2021.05.03 21:44 "Re: [Tiff] SIMD optimizations", by Larry Bank

Bob,

It was probably me who reported my improved CCITT G4 compressor years ago. I gave up trying to earn money from licensing my code and have been turning it into open source projects for the last few years.

There's nothing specific to the Apple M1 in my improvements to libtiff - they will run on any Arm CPU.

My level of interest in optimizing this code is directly related to the level of interest of the consumers/users. If no one cares about more speed, then I'm not going to bother investing time in it. I'm humble enough to know that other people have the same or better code optimization skills as me and that if I write some SIMD code in libtiff, there are literally thousands of others in the world who could maintain/improve the code if necessary. Luckily the things that I intend to improve are stable functions that have a fixed purpose. As long as they work with the test images, they should continue to work unchanged.

Your point about risks is why I suggested having the SIMD code default to disabled. The people that care about extra speed can make the informed choice of enabling them if they choose.

Jeff,

The improvement for the G4 encoder doesn't involve SIMD, just better memory management. I can do some testing and see how it performs on modern x86/Arm hardware and report back here.

Larry B.