ugrep v6.0

[genivia-inc]

What's new?

• quicker! 🚀 much faster search speeds on ARM CPUs and also faster overall on other CPUs using essentially the same SIMD algorithms but better optimized [FR] 90% speed up by refactoring and optimizing some code #385 see latest ugrep benchmarks for results
• more good stuff! 🎉 includes ugrep-indexer, which is installed together with ugrep, see ugrep-indexer release notes v1.0.0

Please note

• Distribution maintainers please check if ugrep-indexer and its manual page is indeed also included in your ugrep distribution package and will be installed on user systems. Thank you! 🙏❤️

---

This discussion was created from the release ugrep v6.0.

[genivia-inc]

Happy that the benchmark results were already looking good in the past and are now even better. Ugrep is certainly quicker than before.

As a side note, I have to emphasize speed as being important for this project because I get unfairly attacked by some people including the author of ripgrep when I post about ugrep updates, with them saying that ugrep is not faster than ripgrep and then they show some anecdotal test case as "proof". He's been trying to keep up with me ever since ugrep was first released. He then claims that he "was never able to replicate ugrep performance" when in fact he had admitted to me that ugrep was faster for a benchmark set of cases that I showed him when I initially released ugrep. He then released ripgrep v13 months later to speed up rg to show that "he was right" and blames me for being dishonest and a bad guy. Good for him to improve ripgrep, but thanks to me I suppose...?

I didn't write ugrep for a speed competition at all, but rather for new features while making sure ugrep is very fast.

So right now ugrep is quicker in many cases than rg v14 which appears to run very slow on some tests, such as when searching for words, taking over 6 minutes (375 seconds) when ugrep takes half a second on x64 for this test:

grepping -fwords/4.txt elapsed real time (s)

search		-n	-no	-wn	-win	-wino	-c	-wic	-l	-wil
ugrep	0.23	0.23	0.23	0.24	0.45	0.46	0.22	0.42	0.01	0.02
rg	0.28	0.29	0.29	4.07	374.86	380.29	0.27	376.54	0.01	0.11

and rg 14 is also very slow on arm64 for this test:

grepping -fwords/4.txt elapsed real time (s)

search		-n	-no	-wn	-win	-wino	-c	-wic	-l	-wil
ugrep	0.18	0.18	0.18	0.18	0.32	0.33	0.17	0.31	0.00	0.01
rg	0.22	0.22	0.23	2.43	176.62	179.72	0.21	174.58	0.00	0.04

Recursive searching on MacOS M1 in the Swift source code repo is also showing a consistent speed advantage for ugrep over rg:

grepping String|Int|Double|Array|Dictionary elapsed real time (s)

search	-n	-nr	-wn	-wnr	-win	-winr	-wino	-winor	-wic	-wicr	-wil	-wilr
ugrep	0.15		0.14		0.17		0.16		0.17		0.12
rg	0.32		0.31		0.32		0.35		0.31		0.33

Let me add that there is no grep that is always consistently faster than another grep. There are cases when ugrep is slower than rg just as there are cases the other way round. I never claimed that ugrep is always faster than other grep. But I have implemented my own ideas for new methods and algos to make it pretty fast.