[webkit-dev] [Block Pointer] Deterministic Region Based Memory Manager

[Phil Bouchard]

On 03/06/2016 10:36 PM, Ryosuke Niwa wrote:
>
> No :(  It's pretty relevant that GC in JavaScriptCore was written for
> JavaScriptCore.  In particular, JSC doesn't focus on optimizing binary
> size, memory usage, etc... for embeddable devices so I would expect
> the performance characteristics to be dramatically different from that
> of Duktape or XS6.

True but if block_ptr<> is 10x faster than the Mark & Sweep GC then I 
think we got something perhaps worth investigating.

> Analyzing the performance of a computer software is really hard.  I've
> been working on WebKit for six and half years and I'm still surprised
> by the kind of performance issues we encounter and the difference in
> the performance characteristics between different CPU architectures,
> the number of cores, and the spec of a particular CPU.

I hear you.  I am the one porting WebKit and all the dependencies on 
different CPUs where I work.

But my point is if I can't get something notable then it is what it is.


-Phil