tiny-skia VS json-benchmark

You can even check https://github.com/linebender/piet and https://github.com/RazrFalcon/tiny-skia. They are pure rust libraries. Skia as a whole is a big binary to add.

Looking through the source code. It looks like it's using tiny-skia for rendering.

Tiny Skia (https://github.com/RazrFalcon/tiny-skia) which is a slower Rust port of Skia and does all software rendering.

https://github.com/RazrFalcon/tiny-skia is pretty great.

I understand that this may sound harsh, but I also ported two (far bigger) codebases from C++ to Rust: rustybuzz and tiny-skia. Both of which are production -ready and not just prototypes. And mine not only do not use pointers, but also barely use unsafe in general.

Here is a real world implementation in tiny-skia: https://github.com/RazrFalcon/tiny-skia/blob/master/src/pipeline/highp.rs

How about tiny-skia?

How about tiny-skia? Almost the same performance as C, no unsafe, a lot of explicit SIMD.

Now I would love to see a universal Bezier path crate, storage + common utilities like length, point at offset, some low-level stuff, etc. So it could be used as a foundation for other crates. Currently, there is a lot of friction because each crate has its own implementation.

Nice to see tiny-skia in use.

I don’t really understand what this is trying to prove:

- you don’t seem to specify the size of the input. This is the most important omission

- you are constructing an optimised representation (in this case, strict with fields in the right places) instead of a generic ‘dumb’ representation that is more like a tree of python dicts

- rust is not a ‘moderately fast language’ imo (though this is not a very important point. It’s more about how optimised the parser is, and I suspect that serde_json is written in an optimised way, but I didn’t look very hard).

I found[1], which gives serde_json to a dom 300-400MB/s on a somewhat old laptop cpu. A simpler implementation runs at 100-200, a very optimised implementation gets 400-800. But I don’t think this does that much to confirm what I said in the comment you replied to. The numbers for simd json are a bit lower than I expected (maybe due to the ‘dom’ part). I think my 50MB/a number was probably a bit off but maybe the python implementation converts json to some C object and then converts that C object to python objects. That might half your throughput (my guess is that this is what the ‘strict parse’ case for rustc_serialise is roughly doing).

[1] https://github.com/serde-rs/json-benchmark

But the code OP posted deserializes JSON without knowing anything about the structure, which is known to be slow in serde-json and doesn't appear to be the focus for the library. The json and json-deserializer crates should perform much better in that scenario.

Of course that doesnt mean that in practice the available libraries are as optimized. Did you try actix? It tends to be faster than rocket. Also json-rust and simd-json are usually faster than serde-json, when you don't deserialize a known structure. Here are some benchmarks: https://github.com/serde-rs/json-benchmark

The data for the test is taken from here: https://github.com/serde-rs/json-benchmark/tree/master/data

Here are some benchmarks https://github.com/serde-rs/json-benchmark

I found two projects: * https://github.com/djkoloski/rust_serialization_benchmark - doesn't use Serde zero copy * https://github.com/serde-rs/json-benchmark - has copy vs borrowed, but the results were the same for both, so something's off there

Rust has a good enough standard library (I’d say comparable to C++), that you don’t really need packages for a lot of stuff. Most of my projects have 1 or 2 dependencies. Most of the time I am pulling in a JS parser (serde) and a parallelization library (rayon). These are both high performance libraries that make writing very fast (serde can handle 850 MB/s on a 5 year old laptop cpu per their benchmarks). Rayon is one of the best parallelism libraries I’ve worked with.