RFR: 8347009: Speed ​​up parseInt and parseLong [v11]

[Shaojin Wen]

On Wed, 29 Jan 2025 18:47:25 GMT, Johannes Graham <duke at openjdk.org> wrote:

>> Shaojin Wen has updated the pull request incrementally with one additional commit since the last revision:
>> 
>>   copyright
>
> src/java.base/share/classes/jdk/internal/util/DecimalDigits.java line 181:
> 
>> 179:             return -1;
>> 180:         }
>> 181:         return ((d & 0xF) << 3) + ((d & 0xF) << 1)  // (d & 0xF) * 10
> 
> Is simply using `(d & 0xF) * 10` any worse? I expect the compiler knows this trick and would use it when appropriate.

The performance results are identical across x64 and AArch64 platforms.

## 1. Script

git remote add wenshao git at github.com:wenshao/jdk.git
git fetch wenshao

#baseline (shift)
git checkout 0b04a70609b3195aca6481c6f1c88b27c872a65b
make test TEST="micro:java.lang.Longs.parseLong"

# current (muliti 10)
git checkout a6d9846809d6213fd70e852773ff2f57b5bf0201
make test TEST="micro:java.lang.Longs.parseLong"


## 2. MacBook M1 Pro (aarch64)

Benchmark        (size)  Mode  Cnt  Score   Error  Units (0b04a70609b)
Longs.parseLong     500  avgt   15  2.237 ? 0.029  us/op


Benchmark        (size)  Mode  Cnt  Score   Error  Units (a6d9846809d)
Longs.parseLong     500  avgt   15  2.225 ? 0.021  us/op


## 3. aliyun_ecs_c8a_x64 (CPU AMD EPYC™ Genoa)

Benchmark        (size)  Mode  Cnt  Score   Error  Units (0b04a70609b)
Longs.parseLong     500  avgt   15  2.920 ± 0.018  us/op

Benchmark        (size)  Mode  Cnt  Score   Error  Units (a6d9846809d)
Longs.parseLong     500  avgt   15  2.912 ± 0.020  us/op

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

PR Review Comment: https://git.openjdk.org/jdk/pull/22919#discussion_r1935027236