dsp - CMSIS DSP Library from CMSIS 2.0. See http://www.…

Users » simon » Code » dsp

CMSIS DSP Library from CMSIS 2.0. See http://www.onarm.com/cmsis/ for full details

Dependents: K22F_DSP_Matrix_least_square BNO055-ELEC3810 1BNO055 ECE4180Project--Slave2 ... more

src/Cortex-M4-M3/FilteringFunctions/arm_fir_sparse_q15.c@0:1014af42efd9, 2011-03-10 (annotated)

Committer:: simon
Date:: Thu Mar 10 15:07:50 2011 +0000
Revision:: 0:1014af42efd9

Who changed what in which revision?

User	Revision	Line number	New contents of line
simon	0:1014af42efd9	1	/* ----------------------------------------------------------------------
simon	0:1014af42efd9	2	* Copyright (C) 2010 ARM Limited. All rights reserved.
simon	0:1014af42efd9	3	*
simon	0:1014af42efd9	4	* $Date: 29. November 2010
simon	0:1014af42efd9	5	* $Revision: V1.0.3
simon	0:1014af42efd9	6	*
simon	0:1014af42efd9	7	* Project: CMSIS DSP Library
simon	0:1014af42efd9	8	* Title: arm_fir_sparse_q15.c
simon	0:1014af42efd9	9	*
simon	0:1014af42efd9	10	* Description: Q15 sparse FIR filter processing function.
simon	0:1014af42efd9	11	*
simon	0:1014af42efd9	12	* Target Processor: Cortex-M4/Cortex-M3
simon	0:1014af42efd9	13	*
simon	0:1014af42efd9	14	* Version 1.0.3 2010/11/29
simon	0:1014af42efd9	15	* Re-organized the CMSIS folders and updated documentation.
simon	0:1014af42efd9	16	*
simon	0:1014af42efd9	17	* Version 1.0.2 2010/11/11
simon	0:1014af42efd9	18	* Documentation updated.
simon	0:1014af42efd9	19	*
simon	0:1014af42efd9	20	* Version 1.0.1 2010/10/05
simon	0:1014af42efd9	21	* Production release and review comments incorporated.
simon	0:1014af42efd9	22	*
simon	0:1014af42efd9	23	* Version 1.0.0 2010/09/20
simon	0:1014af42efd9	24	* Production release and review comments incorporated
simon	0:1014af42efd9	25	*
simon	0:1014af42efd9	26	* Version 0.0.7 2010/06/10
simon	0:1014af42efd9	27	* Misra-C changes done
simon	0:1014af42efd9	28	* ------------------------------------------------------------------- */
simon	0:1014af42efd9	29	#include "arm_math.h"
simon	0:1014af42efd9	30
simon	0:1014af42efd9	31	/**
simon	0:1014af42efd9	32	* @addtogroup FIR_Sparse
simon	0:1014af42efd9	33	* @{
simon	0:1014af42efd9	34	*/
simon	0:1014af42efd9	35
simon	0:1014af42efd9	36	/**
simon	0:1014af42efd9	37	* @brief Processing function for the Q15 sparse FIR filter.
simon	0:1014af42efd9	38	* @param[in] *S points to an instance of the Q15 sparse FIR structure.
simon	0:1014af42efd9	39	* @param[in] *pSrc points to the block of input data.
simon	0:1014af42efd9	40	* @param[out] *pDst points to the block of output data
simon	0:1014af42efd9	41	* @param[in] *pScratchIn points to a temporary buffer of size blockSize.
simon	0:1014af42efd9	42	* @param[in] *pScratchOut points to a temporary buffer of size blockSize.
simon	0:1014af42efd9	43	* @param[in] blockSize number of input samples to process per call.
simon	0:1014af42efd9	44	* @return none.
simon	0:1014af42efd9	45	*
simon	0:1014af42efd9	46	* <b>Scaling and Overflow Behavior:</b>
simon	0:1014af42efd9	47	* \par
simon	0:1014af42efd9	48	* The function is implemented using an internal 32-bit accumulator.
simon	0:1014af42efd9	49	* The 1.15 x 1.15 multiplications yield a 2.30 result and these are added to a 2.30 accumulator.
simon	0:1014af42efd9	50	* Thus the full precision of the multiplications is maintained but there is only a single guard bit in the accumulator.
simon	0:1014af42efd9	51	* If the accumulator result overflows it will wrap around rather than saturate.
simon	0:1014af42efd9	52	* After all multiply-accumulates are performed, the 2.30 accumulator is truncated to 2.15 format and then saturated to 1.15 format.
simon	0:1014af42efd9	53	* In order to avoid overflows the input signal or coefficients must be scaled down by log2(numTaps) bits.
simon	0:1014af42efd9	54	*/
simon	0:1014af42efd9	55
simon	0:1014af42efd9	56
simon	0:1014af42efd9	57	void arm_fir_sparse_q15(
simon	0:1014af42efd9	58	arm_fir_sparse_instance_q15 * S,
simon	0:1014af42efd9	59	q15_t * pSrc,
simon	0:1014af42efd9	60	q15_t * pDst,
simon	0:1014af42efd9	61	q15_t * pScratchIn,
simon	0:1014af42efd9	62	q31_t * pScratchOut,
simon	0:1014af42efd9	63	uint32_t blockSize)
simon	0:1014af42efd9	64	{
simon	0:1014af42efd9	65
simon	0:1014af42efd9	66	q15_t pState = S->pState; / State pointer */
simon	0:1014af42efd9	67	q15_t pIn = (q15_t ) pSrc; /* Working pointer for input */
simon	0:1014af42efd9	68	q15_t pOut = pDst; / Working pointer for output */
simon	0:1014af42efd9	69	q15_t pCoeffs = S->pCoeffs; / Coefficient pointer */
simon	0:1014af42efd9	70	q15_t px; / Temporary pointers for scratch buffer */
simon	0:1014af42efd9	71	q15_t pb = pScratchIn; / Temporary pointers for scratch buffer */
simon	0:1014af42efd9	72	q15_t py = pState; / Temporary pointers for state buffer */
simon	0:1014af42efd9	73	int32_t pTapDelay = S->pTapDelay; / Pointer to the array containing offset of the non-zero tap values. */
simon	0:1014af42efd9	74	uint32_t delaySize = S->maxDelay + blockSize; /* state length */
simon	0:1014af42efd9	75	uint16_t numTaps = S->numTaps; /* Filter order */
simon	0:1014af42efd9	76	int32_t readIndex; /* Read index of the state buffer */
simon	0:1014af42efd9	77	uint32_t tapCnt, blkCnt; /* loop counters */
simon	0:1014af42efd9	78	q15_t coeff = pCoeffs++; / Read the first coefficient value */
simon	0:1014af42efd9	79	q31_t pScr2 = pScratchOut; / Working pointer for pScratchOut */
simon	0:1014af42efd9	80	q31_t in1, in2; /* Temporary variables */
simon	0:1014af42efd9	81
simon	0:1014af42efd9	82
simon	0:1014af42efd9	83
simon	0:1014af42efd9	84	/* BlockSize of Input samples are copied into the state buffer */
simon	0:1014af42efd9	85	/* StateIndex points to the starting position to write in the state buffer */
simon	0:1014af42efd9	86	arm_circularWrite_q15(py, delaySize, &S->stateIndex, 1, pIn, 1, blockSize);
simon	0:1014af42efd9	87
simon	0:1014af42efd9	88	/* Loop over the number of taps. */
simon	0:1014af42efd9	89	tapCnt = numTaps;
simon	0:1014af42efd9	90
simon	0:1014af42efd9	91	/* Read Index, from where the state buffer should be read, is calculated. */
simon	0:1014af42efd9	92	readIndex = (S->stateIndex - blockSize) - *pTapDelay++;
simon	0:1014af42efd9	93
simon	0:1014af42efd9	94	/* Wraparound of readIndex */
simon	0:1014af42efd9	95	if(readIndex < 0)
simon	0:1014af42efd9	96	{
simon	0:1014af42efd9	97	readIndex += (int32_t) delaySize;
simon	0:1014af42efd9	98	}
simon	0:1014af42efd9	99
simon	0:1014af42efd9	100	/* Working pointer for state buffer is updated */
simon	0:1014af42efd9	101	py = pState;
simon	0:1014af42efd9	102
simon	0:1014af42efd9	103	/* blockSize samples are read from the state buffer */
simon	0:1014af42efd9	104	arm_circularRead_q15(py, delaySize, &readIndex, 1,
simon	0:1014af42efd9	105	pb, pb, blockSize, 1, blockSize);
simon	0:1014af42efd9	106
simon	0:1014af42efd9	107	/* Working pointer for the scratch buffer of state values */
simon	0:1014af42efd9	108	px = pb;
simon	0:1014af42efd9	109
simon	0:1014af42efd9	110	/* Working pointer for scratch buffer of output values */
simon	0:1014af42efd9	111	pScratchOut = pScr2;
simon	0:1014af42efd9	112
simon	0:1014af42efd9	113	/* Loop over the blockSize. Unroll by a factor of 4.
simon	0:1014af42efd9	114	* Compute 4 multiplications at a time. */
simon	0:1014af42efd9	115	blkCnt = blockSize >> 2;
simon	0:1014af42efd9	116
simon	0:1014af42efd9	117	while(blkCnt > 0u)
simon	0:1014af42efd9	118	{
simon	0:1014af42efd9	119	/* Perform multiplication and store in the scratch buffer */
simon	0:1014af42efd9	120	pScratchOut++ = ((q31_t) px++ * coeff);
simon	0:1014af42efd9	121	pScratchOut++ = ((q31_t) px++ * coeff);
simon	0:1014af42efd9	122	pScratchOut++ = ((q31_t) px++ * coeff);
simon	0:1014af42efd9	123	pScratchOut++ = ((q31_t) px++ * coeff);
simon	0:1014af42efd9	124
simon	0:1014af42efd9	125	/* Decrement the loop counter */
simon	0:1014af42efd9	126	blkCnt--;
simon	0:1014af42efd9	127	}
simon	0:1014af42efd9	128
simon	0:1014af42efd9	129	/* If the blockSize is not a multiple of 4,
simon	0:1014af42efd9	130	* compute the remaining samples */
simon	0:1014af42efd9	131	blkCnt = blockSize % 0x4u;
simon	0:1014af42efd9	132
simon	0:1014af42efd9	133	while(blkCnt > 0u)
simon	0:1014af42efd9	134	{
simon	0:1014af42efd9	135	/* Perform multiplication and store in the scratch buffer */
simon	0:1014af42efd9	136	pScratchOut++ = ((q31_t) px++ * coeff);
simon	0:1014af42efd9	137
simon	0:1014af42efd9	138	/* Decrement the loop counter */
simon	0:1014af42efd9	139	blkCnt--;
simon	0:1014af42efd9	140	}
simon	0:1014af42efd9	141
simon	0:1014af42efd9	142	/* Load the coefficient value and
simon	0:1014af42efd9	143	* increment the coefficient buffer for the next set of state values */
simon	0:1014af42efd9	144	coeff = *pCoeffs++;
simon	0:1014af42efd9	145
simon	0:1014af42efd9	146	/* Read Index, from where the state buffer should be read, is calculated. */
simon	0:1014af42efd9	147	readIndex = (S->stateIndex - blockSize) - *pTapDelay++;
simon	0:1014af42efd9	148
simon	0:1014af42efd9	149	/* Wraparound of readIndex */
simon	0:1014af42efd9	150	if(readIndex < 0)
simon	0:1014af42efd9	151	{
simon	0:1014af42efd9	152	readIndex += (int32_t) delaySize;
simon	0:1014af42efd9	153	}
simon	0:1014af42efd9	154
simon	0:1014af42efd9	155	/* Loop over the number of taps. */
simon	0:1014af42efd9	156	tapCnt = (uint32_t) numTaps - 1u;
simon	0:1014af42efd9	157
simon	0:1014af42efd9	158	while(tapCnt > 0u)
simon	0:1014af42efd9	159	{
simon	0:1014af42efd9	160	/* Working pointer for state buffer is updated */
simon	0:1014af42efd9	161	py = pState;
simon	0:1014af42efd9	162
simon	0:1014af42efd9	163	/* blockSize samples are read from the state buffer */
simon	0:1014af42efd9	164	arm_circularRead_q15(py, delaySize, &readIndex, 1,
simon	0:1014af42efd9	165	pb, pb, blockSize, 1, blockSize);
simon	0:1014af42efd9	166
simon	0:1014af42efd9	167	/* Working pointer for the scratch buffer of state values */
simon	0:1014af42efd9	168	px = pb;
simon	0:1014af42efd9	169
simon	0:1014af42efd9	170	/* Working pointer for scratch buffer of output values */
simon	0:1014af42efd9	171	pScratchOut = pScr2;
simon	0:1014af42efd9	172
simon	0:1014af42efd9	173	/* Loop over the blockSize. Unroll by a factor of 4.
simon	0:1014af42efd9	174	* Compute 4 MACS at a time. */
simon	0:1014af42efd9	175	blkCnt = blockSize >> 2;
simon	0:1014af42efd9	176
simon	0:1014af42efd9	177	while(blkCnt > 0u)
simon	0:1014af42efd9	178	{
simon	0:1014af42efd9	179	/* Perform Multiply-Accumulate */
simon	0:1014af42efd9	180	pScratchOut++ += (q31_t) px++ * coeff;
simon	0:1014af42efd9	181	pScratchOut++ += (q31_t) px++ * coeff;
simon	0:1014af42efd9	182	pScratchOut++ += (q31_t) px++ * coeff;
simon	0:1014af42efd9	183	pScratchOut++ += (q31_t) px++ * coeff;
simon	0:1014af42efd9	184
simon	0:1014af42efd9	185	/* Decrement the loop counter */
simon	0:1014af42efd9	186	blkCnt--;
simon	0:1014af42efd9	187	}
simon	0:1014af42efd9	188
simon	0:1014af42efd9	189	/* If the blockSize is not a multiple of 4,
simon	0:1014af42efd9	190	* compute the remaining samples */
simon	0:1014af42efd9	191	blkCnt = blockSize % 0x4u;
simon	0:1014af42efd9	192
simon	0:1014af42efd9	193	while(blkCnt > 0u)
simon	0:1014af42efd9	194	{
simon	0:1014af42efd9	195	/* Perform Multiply-Accumulate */
simon	0:1014af42efd9	196	pScratchOut++ += (q31_t) px++ * coeff;
simon	0:1014af42efd9	197
simon	0:1014af42efd9	198	/* Decrement the loop counter */
simon	0:1014af42efd9	199	blkCnt--;
simon	0:1014af42efd9	200	}
simon	0:1014af42efd9	201
simon	0:1014af42efd9	202	/* Load the coefficient value and
simon	0:1014af42efd9	203	* increment the coefficient buffer for the next set of state values */
simon	0:1014af42efd9	204	coeff = *pCoeffs++;
simon	0:1014af42efd9	205
simon	0:1014af42efd9	206	/* Read Index, from where the state buffer should be read, is calculated. */
simon	0:1014af42efd9	207	readIndex = (S->stateIndex - blockSize) - *pTapDelay++;
simon	0:1014af42efd9	208
simon	0:1014af42efd9	209	/* Wraparound of readIndex */
simon	0:1014af42efd9	210	if(readIndex < 0)
simon	0:1014af42efd9	211	{
simon	0:1014af42efd9	212	readIndex += (int32_t) delaySize;
simon	0:1014af42efd9	213	}
simon	0:1014af42efd9	214
simon	0:1014af42efd9	215	/* Decrement the tap loop counter */
simon	0:1014af42efd9	216	tapCnt--;
simon	0:1014af42efd9	217	}
simon	0:1014af42efd9	218
simon	0:1014af42efd9	219	/* All the output values are in pScratchOut buffer.
simon	0:1014af42efd9	220	Convert them into 1.15 format, saturate and store in the destination buffer. */
simon	0:1014af42efd9	221	/* Loop over the blockSize. */
simon	0:1014af42efd9	222	blkCnt = blockSize >> 2;
simon	0:1014af42efd9	223
simon	0:1014af42efd9	224	while(blkCnt > 0u)
simon	0:1014af42efd9	225	{
simon	0:1014af42efd9	226	in1 = *pScr2++;
simon	0:1014af42efd9	227	in2 = *pScr2++;
simon	0:1014af42efd9	228	*__SIMD32(pOut)++ =
simon	0:1014af42efd9	229	__PKHBT((q15_t) __SSAT(in1 >> 15, 16), (q15_t) __SSAT(in2 >> 15, 16),
simon	0:1014af42efd9	230	16);
simon	0:1014af42efd9	231
simon	0:1014af42efd9	232	in1 = *pScr2++;
simon	0:1014af42efd9	233	in2 = *pScr2++;
simon	0:1014af42efd9	234	*__SIMD32(pOut)++ =
simon	0:1014af42efd9	235	__PKHBT((q15_t) __SSAT(in1 >> 15, 16), (q15_t) __SSAT(in2 >> 15, 16),
simon	0:1014af42efd9	236	16);
simon	0:1014af42efd9	237
simon	0:1014af42efd9	238	blkCnt--;
simon	0:1014af42efd9	239
simon	0:1014af42efd9	240	}
simon	0:1014af42efd9	241
simon	0:1014af42efd9	242	/* If the blockSize is not a multiple of 4,
simon	0:1014af42efd9	243	remaining samples are processed in the below loop */
simon	0:1014af42efd9	244	blkCnt = blockSize % 0x4u;
simon	0:1014af42efd9	245
simon	0:1014af42efd9	246	while(blkCnt > 0u)
simon	0:1014af42efd9	247	{
simon	0:1014af42efd9	248	pOut++ = (q15_t) __SSAT(pScr2++ >> 15, 16);
simon	0:1014af42efd9	249	blkCnt--;
simon	0:1014af42efd9	250	}
simon	0:1014af42efd9	251	}
simon	0:1014af42efd9	252
simon	0:1014af42efd9	253	/**
simon	0:1014af42efd9	254	* @} end of FIR_Sparse group
simon	0:1014af42efd9	255	*/

Repository toolbox

Export to desktop IDE

Repository details

Type:	Library
Created:	10 Mar 2011
Imports:	907
Forks:	1
Commits:	3
Dependents:	5
Dependencies:	0
Followers:	35

src/Cortex-M4-M3/FilteringFunctions/arm_fir_sparse_q15.c@0:1014af42efd9, 2011-03-10 (annotated)

Who changed what in which revision?

Repository toolbox

Repository details

Important Information for this Arm website

Access Warning