RFR (XS) JDK-6961433: Revisit need to disable Windows C++ compiler optimisation of sharedRuntimeTrig.cpp

Daniel D. Daugherty daniel.daugherty at oracle.com
Thu May 29 22:13:54 UTC 2014


On 5/29/14 12:25 PM, Lois Foltan wrote:
> Hello,
>
> Please review the following fix:
>
> Webrev:
>     http://cr.openjdk.java.net/~lfoltan/bug_jdk6961433/

src/share/vm/runtime/sharedRuntimeTrig.cpp
     Cleanly removed. The udiffs link is the way to review this!

Thumbs up.

Dan


>
> Bug: Revisit need to disable Windows C++ compiler optimisation of 
> sharedRuntimeTrig.cpp
>     https://bugs.openjdk.java.net/browse/JDK-6961433
>
> Summary of fix:
> Remove WIN32 specific pragma optimize "off" within 
> sharedRuntimeTrig.cpp which resulted in the file being compiled with 
> no optimizations.  The SAFEBUF macro definition is also being 
> removed.  It was a workaround caused by the pragma being in effect. 
> The problem report indicates that this pragma was added in the 
> VS2003/VS2005 time frame.  Where or how the C++ compiler optimization 
> manifested itself could not be located.  Thank you to Dan Daugherty 
> for completing a historical search to try to track this down.  His 
> search is posted in the problem report.  Also, thank you to Vladimir 
> Koslov for proposing ideas on how this should be tested.  Since VS2010 
> & higher should not have this optimization issue, the pragma is being 
> removed early in JDK 9 so it can benefit from continual full testing.
>
> Tests:
> JPRT build & test
> vm.quick.testlist via Adhoc Aurora testing on Windows 32 & 64 bit - 
> runs with/without -Xint
> JDK java/lang & java/util - runs with/without -Xint
> Hotspot JTREG - runs with/without -Xint
> Built full JDK 9 fastdebug and production images with this Hotspot 
> change, resulting image used for testing.



More information about the hotspot-runtime-dev mailing list