On Sat, 29 Mar 2025 01:32:19 GMT, Vladimir Ivanov <vliva...@openjdk.org> wrote:
>> Hi all,
>>
>> The return type of function `const __m256i &perm` is `__m256i`, so `const
>> __m256i &perm` should be replaced as 'const __m256i perm'.
>>
>> The function implementation in gcc/clang compiler header:
>>
>> 1. gcc: lib/gcc/x86_64-pc-linux-gnu/14.2.0/include/avxintrin.h
>>
>>
>> extern __inline __m256i __attribute__((__gnu_inline__, __always_inline__,
>> __artificial__))
>> _mm256_loadu_si256 (__m256i_u const *__P)
>> {
>> return *__P;
>> }
>>
>>
>> 2. clang: lib64/clang/17/include/avxintrin.h
>>
>>
>> static __inline __m256i __DEFAULT_FN_ATTRS
>> _mm256_loadu_si256(__m256i_u const *__p)
>> {
>> struct __loadu_si256 {
>> __m256i_u __v;
>> } __attribute__((__packed__, __may_alias__));
>> return ((const struct __loadu_si256*)__p)->__v;
>> }
>>
>>
>> Additional testing:
>>
>> - [x] jtreg tests(include tier1/2/3 etc.) on linux-x64(AMD EPYC 9T24 96-Core
>> Processor) with release build
>> - [x] jtreg tests(include tier1/2/3 etc.) on linux-x64(AMD EPYC 9T24 96-Core
>> Processor) with fastdebug build
>
> Testing results (hs-tier1 - hs-tier4) are clean.
GHA test failures are unrelated to this PR. Thanks for the reviews @iwanowww
-------------
PR Comment: https://git.openjdk.org/jdk/pull/23925#issuecomment-2764554842
