jmn (07/09/82)
To provide a reference for uiucdes!richards first crude timings of the 432, I ran similar runs on a pdp11/70. Programs were main routines written in C, and timings obtained from UNIX (tm) per process accounting. Statement (repeated 1,000,000 times) CPU time Statement Time null 7.0 sec --------- z = x + y (All 32 bit integers) 15.1 sec 8.1 us z = x * y " 58.2 51.2 z = x / y " 1:56 109. z = x + y (All 32 bit floating point) 19.9 sec 12.9 us z = x * y " 21.6 14.6 z = x / y " 26.7 19.7 z = x + y (All 16 bit integers) 11.1 sec 4.1 us z = x * y " 15.1 8.1 z = x / y " 19.4 12.4 Sorry that soo much data is provided, but simply comparing a 32 bit machine performing 32 bit arithmetic with a 16 bit machine performing 32 bit arithmetic can be misleading. An important difference between richards data and the above is the time required to execute the null statement; i.e., the loop code. That is a measure of the ADA compiler as well as a measure of the machine.