Guillaume

A quick heads up as I've been working on this quite a bit these days. I think I have found an acceptable solution to the problem which also has the...

A quick info/tracking on this, I had made good progress in that branch : https://github.com/gpnuma/density/tree/adaptative-engine. But a problem showed up a few days ago : during testing the gcc compiled...

Yes I was honestly very surprised that these __builtin_memcpy are not properly optimized in gcc, and it's even worse in icc apparently. I tried to look for the slowdown problem...

The first thing that looks worrying is what happens if `pointer` is not aligned ?, but in the case of the above construct, maybe the compiler is clever and will...

> If there are millions of 8 bytes to copy, the alignment matters, but I don't think there will be performance impact if there are only 8 bytes. Well in...

I've just promoted a build to the dev branch improving the chameleon algorithm in the following ways : - adaptive dictionary sizes - better compression ratio for small datasets (see...

@191919 moved our discussions to new, separate issues as this is one is getting too cluttered.

You want to compile with which make system exactly ? Le 17 janv. 2018 à 19:42 +0100, dumblob , a écrit : > ping > — > You are receiving...

Premake actually creates makefiles that are compatible with gnu make. They should be compatible with posix make as well I suppose.

@191919 how was 1.buf compressed exactly ? Was it with the same density version (the one in the dev branch) ? Do you have the original (ie uncompressed) data so...