brad@kontron.UUCP (Brad Yearwood) (01/15/87)
Thank you to the several people who pointed out that I must use one or another form of binary option on an open in order to successfully read binary (in my case, Neochrome) files. A case of "We're from your I/O library, and we're here to help you". Just another reminder that when I get home, I am _not_ in a Unix world (I gather that the binary open distinction reflects on distant CP/M ancestry, and that it is also shared by the PC world). I decided to buy a Mark Williams C compiler, and am beginning to work with it some. If anybody is interested in hearing how I fare with it, please let me know. If there is sufficient interest, I might post a more extended review. So far, I appreciate the package of utilities included with MWC, find the compile speed to be fair, and the code to show at least some signs of cleverness. Detractors include not being able to declare immense arrays (supposedly to be fixed soon - a little kludgery with lmalloc() and #define can work around it), and discrepancies in floating point results between MWC and Alcyon (in whose favor, I do not yet know for certain - I suspect that the Alcyon/DRI library wins here though, pending closer analysis). Does anyone have a few floating point accuracy benchmarks they might like to contribute to the comparison task? Brad Yearwood Kontron Electronics {voder, pyramid}!kontron!brad Mountain View, California