Rounding to the next power of two

[❤x1]

Code: Select all

Procedure RoundPow(Number)
	Number - 1
	Number = (Number | Number >> 1)
	Number = (Number | Number >> 2)
	Number = (Number | Number >> 4)
	Number = (Number | Number >> 8)
	Number = (Number | Number >> 16)
	
	CompilerIf #PB_Compiler_Processor = #PB_Processor_x64
		Number = (Number | Number >> 32)
	CompilerEndIf
	
	ProcedureReturn Number + 1
EndProcedure

Debug RoundPow(67)

[eck49]

Clever!

A very elegant application of binary arithmetic.

[BarryG]

Since I'm a bit dumb, when would you need this type of trick? Can you give an example? I like to learn. Thanks.

[Keya]

I think you can also use the Intel assembly instruction "bsr" (bitscan reverse) for this

[skywalk]

This is faster than the direct way with pow() function. I will use it.
Useful when sampling discretely and FFT samples must be powers of 2.

[idle]

Yes a nice tip if you need to call it frequently the assembly equivalent would be like this.

Code: Select all

Procedure RoundPowUP(Number) 
  CompilerIf SizeOf(Integer) = 8
    !mov rdx,[p.v_Number] 
    !sub rdx, 1
    !bsr qword rcx , rdx
    !add rcx,1
    !shl rax,cl 
  CompilerElse
    !mov edx,[p.v_Number] 
    !sub edx, 1
    !bsr qword ecx , edx
    !add ecx,1
    !shl eax,cl 
  CompilerEndIf   
  ProcedureReturn 
EndProcedure