On 09/30/2014 02:51 AM, Florian Weimer wrote:
On 09/29/2014 10:33 PM, Anthony Scarpino wrote:
Please review the below JEP for JVM Hardware Crypto Acceleration. The
focus is to use ISA crypto instructions available in x86 and sparc
platforms and to help access to use them.
https://bugs.openjdk.java.net/browse/JDK-8046943
Can you show what you actually benchmarked? Just GHASH, or the full
cipher suite?
The actual impact of this change will be less than expected for TLS
because after fixing GHASH (for which I had already posted a patch), the
TLS implementation is largely GC-bound.
As this being a work in progress, much isn't done. That being said, the
prototype using PCLMULQDQ with AES-GCM performed roughly 34x better.
Tony
