Hi Corey,
If there's a performance issue, that's an option.
Though its tempting to add intrinsics for each supported instruction
set, it adds code
that has to be maintained forever, so it best to wait till its
determined to be a bottleneck.
Most of the cases I've seen in the code are fairly small, and in
non-performance critical data paths,
for example, the toString method of a certificate.
Thanks, Roger
On 8/24/20 5:11 PM, Corey Ashford wrote:
Just as a general comment, an SIMD-based encode() intrinsic would be
an easy performance upgrade for this Hex class. I'd guess you'd get a
5x-10x improvement, depending on the specific CPU/arch being used.
That said, the intrinsic would do better on larger buffers, and I
don't know what the average length is for the use cases you have in mind.
- Corey
On 8/19/20 2:14 PM, Roger Riggs wrote:
Please review a java.util.Hex API to encode and decode hexadecimal
strings to and from byte arrays.
Within the JDK and JDK tests there are multiple implementations to
encode and decode
hexadecimal strings to byte arrays. Hex encoders and decoders support
upper or lower case hexadecimal characters, delimiters, prefix, and
suffix.
The API is modeled after the java.util.Base64 API providing static
factories,
immutable threadsafe instances with methods to encode to and decode from
string and StringBuilder.
JavaDoc:
http://cr.openjdk.java.net/~rriggs/hex-javadoc/java.base/java/util/Hex.html
Webrev for Hex encoder and decoder:
http://cr.openjdk.java.net/~rriggs/webrev-hex-encoder-8251989
Webrev for applying to java.security:
http://cr.openjdk.java.net/~rriggs/webrev-hex-security-8252055
CSR:
https://bugs.openjdk.java.net/browse/JDK-8251991
Issue for API and a few uses:
https://bugs.openjdk.java.net/browse/JDK-8251989
Issue for Use in java.security and tests:
https://bugs.openjdk.java.net/browse/JDK-8252055
