CODEWITHSUNDEEP
cgcc
user555850
user555850

Reputation: 51

GCC gives incorrect answer with optimization enabled

I have the following C program

double d = 1.4;
int x;
x = d * 10;
printf("\n\n VALUE = %d " ,x);

I have gcc 4.3.3 which comes with Ubuntu 9.04

I get answer as 13 with -O0 but get correct answer i.e 14 with higher levels of optimization

Is this a known issue or something wrong with my code?

Upvotes: 5

Views: 400

Answers (4)

F'x
F'x

Reputation: 12298

What Every Computer Scientist Should Know About Floating-Point Arithmetic

It really is what the title says, and it also applies to programmers in general.

Upvotes: 2

ismail
ismail

Reputation: 47572

This is gcc bug #323 , in reality this is not a bug but an implementation detail.

Upvotes: 9

Chris Lutz
Chris Lutz

Reputation: 75389

At higher levels of optimization, GCC probably optimizes both variables away and precomputes the value to print. At no optimization, the value of d (and thus the printing value) will be subject to floating point representation, and may not be exactly 1.4. Try this:

double d = 1.4;
int x;
x = d * 10;
printf("Old = %lf, New = %d\n", d, x);

Upvotes: 4

crazylammer
crazylammer

Reputation: 1172

You can't represent 1.4 exactly using double, the value is actually a bit lagrer or a bit smaller (see this). So there is no "correct" answer - use round() instead of implictly truncating.

Upvotes: 9

Related Questions