Log2 FPU operation

I see only one way - create your own look-up table in memory, and then approximate log2 function by using Taylor series.

Alexey Tsirlin said:

Is there an efficient way to calculate ceil log2 operation (extraction of the exponent out of the floating point number) using FPU math on C28 Delfino MCU?

Yaminov Kamil said:

I see only one way - create your own look-up table in memory, and then approximate log2 function by using Taylor series.

There is a fast atan2 function in the C28x FPU Fast RTS Library - it uses a lookup table method.  All source code is included so you could modify it to suit your needs.

http://focus.ti.com/docs/toolsw/folders/print/sprc664.html

Cheers

Lori

Heh. Extremely fast way - is to use only lookup table. But it costs memory.

Yaminov Kamil said:

Heh. Extremely fast way - is to use only lookup table. But it costs memory.

True :)

One clarification - the lookup table used by the fastRTS library algorithum is in the ROM of the device so it won't use up extra RAM memory.

Cheers

Lori

Alexey,

I know it's an old post, but  how did you resolve this issue?  I have the same requirement.  I need a ceil(log2()) function.

John

John,

I didn't find good way to calculate log2 function.... I just had to update my algorithms so I won't need it....

Alexey.

PS. Let me know if you will find suitable solution.

Try to look into CORDIC alghorytm. http://3month-of-life.blogspot.com/2011/02/cordic.html