Sign in to follow this  
Jason Goepel

Power Operator

Recommended Posts

Jason Goepel    797

Many of my users write math-oriented scripts.  Do you have any opposition to a binary operator that computes power?

 

I was thinking that ** wouldn't cause any conflicts since AngelScript doesn't support * for pointer dereferencing.

Share this post


Link to post
Share on other sites
Jason Goepel    797

I was going to either implement it in the compiler or as a pre-processing feature, depending on whether you would accept it in the compiler (I didn't want to maintain a fork).

 

The C-runtime "pow" function always returns a floating-point number.  Do we want this behavior even if the operator is used between two integers (i.e.  3 ** 4 ==> 81.0 )?

Share this post


Link to post
Share on other sites
Jason Goepel    797

Of course, this operator would be able to be overloaded like any of the other math operators.

 

Regarding my earlier post, I am inclined to make ** not always return a double.  I am inclined to think it should behave like the other math operators and return the same type as the operands when possible.  I will look into an integer implementation that is better than a simple multiply loop.

Share this post


Link to post
Share on other sites
WitchLord    4678

If the power-of operator is implemented as a built-in operator it should keep the type of the operands (with appropriate implicit casts when they are not equal). Much like how it is done for other math operators.

 

And yes, it should be possible to overload it as well for, e.g. by registering/implementing the opPow/opPow_r methods.

 

You'll need to implement new bytecode instructions for this too. Internally in the bytecode instructions can possibly use the pow() function and then convert the result to the correct type. This is of course if you don't find a more efficient way of calculating power-of. :)

Share this post


Link to post
Share on other sites
Jason Goepel    797

I think I will implement my own integer power function.  I've done some limited testing and found that it is up to ten times faster to do the calculation with integers than to pass integers through the floating-point pow() function.

 

Where should I put the powi and powi64 functions?  Should I create a new file in the project?

 

Also, it is fairly easy for powi to overflow.  Should I detect overflows and set an exception?  Overflow detection is pretty cheap.  I can pre-calculate a table with 64 entries and assign the maximum value of a base for any given exponent.  For my users I think it would be helpful to see an overflow exception.

Share this post


Link to post
Share on other sites
WitchLord    4678

You can just add the powi and powi64 functions to one of the existing modules, e.g. as_context.cpp where I assume they will be most frequently called. Once I merge the code into the SVN I'll consider creating separate file for these utility functions or not.

 

Try to keep the way the operators work the same regardless on the type they're working on. I agree that having an overflow exception can be useful (similar to the division by zero exception), but the float and double versions should also throw the same exception.

Share this post


Link to post
Share on other sites
Jason Goepel    797

I've done some more thinking about the overflow exceptions.  The C-runtime implementations vary on whether or not they set errno to ERANGE.  They do all seem to return HUGE_VAL though.  Perhaps the integer implementation should not overflow, but instead return INT_MAX.

 

There are also inconsistent implementations of 0**0.  Microsoft seems to return 1.

Share this post


Link to post
Share on other sites
WitchLord    4678

The most important part is to be consistent. Perhaps you're right not to raise an exception in this case and instead return INT_MAX, but then the float and double variants my behave consistently and return the largest possible value.

Share this post


Link to post
Share on other sites
WitchLord    4678

I took a quick look at the patch. It looks like everything is OK. You even took the time to update the documentation. Nice!

 

I'll be doing a bit of testing before I check in the changes, but rest assured that this will be part of the 2.28.1 release. :)

Share this post


Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now

Sign in to follow this