RFR: 8275863: Use encodeASCII for ASCII-compatible DoubleByte encodings
Enhance ASCII-compatible `DoubleByte` encodings to make use of the `StringCoding.implEncodeAsciiArray` intrinsic, which makes many such `CharsetEncoder`s encode ASCII text at speeds comparable to most single-byte encoders - and also more in line with how well these charsets behave when calling `String.getBytes`: Before: Benchmark (size) (type) Mode Cnt Score Error Units CharsetEncodeDecode.encode 16384 ISO-8859-1 avgt 30 3.021 ± 0.120 us/op CharsetEncodeDecode.encode 16384 Shift-JIS avgt 30 47.793 ± 1.942 us/op CharsetEncodeDecode.encode 16384 GB2312 avgt 30 49.598 ± 2.006 us/op CharsetEncodeDecode.encode 16384 EUC-JP avgt 30 68.709 ± 5.019 us/op CharsetEncodeDecode.encode 16384 EUC-KR avgt 30 48.033 ± 1.651 us/op Patched: Benchmark (size) (type) Mode Cnt Score Error Units CharsetEncodeDecode.encode 16384 ISO-8859-1 avgt 30 2.856 ± 0.078 us/op CharsetEncodeDecode.encode 16384 Shift-JIS avgt 30 5.287 ± 0.209 us/op CharsetEncodeDecode.encode 16384 GB2312 avgt 30 5.490 ± 0.251 us/op CharsetEncodeDecode.encode 16384 EUC-JP avgt 30 7.657 ± 0.368 us/op CharsetEncodeDecode.encode 16384 EUC-KR avgt 30 5.718 ± 0.267 us/op The `isASCIICompatible` predicate on `DoubleByte` was added in JEP 254 for the purpose of implementing such ASCII fast-paths, but is only used in what is now `String.encodeWithEncoder` to speed up `String.getBytes(...)` operations. Testing: tier1-3 ------------- Commit messages: - Use field directly - Grant accessClassInPackage.jdk.internal.access permission to jdk.charsets module - 8275863: Use encodeASCII for ASCII-compatible DoubleByte encodings Changes: https://git.openjdk.java.net/jdk/pull/6102/files Webrev: https://webrevs.openjdk.java.net/?repo=jdk&pr=6102&range=00 Issue: https://bugs.openjdk.java.net/browse/JDK-8275863 Stats: 35 lines in 5 files changed: 24 ins; 4 del; 7 mod Patch: https://git.openjdk.java.net/jdk/pull/6102.diff Fetch: git fetch https://git.openjdk.java.net/jdk pull/6102/head:pull/6102 PR: https://git.openjdk.java.net/jdk/pull/6102
On Mon, 25 Oct 2021 10:16:23 GMT, Claes Redestad <redestad@openjdk.org> wrote:
Enhance ASCII-compatible `DoubleByte` encodings to make use of the `StringCoding.implEncodeAsciiArray` intrinsic, which makes many such `CharsetEncoder`s encode ASCII text at speeds comparable to most single-byte encoders - and also more in line with how well these charsets behave when calling `String.getBytes`:
Before:
Benchmark (size) (type) Mode Cnt Score Error Units CharsetEncodeDecode.encode 16384 ISO-8859-1 avgt 30 3.021 ± 0.120 us/op CharsetEncodeDecode.encode 16384 Shift-JIS avgt 30 47.793 ± 1.942 us/op CharsetEncodeDecode.encode 16384 GB2312 avgt 30 49.598 ± 2.006 us/op CharsetEncodeDecode.encode 16384 EUC-JP avgt 30 68.709 ± 5.019 us/op CharsetEncodeDecode.encode 16384 EUC-KR avgt 30 48.033 ± 1.651 us/op
Patched:
Benchmark (size) (type) Mode Cnt Score Error Units CharsetEncodeDecode.encode 16384 ISO-8859-1 avgt 30 2.856 ± 0.078 us/op CharsetEncodeDecode.encode 16384 Shift-JIS avgt 30 5.287 ± 0.209 us/op CharsetEncodeDecode.encode 16384 GB2312 avgt 30 5.490 ± 0.251 us/op CharsetEncodeDecode.encode 16384 EUC-JP avgt 30 7.657 ± 0.368 us/op CharsetEncodeDecode.encode 16384 EUC-KR avgt 30 5.718 ± 0.267 us/op
The `isASCIICompatible` predicate on `DoubleByte` was added in JEP 254 for the purpose of implementing such ASCII fast-paths, but is only used in what is now `String.encodeWithEncoder` to speed up `String.getBytes(...)` operations.
Testing: tier1-3
Looks good. ------------- Marked as reviewed by naoto (Reviewer). PR: https://git.openjdk.java.net/jdk/pull/6102
On Mon, 25 Oct 2021 10:16:23 GMT, Claes Redestad <redestad@openjdk.org> wrote:
Enhance ASCII-compatible `DoubleByte` encodings to make use of the `StringCoding.implEncodeAsciiArray` intrinsic, which makes many such `CharsetEncoder`s encode ASCII text at speeds comparable to most single-byte encoders - and also more in line with how well these charsets behave when calling `String.getBytes`:
Before:
Benchmark (size) (type) Mode Cnt Score Error Units CharsetEncodeDecode.encode 16384 ISO-8859-1 avgt 30 3.021 ± 0.120 us/op CharsetEncodeDecode.encode 16384 Shift-JIS avgt 30 47.793 ± 1.942 us/op CharsetEncodeDecode.encode 16384 GB2312 avgt 30 49.598 ± 2.006 us/op CharsetEncodeDecode.encode 16384 EUC-JP avgt 30 68.709 ± 5.019 us/op CharsetEncodeDecode.encode 16384 EUC-KR avgt 30 48.033 ± 1.651 us/op
Patched:
Benchmark (size) (type) Mode Cnt Score Error Units CharsetEncodeDecode.encode 16384 ISO-8859-1 avgt 30 2.856 ± 0.078 us/op CharsetEncodeDecode.encode 16384 Shift-JIS avgt 30 5.287 ± 0.209 us/op CharsetEncodeDecode.encode 16384 GB2312 avgt 30 5.490 ± 0.251 us/op CharsetEncodeDecode.encode 16384 EUC-JP avgt 30 7.657 ± 0.368 us/op CharsetEncodeDecode.encode 16384 EUC-KR avgt 30 5.718 ± 0.267 us/op
The `isASCIICompatible` predicate on `DoubleByte` was added in JEP 254 for the purpose of implementing such ASCII fast-paths, but is only used in what is now `String.encodeWithEncoder` to speed up `String.getBytes(...)` operations.
Testing: tier1-3
Marked as reviewed by rriggs (Reviewer). ------------- PR: https://git.openjdk.java.net/jdk/pull/6102
On Mon, 25 Oct 2021 10:16:23 GMT, Claes Redestad <redestad@openjdk.org> wrote:
Enhance ASCII-compatible `DoubleByte` encodings to make use of the `StringCoding.implEncodeAsciiArray` intrinsic, which makes many such `CharsetEncoder`s encode ASCII text at speeds comparable to most single-byte encoders - and also more in line with how well these charsets behave when calling `String.getBytes`:
Before:
Benchmark (size) (type) Mode Cnt Score Error Units CharsetEncodeDecode.encode 16384 ISO-8859-1 avgt 30 3.021 ± 0.120 us/op CharsetEncodeDecode.encode 16384 Shift-JIS avgt 30 47.793 ± 1.942 us/op CharsetEncodeDecode.encode 16384 GB2312 avgt 30 49.598 ± 2.006 us/op CharsetEncodeDecode.encode 16384 EUC-JP avgt 30 68.709 ± 5.019 us/op CharsetEncodeDecode.encode 16384 EUC-KR avgt 30 48.033 ± 1.651 us/op
Patched:
Benchmark (size) (type) Mode Cnt Score Error Units CharsetEncodeDecode.encode 16384 ISO-8859-1 avgt 30 2.856 ± 0.078 us/op CharsetEncodeDecode.encode 16384 Shift-JIS avgt 30 5.287 ± 0.209 us/op CharsetEncodeDecode.encode 16384 GB2312 avgt 30 5.490 ± 0.251 us/op CharsetEncodeDecode.encode 16384 EUC-JP avgt 30 7.657 ± 0.368 us/op CharsetEncodeDecode.encode 16384 EUC-KR avgt 30 5.718 ± 0.267 us/op
The `isASCIICompatible` predicate on `DoubleByte` was added in JEP 254 for the purpose of implementing such ASCII fast-paths, but is only used in what is now `String.encodeWithEncoder` to speed up `String.getBytes(...)` operations.
Testing: tier1-3
Good work ------------- Marked as reviewed by alanb (Reviewer). PR: https://git.openjdk.java.net/jdk/pull/6102
On Mon, 25 Oct 2021 10:16:23 GMT, Claes Redestad <redestad@openjdk.org> wrote:
Enhance ASCII-compatible `DoubleByte` encodings to make use of the `StringCoding.implEncodeAsciiArray` intrinsic, which makes many such `CharsetEncoder`s encode ASCII text at speeds comparable to most single-byte encoders - and also more in line with how well these charsets behave when calling `String.getBytes`:
Before:
Benchmark (size) (type) Mode Cnt Score Error Units CharsetEncodeDecode.encode 16384 ISO-8859-1 avgt 30 3.021 ± 0.120 us/op CharsetEncodeDecode.encode 16384 Shift-JIS avgt 30 47.793 ± 1.942 us/op CharsetEncodeDecode.encode 16384 GB2312 avgt 30 49.598 ± 2.006 us/op CharsetEncodeDecode.encode 16384 EUC-JP avgt 30 68.709 ± 5.019 us/op CharsetEncodeDecode.encode 16384 EUC-KR avgt 30 48.033 ± 1.651 us/op
Patched:
Benchmark (size) (type) Mode Cnt Score Error Units CharsetEncodeDecode.encode 16384 ISO-8859-1 avgt 30 2.856 ± 0.078 us/op CharsetEncodeDecode.encode 16384 Shift-JIS avgt 30 5.287 ± 0.209 us/op CharsetEncodeDecode.encode 16384 GB2312 avgt 30 5.490 ± 0.251 us/op CharsetEncodeDecode.encode 16384 EUC-JP avgt 30 7.657 ± 0.368 us/op CharsetEncodeDecode.encode 16384 EUC-KR avgt 30 5.718 ± 0.267 us/op
The `isASCIICompatible` predicate on `DoubleByte` was added in JEP 254 for the purpose of implementing such ASCII fast-paths, but is only used in what is now `String.encodeWithEncoder` to speed up `String.getBytes(...)` operations.
Testing: tier1-3
Thanks for reviewing! ------------- PR: https://git.openjdk.java.net/jdk/pull/6102
On Mon, 25 Oct 2021 10:16:23 GMT, Claes Redestad <redestad@openjdk.org> wrote:
Enhance ASCII-compatible `DoubleByte` encodings to make use of the `StringCoding.implEncodeAsciiArray` intrinsic, which makes many such `CharsetEncoder`s encode ASCII text at speeds comparable to most single-byte encoders - and also more in line with how well these charsets behave when calling `String.getBytes`:
Before:
Benchmark (size) (type) Mode Cnt Score Error Units CharsetEncodeDecode.encode 16384 ISO-8859-1 avgt 30 3.021 ± 0.120 us/op CharsetEncodeDecode.encode 16384 Shift-JIS avgt 30 47.793 ± 1.942 us/op CharsetEncodeDecode.encode 16384 GB2312 avgt 30 49.598 ± 2.006 us/op CharsetEncodeDecode.encode 16384 EUC-JP avgt 30 68.709 ± 5.019 us/op CharsetEncodeDecode.encode 16384 EUC-KR avgt 30 48.033 ± 1.651 us/op
Patched:
Benchmark (size) (type) Mode Cnt Score Error Units CharsetEncodeDecode.encode 16384 ISO-8859-1 avgt 30 2.856 ± 0.078 us/op CharsetEncodeDecode.encode 16384 Shift-JIS avgt 30 5.287 ± 0.209 us/op CharsetEncodeDecode.encode 16384 GB2312 avgt 30 5.490 ± 0.251 us/op CharsetEncodeDecode.encode 16384 EUC-JP avgt 30 7.657 ± 0.368 us/op CharsetEncodeDecode.encode 16384 EUC-KR avgt 30 5.718 ± 0.267 us/op
The `isASCIICompatible` predicate on `DoubleByte` was added in JEP 254 for the purpose of implementing such ASCII fast-paths, but is only used in what is now `String.encodeWithEncoder` to speed up `String.getBytes(...)` operations.
Testing: tier1-3
This pull request has now been integrated. Changeset: 6c05cc9d Author: Claes Redestad <redestad@openjdk.org> URL: https://git.openjdk.java.net/jdk/commit/6c05cc9d15fb6014b8293a66ef132f3461ba... Stats: 35 lines in 5 files changed: 24 ins; 4 del; 7 mod 8275863: Use encodeASCII for ASCII-compatible DoubleByte encodings Reviewed-by: naoto, rriggs, alanb ------------- PR: https://git.openjdk.java.net/jdk/pull/6102
participants (4)
-
Alan Bateman
-
Claes Redestad
-
Naoto Sato
-
Roger Riggs