Hello,
Since Tizen is a supported platform, I've been performing some
benchmarks, comparing Crosswalk (XW) with the standard Tizen runtime
(WRT).
Two tests were performed:
1. Create a file, and then immediately remove it.
2. Open a ~831KiB text file, open, read it all, close.
Test (1) tests how many simple operations can be performed. On my
implementation of Tizen's Filesystem API, this consists basically of
three messages: one to resolve the root directory, one to create the
file, and one to delete it.
Test (2) tests how much reading a large-ish file from disk will impact
serialization and transfer in the IPC channels. This sends four
messages: one to resolve the file to be read, one to open the stream,
one to read the file, and one to close the stream.
Both tests were performed using a simple script that performs as much of
these operations in a fixed amount of time as possible, counts how many
of these succeeded or failed, and then calculates the amount of
operations per second. The script is available here[1]. Tests were
performed 5 times, and the results below are the average:
** Reading/writing from flash **
Test Crosswalk Tizen WRT XW/WRT Ratio
1 242ops/s 47ops/s ~2.95x
2 47ops/s 14ops/s ~3.35x
** Reading/writing from tmpfs **
Test Crosswalk Tizen WRT XW/WRT Ratio
1 263ops/s 75ops/s ~3.50x
2 50ops/s 13ops/s ~3.85x
[1] https://gist.github.com/lpereira/28513ca299de3429d898
Crosswalk performs lots of IPC calls and serializes the information
using JSON. I was expecting these things to be serious bottlenecks, but
as it turns out, Crosswalk is around 3x faster than Tizen WRT for this
particular benchmark.
Cheers,
Leandro
_______________________________________________
Crosswalk-dev mailing list
Crosswalk-dev@lists.crosswalk-project.org
https://lists.crosswalk-project.org/mailman/listinfo/crosswalk-dev
