This discussion has been locked.

You can no longer post new replies to this discussion. If you have a question you can start a new discussion

speed of arm_sin_f32 function

Note: This was originally posted on 26th April 2013 at http://forums.arm.com

Hello,

I tried to increase the speed of the sin/cos calculation (float) for a Clark/Park tansformation by using the arm_sin_f32 functions of the CMSIS Lib.
My µC is a Cortex M4 with FPU.
I've compared the execution speeds of following functions:

standard from math.h:
sinf(float x)

CMSIS DSP LIB from arm_math.h
arm_sin_f32(float32_t x)

The standard sinf() function is 20% faster then the "fast" arm_sin_f32() function.

Has anyone experience by using arm_sin_f32(), or increasing speed of float sin/cos calculation?

Thanks in Advance

Peter Harris over 12 years ago

Note: This was originally posted on 26th April 2013 at http://forums.arm.com

The usual approach to making sin and cos faster is to exploit the knowledge of your algorithm to have a "maximal error" version rather than a fully accurate fp32 value - i.e. if you only need 18 bits of precision you can early out.
Cancel
Vote up 0 Vote down

Cancel
Thomas Pritzen over 12 years ago

Note: This was originally posted on 26th April 2013 at http://forums.arm.com

I Know that there are several ways to increase the speed of cos/sin functions.
But I do not get why a standard sinf() function from math.h is faster then a "special fast" function from a vendor deliverd ARM CMSIS lib, which is based on look up tables.
Cancel
Vote up 0 Vote down

Cancel