Text encoding speedup

[lucacasonato]

I have sped up the TextEncoder for UTF-8 encoding by utilizing a encoding function from https://github.com/samthor/fast-text-encoding.

The speed improvement is about 4.5x. This is probably because there are way less allocations as the encoded bytes are written right into a Uint8Array. Maybe this can be optimized further, but it is a definite improvement over what we have now.

I did not do this for TextEncoder#encodeInto as I can not guarantee that passed Uint8Array() is long enough for the content. TextEncoder#encodeInto still uses the old encoder.

[ry]

Great! Can you patch deno_website2 so that the benchmark displays and we can observe this improvement?

[lucacasonato]

Great! Can you patch deno_website2 so that the benchmark displays and we can observe this improvement?

Already does. The website picks up new variants for existing benchmarks automatically. As this is just a new variation for the speed benchmarks (Execution time, Thread count, Max Memory Usage, ect.) its on there already.

BTW idk why GitHub is showing that I added the perf benchmark in this PR. I added that in #3589 so the website already has one data point for that benchmarks before this speedup.

[nayeemrmn]

BTW idk why GitHub is showing that I added the perf benchmark in this PR. I added that in #3589 so the website already has one data point for that benchmarks before this speedup.

You've based this on your first PR commit (fbd4bbb) as opposed to the one merged onto master (0a90094). I don't know that it makes a difference, though.

[ry]

LGTM