Hi, I found this C code that...

...I don't quite understand. What does the (A >> 1) part do here and what would you do in C++ for the same effect?

Code:

---

#define Blend_SoftLight(A,B)    ((uint8)((B < 128) ? (2*((A >> 1)+64)) * (B/255):(255 - (2*(255-((A >> 1) + 64))*(255-B)/255))))

---

The above macro works with color values form 1 to 255. Myself I will be working with float values from 0.0 to 1.0.

right bit shift.

I suggest that you read the tutorial on bitwise operators.

Quote:

---

Originally Posted by valaris

right bit shift.

---

Ok so lets say that A is a float value of 0.4, what should happen to that value in this case?

Quote:

---

Ok so lets say that A is a float value of 0.4, what should happen to that value in this case?

---

You will probably get a compile error since the bit shift operators work with integral operands (unless they are overloaded, in which case what it does depends on what the author wrote).

Quote:

---

Originally Posted by laserlight

You will probably get a compile error since the bit shift operators work with integral operands (unless they are overloaded, in which case what it does depends on what the author wrote it to do).

---

Yes sure but I need to convert this code to work with floats. So.. what's the way to do that? Take this small part as an example:

Code:

---

2*((A >> 1)+64)

---

The resulting float should be within 0.0 and 1.0 instead of between 1 and 255 which the code is working with. So I guess 64 stands for 255/4 here which in my case will be 1/4 = 0.25.

Quote:

---

Yes sure but I need to convert this code to work with floats. So.. what's the way to do that?

---

Divide by two instead of right shifting by one.

Actually with the math that you have the equation works like this

A = 255
A /= 2;
2 * A
A + 64
A = 319

Nowhere close to what you've intended.

But there are at least 255 rational numbers between 0 and 1, so I believe the answer would be (64.0 + A) / 255.0

Quote:

---

Originally Posted by laserlight

Divide by two instead of right shifting by one.

---

Ok great , that was what I was looking for. So what if it instead said (A << 8), would that be divide by 8 or 16 or...?

Thanks.

Quote:

---

Originally Posted by citizen

Actually with the math that you have the equation works like this

A = 255
A /= 2;
2 * A
A + 64
A = 319

Nowhere close to what you've intended.

But there are at least 255 rational numbers between 0 and 1, so I believe the answer would be (64.0 + A) / 255.0

---

Yes but the later parts of the code make it between 1 and 255

My code would look like this

2.0f * ((A * 0.5f) + 0.25f)

Woops. Anyway, here's how you can figure out shifts in decimal.
If >>1 is the same as / 2, then >>2 is the same as / 4, >>4 is the same as / 8; see a pattern yet?

Quote:

---

Originally Posted by citizen

Woops. Anyway, here's how you can figure out shifts in decimal.
If >>1 is the same as / 2, then >>2 is the same as / 4, >>4 is the same as / 8; see a pattern yet?

---

Ah yes I think I do. Thanks:)

And the other way round, <<, means multiplying with powers of 2.

Quote:

---

Originally Posted by citizen

If >>1 is the same as / 2, then >>2 is the same as / 4, >>4 is the same as / 8; see a pattern yet?

---

WHOOPS typo! That pattern gives a *much* different result!