RFR: 8265244: Remove useless comparation in LibraryCallKit::inline_vector_convert() [v2]

Andrew Haley aph at openjdk.java.net
Fri Apr 30 08:04:53 UTC 2021

On Fri, 16 Apr 2021 02:10:10 GMT, Wang Huang <whuang at openjdk.org> wrote:

>> * aarch64 can only accept `VectorReinterpret` with 64/128 bits.
>> * I will fix this bug by adding a rule for `VectorReinterpret` in `match_rule_supported_vector`
>> * after changing note with @nsjian and @XiaohongGong , I think that two checks in `inline_vector_conver` is useless now. However, these checks impact other cpus, so I need more reviewers.
> Wang Huang has updated the pull request with a new target base due to a merge or a rebase. The incremental webrev excludes the unrelated changes brought in by the merge/rebase. The pull request contains two additional commits since the last revision:
>  - Merge branch 'master' into JDK-8265244
>  - 8265244: assert(false) failed: bad AD file

src/hotspot/share/opto/vectorIntrinsics.cpp line 1452:

> 1450:       // Since input and output number of elements are not consistent, we need to make sure we
> 1451:       // properly size. Thus, first make a cast that retains the number of elements from source.
> 1452:       // In case the size exceeds the arch size, we do the minimum.

If you're going to remove the MIN2 you must also change this comment.

src/hotspot/share/opto/vectorIntrinsics.cpp line 1471:

> 1469:     } else if (num_elem_from > num_elem_to) {
> 1470:       // Since number elements from input is larger than output, simply reduce size of input (we are supposed to
> 1471:       // drop top elements anyway).

This one too?


PR: https://git.openjdk.java.net/jdk/pull/3507

More information about the hotspot-compiler-dev mailing list