On Wed, 8 Apr 2026 16:37:10 GMT, Sandhya Viswanathan <[email protected]>
wrote:
>> It was segregated because for size_in_bits set to 512 we don't need AVX512VL.
>
> In the instruct rules below the predicate is always handling avx512vlbw
> together. There is no separate consideration for avx512bw alone. e.g. the
> 512-bit vector width in such case goes to vector_slice_avx which would
> assert.
>
> As there is no such platform that supports bw and doesn't support vl there is
> no point in adding complications. Better to simplify it here as:
>
> if (UseAVX > 2 && !VM_Version::supports_avx512vlbw()) {
> return false;
> }
Hi @sviswa7 , Addressed!
-------------
PR Review Comment: https://git.openjdk.org/jdk/pull/24104#discussion_r3052952030
