On Thu, 9 Jun 2022 18:38:18 GMT, Brian Burkhalter <b...@openjdk.org> wrote:
>> Modify native multi-byte read-write code used by the `java.io` classes to
>> limit the size of the allocated native buffer thereby decreasing off-heap
>> memory footprint and increasing throughput.
>
> Brian Burkhalter has updated the pull request with a new target base due to a
> merge or a rebase. The incremental webrev excludes the unrelated changes
> brought in by the merge/rebase. The pull request contains seven additional
> commits since the last revision:
>
> - Merge
> - 6478546: Add break in write loop on ExceptionOccurred
> - Merge
> - 6478546: Clean up io_util.c
> - Merge
> - 6478546: Decrease malloc'ed buffer maximum size to 64 kB
> - 6478546: FileInputStream.read() throws OutOfMemoryError when there is
> plenty available
The following benchmark results are for commit
e6f3b4574b1c49955e5acaca618a3083eb132e05 with 5 warmup iterations of 5 seconds
each, followed by 10 measurement iterations of 10 seconds each on a
MacBookPro16,1.
FileInputStream.read(byte[])
Score (ops/s)
(length) Master Branch
16384 5850.383 ± 220.209 5805.952 ± 187.870
32768 5781.410 ± 345.939 5807.587 ± 515.808
100000 5444.167 ± 409.680 5273.988 ± 159.652
500000 3748.413 ± 88.505 3803.577 ± 256.442
1000000 2647.356 ± 145.245 2700.523 ± 97.717
1048576 2759.337 ± 362.618 2631.779 ± 45.971
10485760 338.840 ± 2.706 392.126 ± 2.59
251503002 10.626 ± 0.024 20.165 ± 0.091
524288000 5.134 ± 0.024 9.826 ± 0.068
1000000000 1.806 ± 0.014 5.194 ± 0.015
FileOutputStream.write(byte[])
Score (ops/s)
(length) Master Branch
16384 2736.489 ± 31.547 2599.203 ± 79.392
32768 2656.471 ± 72.597 2523.470 ± 158.460
100000 2476.514 ± 46.392 2225.597 ± 442.285
500000 1606.205 ± 487.787 1281.550 ± 413.797
1000000 1125.265 ± 13.969 812.184 ± 232.23
1048576 1098.608 ± 13.528 816.657 ± 218.112
10485760 136.867 ± 2.310 153.137 ± 2.93
251503002 7.266 ± 0.026 10.546 ± 0.125
524288000 3.450 ± 0.016 3.880 ± 1.355
1000000000 1.178 ± 0.007 2.637 ± 0.505
The proposed change is approximately at parity with the current code base for
smaller lengths, but has higher throughput for larger lengths, at least for
reading. In all cases where the length is greater than or equal to 65536 (64
KiB) or is a multiple of 8192 (8 KiB), less native memory is allocated.
-------------
PR: https://git.openjdk.org/jdk/pull/8235
