RFR: 8356760: VectorAPI: Optimize VectorMask.fromLong for all-true/all-false cases
Xiaohong Gong
xgong at openjdk.org
Fri Jun 27 06:07:39 UTC 2025
On Fri, 27 Jun 2025 01:36:44 GMT, Xiaohong Gong <xgong at openjdk.org> wrote:
>> If the input long value `l` of `VectorMask.fromLong(SPECIES, l)` would set or unset all lanes, `VectorMask.fromLong(SPECIES, l)` is equivalent to `maskAll(true)` or `maskAll(false)`. But the cost of `maskAll` is relative smaller than that of `fromLong`. This patch does the conversion for these cases if `l` is a compile time constant.
>>
>> And this conversion also enables further optimizations that recognize maskAll patterns, see [1].
>>
>> Some JTReg test cases are added to ensure the optimization is effective.
>>
>> I tried many different ways to write a JMH benchmark, but failed. Since the input of `VectorMask.fromLong(SPECIES, l)` needs to be a specific compile-time constant, the statement will be hoisted out of the loop. If we don't use a loop, the hotspot will become other instructions, and no obvious performance change was observed. However, combined with the optimization of [1], we can observe a performance improvement of about 7% on both aarch64 and x64.
>>
>> The patch was tested on both aarch64 and x64, all of tier1 tier2 and tier3 tests passed.
>>
>> [1] https://github.com/openjdk/jdk/pull/24674
>
> src/hotspot/share/opto/vectorIntrinsics.cpp line 80:
>
>> 78: return false;
>> 79: }
>> 80: long mask = (0xFFFFFFFFFFFFFFFFULL >> (64 - vlen));
>
> The higher bits of long input should be cleared. So we should generate an unsigned right shift instead of the signed one?
I noticed that you used `ULL` suffix. So it should be fine. Please ignore above comment. Thanks!
-------------
PR Review Comment: https://git.openjdk.org/jdk/pull/25793#discussion_r2170871085
More information about the hotspot-compiler-dev
mailing list