35 lines
1.1 KiB
Markdown
35 lines
1.1 KiB
Markdown
|
# xdu
|
||
|
|
|
||
|
|
Just a tool that implements 'du -s' behavior using [OpenMP](https://www.openmp.org).
|
||
|
|
This conforms to [IEEE Std POSIX 1003.1-2024](https://publications.opengroup.org/standards/unix/c243) rather
|
||
|
|
than any specific OS, or distribution behaviors. This shouldn't impact much as most implementations
|
||
|
|
I know of generally conform to POSIX.
|
||
|
|
|
||
|
|
## Building
|
||
|
|
|
||
|
|
Just a standard meson [meson](https://mesonbuild.com) project.
|
||
|
|
|
||
|
|
``` shell
|
||
|
|
meson setup build
|
||
|
|
meson compile -C build
|
||
|
|
```
|
||
|
|
|
||
|
|
## *Very* anecdotal performance findings
|
||
|
|
|
||
|
|
I just ran this, and diskus in my home directory and measured with time.
|
||
|
|
I did it numerous times. Here are just the last results I got.
|
||
|
|
|
||
|
|
``` shell
|
||
|
|
diskus 2.67s user 85.20s system 683% cpu 12.857 total
|
||
|
|
xdu 23.47s user 46.78s system 878% cpu 7.998 total
|
||
|
|
```
|
||
|
|
|
||
|
|
So around about a 40% improvement give or take on each run. I'm not
|
||
|
|
really looking to run an entire benchmark on du implementations here I
|
||
|
|
just wanted to see how OpenMP would perform, and it performs pretty
|
||
|
|
well.
|
||
|
|
|
||
|
|
There are further improvement possibilities with batching tasks and
|
||
|
|
such, but I'll leave that in case I get some crazy idea into making this
|
||
|
|
a full du implementation.
|