[11u] RFR 8240795: [REDO] 8238384 CTW: C2 compilation fails with "assert(store != load->find_exact_control(load->in(0))) failed: dependence cycle found"

[Reingruber, Richard]

Hi Götz,

> I had a look at this change.

Thanks!

> I assume you need it because the follow up calls is_known_instance().

Correct.

> The change looks good.

Thanks.

> "JDK-8238691: C2: turn subtype check into macro node" is not really 
> a candidate for downport. Else I would fear the omitted chunk gets
> forgotten once that change is downported, as this bug will be marked as 
> ported to 11. A solution would be a JBS issue of its own like "Bring parts
> of 8240795 to 11."  But not really necessary here.

I agree. I'll be off-line now. Let's continue this if you don't mind
when I'm back (likely Jan. 4).

Cheers, Richard.

-----Original Message-----
From: Lindenmaier, Goetz <goetz.lindenmaier at sap.com> 
Sent: Dienstag, 22. Dezember 2020 13:26
To: Reingruber, Richard <richard.reingruber at sap.com>; jdk-updates-dev at openjdk.java.net; hotspot-compiler-dev at openjdk.java.net
Subject: RE: [11u] RFR 8240795: [REDO] 8238384 CTW: C2 compilation fails with "assert(store != load->find_exact_control(load->in(0))) failed: dependence cycle found"

Hi Richard, 

I had a look at this change.
I assume you need it because the follow up calls is_known_instance().

The change looks good.

"JDK-8238691: C2: turn subtype check into macro node" is not really 
a candidate for downport. Else I would fear the omitted chunk gets
forgotten once that change is downported, as this bug will be marked as 
ported to 11. A solution would be a JBS issue of its own like "Bring parts
of 8240795 to 11."  But not really necessary here.

Best regards,
  Goetz.



> -----Original Message-----
> From: jdk-updates-dev <jdk-updates-dev-retn at openjdk.java.net> On Behalf
> Of Reingruber, Richard
> Sent: Tuesday, December 22, 2020 12:13 PM
> To: jdk-updates-dev at openjdk.java.net; hotspot-compiler-
> dev at openjdk.java.net
> Subject: [11u] RFR 8240795: [REDO] 8238384 CTW: C2 compilation fails with
> "assert(store != load->find_exact_control(load->in(0))) failed: dependence
> cycle found"
> 
> Hi,
> 
> I would like to downport JDK-8240795 to 11u as prerequisite for the
> downport of
> JDK-8243670.
> 
> I need to be sponsored for this too, please.
> 
> Please review:
> 
> Original bug:
>   https://bugs.openjdk.java.net/browse/JDK-8240795
>   https://hg.openjdk.java.net/jdk/jdk/rev/993974f21271
> 
> 11u webrev:
> 
> http://cr.openjdk.java.net/~rrich/webrevs/8240795_11u_REDO__8238384_
> CTW__C2_compilation_fails_with__assert_dependence_cycle_found/webre
> v.0/
> 
> Modifications:
> 
>   The 2 hunks for loopnode.cpp were removed from the original patch. They
> do not
>   apply because they are based on the enhancement 8238691 which has not
> been
>   downported. These hunks introduced a bug which was fixed with 8252292. I
> don't
>   intend to downport 8252292 but I have verified that the included test case
>   succeeds even without it.
> 
> Testing:
> 
> test/hotspot/jtreg/compiler/escapeAnalysis/TestMissingAntiDependency.jav
> a
>   - from 8252292
>   - succeeds independently of the patch under review
>   - will not be downported
> 
> test/hotspot/jtreg/compiler/escapeAnalysis/TestCopyOfBrokenAntiDepende
> ncy.java
>   - included in the patch under review
>   - fails without patch (assertion "dependence cycle found")
>   - succeeds with patch
> 
> 
> CI testing @SAP: JCK and JTREG, also in Xcomp mode, SPECjvm2008,
> SPECjbb2015, Renaissance Suite,
> SAP specific tests with fastdebug and release builds on all platforms
> 
> Thanks, Richard.