RFR: 8318650: Optimized subword gather for x86 targets. [v3]

[Jatin Bhateja]

On Sun, 5 Nov 2023 12:58:33 GMT, Jatin Bhateja <jbhateja at openjdk.org> wrote:

>> src/hotspot/cpu/x86/x86.ad line 4074:
>> 
>>> 4072:     BasicType elem_bt = Matcher::vector_element_basic_type(this);
>>> 4073:     assert(!is_subword_type(elem_bt), "sanity"); // T_INT, T_LONG, T_FLOAT, T_DOUBLE
>>> 4074:     __ vpcmpeqd($mask$$XMMRegister, $mask$$XMMRegister, $mask$$XMMRegister, vlen_enc);
>> 
>> vpcmpeqd is expensive instruction as compared to movdqu and also unrelated to subword type  support.
>
> compare instruction here does not access a memory operand, hence its cheaper compared to memory loads.

compare instruction here does not access a memory operand, hence its cheaper compared to memory loads.

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

PR Review Comment: https://git.openjdk.org/jdk/pull/16354#discussion_r1382571224