C# Performance penalty for Int32 literals to floats

Question

In c#, you can define a number literal as either int or double

    Double var1 = 56.1;
    int var2 = 51;

These are the default types the literals are assigned. However, the game engine I'm working on uses floats for position, rotation, etc. When the float is assigned a literal double, ie, float varFloat = 75.4; the compiler throws an error saying that the double literal needs to be a float, which is correct. So one needs to turn the double literal into a float ie. float varFloat = 75.4f;. However, when given an int literal, the int is implicitly converted to a float. Ie,

    float varFloat = 44; // This is fine.

My question is is the compiler smart enough to realize that 44 should be a float literal? If not, that means that every time the literal is accessed, it's also performing a conversion. In most cases, this really doesn't matter. But with high performance code, it could potentially become an issue (even if it's a minor one) if ints are used all over the place instead of floats. There is no way as far as I know to change these literals into floats without going through lines of source code which really isn't time well spent at all.

So, does the compiler convert the int literal into a float literal? If not, what can be done about this waste of processing power other then trying to avoid it?

"Waste in processing power" is a pretty bold statement regarding a language construct like literals. It would be worth your time to read through the C# [Language Specification](http://msdn.microsoft.com/en-us/library/vstudio/ms228593.aspx) to get a feel for how C# actually works. — user7116, Apr 23 '13 at 13:25

score 9 · Accepted Answer · answered Apr 23 '13 at 13:15

9

The answer is yes. It is smart enough to know that.

This is the disassembled IL (courtesy of ILSpy) of a simple program which has such an assignment. As you can see from the ldc.r4 instruction, there is no conversion taking place:

.method private hidebysig static 
    void Main (
        string[] args
    ) cil managed 
{
    // Method begins at RVA 0x2050
    // Code size 8 (0x8)
    .maxstack 1
    .entrypoint
    .locals init (
        [0] float32 x
    )

    IL_0000: nop
    IL_0001: ldc.r4 44
    IL_0006: stloc.0
    IL_0007: ret
} // end of method Program::Main

answered Apr 23 '13 at 13:15

Dave Markle

95,573
20
147
170

[I wouldn't call "per the standard" being terribly smart](http://msdn.microsoft.com/en-us/library/vstudio/y5b434w4.aspx). Perhaps "handy". – user7116 Apr 23 '13 at 13:19
4

How about "the standard is smart enough"? :) – Dave Markle Apr 23 '13 at 13:20

score 3 · Answer 2 · edited May 23 '17 at 12:28

3

float is "smaller" than double

when you try to convert float to double you may lose precision

see this:C: convert double to float, preserving decimal point precision

same is true about C#

when you write float x = 5; the compiler will in compile time will convert this to float x = 5.0;

no performance problems

edited May 23 '17 at 12:28

Community

1
1

answered Apr 23 '13 at 13:18

Nahum

6,959
12
48
69

+1 for a plausible explanation of _why_ it converts 5.0 but not 5.1. – D Stanley Apr 23 '13 at 13:30
It will preserve magnitude but not necessarily precision. – user7116 Apr 23 '13 at 13:36
This wasn't a question about converting doubles to floats, but if there is a performance impact about assigning integer literals to floats. Still, useful information for others. – Chris Beamond Apr 24 '13 at 10:46

C# Performance penalty for Int32 literals to floats

2 Answers2