Re: Visualising comparative programming language performance

[Dolphin <dhawkins@...>]

Mike Pall <mikelu-0906 <at> mike.de> writes:

> 
> Well, the binarytrees benchmark *is* intended to measure the
> performance of the memory allocator and garbage collector (if any).
> The name is a bit misleading -- it doesn't really measure how fast
> or easy you can construct a binary tree.
> 
> IIRC there was a controversy about some "optimized" JVM settings,
> which effectively turned off the GC for this benchmark. The
> stricter rules were enforced right after that ... 
> 
> --Mike
> 
> 
I also ran it with the allocator we are using in house using the tlsf 
algorithm with the default collector settings and the speed was about equal to
the default allocator with the collectgarbage("setstepmul", 0).  Turning the
collector off completely improved it a bit, but not a whole lot over
setstepmul=0.