FAQ: What are the major feature of the Math Accelerator on ADSP-CM41x series

Document created by AkashA Employee on Apr 29, 2016Last modified by AkashA Employee on Apr 29, 2016
Version 3Show Document
  • View in full screen mode

Math Accelerator on the ADSP-CM41x series provides a convenient and accurate way to perform:

  • Accurate single precision floating point computation of common transcendental such as trigonometric and exponential and there inverse
  • Faster reciprocals and square roots than provided by the Cortex-M4
  • Accelerated functions to convert Rectangular and polar coordinates


The following features

  • Accepts and produces IEEE-754 single precision floating point operands and results for all functions
  • Accelerator unit requires no additional handshaking after the input values are programmed in the corresponding input registers
  • The output result can be obtained whenever required from the result registers
  • It does not use the Core Rx or Sx Register, so the stack operations can proceed without any problems
  • Provides status register which provide the status and error on last  operation
  • Capability to generate an interrupt on Underflow, overflow, Divide by Zero or Invalid operation


Math Accelerator Unit supports the following operation

  • Trigonometric functions and their inverses: sinf, cosf, tanf, asinf, acosf, atanf, with angles in radians
  • Exponential functions expf (ex) and exp2f (2x)
  • Logarithmic functions: lnf and log2f
  • Square root
  • Reciprocal
  • Two-operand functions: atan2f(y/x) and hypotf(x,y) = v(x2 + y2)
  • Rectangular and polar coordinate conversions: (x,y)=ptor(r,a), (r,a) = rtop(x,y), with angles in radians


Details of the features are provided in the CM41x Hardware Reference Manual Refer to Math Accelerator Unit chapter