U16 takes more time than my U32 for sine operation

Question

I was doing a performance monitoring for my sine function and interestingly when I use the variable a in the sine as 16 bit unsigned types it takes more time to calculate than when I use the same variable as a 32 bit unsigned. Could this be because inside the sine algorithm in math.h/.c it takes more time to type caste the 16 bit type ??

This is how my operation looks like, start_performance, for(50 times) { a=sin(a*3.14/180); } stop_performance_and_read values

My performance monitoring unit works fine , its just that I dont understand why the result is odd.

Answer 1

You don't say what processor this is running on. If it's a 32 bit machine, then calculations are most likely done on 32 bit values, and converting to/from 16 bit takes extra cycles.

I once did a benchmark test on a IBM workstation that had double precision floating point hardware. Double precision (64 bit) floating point calculations took longer than 32 bit.