RFR: 8189103 - AARCH64: optimize String indexOf intrinsic

Dmitrij Pochepko dmitrij.pochepko at bell-sw.com
Fri Apr 20 19:37:28 UTC 2018


I've updated comment and uploaded new webrev(it's the same, except BM 
algorithm C-code desciption).




On 20.04.2018 13:30, Dmitrij Pochepko wrote:
> On 19.04.2018 20:08, Andrew Haley wrote:
>> On 04/18/2018 05:09 PM, Dmitrij Pochepko wrote:
>>> 2) large update of Boyer Moore algorithm implementation by increasing
>>> search table size from 128 to 256 to remove few branches for Latin1
>>> encoding cases and also improve performance for cases when Latin1
>>> symbols with values >128 are met. This patch also significantly 
>>> improves
>>> search of Latin1 string inside UTF string by upgrading algorithm
>>> logic(the idea is to skip <pattern length> symbols in case pure UTF
>>> symbol is met in source string).
>> I don't see a change to the comment which describes the algorithm.
>> Does that comment need to be changed?
> Do you mean C pseudo-code which describes algorithm in general? Well, 
> algorithm main idea wasn't changed, so, I left pseudo-code untouched 
> to keep it more readable(my patch adds few "if" operators depending on 
> LL/UU/UL cases). I can update this comment to reflect exact changes it 
> if you think it'll make things clearer.
> Thanks,
> Dmitrij

More information about the hotspot-compiler-dev mailing list