The Mana World Forums

Posted: **18 Aug 2008, 20:54**

just looked at the code...

SIN45 is already defined in <cmath>
----> #define SIN45 M_SQRT1_2

in addTextSplashEffect
three call to rand and three modulos... can be done more efficiently:
instead of

Code: Select all

newParticle->setVelocity(((rand() % 100) - 50) / 200.0f,    // X
                         ((rand() % 100) - 50) / 200.0f,    // Y
                         ((rand() % 100) / 200.0f) + 4.0f); // Z

I would recommand

Code: Select all

int r=rand()>>4;
newParticle->setVelocity((( r      & 127) - 64) / 256.0f,    // X
                         (((r>>8 ) & 127) - 64) / 256.0f,    // Y
                         (((r>>16) & 127) / 256.0f) + 4.0f); // Z

Ok, you're right, this is far from critical... cf. the subject of my post

Posted: **18 Aug 2008, 21:36**

In case of non-performance critical code (this particular piece of code is executed only once everytime a damage number is generated) code readability is more important than efficiency. "(rand() % 100) - 50" immediately tells you "random number between -50 and 50" while the meaning of "((r>>8 ) & 127) - 64" is not that apparent on the first look.

And although slow divisions by powers of two can be optimized by the compiler by substituting them with fast bitwise shifts in integer arithmetic this is not the case when dealing with floating point numbers.

Posted: **18 Aug 2008, 22:02**

the point was not to remove the slow division, but the slow modulo

and most of all replace three calls to rand with one.

But I agree with you, it might be less readable.

The Mana World Forums

some (useless) remarks about particle.cpp

some (useless) remarks about particle.cpp

Re: some (useless) remarks about particle.cpp

Re: some (useless) remarks about particle.cpp