cu26mawy at rbg.informatik.tu-darmstadt.de
Wed Jul 22 12:48:08 UTC 2015
an important disclaimer is that the scope of our evaluation is still
rather limited which is also why we did not include any performance
numbers to the announcement of Reactive Ruby. However, we are happy to
share our current results.
The evaluation is based on benchmarks which measure the time to
propagate changes through a dependency graph. The result is compared to
an Observer design pattern implementation that mimics the same
We consider graphs shaped like a chain, fan and reverse fan. I attached
pics that visualize the graph structure of these benchmarks. In all
benchmarks the nodes perform simple calculations, like adding a constant
value to the value of their predecessor node or adding the values of all
predecessors together. In the chain and fan benchmarks 200k change
events are propagated and in the reverse fan benchmarks 1.6kk change
events are propagated.
The measured time in these benchmarks is the following:
Reverse fan A
Reverse fan B
We think that these initial results are promising as they show that --
at least in these benchmarks -- Reactive Ruby can reach around the same
performance than the hand-made observer pattern. We also evaluated the
similar benchmark setup. In that evaluation Rx.JS was on average around
a factor 10 slower than the handmade Observer pattern, Rx.JS was the
fasted evaluated RP languages in JS.
Am 21.07.2015 um 12:17 schrieb Stefan Marr:
> Hi Malte:
>> On 20 Jul 2015, at 22:40, Malte Viering <cu26mawy at rbg.informatik.tu-darmstadt.de> wrote:
>> Despite its increasing popularity Reactive Programming still suffers
>> from a significant performance overhead. In my thesis, I looked into
>> the option of using Truffle to provide an efficient runtime system for
>> Reactive Programming.
> Very interesting project, thanks for sharing!
> Do you perhaps have some numbers to share with respect to performance?
More information about the graal-dev