On Thu, 10 Apr 2025 13:19:05 GMT, Ferenc Rakoczi <d...@openjdk.org> wrote:
>> By using the aarch64 vector registers the speed of the computation of the
>> ML-KEM algorithms (key generation, encapsulation, decapsulation) can be
>> approximately doubled.
>
> Ferenc Rakoczi has updated the pull request incrementally with two additional
> commits since the last revision:
>
> - Code rearrange, some renaming, fixing comments
> - Changes suggested by Andrew Dinn.
src/hotspot/cpu/aarch64/stubGenerator_aarch64.cpp line 5590:
> 5588: __ add(tmpAddr, coeffs, 0);
> 5589: store64shorts(vs2, tmpAddr);
> 5590:
I'd like to make explicit the fact that we have avoided doing an add here (and
in the next two cases) by adding a commented out generation step i.e. at this
line insert
// __ add(tmpAddr, coeffs, 128); // unneeded as implied by preceding load
src/hotspot/cpu/aarch64/stubGenerator_aarch64.cpp line 5595:
> 5593: __ add(tmpAddr, coeffs, 128);
> 5594: store64shorts(vs2, tmpAddr);
> 5595:
Likewise insert:
// __ add(tmpAddr, coeffs, 256); // unneeded as implied by preceding load
src/hotspot/cpu/aarch64/stubGenerator_aarch64.cpp line 5601:
> 5599: store64shorts(vs2, tmpAddr);
> 5600:
> 5601: load64shorts(vs1, tmpAddr);
Likewise insert:
// __ add(tmpAddr, coeffs, 384); // unneeded as implied by preceding load
-------------
PR Review Comment: https://git.openjdk.org/jdk/pull/23663#discussion_r2037607688
PR Review Comment: https://git.openjdk.org/jdk/pull/23663#discussion_r2037609104
PR Review Comment: https://git.openjdk.org/jdk/pull/23663#discussion_r2037611049
