for review (S): 6912065: final fields in objects need to support inlining optimizations for JSR 292
Vladimir.Kozlov at Sun.COM
Tue Jan 12 19:30:20 PST 2010
Looks good to me.
PS: please, point updated webrev (so I don't need to search previous mails and guess which version).
John Rose wrote:
> I've updated the webrev for this bug to remove the string-valued option, in favor of a boolean switch.
> -- John
> On Jan 8, 2010, at 3:36 PM, John Rose wrote:
>> On Jan 7, 2010, at 10:06 AM, Vladimir Kozlov wrote:
>>> I don't like the idea to have list of classes in the global flag.
>>> I would prefer to have it in systemDictionary like we have well known classes there.
>> I would really like this optimization to apply to all code, with the necessary extra analysis to catch corner cases. Corner cases are found in (a) code which is reached from constructors, (b) reflection, (c) JNI.
>> The motivation for making it a string option is so that dynamic language people like the JRuby folks can turn on the optimization specifically for their code.
>> The current proposed change is therefore a temporary thing for power users, until we get a principled optimization in place.
>> -- John
More information about the hotspot-compiler-dev