for review (M): 6912063: inlining parameters need to be adjusted for some uses of the JVM

[Ulf Zibis]

Please note Bug Id: 6908238 on the way.

Thanks,

-Ulf


Am 04.01.2010 10:46, Christian Thalinger schrieb:
> On Sat, 2009-12-19 at 00:46 -0800, John Rose wrote:
>   
>> Currently, the inlining heuristics cannot be tuned in product builds.
>> This makes it impossible to tune them meaningfully, especially for
>> uses of the JVM for languages other than Java, which may have
>> radically different call graphs.
>>
>> We should move the inlining heuristic flags into product mode (as
>> hidden diagnostics, at least).
>>
>> Also, call sites which serves a "glue" for JSR 292 should not be
>> charged to the inlining depth heuristic.
>>
>> http://cr.openjdk.java.net/~jrose/6912063/webrev.00/
>>     
>
>  489     if (new_depth_adjust != 0 && PrintInlining) {
>  490       stringStream nm1; caller_jvms->method()->print_name(&nm1);
>  491       stringStream nm2; callee_method->print_name(&nm1);
>                                                         ^^^ should be nm2
>  492       tty->print_cr("discounting inlining depth from %s to %s", nm1.base(), nm2.base());
>
> Otherwise looks good.
>
> -- Christian
>
>
>