RFR: 8254078: DataOutputStream is very slow post-disabling of Biased Locking [v4]

[Aleksey Shipilev]

On Mon, 12 Oct 2020 18:09:12 GMT, Andrew Haley <aph at openjdk.org> wrote:

>> test/micro/org/openjdk/bench/java/io/DataOutputStreamTest.java line 117:
>> 
>>> 115:             case STRING: writeString(state, dataOutput); break;
>>> 116:         }
>>> 117:     }
>> 
>> So, the way I see it, this can be replaced by:
>> 
>>     public void write(DataOutput dataOutput) {
>>         switch (basicType) {
>>             case CHAR:
>>                 for (int i = 0; i < SIZE; i += 2) {
>>                     dataOutput.writeChar(i);
>>                 }
>>                 break;
>>             case SHORT:
>>                 for (int i = 0; i < SIZE; i += 2) {
>>                     dataOutput.writeShort(i);
>>                 }
>>                 break;
>>             case INT:
>>                 for (int i = 0; i < SIZE; i += 4) {
>>                     dataOutput.writeInt(i);
>>                 }
>>                 break;
>>             case STRING:
>>                 dataOutput.writeChars(state.outputString);
>>                 break;
>>         }
>>     }
>> 
>> ...thus keeping all the code in one place, and simplifying the benchmark.
>
> I didn't quite do that, preferring to keep the writeX functions separate.

No problem, this is a question of personal style. If this was a very fine-grained benchmark, we would worry about
inlining/dispatch costs.

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

PR: https://git.openjdk.java.net/jdk/pull/542