RFR(XS) 8227632: Incorrect PrintCompilation message: made not compilable on levels 0 1 2 3 4
dean.long at oracle.com
dean.long at oracle.com
Mon Jul 15 20:27:24 UTC 2019
OK, and webrev.01 looks good.
dl
On 7/15/19 9:25 AM, Vladimir Kozlov wrote:
> I think we should use level number as Boris did because in a future it
> could be Graal JIT with different levels.
>
> Boris, webrev.01 looks good to me.
>
> Thanks,
> Vladimir
>
> On 7/15/19 5:42 AM, Boris Ulasevich wrote:
>> Hi Dean,
>>
>> Yes, in practice c1/c2/all can be managed, but not all levels.
>>
>> It is not good by the way because with disabled C2 for certain module
>> I get compilation to C1_full (what for?) followed by recompilation to
>> C1_simple. I think it would be good either to enable
>> CompilerDirectives to manage all compiler levels or to teach
>> CompilerBroker to skip C1_full when C2 is disabled for the method.
>>
>> And I think it is better to have a single print in one place.
>>
>> regards,
>> Boris
>>
>> On 12.07.2019 21:28, dean.long at oracle.com wrote:
>>> In practice, aren't the only "not compilable" settings "all", "c1",
>>> and "c2"? If so, and there aren't any tests that check for these
>>> diagnostic messages, it might be less confusing to only print "c1"
>>> or "c2" and move the print into the
>>> set_not_c1_compilable()/set_not_c2_compilable()/set_not_c1_osr_compilable()/set_not_c2_osr_compilable()
>>> functions.
>>>
>>> dl
>>>
>>> On 7/12/19 10:02 AM, Vladimir Kozlov wrote:
>>>> Hi, Boris
>>>>
>>>> I agree that printing all levels values when particular one is
>>>> passed is wrong.
>>>>
>>>> I think you should simply print comp_level value because it should
>>>> be specific in all cases (leaving CompLevel_all check as it is and
>>>> may be adding missing AOT check).
>>>>
>>>> thanks,
>>>> Vladimir
>>>>
>>>> On 7/12/19 5:33 AM, Boris Ulasevich wrote:
>>>>> Hi,
>>>>>
>>>>> Can I please have a review for the simple logging issue. Current
>>>>> PrintCompilation message looks confusing, it reports "levels 0 1 2
>>>>> 3 4" when level 4 only was disabled.
>>>>> Example:
>>>>> $ echo [{ match: [\"*.*\"], c2: {Exclude:true} }] > cdf.txt
>>>>> $ java -XX:+PrintCompilation -XX:+UnlockDiagnosticVMOptions
>>>>> -XX:CompilerDirectivesFile=cdf.txt HelloWorld
>>>>> ### Excluding compile: java.io.OutputStreamWriter::write
>>>>> made not compilable on levels 0 1 2 3 4
>>>>> java.io.OutputStreamWriter::write (11 bytes) excluded by
>>>>> CompileCommand
>>>>> ### Excluding compile: java.lang.String::equals
>>>>> made not compilable on level 0 1 2 3 4 java.lang.String::equals
>>>>> (50 bytes) excluded by CompileCommand
>>>>>
>>>>> http://bugs.openjdk.java.net/browse/JDK-8227632
>>>>> http://cr.openjdk.java.net/~avoitylov/8227632/webrev.00
>>>>>
>>>>> thanks,
>>>>> Boris
>>>
More information about the hotspot-compiler-dev
mailing list