECIES_512/IntVector.SPECIES_512, VectorShape.S_256_BIT/S_512_BIT is the actual length of indexMap vector respectively.
> 
> IntSpecies species(VectorShape s)
> 
> returns the corresponding IntSpecies by Switch on Enum type "VectorShape". [1] 
> 
> With this change introduced, elements in the SwitchMap array (initialized in clinit) can be constant-folded so that determined IntSpecies can be acquired for a constant VectorShape.
> 
> jtreg test passed without new failure.
> Please help review this change and let me know if any comments.
> 
> [1] https://github.com/openjdk/jdk/blob/894ffb098c80bfeb4209038c017d01dbf53fac0f/src/jdk.incubator.vector/share/classes/jdk/incubator/vector/IntVector.java#L4043

Unfortunately the `@Stable` annotation can only be used by classes in the `java.base` module and other select modules. It cannot be used generally by application code given its unsafe nature.

We need read-only/frozen/constant arrays to do this properly in code outside of the JDK. 

Perhaps there is a clever alternative strategy javac could use, e.g. using a bootstrap method although nothing specific comes to mind at this moment. I would hold off on that until the code generation strategy for patterns in switch settles down.

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

PR: https://git.openjdk.java.net/jdk/pull/7721