RFR: 8215473: [lworld] Support for substitutability checks in C2

[Tobias Hartmann]

Hi Roland,

Looking at the changes in sharedRuntime.cpp, wouldn't you also need those in
SharedRuntime::resolve_opt_virtual_call_C()?

I would prefer moving this code into SharedRuntime::resolve_sub_helper() and guard it by the
is_virtual/is_optimized arguments. There you already lookup caller_frame/caller_cb/...

Otherwise it looks good to me. Some comments in Parse::do_acmp would be nice but we can refactor
that later (this acmp stuff will probably evolve further anyway).

Thanks,
Tobias

On 24.01.19 15:53, Roland Westrelin wrote:
> 
> http://cr.openjdk.java.net/~roland/8215473/webrev.00/
> 
> This gives the option of turning acmp into a substitutability test in c2
> code. The actual test on values is performed by always calling
> ValueBootstrapMethods.isSubstitutable().
> 
> Given this change doesn't include interpreter support and the code path
> for value comparison is not optimized, the main purpose of this change,
> at this point, is to enable performance evaluation for workloads that
> make no use of values.
> 
> Roland.
>