RFR (S): Combine matching rules for comparison operations on x86

[Viswanathan, Sandhya]

Hi Kishor,

Thanks a lot for looking into this. What is the savings that you get? How many rules reduced and also savings in binary size?

instruct vcmpCC16F, vcmpCC8D, vcmpCC16I, vcmpCC32S
   I think TEMP dst is not required here in effects.
 
Rest looks good. 

Best Regards,
Sandhya

-----Original Message-----
From: panama-dev [mailto:panama-dev-bounces at openjdk.java.net] On Behalf Of Kharbas, Kishor
Sent: Friday, July 26, 2019 2:15 PM
To: panama-dev at openjdk.java.net
Subject: RFR (S): Combine matching rules for comparison operations on x86

Hi,
Requesting review for the patch to reduce the number of rules in x86 ad file for compares.
This is done by taking the comparison operations as input to the vector node.

http://cr.openjdk.java.net/~kkharbas/vector-api/compares_reduction/webrev.00/webrev/

All tests pass on Intel CLX for AVX=0,1,2,3.

Thanks,
Kishor