On 6/07/2018 3:29 PM, Mikael Vidstedt wrote:
Looks good. Nice speedup!
Thanks for looking at it Mikael!
Still the second longest test in com/sun/jdi at 21 minutes!!!
David
Cheers,
Mikael
On Jul 5, 2018, at 10:21 PM, David Holmes <[email protected]> wrote:
Bug: https://bugs.openjdk.java.net/browse/JDK-8205966
webrev: http://cr.openjdk.java.net/~dholmes/8205966/webrev/
One of the @run variants was taking around 15x longer to execute. That variant
uses the InMemoryJavaCompiler which involves a lot of classes and code
execution. The test was enabling method entry event generation for all of main,
resulting in the massive slowdown.
The fix is to add a new breakpoint() function that gets called after the
in-memory compilation setup is done, and we initially run the test to that
point before enabling the events.
The problem @run now only takes 2x the other tests and so should avoid the
timeouts.
Testing: mach5 tier4 solaris-sparc
mach5 tier 1-3
Thanks,
David
