RFR(S): 8205999: C2 compilation fails with "assert(store->find_edge(load) != -1) failed: missing precedence edge"

[Doerr, Martin]

Hi Nils,

thanks for reverting. We had seen missing precedence edge between load
 1395	loadN	===  1359  1098  1380  [[ 1393  2436  1396  1402  1990  1356 ]] narrowoop: narrowoop: java/util/concurrent/ConcurrentHashMap$Node *[int:>=0] * !jvms: StandardGlyphVector$GlyphStrike::getGlyphOutlineBounds @ bci:10 StandardGlyphVector::getGlyphOutlineBounds @ bci:37
and store
 1385	Phi	===  1359  1056  1386  [[ 1384  1414  1589  1986 ]]  #memory  Memory: @BotPTR *+bot, idx=Bot; !jvms: StandardGlyphVector::getDefaultStrike @ bci:43 StandardGlyphVector::getGlyphStrike @ bci:8 StandardGlyphVector::getGlyphOutlineBounds @ bci:10
in our tests.

Looks good to me.

@Tobias: Thanks for pointing me to the bug.

Best regards,
Martin


-----Original Message-----
From: hotspot-compiler-dev [mailto:hotspot-compiler-dev-bounces at openjdk.java.net] On Behalf Of Tobias Hartmann
Sent: Dienstag, 3. Juli 2018 09:50
To: Nils Eliasson <nils.eliasson at oracle.com>; hotspot-compiler-dev at openjdk.java.net compiler <hotspot-compiler-dev at openjdk.java.net>
Subject: Re: RFR(S): 8205999: C2 compilation fails with "assert(store->find_edge(load) != -1) failed: missing precedence edge"

Hi Nils,

this looks good to me. Can you please file a new bug for the original failure?

Thanks,
Tobias

On 03.07.2018 09:35, Nils Eliasson wrote:
> Hi,
> 
> This change backs out 8204157 and 8192992 so we get back to the state pre 8192992. Total diff is
> just one line.
> 
> This gets us back to a state when the problem only reproduces rarely with
> -XX:+AlwaysInlineIncrementally on x64 and is handled gracefully in production builds.
> 
> Bug: https://bugs.openjdk.java.net/browse/JDK-8205999
> 
> Webrev: http://cr.openjdk.java.net/~neliasso/8205999/webrev.01/
> 
> Regards,
> 
> Nils
> 
> 
> 
>