[OpenJDK Rasterizer] RFR: Marlin renderer #2

[Jim Graham]

Hi Laurent,

On 6/10/15 2:00 PM, Laurent Bourgès wrote:
>     Because a nanosecond on an operation that takes several milliseconds
>     is worth making the code obscure?  ;)
>
>     Also, factoring out the subtraction has a side affect of requiring
>     you to insert a new "+1" that didn't use to be there.
>
>     I appreciate the attention to detail on some of these calculations,
>     but I think there are a lot of opportunities to simplify the
>     algorithms to make even bigger impacts.  If a calculation was being
>     factored out of a few hundred iterations then it might be worth some
>     level of obscurity in the code, but saving 1 instruction per
>     rendering sequence doesn't seem worth any amount of disruption to
>     the setup code.
>
>
> Fixed but I kept the local variable as it has a measurable impact (760ms
> vs 800ms) with a medium complex map (135 000 shapes).

Please confirm that I am reading this correctly.

You reverted the 2 variable loads at lines 1258,1259 in Renderer.java 
and made no other changes and saw a 5% performance drop on the total 
runtime of a benchmark that renders 135,000 shapes?

			...jim