diff options
Diffstat (limited to 'Drivers/CMSIS/DSP/Source/FilteringFunctions/arm_biquad_cascade_df1_fast_q31.c')
-rw-r--r-- | Drivers/CMSIS/DSP/Source/FilteringFunctions/arm_biquad_cascade_df1_fast_q31.c | 174 |
1 files changed, 89 insertions, 85 deletions
diff --git a/Drivers/CMSIS/DSP/Source/FilteringFunctions/arm_biquad_cascade_df1_fast_q31.c b/Drivers/CMSIS/DSP/Source/FilteringFunctions/arm_biquad_cascade_df1_fast_q31.c index 00dbae1f7..586296b32 100644 --- a/Drivers/CMSIS/DSP/Source/FilteringFunctions/arm_biquad_cascade_df1_fast_q31.c +++ b/Drivers/CMSIS/DSP/Source/FilteringFunctions/arm_biquad_cascade_df1_fast_q31.c @@ -3,13 +3,13 @@ * Title: arm_biquad_cascade_df1_fast_q31.c * Description: Processing function for the Q31 Fast Biquad cascade DirectFormI(DF1) filter * - * $Date: 27. January 2017 - * $Revision: V.1.5.1 + * $Date: 18. March 2019 + * $Revision: V1.6.0 * * Target Processor: Cortex-M cores * -------------------------------------------------------------------- */ /* - * Copyright (C) 2010-2017 ARM Limited or its affiliates. All rights reserved. + * Copyright (C) 2010-2019 ARM Limited or its affiliates. All rights reserved. * * SPDX-License-Identifier: Apache-2.0 * @@ -29,55 +29,52 @@ #include "arm_math.h" /** - * @ingroup groupFilters + @ingroup groupFilters */ /** - * @addtogroup BiquadCascadeDF1 - * @{ + @addtogroup BiquadCascadeDF1 + @{ */ /** - * @details - * - * @param[in] *S points to an instance of the Q31 Biquad cascade structure. - * @param[in] *pSrc points to the block of input data. - * @param[out] *pDst points to the block of output data. - * @param[in] blockSize number of samples to process per call. - * @return none. - * - * <b>Scaling and Overflow Behavior:</b> - * \par - * This function is optimized for speed at the expense of fixed-point precision and overflow protection. - * The result of each 1.31 x 1.31 multiplication is truncated to 2.30 format. - * These intermediate results are added to a 2.30 accumulator. - * Finally, the accumulator is saturated and converted to a 1.31 result. - * The fast version has the same overflow behavior as the standard version and provides less precision since it discards the low 32 bits of each multiplication result. - * In order to avoid overflows completely the input signal must be scaled down by two bits and lie in the range [-0.25 +0.25). Use the intialization function - * arm_biquad_cascade_df1_init_q31() to initialize filter structure. - * - * \par - * Refer to the function <code>arm_biquad_cascade_df1_q31()</code> for a slower implementation of this function which uses 64-bit accumulation to provide higher precision. Both the slow and the fast versions use the same instance structure. - * Use the function <code>arm_biquad_cascade_df1_init_q31()</code> to initialize the filter structure. + @brief Processing function for the Q31 Biquad cascade filter (fast variant). + @param[in] S points to an instance of the Q31 Biquad cascade structure + @param[in] pSrc points to the block of input data + @param[out] pDst points to the block of output data + @param[in] blockSize number of samples to process per call + @return none + + @par Scaling and Overflow Behavior + This function is optimized for speed at the expense of fixed-point precision and overflow protection. + The result of each 1.31 x 1.31 multiplication is truncated to 2.30 format. + These intermediate results are added to a 2.30 accumulator. + Finally, the accumulator is saturated and converted to a 1.31 result. + The fast version has the same overflow behavior as the standard version and provides less precision since it discards the low 32 bits of each multiplication result. + In order to avoid overflows completely the input signal must be scaled down by two bits and lie in the range [-0.25 +0.25). Use the intialization function + arm_biquad_cascade_df1_init_q31() to initialize filter structure. + @remark + Refer to \ref arm_biquad_cascade_df1_q31() for a slower implementation of this function + which uses 64-bit accumulation to provide higher precision. Both the slow and the fast versions use the same instance structure. + Use the function \ref arm_biquad_cascade_df1_init_q31() to initialize the filter structure. */ void arm_biquad_cascade_df1_fast_q31( const arm_biquad_casd_df1_inst_q31 * S, - q31_t * pSrc, - q31_t * pDst, - uint32_t blockSize) + const q31_t * pSrc, + q31_t * pDst, + uint32_t blockSize) { - q31_t acc = 0; /* accumulator */ - q31_t Xn1, Xn2, Yn1, Yn2; /* Filter state variables */ - q31_t b0, b1, b2, a1, a2; /* Filter coefficients */ - q31_t *pIn = pSrc; /* input pointer initialization */ - q31_t *pOut = pDst; /* output pointer initialization */ - q31_t *pState = S->pState; /* pState pointer initialization */ - q31_t *pCoeffs = S->pCoeffs; /* coeff pointer initialization */ - q31_t Xn; /* temporary input */ - int32_t shift = (int32_t) S->postShift + 1; /* Shift to be applied to the output */ - uint32_t sample, stage = S->numStages; /* loop counters */ - + const q31_t *pIn = pSrc; /* Source pointer */ + q31_t *pOut = pDst; /* Destination pointer */ + q31_t *pState = S->pState; /* pState pointer */ + const q31_t *pCoeffs = S->pCoeffs; /* Coefficient pointer */ + q31_t acc = 0; /* Accumulator */ + q31_t b0, b1, b2, a1, a2; /* Filter coefficients */ + q31_t Xn1, Xn2, Yn1, Yn2; /* Filter pState variables */ + q31_t Xn; /* Temporary input */ + int32_t shift = (int32_t) S->postShift + 1; /* Shift to be applied to the output */ + uint32_t sample, stage = S->numStages; /* Loop counters */ do { @@ -88,22 +85,23 @@ void arm_biquad_cascade_df1_fast_q31( a1 = *pCoeffs++; a2 = *pCoeffs++; - /* Reading the state values */ + /* Reading the pState values */ Xn1 = pState[0]; Xn2 = pState[1]; Yn1 = pState[2]; Yn2 = pState[3]; +#if defined (ARM_MATH_LOOPUNROLL) + /* Apply loop unrolling and compute 4 output values simultaneously. */ - /* The variables acc ... acc3 hold output values that are being computed: + /* Variables acc ... acc3 hold output values that are being computed: * - * acc = b0 * x[n] + b1 * x[n-1] + b2 * x[n-2] + a1 * y[n-1] + a2 * y[n-2] + * acc = b0 * x[n] + b1 * x[n-1] + b2 * x[n-2] + a1 * y[n-1] + a2 * y[n-2] */ + /* Loop unrolling: Compute 4 outputs at a time */ sample = blockSize >> 2U; - /* First part of the processing with loop unrolling. Compute 4 outputs at a time. - ** a second loop below computes the remaining 1 to 3 samples. */ while (sample > 0U) { /* Read the input */ @@ -111,19 +109,19 @@ void arm_biquad_cascade_df1_fast_q31( /* acc = b0 * x[n] + b1 * x[n-1] + b2 * x[n-2] + a1 * y[n-1] + a2 * y[n-2] */ /* acc = b0 * x[n] */ - /*acc = (q31_t) (((q63_t) b1 * Xn1) >> 32);*/ + /* acc = (q31_t) (((q63_t) b1 * Xn1) >> 32);*/ mult_32x32_keep32_R(acc, b1, Xn1); /* acc += b1 * x[n-1] */ - /*acc = (q31_t) ((((q63_t) acc << 32) + ((q63_t) b0 * (Xn))) >> 32);*/ + /* acc = (q31_t) ((((q63_t) acc << 32) + ((q63_t) b0 * (Xn))) >> 32);*/ multAcc_32x32_keep32_R(acc, b0, Xn); /* acc += b[2] * x[n-2] */ - /*acc = (q31_t) ((((q63_t) acc << 32) + ((q63_t) b2 * (Xn2))) >> 32);*/ + /* acc = (q31_t) ((((q63_t) acc << 32) + ((q63_t) b2 * (Xn2))) >> 32);*/ multAcc_32x32_keep32_R(acc, b2, Xn2); /* acc += a1 * y[n-1] */ - /*acc = (q31_t) ((((q63_t) acc << 32) + ((q63_t) a1 * (Yn1))) >> 32);*/ + /* acc = (q31_t) ((((q63_t) acc << 32) + ((q63_t) a1 * (Yn1))) >> 32);*/ multAcc_32x32_keep32_R(acc, a1, Yn1); /* acc += a2 * y[n-2] */ - /*acc = (q31_t) ((((q63_t) acc << 32) + ((q63_t) a2 * (Yn2))) >> 32);*/ + /* acc = (q31_t) ((((q63_t) acc << 32) + ((q63_t) a2 * (Yn2))) >> 32);*/ multAcc_32x32_keep32_R(acc, a2, Yn2); /* The result is converted to 1.31 , Yn2 variable is reused */ @@ -137,19 +135,19 @@ void arm_biquad_cascade_df1_fast_q31( /* acc = b0 * x[n] + b1 * x[n-1] + b2 * x[n-2] + a1 * y[n-1] + a2 * y[n-2] */ /* acc = b0 * x[n] */ - /*acc = (q31_t) (((q63_t) b0 * (Xn2)) >> 32);*/ + /* acc = (q31_t) (((q63_t) b0 * (Xn2)) >> 32);*/ mult_32x32_keep32_R(acc, b0, Xn2); /* acc += b1 * x[n-1] */ - /*acc = (q31_t) ((((q63_t) acc << 32) + ((q63_t) b1 * (Xn))) >> 32);*/ + /* acc = (q31_t) ((((q63_t) acc << 32) + ((q63_t) b1 * (Xn))) >> 32);*/ multAcc_32x32_keep32_R(acc, b1, Xn); /* acc += b[2] * x[n-2] */ - /*acc = (q31_t) ((((q63_t) acc << 32) + ((q63_t) b2 * (Xn1))) >> 32);*/ + /* acc = (q31_t) ((((q63_t) acc << 32) + ((q63_t) b2 * (Xn1))) >> 32);*/ multAcc_32x32_keep32_R(acc, b2, Xn1); /* acc += a1 * y[n-1] */ - /*acc = (q31_t) ((((q63_t) acc << 32) + ((q63_t) a1 * (Yn2))) >> 32);*/ + /* acc = (q31_t) ((((q63_t) acc << 32) + ((q63_t) a1 * (Yn2))) >> 32);*/ multAcc_32x32_keep32_R(acc, a1, Yn2); /* acc += a2 * y[n-2] */ - /*acc = (q31_t) ((((q63_t) acc << 32) + ((q63_t) a2 * (Yn1))) >> 32);*/ + /* acc = (q31_t) ((((q63_t) acc << 32) + ((q63_t) a2 * (Yn1))) >> 32);*/ multAcc_32x32_keep32_R(acc, a2, Yn1); /* The result is converted to 1.31, Yn1 variable is reused */ @@ -163,19 +161,19 @@ void arm_biquad_cascade_df1_fast_q31( /* acc = b0 * x[n] + b1 * x[n-1] + b2 * x[n-2] + a1 * y[n-1] + a2 * y[n-2] */ /* acc = b0 * x[n] */ - /*acc = (q31_t) (((q63_t) b0 * (Xn1)) >> 32);*/ + /* acc = (q31_t) (((q63_t) b0 * (Xn1)) >> 32);*/ mult_32x32_keep32_R(acc, b0, Xn1); /* acc += b1 * x[n-1] */ - /*acc = (q31_t) ((((q63_t) acc << 32) + ((q63_t) b1 * (Xn2))) >> 32);*/ + /* acc = (q31_t) ((((q63_t) acc << 32) + ((q63_t) b1 * (Xn2))) >> 32);*/ multAcc_32x32_keep32_R(acc, b1, Xn2); /* acc += b[2] * x[n-2] */ - /*acc = (q31_t) ((((q63_t) acc << 32) + ((q63_t) b2 * (Xn))) >> 32);*/ + /* acc = (q31_t) ((((q63_t) acc << 32) + ((q63_t) b2 * (Xn))) >> 32);*/ multAcc_32x32_keep32_R(acc, b2, Xn); /* acc += a1 * y[n-1] */ - /*acc = (q31_t) ((((q63_t) acc << 32) + ((q63_t) a1 * (Yn1))) >> 32);*/ + /* acc = (q31_t) ((((q63_t) acc << 32) + ((q63_t) a1 * (Yn1))) >> 32);*/ multAcc_32x32_keep32_R(acc, a1, Yn1); /* acc += a2 * y[n-2] */ - /*acc = (q31_t) ((((q63_t) acc << 32) + ((q63_t) a2 * (Yn2))) >> 32);*/ + /* acc = (q31_t) ((((q63_t) acc << 32) + ((q63_t) a2 * (Yn2))) >> 32);*/ multAcc_32x32_keep32_R(acc, a2, Yn2); /* The result is converted to 1.31, Yn2 variable is reused */ @@ -190,7 +188,7 @@ void arm_biquad_cascade_df1_fast_q31( /* acc = b0 * x[n] + b1 * x[n-1] + b2 * x[n-2] + a1 * y[n-1] + a2 * y[n-2] */ /* acc = b0 * x[n] */ - /*acc = (q31_t) (((q63_t) b0 * (Xn)) >> 32);*/ + /* acc = (q31_t) (((q63_t) b0 * (Xn)) >> 32);*/ mult_32x32_keep32_R(acc, b0, Xn); /* acc += b1 * x[n-1] */ /*acc = (q31_t) ((((q63_t) acc << 32) + ((q63_t) b1 * (Xn1))) >> 32);*/ @@ -207,47 +205,53 @@ void arm_biquad_cascade_df1_fast_q31( /* Every time after the output is computed state should be updated. */ /* The states should be updated as: */ - /* Xn2 = Xn1 */ + /* Xn2 = Xn1 */ Xn2 = Xn1; /* The result is converted to 1.31, Yn1 variable is reused */ Yn1 = acc << shift; - /* Xn1 = Xn */ + /* Xn1 = Xn */ Xn1 = Xn; /* Store the output in the destination buffer. */ *(pOut + 3U) = Yn1; pOut += 4U; - /* decrement the loop counter */ + /* decrement loop counter */ sample--; } - /* If the blockSize is not a multiple of 4, compute any remaining output samples here. - ** No loop unrolling is used. */ + /* Loop unrolling: Compute remaining outputs */ sample = (blockSize & 0x3U); - while (sample > 0U) - { +#else + + /* Initialize blkCnt with number of samples */ + sample = blockSize; + +#endif /* #if defined (ARM_MATH_LOOPUNROLL) */ + + while (sample > 0U) + { /* Read the input */ Xn = *pIn++; /* acc = b0 * x[n] + b1 * x[n-1] + b2 * x[n-2] + a1 * y[n-1] + a2 * y[n-2] */ /* acc = b0 * x[n] */ - /*acc = (q31_t) (((q63_t) b0 * (Xn)) >> 32);*/ + /* acc = (q31_t) (((q63_t) b0 * (Xn)) >> 32);*/ mult_32x32_keep32_R(acc, b0, Xn); /* acc += b1 * x[n-1] */ - /*acc = (q31_t) ((((q63_t) acc << 32) + ((q63_t) b1 * (Xn1))) >> 32);*/ + /* acc = (q31_t) ((((q63_t) acc << 32) + ((q63_t) b1 * (Xn1))) >> 32);*/ multAcc_32x32_keep32_R(acc, b1, Xn1); /* acc += b[2] * x[n-2] */ - /*acc = (q31_t) ((((q63_t) acc << 32) + ((q63_t) b2 * (Xn2))) >> 32);*/ + /* acc = (q31_t) ((((q63_t) acc << 32) + ((q63_t) b2 * (Xn2))) >> 32);*/ multAcc_32x32_keep32_R(acc, b2, Xn2); /* acc += a1 * y[n-1] */ - /*acc = (q31_t) ((((q63_t) acc << 32) + ((q63_t) a1 * (Yn1))) >> 32);*/ + /* acc = (q31_t) ((((q63_t) acc << 32) + ((q63_t) a1 * (Yn1))) >> 32);*/ multAcc_32x32_keep32_R(acc, a1, Yn1); /* acc += a2 * y[n-2] */ - /*acc = (q31_t) ((((q63_t) acc << 32) + ((q63_t) a2 * (Yn2))) >> 32);*/ + /* acc = (q31_t) ((((q63_t) acc << 32) + ((q63_t) a2 * (Yn2))) >> 32);*/ multAcc_32x32_keep32_R(acc, a2, Yn2); /* The result is converted to 1.31 */ @@ -255,10 +259,10 @@ void arm_biquad_cascade_df1_fast_q31( /* Every time after the output is computed state should be updated. */ /* The states should be updated as: */ - /* Xn2 = Xn1 */ - /* Xn1 = Xn */ - /* Yn2 = Yn1 */ - /* Yn1 = acc */ + /* Xn2 = Xn1 */ + /* Xn1 = Xn */ + /* Yn2 = Yn1 */ + /* Yn1 = acc */ Xn2 = Xn1; Xn1 = Xn; Yn2 = Yn1; @@ -267,18 +271,18 @@ void arm_biquad_cascade_df1_fast_q31( /* Store the output in the destination buffer. */ *pOut++ = acc; - /* decrement the loop counter */ + /* decrement loop counter */ sample--; - } + } - /* The first stage goes from the input buffer to the output buffer. */ - /* Subsequent stages occur in-place in the output buffer */ + /* The first stage goes from the input buffer to the output buffer. */ + /* Subsequent stages occur in-place in the output buffer */ pIn = pDst; /* Reset to destination pointer */ pOut = pDst; - /* Store the updated state variables back into the pState array */ + /* Store the updated state variables back into the pState array */ *pState++ = Xn1; *pState++ = Xn2; *pState++ = Yn1; @@ -288,5 +292,5 @@ void arm_biquad_cascade_df1_fast_q31( } /** - * @} end of BiquadCascadeDF1 group - */ + @} end of BiquadCascadeDF1 group + */ |