RFR(S): 8178811: Minimize the AVX <-> SSE transition penalty on x86

[Tobias Hartmann]

Hi,

this issue slipped through our triaging process because the subcomponent was not set to "compiler". I converted it to an enhancement and targeted it to JDK 10.

As Vladimir said, we can later backport it to a JDK 9 update release.

Thanks,
Tobias

On 17.04.2017 06:30, Vladimir Kozlov wrote:
> What is important is 'Fix version'. 'Affected version' could be all current and previous releases.
> 
> From the beginning I thought it is targeting JDK 10. To avoid confusion I set 'Fix version' to 10.
> 
> It may go into 9 Update as other small enchantments.
> 
> Regards,
> Vladimir
> 
> On 4/16/17 3:03 AM, Andrew Haley wrote:
>> Hi,
>>
>> On 14/04/17 19:17, Deshpande, Vivek R wrote:
>>
>>> This fix minimizes the AVX to SSE and SSE to AVX transition penalty
>>> through generation of vzeroupper instruction.
>>
>> This looks like a performance enhancement, rather than a bug.  Is JDK
>> 9 still open for performance enhancement?  I thought we were in
>> rampdown, but the report is targeted to 9.
>>
>> Andrew.
>>