Re: [rvm-research] Optimizing modulo operation

[Michael B. <mik...@cs...>]

Hi Anthony,

It might not be a significant performance difference.  What's your 
execution methodology (e.g., adaptive vs. replay; how many trials)?  The 
bloat benchmark in particular can have lots of run-to-run performance 
variation (kind of looks bi-modal?) even with replay methodology.  So you 
might try all the other DaCapo benchmarks, and also try bloat with lots of 
trials and even look at the distribution of run times across trials.

I agree that the two increment-like operations should add similar (and 
low) overheads.

Cheers,
Mike


On Fri, 4 Dec 2009, Anthony Hocquet wrote:

> Hi all,
> 
> I'm doing some tests for the write barrier. I added a simple operation in both fastPath() methods :
>     idx = (idx + 1) % 64;
> 
> The variable idx being a simple int field from the class GenMutator.
> 
> Dacapo bloat bench without this instruction computes in 5600ms. But when I add this instruction, it grows up to 6300ms.
> Though, is there a way to keep the cost of this modulo operation as little as possible ?
> 
> I tried with a simple instruction such as "idx++", and I didn't notice any cost for that.
> 
> Best regards,
> --
> Anthony Hocquet
> 
>