When are HLSL literals taken as a float and integer?

Question

float2 uv = float2(fX % 1, fY % 1) * 1/16;

In the above example 1/16 is treated as a float, multiplying the float2 by 0.0625, but in:

float2 uv = float2(fX % 1, fY % 1) * 1/16 + float2(1/16,0);

The second 1/16 is taken as an integer and the second float2 becomes (0,0).

What's the order in which a specific type is used. And is there any suffix to force a float? I know 1.0/16.0 would work.

Answer 1

Use a decimal to mark a type as a float. 1.0 is always floating point. Hence 1.0/16 or 1/16.0 will always be floating point.

Numeric literals are treated as floats if used in a floating-point context. Remember that operators like * and / are left-associative. So in your first example, the code is equivalent to (float(...) * 1) / 16. If you put the 1/16 in parentheses then it would be equal to 0 and that would be multiplied with the vector, e.g float2(...) * (1/16) == float2(0,0)

In your second example, the 1/16 is seen by itself, computed (as integers, because there's no float involved), and then the result of the expression is converted to float for the float2 construction. You need to cause either the 1 or the 16 to be a float in order for the / operator to be in a float context and force the other value to be treated as a float.