ARM: Fix handling of special methods

[Andrew Dinn]

On 08/06/12 15:28, Andrew Haley wrote:
> Two bugs:
> 
> Firstly, special-case inlines aren't being handled at all because
> of a thinko when merging the big patch.
> 
> Secondly, it may be just as well because compareAndSwapLong is being
> incorrectly compiled: we need to save all locals in registers, not
> just those of ref type.

This all looked fine by eyeball and the patched code built ok.

I don't have a test case to exercise the compare and swap code.

I ran a test case which repeatedly exercised the maths functions in a
tight loop (see attached). This brought execution time down from 33 secs
to 11 secs. I also eyeballed the generated code to check that it was
doing the right thing.

regards,


Andrew Dinn
-----------
-------------- next part --------------
A non-text attachment was scrubbed...
Name: Test9.java
Type: text/x-java
Size: 586 bytes
Desc: not available
Url : http://mail.openjdk.java.net/pipermail/distro-pkg-dev/attachments/20120611/18cc2993/Test9.java