2021.05.05 01:14 "Re: [Tiff] SIMD optimizations", by Steve Underwood

The 2D encoder in the fax engine in my spandsp library applies CLZ, and other ways of finding the top bit quickly, to 32 bit words to reformulate the rows into a series of run lengths, and then compresses the vectors of run lengths. I originally wrote that when 64 bit machines were less common. It might make sense to scan 64 bit words with CLZ these days.

Regards,
Steve