[9] RFR (S) 8178723: Workaround for failure of CRC32C intrinsic on x86 machines without CLMUL support (JDK-8178720)

[Zoltán Majó]

Hi Vladimir,


On 04/13/2017 07:28 PM, Vladimir Kozlov wrote:
> Hi Zoltan,
>
> The fix looks good.

thank you!

> Please, add label and fix request comment according to RDP2 rules for 
> JDK 9.

I did so. Once the fix is approved, I'll push it.

Best regards,


Zoltan

>
> Thanks,
> Vladimir
>
> On 4/13/17 7:54 AM, Zoltán Majó wrote:
>> Hi,
>>
>>
>> please review the following fix for 8178723. (8178723 is a workaround 
>> for 8178720.)
>> https://bugs.openjdk.java.net/browse/JDK-8178723
>> http://cr.openjdk.java.net/~zmajo/8178723/webrev.00/
>>
>> The problem is that the CRC32C intrinsic is apparently incorrect on 
>> x86 machines without CLMUL support. The tests
>> provided by Lutz Schmidt (also part of this change set) have 
>> uncovered that problem.
>>
>> Given the state of JDK 9, I think it's best to split the problem into 
>> two parts:
>> - (1) this workaround (8178723), which disables the CRC32C intrinsics 
>> on x86 machines without CLMUL support; the
>> workaround also the tests provided by Lutz (with proper credit in the 
>> final change set);
>> - (2) the "proper" fix for the problem is handled in 8178720; I asked 
>> Michael Berg to look into that problem.
>>
>> If (and once) the workaround is fine with you, I'll take care of the 
>> fix-request process.
>>
>> JPRT testing is running at the moment.
>>
>> Thank you!
>>
>> Best regards,
>>
>>
>> Zoltan
>>