RFR 8135028 - support for vectorizing double precision sqrt

[Berg, Michael C]

I believe access is through Math.sqrt(…) for double precision, at least my test case follows that usage model.

Thanks,
-Michael

From: Christian Thalinger [mailto:christian.thalinger at oracle.com]
Sent: Tuesday, September 08, 2015 10:19 AM
To: Berg, Michael C
Cc: hotspot-compiler-dev at openjdk.java.net
Subject: Re: RFR 8135028 - support for vectorizing double precision sqrt


On Sep 4, 2015, at 10:34 AM, Berg, Michael C <michael.c.berg at intel.com<mailto:michael.c.berg at intel.com>> wrote:


Hi Vladimir, as requested I have made a separate review for the sqrt vector changes.

I would like to contribute support for vectorizing double precision sqrt on Intel architectures. I need two reviewers to examine this patch and comment as needed:

Bug-id: https://bugs.openjdk.java.net/browse/JDK-8135028

webrev:
http://cr.openjdk.java.net/~mcberg/8135028/webrev.01/


These changes are made to add support to vectorize double precision sqrt which is available through the math library in scalar form.

The change looks good but I don’t understand the “math lib dependency” comments.


The changes are for current and future Intel architectures.

Thanks,
Michael


-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.openjdk.java.net/pipermail/hotspot-compiler-dev/attachments/20150908/de6530dd/attachment.html>