Googles appar
Huvudmeny

Post a Comment On: cbloom rants

"Oodle Selkie"

3 Comments -

1 – 3 of 3
Blogger Sanmayce said...

Mr. Bloom, you seem to have a blind spot for LzTurbo, why is that?

> I'm not picking on LZ4 here because it's bad; I'm comparing to it because it's the best of the rest, and it's widely known.

Yann's decompressor is fantastic but Hamid's one is definitely better, as for Conor's LZSSE2 - on text it is the best, why do you say LZ4 is the best?

In my textual '88' benchmark LzTurbo 39 is TURBO-AMAZING, why don't you put it side by side with Selkie/Kraken, 19 also:
https://sanmayce.wordpress.com/2016/05/23/the-88-benchmark/

I would like to see your decompressors compared to LzTurbo 39 and LZSSE2 17.

July 26, 2016 at 11:35 AM

Blogger cbloom said...

I don't generally test on text. On the binary game files I test on, LZSSE8 is much better than LZSSE2. I do run LZSSE8 at level 17, and you're right it is generally slightly better than LZ4 (thought that's largely implementation; my LZB16 which is an LZ4-ish is often faster than LZSSE), though the difference is small on PC's.

I compare to LZ4 instead of LZSSE generally because it's widely known and easy to build on lots of platforms. It's a de-facto standard for fast compression.

As for LZTurbo, you can feel free to run tests with that yourself. I don't generally test other people's exes, there are too many funny issues, I like to test code or libs.

July 26, 2016 at 11:43 AM

Blogger Sanmayce said...

>I don't generally test on text.

Okay, textual realm may be not as profitable as game-oriented data however it is more ... cultural. Accessing fast textual data trumps all other scenarios, literature-wise, even human-wise I dare to say.

>As for LZTurbo, you can feel free to run tests with that yourself. I don't generally test other people's exes, there are too many funny issues, I like to test code or libs.

Fair enough, yet, my feedback is to remind you that the process is two-way - the other way around is also valid - I mean, both Kraken and LzTurbo are special entities defying stupid status-quo, however your performer is not presented even in .exe format let alone library or source - nothing wrong, but that makes your best results kinda ifY (if-adjective used in 'Big Bang Theory' comedy). My point, when one speaks of best there is no room for personal preferences, platform mumbo-jumbo-isms, openness or whatever - just PERFORMANCE, best is either best or not.

>...there are too many funny issues...

Maybe, but, to me, TurboBench is a workinghorse, I see no better benchmark than it, I trust its output.

July 27, 2016 at 1:16 PM

You can use some HTML tags, such as <b>, <i>, <a>

This blog does not allow anonymous comments.

Comment moderation has been enabled. All comments must be approved by the blog author.

You will be asked to sign in after submitting your comment.