RFR(S) 8193009: compiler/c2/Test7029152.java crashes with SIGILL in java.lang.StringLatin1.indexOf with -XX:+UseJVMCICompiler

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

RFR(S) 8193009: compiler/c2/Test7029152.java crashes with SIGILL in java.lang.StringLatin1.indexOf with -XX:+UseJVMCICompiler

dean.long
https://bugs.openjdk.java.net/browse/JDK-8193009
http://cr.openjdk.java.net/~dlong/8193009/webrev/

I tested all 16 base register combinations to make sure we handle
anything Graal might generate.

dl
Reply | Threaded
Open this post in threaded view
|

Re: RFR(S) 8193009: compiler/c2/Test7029152.java crashes with SIGILL in java.lang.StringLatin1.indexOf with -XX:+UseJVMCICompiler

Igor Veresov
Looks good.

igor

> On Dec 5, 2017, at 2:12 PM, [hidden email] wrote:
>
> https://bugs.openjdk.java.net/browse/JDK-8193009
> http://cr.openjdk.java.net/~dlong/8193009/webrev/
>
> I tested all 16 base register combinations to make sure we handle anything Graal might generate.
>
> dl

Reply | Threaded
Open this post in threaded view
|

Re: RFR(S) 8193009: compiler/c2/Test7029152.java crashes with SIGILL in java.lang.StringLatin1.indexOf with -XX:+UseJVMCICompiler

dean.long
Thanks Igor.

dl


On 12/5/17 2:17 PM, Igor Veresov wrote:

> Looks good.
>
> igor
>
>> On Dec 5, 2017, at 2:12 PM, [hidden email] wrote:
>>
>> https://bugs.openjdk.java.net/browse/JDK-8193009
>> http://cr.openjdk.java.net/~dlong/8193009/webrev/
>>
>> I tested all 16 base register combinations to make sure we handle anything Graal might generate.
>>
>> dl

Reply | Threaded
Open this post in threaded view
|

Re: RFR(S) 8193009: compiler/c2/Test7029152.java crashes with SIGILL in java.lang.StringLatin1.indexOf with -XX:+UseJVMCICompiler

Vladimir Kozlov
In reply to this post by dean.long
Should we put new check code under #ifdef INCLUDE_JVMCI and if (UseJVMCICompiler)? Stubs code
generation happens after flags parsing.

Thanks,
Vladimir

On 12/5/17 2:12 PM, [hidden email] wrote:
> https://bugs.openjdk.java.net/browse/JDK-8193009
> http://cr.openjdk.java.net/~dlong/8193009/webrev/
>
> I tested all 16 base register combinations to make sure we handle anything Graal might generate.
>
> dl
Reply | Threaded
Open this post in threaded view
|

Re: RFR(S) 8193009: compiler/c2/Test7029152.java crashes with SIGILL in java.lang.StringLatin1.indexOf with -XX:+UseJVMCICompiler

dean.long
On 12/6/17 5:39 PM, Vladimir Kozlov wrote:

> Should we put new check code under #ifdef INCLUDE_JVMCI and if
> (UseJVMCICompiler)? Stubs code generation happens after flags parsing.
>

We could, but I don't think it's worth it.  I would rather allow C1 and
C2 to emit the same range of encodings as Graal.  I think it would help
avoid spills.

dl

> Thanks,
> Vladimir
>
> On 12/5/17 2:12 PM, [hidden email] wrote:
>> https://bugs.openjdk.java.net/browse/JDK-8193009
>> http://cr.openjdk.java.net/~dlong/8193009/webrev/
>>
>> I tested all 16 base register combinations to make sure we handle
>> anything Graal might generate.
>>
>> dl

Reply | Threaded
Open this post in threaded view
|

Re: RFR(S) 8193009: compiler/c2/Test7029152.java crashes with SIGILL in java.lang.StringLatin1.indexOf with -XX:+UseJVMCICompiler

Vladimir Kozlov
On 12/6/17 5:56 PM, [hidden email] wrote:
> On 12/6/17 5:39 PM, Vladimir Kozlov wrote:
>
>> Should we put new check code under #ifdef INCLUDE_JVMCI and if (UseJVMCICompiler)? Stubs code
>> generation happens after flags parsing.
>>
>
> We could, but I don't think it's worth it.  I would rather allow C1 and C2 to emit the same range of
> encodings as Graal.  I think it would help avoid spills.

Okay, it make sense. Please, file follow up RFE for C1 and C2.

Thanks,
Vladimir

>
> dl
>
>> Thanks,
>> Vladimir
>>
>> On 12/5/17 2:12 PM, [hidden email] wrote:
>>> https://bugs.openjdk.java.net/browse/JDK-8193009
>>> http://cr.openjdk.java.net/~dlong/8193009/webrev/
>>>
>>> I tested all 16 base register combinations to make sure we handle anything Graal might generate.
>>>
>>> dl
>
Reply | Threaded
Open this post in threaded view
|

Re: RFR(S) 8193009: compiler/c2/Test7029152.java crashes with SIGILL in java.lang.StringLatin1.indexOf with -XX:+UseJVMCICompiler

dean.long
On 12/6/17 6:19 PM, Vladimir Kozlov wrote:

> On 12/6/17 5:56 PM, [hidden email] wrote:
>> On 12/6/17 5:39 PM, Vladimir Kozlov wrote:
>>
>>> Should we put new check code under #ifdef INCLUDE_JVMCI and if
>>> (UseJVMCICompiler)? Stubs code generation happens after flags parsing.
>>>
>>
>> We could, but I don't think it's worth it.  I would rather allow C1
>> and C2 to emit the same range of encodings as Graal.  I think it
>> would help avoid spills.
>
> Okay, it make sense. Please, file follow up RFE for C1 and C2.
>

Thanks Vladimir.

dl

> Thanks,
> Vladimir
>
>>
>> dl
>>
>>> Thanks,
>>> Vladimir
>>>
>>> On 12/5/17 2:12 PM, [hidden email] wrote:
>>>> https://bugs.openjdk.java.net/browse/JDK-8193009
>>>> http://cr.openjdk.java.net/~dlong/8193009/webrev/
>>>>
>>>> I tested all 16 base register combinations to make sure we handle
>>>> anything Graal might generate.
>>>>
>>>> dl
>>