[Solved] Different results on using sqrt() and pow() functions

Question

First of all, sqrt(x) should be faster and more accurate than pow(x,0.5), why do you think it’s in the library? Second, you’re probably getting a wrong answer because your loop termination condition is testing a floating-point number. A tiny round-off somewhere in one of those 2 million loops is probably enough to throw off the total. Finally, you’re calculating a floating point sqrt() twice inside each iteration of the loop, which is terribly slow.

Instead of

for (j = 0; j <= sqrt(i); ++j)  . . .

try

limit = (int)(sqrt(i) + 0.5);
for (j = 0; j <= limit; ++j) . . .

Accepted Answer

First of all, sqrt(x) should be faster and more accurate than pow(x,0.5), why do you think it’s in the library? Second, you’re probably getting a wrong answer because your loop termination condition is testing a floating-point number. A tiny round-off somewhere in one of those 2 million loops is probably enough to throw off the total. Finally, you’re calculating a floating point sqrt() twice inside each iteration of the loop, which is terribly slow.

Instead of

for (j = 0; j <= sqrt(i); ++j)  . . .

try

limit = (int)(sqrt(i) + 0.5);
for (j = 0; j <= limit; ++j) . . .