![[personal profile]](https://www.dreamwidth.org/img/silk/identity/user.png){kind=small}
The rain stopped, the running water didn't. Pastures still flooded today.
In between work and trying to sleep and eat, I squeezed in more math and computer geeking. Replacing my trapezoidal algorithm for numerical integration with Simpson's rule, I achieved 15 decimal places for PI in 8.5 hours on the Alpha as opposed to 30+ with the trapezoidal. It took me longer to figure out what was wrong with the example program (written in FORTRAN IV) than it did for my test to run once it was debugged. This program is more than 16 lines now, but still less than 40. I tried the same code on the Hercules IBM mainframe emulator, with IBM's FORTRAN H optimizing compiler, but couldn't duplicate the degree of precision possible on the Alpha. Granted the program ran quickly, but on the 32-bit System 370 it only reached 9 decimal places max.
Next? Romberg's method, which is reputed to be even faster. ;p And I shall make a real pie this weekend as well. The rhubarb is going crazy from the rain.
In between work and trying to sleep and eat, I squeezed in more math and computer geeking. Replacing my trapezoidal algorithm for numerical integration with Simpson's rule, I achieved 15 decimal places for PI in 8.5 hours on the Alpha as opposed to 30+ with the trapezoidal. It took me longer to figure out what was wrong with the example program (written in FORTRAN IV) than it did for my test to run once it was debugged. This program is more than 16 lines now, but still less than 40. I tried the same code on the Hercules IBM mainframe emulator, with IBM's FORTRAN H optimizing compiler, but couldn't duplicate the degree of precision possible on the Alpha. Granted the program ran quickly, but on the 32-bit System 370 it only reached 9 decimal places max.
Next? Romberg's method, which is reputed to be even faster. ;p And I shall make a real pie this weekend as well. The rhubarb is going crazy from the rain.
