RFR: 8320649: C2: Optimize scoped values [v7]

[Andrew Haley]

On Thu, 8 Feb 2024 12:37:13 GMT, Emanuel Peter <epeter at openjdk.org> wrote:

> This really makes me nervous. Can you state a proof/argument in the comments?

I presume this refers to range checks on the cache indexes.

The cache is created `CACHE_TABLE_SIZE` objects in size. `CACHE_TABLE_SIZE` is an integer power of two, by construction.
The source for cache indexes is `primarySlot()` and `secondarySlot()`. Both of these methods, and any intrinsic code that duplicates their functions, must clip the range of the returned hash value with  `& SLOT_MASK.`
We know that `CACHE_TABLE_SIZE` is an integer power of two, and `SLOT_MASK` is `cacheSize - 1`.
Therefore the resulting index is in range.

Is this the question you're asking?

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

PR Review Comment: https://git.openjdk.org/jdk/pull/16966#discussion_r1487480092