Aarni Perko, cont. 1
A more difficult part was to compute volumes and moments for stability calculations when the ship was in an inclined position. There were plenty of special cases and problems with Simpson's rule. These computations took 3 hours. Valter Kostilainen and his group later also programmed computations of tank volumes.I had my first night with Wegematic 1000 1960 in Örebro where I tested the above programs. Carl Björk and Åke Crantz from ABN tutored a stupid green programmer. Conducted by professor Kalervo Laurikainen I made programs for solving differential equations, especially the Schrödinger equation for the deuteron problem. Integration was made by using the Runge-Kutta-Gill method and simple shooting. One shot from the asymptotic solution took about 20 min.
Other programs connected with nuclear physics were tabulating the Clebsch-Gordan coefficients and other minor tasks connected with my licentiate thesis.
Factor analysis was a hot topic in 1960 in mathematical statistics. Initiated by professor Magnus von Wright and advised by FM Kalimo I programmed some routines from this area. The correlation matrix was diagonalized using Jacobi rotations. Matrices were stored in row major order with one block per row, permitting 32 x 32 as the maximum size of a matrix. Diagonalization of such matrices took 4 hours. In Wegematic access to the drum was done using a relay matrix. From ticks of these relays you could follow the progress of the calculations. When the Jacobi iteration changed from row scan to column scan then a different kind of ticking was heard.
In 1961 the Town of Turku did a large traffic survey. Advised by Nyyro Koskela and Pentti Lehvonen from the Street Building Department I did programs for data analysis of the survey. Punched card machines were also used in this task. Two-color graphical representations of traffic intensities were made by using a Flexowriter for offline printing of paper tape output from Wegematic. Input was paper tape prepared with programmable punching machines, which calculated and printed control sums for a block of data. Data from questionnaires had plenty of errors in clock times and dates. Luckily, I had programmed complete validity checks. Erroneous blocks were simply cut out from the tape, and corrected afterwards. The valid blocks could just be joined by cellotape, because our tape reader was so slow and robust that such spliced tape could be successfully read.
To some extent Wegematic was used to solve problems in number theory, and for curve fitting etc. in physics. Programming in hexadecimal code was tricky. One example of this was a course in programming which I gave. All those which did the course acceptably later became professors in mathematics. Wegematic was also in use so short time that programming standards did not emerge.