I've just finished version 1.1. It fixes the memory errors during decompression and now loads either the 32 bit or 64 bit version of DivSufSort based on the operating system. The algorithm remains unchanged.
I should also note that the transform that is used before the BWT stage is Eugene Shelwein's BWT_reorder_v2. I apply the transform before the BWT, then use the inverse directly afterwards.
The results of the new version on my 64-bit Windows 7 laptop (4gb RAM, 2.33ghz with 4 cores):
FILE BLOCK SIZE COMPRESSED SIZE COMPRESS TIME DECOMPRESS TIME MEMORY
enwik8 100000000 22,554,133 41.50 39.74 507843kb
enwik9 333333334 185,975,548 451.17 415.08 1647166kb