RFR: 8364305: Support AVX10 saturating floating point conversion instructions [v3]

[Jatin Bhateja]

On Mon, 1 Sep 2025 07:50:57 GMT, Jatin Bhateja <jbhateja at openjdk.org> wrote:

>> Mohamed Issa has updated the pull request incrementally with one additional commit since the last revision:
>> 
>>   Fix input size enum values for AVX 10.2 conversion instructions that take memory as the source
>
> src/hotspot/cpu/x86/x86.ad line 7804:
> 
>> 7802:   predicate(VM_Version::supports_avx10_2() &&
>> 7803:             is_integral_type(Matcher::vector_element_basic_type(n)));
>> 7804:   match(Set dst (VectorCastD2X src));
> 
> I assume your intent here is to feed the memory operand to the vector cast IR, a memory operand is first loaded into register using LoadVector IR, so a CISC / memory variant of pattern should consume the Load IR such that the operand is directly exposed to the instruction. Checkout https://github.com/openjdk/jdk/blob/master/src/hotspot/cpu/x86/x86.ad#L8986

Make a similar change in all the newly added memory patterns.

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

PR Review Comment: https://git.openjdk.org/jdk/pull/26919#discussion_r2313167705