- https://bugs.openjdk.org/browse/JDK-8321271
- Diff: https://github.com/vlsi/jdk/compare/ce108446ca1fe604ecc24bbefb0bf1c6318271c7...write_bytebuffer
- https://github.com/vlsi/jdk/tree/write_bytebuffer
- https://github.com/vlsi/jdk/blob/write_bytebuffer/test/micro/org/openjdk/bench/java/io/ByteArrayInputStreamTransfer.java
ByteArrayInputStream.transferTo(BufferedOutputStream) no longer allocates temporary arrays, and it improves performance ~2x for 1MiB inputs.
The current ByteArrayInputStream always allocates 128KiB buffer (it allocates min(bufferedData.length, 128KiB) copies the data for untrusted streams which is likely the reason the suggested in JDK-8321271 approach yields ~5x improvement for 10KiB input.
ByteArrayInputStream.transferTo(FileOutputStream) was special-cased previously, so the performance is the same.
However, ByteArrayInputStream.transferTo(DataOutputStream(FileOutputStream)) can not be special-cased (see transferToDataFile), and the suggested in JDK-8321271 approach again reduces allocations and add improves performance.
Apple M1, openjdk ce108446ca1fe604ecc24bbefb0bf1c6318271c7
$ make test TEST="micro:ByteArrayInputStreamTransfer" MICRO="FORK=2;WARMUP_ITER=2;OPTIONS=-prof gc"
UDP 2023-12-16: ZIP-related failures are fixed, so benchmark results are valid now
Benchmark (size) Mode Cnt Score Error Units
transferToBAOS 1024 avgt 12 0,046 ± 0,004 us/op
transferToBAOS 102400 avgt 12 1,637 ± 0,395 us/op
transferToBAOS 1024000 avgt 12 20,467 ± 1,614 us/op
transferToBufferedBAOS 1024 avgt 12 0,109 ± 0,002 us/op
transferToBufferedBAOS 102400 avgt 12 7,012 ± 0,043 us/op
transferToBufferedBAOS 1024000 avgt 12 47,927 ± 0,948 us/op
transferToDataFile 1024 avgt 12 49,816 ± 1,173 us/op
transferToDataFile 102400 avgt 12 79,839 ± 2,705 us/op
transferToDataFile 1024000 avgt 12 442,584 ± 24,911 us/op
transferToFile 1024 avgt 12 50,021 ± 1,661 us/op
transferToFile 102400 avgt 12 75,704 ± 2,429 us/op
transferToFile 1024000 avgt 12 415,853 ± 32,296 us/op
transferToGZIP 1024 avgt 12 18,019 ± 0,828 us/op
transferToGZIP 102400 avgt 12 1783,493 ± 21,657 us/op
transferToGZIP 1024000 avgt 12 19647,320 ± 157,758 us/op
Benchmark (size) Mode Cnt Score Error Units
transferToBAOS 1024 avgt 12 0,046 ± 0,008 us/op
transferToBAOS 102400 avgt 12 1,539 ± 0,381 us/op
transferToBAOS 1024000 avgt 12 20,561 ± 2,254 us/op
transferToBufferedBAOS 1024 avgt 12 0,071 ± 0,002 us/op
transferToBufferedBAOS 102400 avgt 12 1,596 ± 0,346 us/op
transferToBufferedBAOS 1024000 avgt 12 18,198 ± 3,705 us/op
transferToDataFile 1024 avgt 12 48,778 ± 2,007 us/op
transferToDataFile 102400 avgt 12 75,509 ± 1,703 us/op
transferToDataFile 1024000 avgt 12 389,400 ± 20,381 us/op
transferToFile 1024 avgt 12 49,936 ± 2,021 us/op
transferToFile 102400 avgt 12 76,250 ± 4,523 us/op
transferToFile 1024000 avgt 12 553,221 ± 119,340 us/op
transferToGZIP 1024 avgt 12 19,963 ± 1,815 us/op
transferToGZIP 102400 avgt 12 1786,237 ± 10,187 us/op
transferToGZIP 1024000 avgt 12 19593,844 ± 55,469 us/op
Benchmark (size) Mode Cnt Score Error Units
transferToBAOS 1024 avgt 12 0,046 ± 0,004 us/op
transferToBAOS:gc.alloc.rate 1024 avgt 12 0,009 ± 0,001 MB/sec
transferToBAOS:gc.alloc.rate.norm 1024 avgt 12 0,001 ± 0,001 B/op
transferToBAOS:gc.count 1024 avgt 12 ~ 0 counts
transferToBAOS 102400 avgt 12 1,637 ± 0,395 us/op
transferToBAOS:gc.alloc.rate 102400 avgt 12 0,105 ± 0,001 MB/sec
transferToBAOS:gc.alloc.rate.norm 102400 avgt 12 0,185 ± 0,044 B/op
transferToBAOS:gc.count 102400 avgt 12 ~ 0 counts
transferToBAOS 1024000 avgt 12 20,467 ± 1,614 us/op
transferToBAOS:gc.alloc.rate 1024000 avgt 12 0,978 ± 0,002 MB/sec
transferToBAOS:gc.alloc.rate.norm 1024000 avgt 12 21,077 ± 1,659 B/op
transferToBAOS:gc.count 1024000 avgt 12 ~ 0 counts
transferToBufferedBAOS 1024 avgt 12 0,109 ± 0,002 us/op
transferToBufferedBAOS:gc.alloc.rate 1024 avgt 12 7028,938 ± 104,401 MB/sec
transferToBufferedBAOS:gc.alloc.rate.norm 1024 avgt 12 1040,001 ± 0,001 B/op
transferToBufferedBAOS:gc.count 1024 avgt 12 96,000 counts
transferToBufferedBAOS:gc.time 1024 avgt 12 68,000 ms
transferToBufferedBAOS 102400 avgt 12 7,012 ± 0,043 us/op
transferToBufferedBAOS:gc.alloc.rate 102400 avgt 12 13857,300 ± 85,362 MB/sec
transferToBufferedBAOS:gc.alloc.rate.norm 102400 avgt 12 102416,776 ± 0,004 B/op
transferToBufferedBAOS:gc.count 102400 avgt 12 142,000 counts
transferToBufferedBAOS:gc.time 102400 avgt 12 100,000 ms
transferToBufferedBAOS 1024000 avgt 12 47,927 ± 0,948 us/op
transferToBufferedBAOS:gc.alloc.rate 1024000 avgt 12 2602,958 ± 52,568 MB/sec
transferToBufferedBAOS:gc.alloc.rate.norm 1024000 avgt 12 131137,294 ± 0,972 B/op
transferToBufferedBAOS:gc.count 1024000 avgt 12 53,000 counts
transferToBufferedBAOS:gc.time 1024000 avgt 12 32,000 ms
transferToDataFile 1024 avgt 12 49,816 ± 1,173 us/op
transferToDataFile:gc.alloc.rate 1024 avgt 12 24,180 ± 0,567 MB/sec
transferToDataFile:gc.alloc.rate.norm 1024 avgt 12 1264,461 ± 0,017 B/op
transferToDataFile:gc.count 1024 avgt 12 2,000 counts
transferToDataFile:gc.time 1024 avgt 12 2,000 ms
transferToDataFile 102400 avgt 12 79,839 ± 2,705 us/op
transferToDataFile:gc.alloc.rate 102400 avgt 12 1225,219 ± 40,589 MB/sec
transferToDataFile:gc.alloc.rate.norm 102400 avgt 12 102648,813 ± 0,293 B/op
transferToDataFile:gc.count 102400 avgt 12 26,000 counts
transferToDataFile:gc.time 102400 avgt 12 18,000 ms
transferToDataFile 1024000 avgt 12 442,584 ± 24,911 us/op
transferToDataFile:gc.alloc.rate 1024000 avgt 12 283,786 ± 15,741 MB/sec
transferToDataFile:gc.alloc.rate.norm 1024000 avgt 12 131766,626 ± 26,260 B/op
transferToDataFile:gc.count 1024000 avgt 12 6,000 counts
transferToDataFile:gc.time 1024000 avgt 12 2,000 ms
transferToFile 1024 avgt 12 50,021 ± 1,661 us/op
transferToFile:gc.alloc.rate 1024 avgt 12 2,752 ± 0,090 MB/sec
transferToFile:gc.alloc.rate.norm 1024 avgt 12 144,466 ± 0,012 B/op
transferToFile:gc.count 1024 avgt 12 ~ 0 counts
transferToFile 102400 avgt 12 75,704 ± 2,429 us/op
transferToFile:gc.alloc.rate 102400 avgt 12 1,920 ± 0,058 MB/sec
transferToFile:gc.alloc.rate.norm 102400 avgt 12 152,519 ± 0,602 B/op
transferToFile:gc.count 102400 avgt 12 ~ 0 counts
transferToFile 1024000 avgt 12 415,853 ± 32,296 us/op
transferToFile:gc.alloc.rate 1024000 avgt 12 1,309 ± 0,025 MB/sec
transferToFile:gc.alloc.rate.norm 1024000 avgt 12 571,616 ± 33,224 B/op
transferToFile:gc.count 1024000 avgt 12 ~ 0 counts
transferToGZIP 1024 avgt 12 18,019 ± 0,828 us/op
transferToGZIP:gc.alloc.rate 1024 avgt 12 95,582 ± 4,362 MB/sec
transferToGZIP:gc.alloc.rate.norm 1024 avgt 12 1808,167 ± 0,010 B/op
transferToGZIP:gc.count 1024 avgt 12 2,000 counts
transferToGZIP:gc.time 1024 avgt 12 2,000 ms
transferToGZIP 102400 avgt 12 1783,493 ± 21,657 us/op
transferToGZIP:gc.alloc.rate 102400 avgt 12 55,247 ± 0,662 MB/sec
transferToGZIP:gc.alloc.rate.norm 102400 avgt 12 103380,407 ± 2,432 B/op
transferToGZIP:gc.count 102400 avgt 12 ~ 0 counts
transferToGZIP 1024000 avgt 12 19647,320 ± 157,758 us/op
transferToGZIP:gc.alloc.rate 1024000 avgt 12 7,358 ± 0,050 MB/sec
transferToGZIP:gc.alloc.rate.norm 1024000 avgt 12 151838,119 ± 246,437 B/op
transferToGZIP:gc.count 1024000 avgt 12 2,000 counts
transferToGZIP:gc.time 1024000 avgt 12 2,000 ms
Benchmark (size) Mode Cnt Score Error Units
transferToBAOS 1024 avgt 12 0,046 ± 0,008 us/op
transferToBAOS:gc.alloc.rate 1024 avgt 12 0,009 ± 0,001 MB/sec
transferToBAOS:gc.alloc.rate.norm 1024 avgt 12 0,001 ± 0,001 B/op
transferToBAOS:gc.count 1024 avgt 12 ~ 0 counts
transferToBAOS 102400 avgt 12 1,539 ± 0,381 us/op
transferToBAOS:gc.alloc.rate 102400 avgt 12 35,240 ± 8,539 MB/sec
transferToBAOS:gc.alloc.rate.norm 102400 avgt 12 56,173 ± 0,042 B/op
transferToBAOS:gc.count 102400 avgt 12 2,000 counts
transferToBAOS:gc.time 102400 avgt 12 3,000 ms
transferToBAOS 1024000 avgt 12 20,561 ± 2,254 us/op
transferToBAOS:gc.alloc.rate 1024000 avgt 12 0,978 ± 0,002 MB/sec
transferToBAOS:gc.alloc.rate.norm 1024000 avgt 12 21,170 ± 2,325 B/op
transferToBAOS:gc.count 1024000 avgt 12 ~ 0 counts
transferToBufferedBAOS 1024 avgt 12 0,071 ± 0,002 us/op
transferToBufferedBAOS:gc.alloc.rate 1024 avgt 12 0,009 ± 0,001 MB/sec
transferToBufferedBAOS:gc.alloc.rate.norm 1024 avgt 12 0,001 ± 0,001 B/op
transferToBufferedBAOS:gc.count 1024 avgt 12 ~ 0 counts
transferToBufferedBAOS 102400 avgt 12 1,596 ± 0,346 us/op
transferToBufferedBAOS:gc.alloc.rate 102400 avgt 12 33,768 ± 7,426 MB/sec
transferToBufferedBAOS:gc.alloc.rate.norm 102400 avgt 12 56,179 ± 0,038 B/op
transferToBufferedBAOS:gc.count 102400 avgt 12 2,000 counts
transferToBufferedBAOS:gc.time 102400 avgt 12 3,000 ms
transferToBufferedBAOS 1024000 avgt 12 18,198 ± 3,705 us/op
transferToBufferedBAOS:gc.alloc.rate 1024000 avgt 12 0,978 ± 0,001 MB/sec
transferToBufferedBAOS:gc.alloc.rate.norm 1024000 avgt 12 18,732 ± 3,817 B/op
transferToBufferedBAOS:gc.count 1024000 avgt 12 ~ 0 counts
transferToDataFile 1024 avgt 12 48,778 ± 2,007 us/op
transferToDataFile:gc.alloc.rate 1024 avgt 12 7,669 ± 0,310 MB/sec
transferToDataFile:gc.alloc.rate.norm 1024 avgt 12 392,454 ± 0,018 B/op
transferToDataFile:gc.count 1024 avgt 12 2,000 counts
transferToDataFile:gc.time 1024 avgt 12 2,000 ms
transferToDataFile 102400 avgt 12 75,509 ± 1,703 us/op
transferToDataFile:gc.alloc.rate 102400 avgt 12 5,404 ± 0,498 MB/sec
transferToDataFile:gc.alloc.rate.norm 102400 avgt 12 428,315 ± 37,443 B/op
transferToDataFile:gc.count 102400 avgt 12 2,000 counts
transferToDataFile:gc.time 102400 avgt 12 2,000 ms
transferToDataFile 1024000 avgt 12 389,400 ± 20,381 us/op
transferToDataFile:gc.alloc.rate 1024000 avgt 12 3,035 ± 0,111 MB/sec
transferToDataFile:gc.alloc.rate.norm 1024000 avgt 12 1240,317 ± 21,035 B/op
transferToDataFile:gc.count 1024000 avgt 12 ~ 0 counts
transferToFile 1024 avgt 12 49,936 ± 2,021 us/op
transferToFile:gc.alloc.rate 1024 avgt 12 6,495 ± 0,760 MB/sec
transferToFile:gc.alloc.rate.norm 1024 avgt 12 340,219 ± 37,149 B/op
transferToFile:gc.count 1024 avgt 12 2,000 counts
transferToFile:gc.time 1024 avgt 12 2,000 ms
transferToFile 102400 avgt 12 76,250 ± 4,523 us/op
transferToFile:gc.alloc.rate 102400 avgt 12 4,009 ± 0,223 MB/sec
transferToFile:gc.alloc.rate.norm 102400 avgt 12 320,410 ± 0,497 B/op
transferToFile:gc.count 102400 avgt 12 2,000 counts
transferToFile:gc.time 102400 avgt 12 2,000 ms
transferToFile 1024000 avgt 12 553,221 ± 119,340 us/op
transferToFile:gc.alloc.rate 1024000 avgt 12 2,325 ± 0,329 MB/sec
transferToFile:gc.alloc.rate.norm 1024000 avgt 12 1328,495 ± 122,318 B/op
transferToFile:gc.count 1024000 avgt 12 ~ 0 counts
transferToGZIP 1024 avgt 12 19,963 ± 1,815 us/op
transferToGZIP:gc.alloc.rate 1024 avgt 12 39,448 ± 3,517 MB/sec
transferToGZIP:gc.alloc.rate.norm 1024 avgt 12 824,184 ± 0,018 B/op
transferToGZIP:gc.count 1024 avgt 12 2,000 counts
transferToGZIP:gc.time 1024 avgt 12 2,000 ms
transferToGZIP 102400 avgt 12 1786,237 ± 10,187 us/op
transferToGZIP:gc.alloc.rate 102400 avgt 12 0,574 ± 0,004 MB/sec
transferToGZIP:gc.alloc.rate.norm 102400 avgt 12 1075,544 ± 2,749 B/op
transferToGZIP:gc.count 102400 avgt 12 ~ 0 counts
transferToGZIP 1024000 avgt 12 19593,844 ± 55,469 us/op
transferToGZIP:gc.alloc.rate 1024000 avgt 12 1,026 ± 0,003 MB/sec
transferToGZIP:gc.alloc.rate.norm 1024000 avgt 12 21123,436 ± 2,204 B/op
transferToGZIP:gc.count 1024000 avgt 12 ~ 0 counts