#ifdef __cplusplus extern "C" { #endif /* (C) Copyright 2009 Signal Essence; All Rights Reserved Module Name - vfixlib.c Author: Hugh McLaughlin Description: Signal Essence vector fixed point library in C definitions. Naming conventions: V -- has one or more vector inputs S -- has a scalar input i16 -- shorthand data type for int16 (signed short) q15 -- shorthand data type for int16 fractional Q15 q14 -- Q14, etc. u16 -- shorthand data type for uint16 (unsigned short) i32 -- shorthand data type for int32 (long) if output is not obvious the the data type is denoted by adding a suffix with _ Nameing: __ V -- if vector operation S -- if 2nd operand is a scalar, otherwise it is implied the 2nd operand is V If all the operands are the same type then only one data type need be listed. If two operands are the same and the result is different, then list the operand type and then the result type. If the operands are different types, then list all types in order of usage. History: hjm - Hugh McLaughlin Mar11,09 hjm created Machine/Compiler: ANSI C ==========================================================================*/ #ifndef ___sevfixlib_h #define ___sevfixlib_h #include "se_types.h" #include "se_platform.h" #include "frmath.h" #include "optimized_routines.h" #include "ext_interface.h" /*----------------------------------------------------------------------- Function Name: VAdd_i16 Description: Vector Add to a Vector. Computes the element by element sum of two vectors. -------------------------------------------------------------------------*/ typedef void (*VAdd_i16_t)( const int16 *src1, const int16 *src2, int16 *out, uint16 N ); extern VAdd_i16_t VAdd_i16; /*----------------------------------------------------------------------= Function Name: VAddVS32 Description: Vector Add. Adds two longword vectors together, element by element. Notes: Can write over a source vector? -----------------------------------------------------------------------*/ void VAdd_i32( const int32 *src1, const int32 *src2, int32 *dest, uint16 N ); void VAddScalar_i32(const int32 *px, int32 k, int32 *py, uint16 N); /*----------------------------------------------------------------------- Function Name: VAvePowerRMS_i16 Description: Vector compute the average power and take the RMS value of the power and return the result in a uint16. -------------------------------------------------------------------------*/ uint16 VAvePowerRMS_i16( const int16 *yptr, uint16 N ); /*======================================================================= Function Name: VComplexPoweri16_i32() Description: For a complex vector where each element is x+iy Computes power[i] = x[i]*x[i] + y[i]*y[i] Returns: void =======================================================================*/ void VComplexPoweri16_i32( const int16 *inptr, // complex vector int32 *outptr, // long word output vector uint16 numOut ); // same as number of complex elements /* ----------------------------------------------------------------------- Function Name: VCMul_16() Description: Multiplies two complex vector where each element is x+iy Computes x[i] = x0[i]*x1[i] - y0[i]*y1[i] y[i] = x0[i]*y1[i] + x1[i]*y0[i] data is stored real, imag, real, imag..... Returns: void ---------------------------------------------------------------------------*/ void VCMul_i16_Win32( const int16 *xptr, // complex vector const int16 *yptr, // complex vector int16 *outptr, // complex output vector uint16 numOut ); // same as number of complex elements /*---------------------------------------------------------------------- Function Name: VCRMulAdd_i16_i32 Description: Multiplies each 16-bit integer element of a complex vector by a Q15 vector, complex element by real element to produce a 32-bit result and add that to an existing 32-bit result. Returns: void -------------------------------------------------------------------------*/ typedef void (*VCRMulAdd_i16_i32_t)( const int16 *cPtr, // complex input const int16 *rPtr, // real coefficient vector const int32 *int32src, // add-to input int32 *destPtr, // result output, allowed to be in-place uint16 N ); // number of complex outputs extern VCRMulAdd_i16_i32_t VCRMulAdd_i16_i32; // for testing void VCRMulAdd_i16_i32_ansi( const int16 *cPtr, // complex input const int16 *rPtr, // real coefficient vector const int32 *int32src, // add-to input int32 *destPtr, // result output, allowed to be in-place uint16 N ); // number of complex outputs /*======================================================================= Function Name: VGenComplexTone_i16() Description: Geneates a complex tone from a phasor and initial state. This is useful for creating linear delay in the frequency domain, or for creating a complex tone in the time domain. data is stored real, imag, real, imag..... Returns: void =======================================================================*/ typedef void (*VGenComplexTone_i16_t)( int16 statePhasor[2], // real and imaginary const int16 deltaPhasor[2], // real and imaginary int16 *destPtr, uint16 numComplexOut ); extern VGenComplexTone_i16_t VGenComplexTone_i16; void VGenComplexTone_i16_float_ansi( int16 statePhasor[2], // real and imaginary const int16 deltaPhasor[2], // real and imaginary int16 *destPtr, uint16 numComplexOut ); /*---------------------------------------------------------------------- Function Name: VGen_Exp_Ramp_i16 Description: Generates an exponential ramp of values by doing a single pole filter with a target and current state. Returns: int16 -------------------------------------------------------------------------*/ void VGen_Exp_Ramp_i16( int16 *gainStatePtr, // gain state -- scalar value int16 alpha, // averaging coefficient int16 target, // target int16 *gptr, // output vector int N ); // num in block void VGenRamp_i16( int16 *px, int len, int16 initVal, int16 incr); /*---------------------------------------------------------------------- Function Name: VGetIndex* Get index of max/min value in vector See vgetindex.c -------------------------------------------------------------------------*/ uint16 VGetIndexMax_i16( const int16 *inPtr, uint16 N ); uint16 VGetIndexMaxAbs_i16( const int16 *inPtr, uint16 N ); uint16 VGetIndexMin_i16( const int16 *inPtr, uint16 N ); uint16 VGetIndexMax_i32( const int32 *pIn, uint16 N ); uint16 VGetIndexMaxAbs_i32( const int32 *inPtr, uint16 N ); // // get max/min value in vector int16 VGetValueMin_i16( const int16 *inPtr, uint16 N ); int16 VGetValueMax_i16( const int16 *inPtr, uint16 N ); uint16 VGetValueMaxAbs_i16( const int16 *inPtr, int32 N ); int32 VGetValueMax_i32( const int32 *pIn, uint16 N ); /*---------------------------------------------------------------------- Function Name: VLimitMin_i32 and VLimitMin_i16 Description: If an element is below the minLimitValue then it is set to the minLimitValue Returns: void -------------------------------------------------------------------------*/ void VLimitMin_i32( const int32 *src, int32 minLimitValue, int32 *dest, uint16 N ); void VLimitMin_i16( const int16 *src, int16 minLimitValue, int16 *dest, uint16 N ); /*---------------------------------------------------------------------- Function Name: VMax_i16 Description: Takes the maximum value between each element in two vectors of 16 bit words. output = max( src1, srcr2 ); Returns: void -------------------------------------------------------------------------*/ void VMax_i16( const int16 *src1Ptr, const int16 *src2Ptr, int16 *outPtr, uint16 N ); /*---------------------------------------------------------------------- Function Name: VMax_i32 Description: Takes the maximum value between each element in two vectors. output = max( src1, srcr2 ); Returns: void -------------------------------------------------------------------------*/ void VMax_i32( const int32 *src1Ptr, const int32 *src2Ptr, int32 *outPtr, uint16 N ); /*---------------------------------------------------------------------- Function Name: VMin_i16 Description: Takes the minimum value between each element in two vectors. output = max( src1, srcr2 ); Returns: void -------------------------------------------------------------------------*/ void VMin_i16( const int16 *src1Ptr, const int16 *src2Ptr, int16 *outPtr, uint16 N ); void VMinWithIndex_i16(const int16* pV0, const int16* pV1, int16* pMinIndex, int16* pMinVal, uint16 N); void VMin3WayWithIndex_i16(const int16* pV0, const int16* pV1, const int16* pV2, int16* pMinIndex, int16* pMinVal, uint16 N); /*---------------------------------------------------------------------- Function Name: VMin_i32 Description: Takes the minimum value between each element in two vectors. output = max( src1, srcr2 ); Returns: void -------------------------------------------------------------------------*/ void VMin_i32( const int32 *src1Ptr, const int32 *src2Ptr, int32 *outPtr, uint16 N ); /*---------------------------------------------------------------------- Function Name: VMMQ15_Add_i32 Description: Multiples a vector by a scaler and multiplies another vector by a scalar and adds them to get a 32 bit result. C = A*a + B*b The scalars are Q15 values. Returns: void -------------------------------------------------------------------------*/ void VMMQ15_Add_i32( const int32 *xptr, int16 a, const int32 *yptr, int16 b, int32 *outPtr, uint16 N ); /*---------------------------------------------------------------------- Function Name: VMul_i32_q15_i32, VMul_i32_q12_i32, VMul_i32_q10_i32 Description: Multiplies each 32-bit integer by a Q15 vector, element by element to produce a 32-bit result. Returns: void -------------------------------------------------------------------------*/ void VMul_i32_q15_i32( const int32 *src, const int16 *q15Ptr, int32 *dest, int32 N ); void VMul_i32_q12_i32( const int32 *src, const int16 *q12Ptr, int32 *dest, uint16 N ); void VMul_i32_q10_i32( const int32 *src, const int16 *q12Ptr, int32 *dest, uint16 N ); /*--------------------------------------------------------------------- Description: Sums all the elements of the input vector containing shorts. Shift result, if desired, so that it fits in a short. The right shift operator (>>) is arithmetic, i.e., negative numbers stay that way. Returns: short -------------------------------------------------------------------------*/ int16 VSumRtShift_i16( const int16 *src, uint16 N, uint16 rt_shift ); /*---------------------------------------------------------------------- Function Name: VSum_i32 Description: Vector Sum signed longwords. Sums the elements in a longword (32 bit signed) vector. Returns: long Notes: ------------------------------------------------------------------------*/ int32 VSum_i32( const int32 *src1, int32 N ); /*---------------------------------------------------------------------- Function Name: VSumRtShifts_i32 Description: Vector Sum signed longwords. Sums the elements in a longword (32 bit signed) vector. Shift each element first to avoid saturation issues. ASM code assumes there are guard bits and the vector is less than 128 elements. Returns: long Notes: ------------------------------------------------------------------------*/ int32 VSumRtShifts_i32( const int32 *src1, uint16 N, uint16 rt_shifts ); /*---------------------------------------------------------------------- Function Name: VSubi16 Description: Vector subtract with saturation. Computes the element by element difference of of a two vectors. Returns: void Notes: -----------------------------------------------------------------------*/ typedef void (*VSub_i16_t)( const int16 *aptr, const int16 *bptr, int16 *dest, uint16 N ); extern VSub_i16_t VSub_i16; /*----------------------------------------------------------------------- Function Name: VSub_i32 Description: Vector Sub. Subtracts one longword vector from another longword vector, element by element. Returns: void -------------------------------------------------------------------------*/ void VSub_i32( const int32 *src1, const int32 *src2, int32 *dest, uint16 N ); /*------------------------------------------------------------------------- Function Name: VTrackUpAveDown Description: Tracks the input in the upward direction and does a leaky average in the down direction. Returns: void -------------------------------------------------------------------------*/ void VTrackUpAveDown_i32(int32 *statePtr, const int32 *inPtr, int16 alpha, int N ); /*------------------------------------------------------------------------- Function Name: VFlip_i16 Description: Vector invert the order of a vector of int16 words. Moves the elements in a vector so that the first is last, second is second to last, ... last is first. Returns: void Notes: -------------------------------------------------------------------------*/ void VFlip_i16( const int16 *src, int16 *out, uint16 N ); /*------------------------------------------------------------------------- Function Name: VAvePower_i16_i32 Description: Computes the average power of a fixed point (integer) vector. S32 = SUM( S[0]..S[N-1] ) / N Returns: long -------------------------------------------------------------------------*/ int32 VAvePower_i16_i32( const int16 *xptr, uint16 N ); /*------------------------------------------------------------------------- Function Name: VDelaySave_i16 Description: Moves a vector to the right by moving element by element starting with the last element. Then retrieves data from a "previous" buffer to fill the area in front. Finally, it saves Returns: void -------------------------------------------------------------------------*/ void VDelaySave_i16( const int16 *src, int16 *dest, int16 *delayBufPtr, int16 blockSize, int16 delaySize ); /*------------------------------------------------------------------------- Function Name: VPower_i16_i32 Description: Computes the power of a fixed point (integer) vector. S32 = SUM( S[0]..S[N-1] ) Returns: long Notes: -------------------------------------------------------------------------*/ typedef uint32 (*VPower_i16_i32_t)( const int16 *inptr, uint16 rshifts, uint16 N ); extern VPower_i16_i32_t VPower_i16_i32; /*------------------------------------------------------------------------- Function Name: VPowerWithExponent_i16_i32 Description: Computes the power of a fixed point (integer) vector. S32 = SUM( S[0]..S[N-1] ) Tries to maximize the precision and determines the exponent, i.e. the number of right shifts taken. The full result would be: retValue*(2^rightShiftsTaken) Returns: long Notes: -------------------------------------------------------------------------*/ typedef uint32 (*VPowerWithExponent_i16_i32_t)( const int16 *inptr, uint16 *expPtr, uint16 N ); extern VPowerWithExponent_i16_i32_t VPowerWithExponent_i16_i32; /*------------------------------------------------------------------------- Function Name: VFill_i32 Description: Vector Fill Long Words. Returns: void Notes: -------------------------------------------------------------------------*/ void VFill_i32(int32* dest, int32 fillWord, int32 N ); /*------------------------------------------------------------------------- Function Name: VFill_i16 Description: Vector Fill int16 Words. Returns: void Notes: -------------------------------------------------------------------------*/ typedef void (*VFill_i16_t)( int16 *dest, int16 fillWord, uint32 N ); extern VFill_i16_t VFill_i16; void VFill_ui16(uint16 *pDest, uint16 val, int N); /*------------------------------------------------------------------------- Function Name: VLeftShifti16 Description: Vector Left Shifts. Shifts each element of a vector left by the numLefts argument. Works for right shifts too (specify negative leftshift) Returns: void Notes: -------------------------------------------------------------------------*/ typedef void (*VLeftShifts_i16_t)( const int16 *src, int16 numLefts, int16 *dest, uint16 N ); extern VLeftShifts_i16_t VLeftShifts_i16; /*------------------------------------------------------------------------- Function Name: VMoveS32 Description: Vector Move. Moves a vector of S32 words from one location to another. Returns: void Notes: -------------------------------------------------------------------------*/ typedef void (*VMove_i32_t)(const int32 *src, int32 *dest, uint32 N ); extern VMove_i32_t VMove_i32; /*------------------------------------------------------------------------- Function Name: VMovei16 Description: Vector Move. Moves a vector from one location to another. Returns: void Notes: VMove_i16_C55 is optimized for C55 platform. See function comments for constraints on usage. The Win32 version contains diagnostic code to identify when non-optimal transfers occur. To identify non-optimal VMove operations on the C55, re-define VMove_i16 to the Win32 version and place breakpoints in the diagnostic code. -------------------------------------------------------------------------*/ typedef void (*VMove_i16_t)( const int16 *src, int16 *dest, uint32 N ); extern VMove_i16_t VMove_i16; /*------------------------------------------------------------------------- Function Name: VMovei16SourceStride Description: Vector Move. Moves a vector of i16 words with a stride between each source word. The destination words are stored in order. Returns: void Notes: -------------------------------------------------------------------------*/ void VMoveSourceStride_i16( const int16 *src, uint16 stride, int16 *dest, int N ); /*------------------------------------------------------------------------- Function Name: VMoveDestStride_i16 Description: Vector Move. Moves a vector of i16 words with a stride between each destination word. Returns: void Notes: -------------------------------------------------------------------------*/ void VMoveDestStride_i16( const int16 *src, int stride, int16 *dest, int N ); /*------------------------------------------------------------------------- Function Name: VShiftRtRnd_i32_i16 Description: Vector Shift Long Word, Round 32 bit word and save 16-bit bits. Returns: void Notes: -------------------------------------------------------------------------*/ typedef void (*VShiftRtRnd_i32_i16_t)( const int32 *lsrc, int16 *sdest, uint16 shifts, uint16 N ); extern VShiftRtRnd_i32_i16_t VShiftRtRnd_i32_i16; /*------------------------------------------------------------------------- Function Name: VShiftLtRnd_i16_i32 Description: Vector Shift short Word left, and save 32-bit bits. Returns: void Notes: -------------------------------------------------------------------------*/ typedef void (*VShiftLtRnd_i16_i32_t)( const int16 *ssrc, int32 *ldest, uint16 shifts, uint16 N ); extern VShiftLtRnd_i16_i32_t VShiftLtRnd_i16_i32; /*------------------------------------------------------------------------- Function Name: VMulQ15Ramp_i16 Description: Vector Multiply by a ramping scalar. Multiply all the elements of a vector by a single scalar that is incremented every sample Returns: the last value of the scalar Notes: -------------------------------------------------------------------------*/ int16 VMulQ15Ramp_i16( const int16 *src, int16 q15val, int16 delta, int16 *dest, uint16 N ); /*------------------------------------------------------------------------- Function Name: VMulSmoothedi16 Description: Vector Multiply by Smoothed Scalar. Multiply all the elements of a vector by a scalar that is ramping like an exponential function from the initial state converging to the target state. Returns: void Notes: -------------------------------------------------------------------------*/ void VMulSmoothed_i16( const int16 *src, int16 *statePtr, const int16 *alpha, int16 target, int16 *dest, uint16 N ); /*------------------------------------------------------------------------- Function Name: VMul_i16_q15_i16, VMul_i16_q12_i16 Description: Element-by-element vector multiply. Multiply each of the elements of a vector by another vector value in Q15 or Q12 format Returns: void Notes: -------------------------------------------------------------------------*/ typedef void (*VMul_i16_q15_i16_t)( const int16 *aptr, const int16 *bptr, int16 *dest, uint16 N ); extern VMul_i16_q15_i16_t VMul_i16_q15_i16; // for testing void VMul_i16_q15_i16_ansi( const int16 *src, const int16 *q15, int16 *dest, uint16 N ); void VMul_i16_q12_i16( const int16 *aptr, const int16 *bptr, int16 *dest, int32 N ); /*----------------------------------------------------------------------= Function Name: VDotProducti16ReturnS32 Description: Computes the dot product (e.g., correlation) of a fixed point (integer) vector. Intended to operate on vectors which are lengths that are a power of 2. The number of right shifts parameter specifies the divisor to compute the average. The number of right shifts must be enough to avoid an overflow. Returns: long Notes: -----------------------------------------------------------------------*/ typedef int32 (*VDotProductRtShifts_i16_i32_t)( const int16 *xptr, const int16 *hptr, uint16 N, uint16 rshifts ); extern VDotProductRtShifts_i16_i32_t VDotProductRtShifts_i16_i32; /*----------------------------------------------------------------------= Function Name: VDotProductQ15_i16_i16 Description: Computes the dot product (e.g., correlation) of a fixed point (integer) vector. Returns: long Notes: -----------------------------------------------------------------------*/ int16 VDotProductQ15_i16_i16_ansi( const int16 *xptr, const int16 *hptr, uint16 N ); typedef void (*VDotProductsLeftShift_q15_i16_t)( const int16 *inPtr, // input const int16 *coefPtr, // other input uint16 leftShifts, // allow upscaling int16 *outPtr, // output vector uint16 numPerDotProd, // number of multiplies per dot product uint16 numOut, // number of output elements uint16 decimationFactor ); // number of input points to skip extern VDotProductsLeftShift_q15_i16_t VDotProductsLeftShift_q15_i16; // // for testing void VDotProductsLeftShift_q15_i16_ansi( const int16 *inPtr, // input const int16 *coefPtr, // other input uint16 leftShifts, // allow upscaling int16 *outPtr, // output vector uint16 numPerDotProd, // number of multiplies per dot product uint16 numOut, // number of output elements uint16 decimationFactor ); // number of input points to skip void VWindowWithPreviousBlock_i16( const int16 *wfPtr, // windowing function, size of window const int16 *newPtr, // new input block int16 *prevPtr, // previous block int16 *outPtr, // windowed block, output is blocksize+overlap int16 overlap, int16 blockSize ); // input blocksize /* ====================================================== Scale a vector by a scalar =========================================================*/ typedef void (*VScale_i32_q15_i32_t)( const int32 *inPtr, int16 q15Scalar, int32 *outPtr, // can be inplace int N ); extern VScale_i32_q15_i32_t VScale_i32_q15_i32; void VScale_i16_q15_i16( const int16 *src, int16 q15val, int16 *dest, int N ); void VScale_i16_q10_i16( const int16 *src, int16 q10val, int16 *dest, int N ); typedef void (*VScale_i16_q12_i16_t)( const int16 *src, int16 q12val, int16 *dest, int N ); extern VScale_i16_q12_i16_t VScale_i16_q12_i16; typedef void (*VScale_i16_q15_i32_t)( const int16 *src, int16 q15val, int32 *dest, int N ); extern VScale_i16_q15_i32_t VScale_i16_q15_i32; //void VMulScalar_i16_i32( int16 *src, int16 i16val, int32 *dest, uint16 N ); void VScale_i16_i16_i32( const int16 *src, int16 i16val, int32 *dest, int N ); typedef void (*VScale_q31_q15_q15_t)(const int32* px, int16 k, int16* py, int N); extern VScale_q31_q15_q15_t VScale_q31_q15_q15; typedef void (*VScale_q31_q14_q15_t)(const int32* px, int16 k_q14, int16* py, int N); extern VScale_q31_q14_q15_t VScale_q31_q14_q15; void VScale_q31_q12_q15(const int32* px, int16 k_q12, int16* py, int N); void VScale_i32_q10_i32(const int32 *src, int16 i16val, int32 *dest, int N ); /* ====================================================== Scale a vector by a scalar, then add another vector! =========================================================*/ void VScaleAdd_i16_q15_i32_i32( const int16 *src, int16 q15val, const int32 *i32src, int32 *dest, uint16 N ); void VScaleAdd_i16_q12_i32_i32( const int16 *src, int16 q12val, const int32 *i32src, int32 *dest, uint16 N ); void VScaleAdd_i16_i16_i32_i32( const int16 *src, int16 i16val, const int32 *i32src, int32 *dest, uint16 N ); void VScaleAdd_i16_q12_i16_i16( const int16 *s1p, int16 i16val, // Q12 const int16 *s2p, int16 *dest, uint16 N ); typedef void (*VScaleAdd_i16_q15_i16_i16_t)( const int16 *s1p, int16 i16val, // Q15 const int16 *s2p, int16 *dest, uint16 N ); extern VScaleAdd_i16_q15_i16_i16_t VScaleAdd_i16_q15_i16_i16; void VScaleAdd_i16_q12_i16_i16( const int16 *s1p, int16 q12val, // Q12 const int16 *s2p, int16 *dest, uint16 N ); float32 VSumPower_i16_flt(const int16 *p, int N); #endif // #ifdef #ifdef __cplusplus } #endif