RFR(M): 8240248 : Extend superword reduction optimizations for x86

Viswanathan, Sandhya sandhya.viswanathan at intel.com
Mon Mar 23 17:24:15 UTC 2020


Hi Yang,

I will remove the limitation.

Thanks for the feedback.

Best Regards,
Sandhya

-----Original Message-----
From: Yang Zhang <Yang.Zhang at arm.com> 
Sent: Monday, March 23, 2020 1:45 AM
To: Viswanathan, Sandhya <sandhya.viswanathan at intel.com>; Tobias Hartmann <tobias.hartmann at oracle.com>; hotspot-compiler-dev at openjdk.java.net
Subject: RE: RFR(M): 8240248 : Extend superword reduction optimizations for x86

Hi Sandhya

test/hotspot/jtreg/compiler/loopopts/superword/RedTest_int.java
28  * @requires os.arch=="x86" | os.arch=="i386" | os.arch=="amd64" | os.arch=="x86_64" | os.arch=="aarch64"

Do you think this limitation is needed? For those platforms without such backend support, scalar implementations can be used.

Regards
Yang



-----Original Message-----
From: hotspot-compiler-dev <hotspot-compiler-dev-bounces at openjdk.java.net> On Behalf Of Viswanathan, Sandhya
Sent: Saturday, March 21, 2020 11:14 AM
To: Tobias Hartmann <tobias.hartmann at oracle.com>; hotspot-compiler-dev at openjdk.java.net
Subject: RE: RFR(M): 8240248 : Extend superword reduction optimizations for x86

Hi Tobias,

Thanks a lot for the review. I have made the style related changes you suggested.

I also noticed that there was an issue with 32-bit build.
I needed to place reduceL, reduce2L, reduce4L and reduce8L under #ifdef _LP64.
I have made that change as well in the following webrev:
http://cr.openjdk.java.net/~sviswanathan/8240248/webrev.02/

Best Regards,
Sandhya

-----Original Message-----
From: Tobias Hartmann <tobias.hartmann at oracle.com>
Sent: Friday, March 20, 2020 2:46 AM
To: Viswanathan, Sandhya <sandhya.viswanathan at intel.com>; hotspot-compiler-dev at openjdk.java.net
Subject: Re: RFR(M): 8240248 : Extend superword reduction optimizations for x86

Hi Sandhya,

Looks good to me.

Noticed some minor style issues:
- vectornode.hpp:642 Comment says "and"
- RedTest_int.java:231 Indentation is wrong
- RedTest_long.java:231 Indentation is wrong

I've submitted some extended testing and will let you know once it passed.

Best regards,
Tobias

On 16.03.20 20:46, Viswanathan, Sandhya wrote:
> Hi All,
>
> Please find below the updated webrev which includes the jtreg, jmh test cases and suggested changes from Vladimir Ivanov:
> http://cr.openjdk.java.net/~sviswanathan/8240248/webrev.01/
>
> Could we please get one more review for this?
>
> Best Regards,
> Sandhya
>
> -----Original Message-----
> From: hotspot-compiler-dev <hotspot-compiler-dev-bounces at openjdk.java.net> On Behalf Of Viswanathan, Sandhya
> Sent: Tuesday, March 03, 2020 9:33 AM
> To: Eric Caspole <eric.caspole at oracle.com>; hotspot-compiler-dev at openjdk.java.net
> Subject: RE: RFR(M): 8240248 : Extend superword reduction optimizations for x86
>
> Hi Eric,
>
> Thanks, we will add the test/micro and send the updated webrev for review.
>
> Best Regards,
> Sandhya
>
> -----Original Message-----
> From: hotspot-compiler-dev <hotspot-compiler-dev-bounces at openjdk.java.net> On Behalf Of Eric Caspole
> Sent: Monday, March 02, 2020 7:04 AM
> To: hotspot-compiler-dev at openjdk.java.net
> Subject: Re: RFR(M): 8240248 : Extend superword reduction optimizations for x86
>
> Hi Sandhya and Shravya,
> Are there already any micros in the test/micro/ that show the benefit of this optimization? If not could you add one to the webrev?
> Thanks,
> Eric
>
>
> On 2/28/20 19:51, Viswanathan, Sandhya wrote:
>> Shravya and I would like to submit a webrev which extends super-word reduction optimizations to support and, or, xor reduction for x86.
>>
>> JBS: https://bugs.openjdk.java.net/browse/JDK-8240248
>> Webrev: http://cr.openjdk.java.net/~sviswanathan/8240248/webrev.00/
>>
>> Patch is extracted from vectorIntrinsics branch of Panama with additional changes to support auto-vectorization.
>>
>> Please review,
>>
>> Best Regards,
>> Sandhya
>>
IMPORTANT NOTICE: The contents of this email and any attachments are confidential and may also be privileged. If you are not the intended recipient, please notify the sender immediately and do not disclose the contents to any other person, use it for any purpose, or store or copy the information in any medium. Thank you.


More information about the hotspot-compiler-dev mailing list