Fabrizio Diversi makes some magical things to make me read } For the rest I do not think is a problen of partition dimension having formatted succesfully a 512Mb and more partition using Mkpart (1.02).
Me too. Partitionning at 512Mb was ok... but using REALLY the file system was not, at least not always in my case. It happened on my Q40, I had put on that big partition the source of SMSQ/E (was 2.98/2.99y++ at that time). Accessing the file was possible, but strangely with the cache enable in either copyback or writethrough, sometime (not always) the full compilation of the damn thing failed on some random files, with silly error reported (just like the accessed file was not the one intended but another). Re-installing the source tree on a 128Mb partition did solve the trouble, whereas reformating a big partition did not. It was also a time where SMSQ/E was using the full 32MB of memory to cache the disk. It is now restricted to less than 1MB, and the only thing still slow is the access to MS-DOS floppy (the longer the file, the slower it get). (1) My guess so far is that the caching system is somehow limited (I do not know the exact math) to identify the cached sectors, and that with a big partition accessed quickly, there is some sectors which share the same cache entry, hence the confusion which occurs when I compiled: the new file was wrongly identified as already in the cache whereas it was another file previously loaded. (1): It would really boost the performance if someone provide a thing similar to 'ram_mdv' for MS-DOS floppy: reading all the sectors in ram, at the real speed of drive, then accessing the image of it. I really think the write speed so far is ok, but the read speed of a big zip is horrible and deserves some speed-up.
