RFR: 8152110: Stabilize C2 loop optimizations

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

RFR: 8152110: Stabilize C2 loop optimizations

Jamsheed C m
Hi,

request for review

jbs: https://bugs.openjdk.java.net/browse/JDK-8152110

webrev: http://cr.openjdk.java.net/~jcm/8152110/webrev.00/

the patch make sure that, there is no heavy loop optimizations performed
after opaque nodes removal. Only simple loop optimizations like

converting single iteration loops to normal code/ removing empty loops
are performed after opaque node removal.

this change is done based on discussion here [1]


mach5/perf report attached in jbs

Best regards,
Jamsheed


[1]

http://mail.openjdk.java.net/pipermail/hotspot-compiler-dev/2016-February/021353.html 

http://mail.openjdk.java.net/pipermail/hotspot-compiler-dev/2016-March/021967.html

Reply | Threaded
Open this post in threaded view
|

Re: RFR: 8152110: Stabilize C2 loop optimizations

Vladimir Kozlov
Thank you, Jamsheed

I would like to re-target this to JDK 11. This change may affect code
generation and we should be careful now for JDK 10. This problem is not
new, we can fix it later.

How you determine which loop optimizations should be skipped?

The check for unswitching is better to do in policy_unswitching() method?

Thanks,
Vladimir

On 12/11/17 6:43 PM, jamsheed wrote:

> Hi,
>
> request for review
>
> jbs: https://bugs.openjdk.java.net/browse/JDK-8152110
>
> webrev: http://cr.openjdk.java.net/~jcm/8152110/webrev.00/
>
> the patch make sure that, there is no heavy loop optimizations performed
> after opaque nodes removal. Only simple loop optimizations like
>
> converting single iteration loops to normal code/ removing empty loops
> are performed after opaque node removal.
>
> this change is done based on discussion here [1]
>
>
> mach5/perf report attached in jbs
>
> Best regards,
> Jamsheed
>
>
> [1]
>
> http://mail.openjdk.java.net/pipermail/hotspot-compiler-dev/2016-February/021353.html 
>
> http://mail.openjdk.java.net/pipermail/hotspot-compiler-dev/2016-March/021967.html 
>
>