We are running a survey to help us improve the experience for all of our members. If you see the survey appear, please take the time to tell us about your experience if you can.
I need a 100% accurate division of any tow 32 bit numbers. So I am using the following code and the (/) operator. However I notice that results are not accurate when the Den (denominator) is shorter that 32 bit. For example (98765432/01234567) gives 120 where as the correct answer is 80. Can anyone tell me why ? Is something in my code incorrect More important, if this is not a bug, how to fix it ?
Incidentally, I cannot use Printf function so I am using a union function to map the long int variables into 4 unsigned char and the Display the char via a serial port. I don't think that process has introduced any errors in the calculations.
Thanks.
Bhal Tulpule
unsigned long int xdata Num,Den,Res; Num = 98765432; Den = 1234567; Res = Num/Den; //Display Res;
As I see it, the Keil toolset is targeted for Embedded Applications. So a microprocessor such as 8051, or ARM processor would be a natural choice, as opposed to a PC however fast it might be. The code to compute an integer divide is trivial and so the question remains as to why valid C wouldn't serve up valid answers. Sometimes the answers are correct but for others they are wrong.
That's all fantastic, but you'd think you could provide some specific and complete test cases, as part of some experimental method? Providing information about the compiler version, and optimization settings. Enumerate the failing numbers so patterns might be apparent.
The 8051 compiler likely makes far more compromises, and optimization choices you might want to focus on. Appropriate casting on an 8-bit microprocessor, where default int size is 16-bit, being particularly important.