RFR: 8342806: Desugar capturing lambda in StringNameTable [v2]
Chen Liang
liach at openjdk.org
Tue Oct 22 11:27:39 UTC 2024
On Tue, 22 Oct 2024 10:14:05 GMT, Claes Redestad <redestad at openjdk.org> wrote:
>> - Desugar `computeIfAbsent` to `get` & `put` to avoid a capturing lambda. Since the map is a non-synchronized `HashMap` this is just as non-thread-safe as before.
>> - Salvage the microbenchmark added as a JBS comment to https://bugs.openjdk.org/browse/JDK-8268622
>> - Microbenchmark show a modest reduction in allocations for tests stressing `StringNameTable` (Macbook M1):
>>
>>
>> Name Cnt Base Error Test Error Unit Change
>> testInternStringTable 10 55,731 ± 3,701 51,083 ± 2,062 ms/op 1,09x (p = 0,000*)
>> :gc.alloc.rate.norm 54705366,956 ± 375197,844 53936224,624 ± 1421298,600 B/op 0,99x (p = 0,031 )
>> testSharedTable 10 47,997 ± 5,802 48,022 ± 4,968 ms/op 1,00x (p = 0,988 )
>> :gc.alloc.rate.norm 56562680,368 ± 1535117,433 53403645,438 ± 218526,779 B/op 0,94x (p = 0,000*)
>> testStringTable 10 45,933 ± 2,542 44,094 ± 1,360 ms/op 1,04x (p = 0,009*)
>> :gc.alloc.rate.norm 56244541,359 ± 1019940,919 53226602,956 ± 510018,314 B/op 0,95x (p = 0,000*)
>> testUnsharedTable 10 60,474 ± 1,912 59,068 ± 1,038 ms/op 1,02x (p = 0,008*)
>> :gc.alloc.rate.norm 66970079,820 ± 1402704,512 67285851,746 ± 883540,307 B/op 1,00x (p = 0,377 )
>> * = significant
>
> Claes Redestad has updated the pull request incrementally with one additional commit since the last revision:
>
> de-this
src/jdk.compiler/share/classes/com/sun/tools/javac/util/StringNameTable.java line 68:
> 66: if (name == null) {
> 67: name = new NameImpl(this, intern ? string.intern() : string);
> 68: nameMap.put(string, name);
Since we are interning, can we use the possibly-interned result as map key when we put the name? This will be another bug fixed by desugaring.
-------------
PR Review Comment: https://git.openjdk.org/jdk/pull/21631#discussion_r1810540329
More information about the compiler-dev
mailing list