[prev in list] [next in list] [prev in thread] [next in thread] 

List:       openjdk-hotspot-compiler-dev
Subject:    Re: [9] RFR(S): 8136914: compiler/loopopts/superword/SumRedSqrt_Double.java times out
From:       Zoltán Majó <zoltan.majo () oracle ! com>
Date:       2015-09-22 11:34:17
Message-ID: 56013CB9.6010001 () oracle ! com
[Download RAW message or body]

Thank you, Vladimir and Dean (who replied elsewhere), for the review!

Best regards,


Zoltan

On 09/22/2015 01:29 PM, Vladimir Kozlov wrote:
> Looks good.
>
> Thanks,
> Vladimir
>
> On 9/22/15 7:27 PM, Zolt=C3=A1n Maj=C3=B3 wrote:
>> Hi,
>>
>>
>> please review the following patch for JDK-8136914.
>>
>> Bug: https://bugs.openjdk.java.net/browse/JDK-8136914
>>
>> Problem: The test SumRedSqrt_Double.java times out on certain non-x86
>> systems in our nightly testing. The test times out even though it is r=
un
>> with a timeout factor of 40 that allocates 80 minutes to a single run =
of
>> the test. On certain non-x86 machines, the test can take between 58 an=
d
>> 108 minutes (depending on the load on the machine). The test is execut=
ed
>> 8 times with different configurations that results in a cumulative
>> execution time of more than 8 hours on these slower systems.
>>
>>
>> Solution: The test was recently added with 8135028: "support for
>> vectorizing double precision sqrt" and is supposed to test the
>> correctness of C2 using Intel-specific sqrt instructions. The test tak=
es
>> between 4 and 50 seconds on Intel systems (depending on machine
>> configuration).
>>
>> As the test is targeting functionality available only Intel systems wi=
th
>> AVX, I propose to execute the test only on Intel-based systems.
>>
>>
>> Webrev: http://cr.openjdk.java.net/~zmajo/8136914/webrev.00/
>>
>> Testing: JPRT.
>>
>> Thank you and best regards,
>>
>>
>> Zoltan
>>

[prev in list] [next in list] [prev in thread] [next in thread]