RFR(XS) 8161292: [JVMCI] missing test files from 8159368

[Tom Rodriguez]

The native calling convention was broken on windows and this also exposed general problems with the AMD64TestAssembler.  Windows native calling only ever uses registers for the first 4 arguments so logic had to be added to AMD64HotSpotRegisterConfig to support this.  Additionally the test assembler wasn’t computing the location of the integer stack slots properly and the argument loading needs to use scratch registers.  I expanded the test to have cases that pass 32 of float, double, int and long to make sure that extreme cases are exercised on all platforms.  This properly passes JPRT.

The changes relative to the original review are http://cr.openjdk.java.net/~never/8161292.diff/webrev/ <http://cr.openjdk.java.net/~never/8161292.diff/webrev/> and the full webrev is http://cr.openjdk.java.net/~never/8161292.v2/webrev/ <http://cr.openjdk.java.net/~never/8161292.v2/webrev/>.

tom

> On Jul 13, 2016, at 4:44 PM, Tom Rodriguez <tom.rodriguez at oracle.com> wrote:
> 
> 
>> On Jul 13, 2016, at 12:19 PM, Tom Rodriguez <tom.rodriguez at oracle.com <mailto:tom.rodriguez at oracle.com>> wrote:
>> 
>> 
>>> On Jul 13, 2016, at 11:49 AM, Christian Thalinger <cthalinger at twitter.com <mailto:cthalinger at twitter.com>> wrote:
>>> 
>>> Looks good.  How did it pass JPRT?
>> 
>> Thanks.  It’s a completely new test.
>> 
>> tom
> 
> Sending it through JPRT exposed a bug where the native calling convention isn’t properly described on windows.  There are only 4 XMM registers used there.  Since the native calling convention is really only used for things which take at most 2 floats, we never noticed.  I’m fixing that now and will send out a new webrev in a bit.
> 
> tom
> 
>> 
>>> 
>>>> On Jul 13, 2016, at 5:55 AM, Tom Rodriguez <tom.rodriguez at oracle.com <mailto:tom.rodriguez at oracle.com>> wrote:
>>>> 
>>>> http://cr.openjdk.java.net/~never/8161292/webrev <http://cr.openjdk.java.net/~never/8161292/webrev>
>>>> 
>>>> I missed some test files file the push for 8159368.  They are unchanged from the review.  Can I get a quick review?
>>>> 
>>>> tom
>>> 
>> 
> 

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.openjdk.java.net/pipermail/hotspot-compiler-dev/attachments/20160718/91cad0b6/attachment-0001.html>