RFR(S): 8191950: assertion failed: no insertions allowed

classic Classic list List threaded Threaded
6 messages Options
Reply | Threaded
Open this post in threaded view
|

RFR(S): 8191950: assertion failed: no insertions allowed

Roland Westrelin-3

http://cr.openjdk.java.net/~roland/8191950/webrev.00/

C2 is trying to incrementally inline a method in a part of the graph
that is dead and for which one the projections of the call to inline is
also an input to the call. We don't have an inexpensive way to check for
unreachable calls when doing incremental inlining so instead I propose
we explicitly check for the corner that fails here.

No test case as triggering incremental inlining reliably is hard.

Roland.
Reply | Threaded
Open this post in threaded view
|

Re: RFR(S): 8191950: assertion failed: no insertions allowed

Tobias Hartmann-2
Hi Roland,

On 01.12.2017 09:24, Roland Westrelin wrote:
> C2 is trying to incrementally inline a method in a part of the graph
> that is dead and for which one the projections of the call to inline is
> also an input to the call. We don't have an inexpensive way to check for
> unreachable calls when doing incremental inlining so instead I propose
> we explicitly check for the corner that fails here.

The "dead loop?" comment is a bit confusing here but I assume the call is part of a loop that became unreachable?

(You can remove the newline in 378)

Was this introduced by a recent change (because the bug still has the integration_blocker label)?

> No test case as triggering incremental inlining reliably is hard.

Please add the 'noreg-hard' label to the bug.

Thanks,
Tobias
Reply | Threaded
Open this post in threaded view
|

Re: RFR(S): 8191950: assertion failed: no insertions allowed

Roland Westrelin-3

Hi Tobias,

Thanks for taking a look at this.

> The "dead loop?" comment is a bit confusing here but I assume the call
> is part of a loop that became unreachable?

Yes.
What about: check for unreachable loop?

> (You can remove the newline in 378)
>
> Was this introduced by a recent change (because the bug still has the integration_blocker label)?

No, not as far I can tell.

Roland.
Reply | Threaded
Open this post in threaded view
|

Re: RFR(S): 8191950: assertion failed: no insertions allowed

Tobias Hartmann-2
Hi Roland,

On 01.12.2017 09:53, Roland Westrelin wrote:
>> The "dead loop?" comment is a bit confusing here but I assume the call
>> is part of a loop that became unreachable?
>
> Yes.
> What about: check for unreachable loop?

Yes, that sounds good to me. I can run some pre-integration testing and sponsor the change if you update the webrev.

>> Was this introduced by a recent change (because the bug still has the integration_blocker label)?
>
> No, not as far I can tell.

Okay, I'll remove the label.

Thanks,
Tobias
Reply | Threaded
Open this post in threaded view
|

Re: RFR(S): 8191950: assertion failed: no insertions allowed

Roland Westrelin-3

> Yes, that sounds good to me. I can run some pre-integration testing
> and sponsor the change if you update the webrev.

Thanks. Here is the updated webrev:

http://cr.openjdk.java.net/~roland/8191950/webrev.01/

Roland.
Reply | Threaded
Open this post in threaded view
|

Re: RFR(S): 8191950: assertion failed: no insertions allowed

Roland Westrelin-3

Thanks for the review and for pushing it.

Roland.