Adding an intrinsic to the interpreter

[Paul Sandoz]

On 14 Sep 2015, at 13:55, Aleksey Shipilev <aleksey.shipilev at oracle.com> wrote:

> Hi Paul,
> 
> Not saying we should actually care about the interpreter performance.
> 
> I would speculate the real trouble is calling multiple native getLong-s
> in getLongUnaligned?
> If so, would re-implementing these Unsafe methods
> in unsafe.cpp help performance?

As a quick experiment I replaced calls in vectorizedMismatch from getLongUnaligned to getLong (safe to do on x86), that makes things slightly better but AFAICT the Unsafe access is still dominating, probably because of JNI.

Paul.

> That will do a single native call into C
> function that you want, and it seems to match your definition of
> "interpreter intrinsic" below.
> 
> -Aleksey
> 
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 841 bytes
Desc: Message signed with OpenPGP using GPGMail
URL: <http://mail.openjdk.java.net/pipermail/hotspot-compiler-dev/attachments/20150914/d69e6b1e/signature.asc>