RFR: 8365205: C2: Optimize popcount value computation using knownbits

[Chen Liang]

On Thu, 25 Sep 2025 15:58:27 GMT, Jatin Bhateja <jbhateja at openjdk.org> wrote:

>> The change looks good, but I wonder:
>> 
>> - if it makes sense to have some kind of IR tests (i.e., it's folded away when unneeded, when the input is a constant, ...)?
>> - whether the explanation could be simplified: Assuming a correct implementation of the KnownBits canonicalization, we can argue
>> 	- `_zeroes` has the bits set that are known to be always 0. So `BitsPer<Type> - popCount(x)` gives you an upper limit of how many bits *might* be 1. And `BitsPer<Type> - popCount(_zeroes)` is equivalent to `popCount(~_zeroes)`.
>> 	- `_ones` has the bits set that are known to be always 1. Trivially, `popCount(_ones)` is a valid lower bound.
>> 	- The rest repeats how `adjust_bits_from_unsigned_bounds` works, but that's not specific to the popcount nodes.
>
> Hi @SirYwell , @chhagedorn , @eme64 , I have addressed your comments. Let me know if this is good to land in.

Hi @jatin-bhateja, sorry for a spurious comment but I wish to ask about the status of [lworld+vector](https://github.com/openjdk/valhalla/tree/lworld%2Bvector) in project Valhalla - It makes use of Unsafe.makePrivateBuffer and finishPrivateBuffer, which is outdated in the current Value Objects model (the larval bit will be gone).  I just wonder if I can proceed with the removal here https://github.com/openjdk/valhalla/pull/1593, or if I should keep these legacy APIs for further vector work. (FYI we can probably migrate to Unsafe.allocateInstance to do the same, as in Method handles)

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

PR Comment: https://git.openjdk.org/jdk/pull/27075#issuecomment-3335032730