CR for RFR 8167987

[Berg, Michael C]

Hi Folks,
Performance on EVEX enabled architectures can be improved by changing the model from merge to clear for mask register usage (by using {z} encoding) as it enables store forwarding.
The enhancement offers as much as 1.5-2x performance increase in some instances (mostly at the micro level), but in general improves most EVEX enabled code performance.

This code was tested as follows: hotspot jreg, SPECjvm2008, SPECjvm98 on hsw, skx and knl targets complete with no issues on 64-bit or 32-bit.

Bug-id: https://bugs.openjdk.java.net/browse/JDK-8167987

webrev:
http://cr.openjdk.java.net/~mcberg/8167987/webrev.01a/

Regards,
Michael
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.openjdk.java.net/pipermail/hotspot-compiler-dev/attachments/20161017/6d0d28a3/attachment-0001.html>