IBM 610 workstation computer 3390

hanchicken4

Well, technically no computer is a "decimal" machine since all eventually get down to binary to do the actual work. Going upward in architecture, binary is used in different character or word coding arrangements. I'm not sure the IBM 1620 could be categorized as a "decimal machine" since I understand the 1620 didn't do math at all. Rather, it looked up stuff in a table, which was presumably more cost efficient circuit-wise than having math logic circuitry.

Yes, the IBM 1620 is a decimal machine. It is true that the 1620 Mod I did use a lookup table for addition, but the 1620 Mod II didn't. Now, it may be that the mod II used a lookup table, it just wasn't visible. The IBM 1620 didn't use binary numbers in any way, just decimal. Not many people want to admit that there are other types of computes besides binary. Gerard S.

As mentioned elsewhere, the IBM System-360 was a decimal machine and a binary machine (and a dessert topping and floor wax but I digress). By design, the S-360 hardware accomodated a variety of math formats and this was a major innovation for its day. (The S-360 also accomodated large and small sizes, another innovation*). For science and engineering applications, binary words were used as well as floating point words. For business applications, packed-decimal was used. Each method had advantages and disadvantages. In all math calculations, the programmer had to understand how the machine would handle the result, especially if the result-field didn't have enough positions to accomodate the precision required. What would get truncated and what would get rounded and how would this occur? This could happen on either side of the decimal point. Some higher level languages did this automatically, others had options to specify it (COBOL has a "ROUNDED" clause). As mentioned, doing business work in Fortran was easily corrected by a slight formula modification. I forgot the details, but it was something like adding .005 which forced rounding if close up; it may have been more complex than that. Plenty of business applications were written in Fortran if that was what the staff was familiar with, such as in an engineering organization. *Until S-360, there were four types of computers--small business, large business, small science, large science. Each CPU design was optimize for that application. Programming languages and hardware peripherals all custom made for each of the four platforms, which was very expensive. Although S-360 didn't completely eliminate the need for specialized high end and low end machines, it did introduce a great deal of valuable standardization. My impression is that S-360 did extremely well in the business market, but not quite as well as in the science market where the compebreastion did better (e.g. Control Data and later Digital PDP on the low end).