Welcome to mirror list, hosted at ThFree Co, Russian Federation.

github.com/Flipper-Zero/STM32CubeWB.git - Unnamed repository; edit this file 'description' to name the repository.
summaryrefslogtreecommitdiff
diff options
context:
space:
mode:
Diffstat (limited to 'Drivers/CMSIS/DSP/Source/FilteringFunctions/arm_fir_sparse_q7.c')
-rw-r--r--Drivers/CMSIS/DSP/Source/FilteringFunctions/arm_fir_sparse_q7.c418
1 files changed, 145 insertions, 273 deletions
diff --git a/Drivers/CMSIS/DSP/Source/FilteringFunctions/arm_fir_sparse_q7.c b/Drivers/CMSIS/DSP/Source/FilteringFunctions/arm_fir_sparse_q7.c
index c1b4ce340..7a2b57f18 100644
--- a/Drivers/CMSIS/DSP/Source/FilteringFunctions/arm_fir_sparse_q7.c
+++ b/Drivers/CMSIS/DSP/Source/FilteringFunctions/arm_fir_sparse_q7.c
@@ -3,13 +3,13 @@
* Title: arm_fir_sparse_q7.c
* Description: Q7 sparse FIR filter processing function
*
- * $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
*
@@ -28,78 +28,70 @@
#include "arm_math.h"
-
/**
- * @ingroup groupFilters
+ @ingroup groupFilters
*/
/**
- * @addtogroup FIR_Sparse
- * @{
+ @addtogroup FIR_Sparse
+ @{
*/
-
/**
- * @brief Processing function for the Q7 sparse FIR filter.
- * @param[in] *S points to an instance of the Q7 sparse FIR structure.
- * @param[in] *pSrc points to the block of input data.
- * @param[out] *pDst points to the block of output data
- * @param[in] *pScratchIn points to a temporary buffer of size blockSize.
- * @param[in] *pScratchOut points to a temporary buffer of size blockSize.
- * @param[in] blockSize number of input samples to process per call.
- * @return none.
- *
- * <b>Scaling and Overflow Behavior:</b>
- * \par
- * The function is implemented using a 32-bit internal accumulator.
- * Both coefficients and state variables are represented in 1.7 format and multiplications yield a 2.14 result.
- * The 2.14 intermediate results are accumulated in a 32-bit accumulator in 18.14 format.
- * There is no risk of internal overflow with this approach and the full precision of intermediate multiplications is preserved.
- * The accumulator is then converted to 18.7 format by discarding the low 7 bits.
- * Finally, the result is truncated to 1.7 format.
+ @brief Processing function for the Q7 sparse FIR filter.
+ @param[in] S points to an instance of the Q7 sparse FIR structure
+ @param[in] pSrc points to the block of input data
+ @param[out] pDst points to the block of output data
+ @param[in] pScratchIn points to a temporary buffer of size blockSize
+ @param[in] pScratchOut points to a temporary buffer of size blockSize
+ @param[in] blockSize number of input samples to process
+ @return none
+
+ @par Scaling and Overflow Behavior
+ The function is implemented using a 32-bit internal accumulator.
+ Both coefficients and state variables are represented in 1.7 format and multiplications yield a 2.14 result.
+ The 2.14 intermediate results are accumulated in a 32-bit accumulator in 18.14 format.
+ There is no risk of internal overflow with this approach and the full precision of intermediate multiplications is preserved.
+ The accumulator is then converted to 18.7 format by discarding the low 7 bits.
+ Finally, the result is truncated to 1.7 format.
*/
void arm_fir_sparse_q7(
- arm_fir_sparse_instance_q7 * S,
- q7_t * pSrc,
- q7_t * pDst,
- q7_t * pScratchIn,
- q31_t * pScratchOut,
- uint32_t blockSize)
+ arm_fir_sparse_instance_q7 * S,
+ const q7_t * pSrc,
+ q7_t * pDst,
+ q7_t * pScratchIn,
+ q31_t * pScratchOut,
+ uint32_t blockSize)
{
-
- q7_t *pState = S->pState; /* State pointer */
- q7_t *pCoeffs = S->pCoeffs; /* Coefficient pointer */
- q7_t *px; /* Scratch buffer pointer */
- q7_t *py = pState; /* Temporary pointers for state buffer */
- q7_t *pb = pScratchIn; /* Temporary pointers for scratch buffer */
- q7_t *pOut = pDst; /* Destination pointer */
- int32_t *pTapDelay = S->pTapDelay; /* Pointer to the array containing offset of the non-zero tap values. */
- uint32_t delaySize = S->maxDelay + blockSize; /* state length */
- uint16_t numTaps = S->numTaps; /* Filter order */
- int32_t readIndex; /* Read index of the state buffer */
- uint32_t tapCnt, blkCnt; /* loop counters */
- q7_t coeff = *pCoeffs++; /* Read the coefficient value */
- q31_t *pScr2 = pScratchOut; /* Working pointer for scratch buffer of output values */
- q31_t in;
-
-
-#if defined (ARM_MATH_DSP)
-
- /* Run the below code for Cortex-M4 and Cortex-M3 */
-
- q7_t in1, in2, in3, in4;
+ q7_t *pState = S->pState; /* State pointer */
+ const q7_t *pCoeffs = S->pCoeffs; /* Coefficient pointer */
+ q7_t *px; /* Scratch buffer pointer */
+ q7_t *py = pState; /* Temporary pointers for state buffer */
+ q7_t *pb = pScratchIn; /* Temporary pointers for scratch buffer */
+ q7_t *pOut = pDst; /* Destination pointer */
+ int32_t *pTapDelay = S->pTapDelay; /* Pointer to the array containing offset of the non-zero tap values. */
+ uint32_t delaySize = S->maxDelay + blockSize; /* state length */
+ uint16_t numTaps = S->numTaps; /* Number of filter coefficients in the filter */
+ int32_t readIndex; /* Read index of the state buffer */
+ uint32_t tapCnt, blkCnt; /* loop counters */
+ q31_t *pScr2 = pScratchOut; /* Working pointer for scratch buffer of output values */
+ q31_t in;
+ q7_t coeff = *pCoeffs++; /* Read the coefficient value */
+
+#if defined (ARM_MATH_LOOPUNROLL)
+ q7_t in1, in2, in3, in4;
+#endif
/* BlockSize of Input samples are copied into the state buffer */
/* StateIndex points to the starting position to write in the state buffer */
- arm_circularWrite_q7(py, (int32_t) delaySize, &S->stateIndex, 1, pSrc, 1,
- blockSize);
+ arm_circularWrite_q7(py, (int32_t) delaySize, &S->stateIndex, 1, pSrc, 1, blockSize);
/* Loop over the number of taps. */
tapCnt = numTaps;
/* Read Index, from where the state buffer should be read, is calculated. */
- readIndex = ((int32_t) S->stateIndex - (int32_t) blockSize) - *pTapDelay++;
+ readIndex = (int32_t) (S->stateIndex - blockSize) - *pTapDelay++;
/* Wraparound of readIndex */
if (readIndex < 0)
@@ -111,8 +103,8 @@ void arm_fir_sparse_q7(
py = pState;
/* blockSize samples are read from the state buffer */
- arm_circularRead_q7(py, (int32_t) delaySize, &readIndex, 1, pb, pb,
- (int32_t) blockSize, 1, blockSize);
+ arm_circularRead_q7(py, (int32_t) delaySize, &readIndex, 1,
+ pb, pb, (int32_t) blockSize, 1, blockSize);
/* Working pointer for the scratch buffer of state values */
px = pb;
@@ -120,32 +112,40 @@ void arm_fir_sparse_q7(
/* Working pointer for scratch buffer of output values */
pScratchOut = pScr2;
- /* Loop over the blockSize. Unroll by a factor of 4.
- * Compute 4 multiplications at a time. */
- blkCnt = blockSize >> 2;
+
+#if defined (ARM_MATH_LOOPUNROLL)
+
+ /* Loop unrolling: Compute 4 outputs at a time. */
+ blkCnt = blockSize >> 2U;
while (blkCnt > 0U)
{
/* Perform multiplication and store in the scratch buffer */
- *pScratchOut++ = ((q31_t) * px++ * coeff);
- *pScratchOut++ = ((q31_t) * px++ * coeff);
- *pScratchOut++ = ((q31_t) * px++ * coeff);
- *pScratchOut++ = ((q31_t) * px++ * coeff);
+ *pScratchOut++ = ((q31_t) *px++ * coeff);
+ *pScratchOut++ = ((q31_t) *px++ * coeff);
+ *pScratchOut++ = ((q31_t) *px++ * coeff);
+ *pScratchOut++ = ((q31_t) *px++ * coeff);
- /* Decrement the loop counter */
+ /* Decrement loop counter */
blkCnt--;
}
- /* If the blockSize is not a multiple of 4,
- * compute the remaining samples */
+ /* Loop unrolling: Compute remaining outputs */
blkCnt = blockSize % 0x4U;
+#else
+
+ /* Initialize blkCnt with number of samples */
+ blkCnt = blockSize;
+
+#endif /* #if defined (ARM_MATH_LOOPUNROLL) */
+
while (blkCnt > 0U)
{
- /* Perform multiplication and store in the scratch buffer */
- *pScratchOut++ = ((q31_t) * px++ * coeff);
+ /* Perform Multiplication and store in the scratch buffer */
+ *pScratchOut++ = ((q31_t) *px++ * coeff);
- /* Decrement the loop counter */
+ /* Decrement loop counter */
blkCnt--;
}
@@ -154,7 +154,7 @@ void arm_fir_sparse_q7(
coeff = *pCoeffs++;
/* Read Index, from where the state buffer should be read, is calculated. */
- readIndex = ((int32_t) S->stateIndex - (int32_t) blockSize) - *pTapDelay++;
+ readIndex = (int32_t) (S->stateIndex - blockSize) - *pTapDelay++;
/* Wraparound of readIndex */
if (readIndex < 0)
@@ -171,8 +171,8 @@ void arm_fir_sparse_q7(
py = pState;
/* blockSize samples are read from the state buffer */
- arm_circularRead_q7(py, (int32_t) delaySize, &readIndex, 1, pb, pb,
- (int32_t) blockSize, 1, blockSize);
+ arm_circularRead_q7(py, (int32_t) delaySize, &readIndex, 1,
+ pb, pb, (int32_t) blockSize, 1, blockSize);
/* Working pointer for the scratch buffer of state values */
px = pb;
@@ -180,9 +180,11 @@ void arm_fir_sparse_q7(
/* Working pointer for scratch buffer of output values */
pScratchOut = pScr2;
- /* Loop over the blockSize. Unroll by a factor of 4.
- * Compute 4 MACS at a time. */
- blkCnt = blockSize >> 2;
+
+#if defined (ARM_MATH_LOOPUNROLL)
+
+ /* Loop unrolling: Compute 4 outputs at a time. */
+ blkCnt = blockSize >> 2U;
while (blkCnt > 0U)
{
@@ -196,21 +198,27 @@ void arm_fir_sparse_q7(
in = *pScratchOut + ((q31_t) * px++ * coeff);
*pScratchOut++ = in;
- /* Decrement the loop counter */
+ /* Decrement loop counter */
blkCnt--;
}
- /* If the blockSize is not a multiple of 4,
- * compute the remaining samples */
+ /* Loop unrolling: Compute remaining outputs */
blkCnt = blockSize % 0x4U;
+#else
+
+ /* Initialize blkCnt with number of samples */
+ blkCnt = blockSize;
+
+#endif /* #if defined (ARM_MATH_LOOPUNROLL) */
+
while (blkCnt > 0U)
{
/* Perform Multiply-Accumulate */
- in = *pScratchOut + ((q31_t) * px++ * coeff);
+ in = *pScratchOut + ((q31_t) *px++ * coeff);
*pScratchOut++ = in;
- /* Decrement the loop counter */
+ /* Decrement loop counter */
blkCnt--;
}
@@ -219,8 +227,7 @@ void arm_fir_sparse_q7(
coeff = *pCoeffs++;
/* Read Index, from where the state buffer should be read, is calculated. */
- readIndex = ((int32_t) S->stateIndex -
- (int32_t) blockSize) - *pTapDelay++;
+ readIndex = (int32_t) (S->stateIndex - blockSize) - *pTapDelay++;
/* Wraparound of readIndex */
if (readIndex < 0)
@@ -228,242 +235,107 @@ void arm_fir_sparse_q7(
readIndex += (int32_t) delaySize;
}
- /* Decrement the tap loop counter */
+ /* Decrement loop counter */
tapCnt--;
}
- /* Compute last tap without the final read of pTapDelay */
-
- /* Working pointer for state buffer is updated */
- py = pState;
-
- /* blockSize samples are read from the state buffer */
- arm_circularRead_q7(py, (int32_t) delaySize, &readIndex, 1, pb, pb,
- (int32_t) blockSize, 1, blockSize);
-
- /* Working pointer for the scratch buffer of state values */
- px = pb;
-
- /* Working pointer for scratch buffer of output values */
- pScratchOut = pScr2;
+ /* Compute last tap without the final read of pTapDelay */
- /* Loop over the blockSize. Unroll by a factor of 4.
- * Compute 4 MACS at a time. */
- blkCnt = blockSize >> 2;
+ /* Working pointer for state buffer is updated */
+ py = pState;
- while (blkCnt > 0U)
- {
- /* Perform Multiply-Accumulate */
- in = *pScratchOut + ((q31_t) * px++ * coeff);
- *pScratchOut++ = in;
- in = *pScratchOut + ((q31_t) * px++ * coeff);
- *pScratchOut++ = in;
- in = *pScratchOut + ((q31_t) * px++ * coeff);
- *pScratchOut++ = in;
- in = *pScratchOut + ((q31_t) * px++ * coeff);
- *pScratchOut++ = in;
+ /* blockSize samples are read from the state buffer */
+ arm_circularRead_q7(py, (int32_t) delaySize, &readIndex, 1,
+ pb, pb, (int32_t) blockSize, 1, blockSize);
- /* Decrement the loop counter */
- blkCnt--;
- }
+ /* Working pointer for the scratch buffer of state values */
+ px = pb;
- /* If the blockSize is not a multiple of 4,
- * compute the remaining samples */
- blkCnt = blockSize % 0x4U;
+ /* Working pointer for scratch buffer of output values */
+ pScratchOut = pScr2;
- while (blkCnt > 0U)
- {
- /* Perform Multiply-Accumulate */
- in = *pScratchOut + ((q31_t) * px++ * coeff);
- *pScratchOut++ = in;
- /* Decrement the loop counter */
- blkCnt--;
- }
+#if defined (ARM_MATH_LOOPUNROLL)
- /* All the output values are in pScratchOut buffer.
- Convert them into 1.15 format, saturate and store in the destination buffer. */
- /* Loop over the blockSize. */
- blkCnt = blockSize >> 2;
+ /* Loop unrolling: Compute 4 outputs at a time. */
+ blkCnt = blockSize >> 2U;
while (blkCnt > 0U)
{
- in1 = (q7_t) __SSAT(*pScr2++ >> 7, 8);
- in2 = (q7_t) __SSAT(*pScr2++ >> 7, 8);
- in3 = (q7_t) __SSAT(*pScr2++ >> 7, 8);
- in4 = (q7_t) __SSAT(*pScr2++ >> 7, 8);
-
- *__SIMD32(pOut)++ = __PACKq7(in1, in2, in3, in4);
-
- /* Decrement the blockSize loop counter */
+ /* Perform Multiply-Accumulate */
+ in = *pScratchOut + ((q31_t) *px++ * coeff);
+ *pScratchOut++ = in;
+ in = *pScratchOut + ((q31_t) *px++ * coeff);
+ *pScratchOut++ = in;
+ in = *pScratchOut + ((q31_t) *px++ * coeff);
+ *pScratchOut++ = in;
+ in = *pScratchOut + ((q31_t) *px++ * coeff);
+ *pScratchOut++ = in;
+
+ /* Decrement loop counter */
blkCnt--;
}
- /* If the blockSize is not a multiple of 4,
- remaining samples are processed in the below loop */
+ /* Loop unrolling: Compute remaining outputs */
blkCnt = blockSize % 0x4U;
- while (blkCnt > 0U)
- {
- *pOut++ = (q7_t) __SSAT(*pScr2++ >> 7, 8);
-
- /* Decrement the blockSize loop counter */
- blkCnt--;
- }
-
#else
- /* Run the below code for Cortex-M0 */
-
- /* BlockSize of Input samples are copied into the state buffer */
- /* StateIndex points to the starting position to write in the state buffer */
- arm_circularWrite_q7(py, (int32_t) delaySize, &S->stateIndex, 1, pSrc, 1,
- blockSize);
-
- /* Loop over the number of taps. */
- tapCnt = numTaps;
-
- /* Read Index, from where the state buffer should be read, is calculated. */
- readIndex = ((int32_t) S->stateIndex - (int32_t) blockSize) - *pTapDelay++;
-
- /* Wraparound of readIndex */
- if (readIndex < 0)
- {
- readIndex += (int32_t) delaySize;
- }
-
- /* Working pointer for state buffer is updated */
- py = pState;
-
- /* blockSize samples are read from the state buffer */
- arm_circularRead_q7(py, (int32_t) delaySize, &readIndex, 1, pb, pb,
- (int32_t) blockSize, 1, blockSize);
-
- /* Working pointer for the scratch buffer of state values */
- px = pb;
-
- /* Working pointer for scratch buffer of output values */
- pScratchOut = pScr2;
-
- /* Loop over the blockSize */
+ /* Initialize blkCnt with number of samples */
blkCnt = blockSize;
+#endif /* #if defined (ARM_MATH_LOOPUNROLL) */
+
while (blkCnt > 0U)
{
- /* Perform multiplication and store in the scratch buffer */
- *pScratchOut++ = ((q31_t) * px++ * coeff);
+ /* Perform Multiply-Accumulate */
+ in = *pScratchOut + ((q31_t) *px++ * coeff);
+ *pScratchOut++ = in;
- /* Decrement the loop counter */
+ /* Decrement loop counter */
blkCnt--;
}
- /* Load the coefficient value and
- * increment the coefficient buffer for the next set of state values */
- coeff = *pCoeffs++;
-
- /* Read Index, from where the state buffer should be read, is calculated. */
- readIndex = ((int32_t) S->stateIndex - (int32_t) blockSize) - *pTapDelay++;
-
- /* Wraparound of readIndex */
- if (readIndex < 0)
- {
- readIndex += (int32_t) delaySize;
- }
+ /* All the output values are in pScratchOut buffer.
+ Convert them into 1.15 format, saturate and store in the destination buffer. */
+#if defined (ARM_MATH_LOOPUNROLL)
- /* Loop over the number of taps. */
- tapCnt = (uint32_t) numTaps - 2U;
+ /* Loop unrolling: Compute 4 outputs at a time. */
+ blkCnt = blockSize >> 2U;
- while (tapCnt > 0U)
+ while (blkCnt > 0U)
{
- /* Working pointer for state buffer is updated */
- py = pState;
-
- /* blockSize samples are read from the state buffer */
- arm_circularRead_q7(py, (int32_t) delaySize, &readIndex, 1, pb, pb,
- (int32_t) blockSize, 1, blockSize);
-
- /* Working pointer for the scratch buffer of state values */
- px = pb;
-
- /* Working pointer for scratch buffer of output values */
- pScratchOut = pScr2;
-
- /* Loop over the blockSize */
- blkCnt = blockSize;
-
- while (blkCnt > 0U)
- {
- /* Perform Multiply-Accumulate */
- in = *pScratchOut + ((q31_t) * px++ * coeff);
- *pScratchOut++ = in;
-
- /* Decrement the loop counter */
- blkCnt--;
- }
-
- /* Load the coefficient value and
- * increment the coefficient buffer for the next set of state values */
- coeff = *pCoeffs++;
-
- /* Read Index, from where the state buffer should be read, is calculated. */
- readIndex =
- ((int32_t) S->stateIndex - (int32_t) blockSize) - *pTapDelay++;
+ in1 = (q7_t) __SSAT(*pScr2++ >> 7, 8);
+ in2 = (q7_t) __SSAT(*pScr2++ >> 7, 8);
+ in3 = (q7_t) __SSAT(*pScr2++ >> 7, 8);
+ in4 = (q7_t) __SSAT(*pScr2++ >> 7, 8);
- /* Wraparound of readIndex */
- if (readIndex < 0)
- {
- readIndex += (int32_t) delaySize;
- }
+ write_q7x4_ia (&pOut, __PACKq7(in1, in2, in3, in4));
- /* Decrement the tap loop counter */
- tapCnt--;
+ /* Decrement loop counter */
+ blkCnt--;
}
- /* Compute last tap without the final read of pTapDelay */
-
- /* Working pointer for state buffer is updated */
- py = pState;
-
- /* blockSize samples are read from the state buffer */
- arm_circularRead_q7(py, (int32_t) delaySize, &readIndex, 1, pb, pb,
- (int32_t) blockSize, 1, blockSize);
-
- /* Working pointer for the scratch buffer of state values */
- px = pb;
-
- /* Working pointer for scratch buffer of output values */
- pScratchOut = pScr2;
-
- /* Loop over the blockSize */
- blkCnt = blockSize;
-
- while (blkCnt > 0U)
- {
- /* Perform Multiply-Accumulate */
- in = *pScratchOut + ((q31_t) * px++ * coeff);
- *pScratchOut++ = in;
+ /* Loop unrolling: Compute remaining outputs */
+ blkCnt = blockSize % 0x4U;
- /* Decrement the loop counter */
- blkCnt--;
- }
+#else
- /* All the output values are in pScratchOut buffer.
- Convert them into 1.15 format, saturate and store in the destination buffer. */
- /* Loop over the blockSize. */
+ /* Initialize blkCnt with number of samples */
blkCnt = blockSize;
+#endif /* #if defined (ARM_MATH_LOOPUNROLL) */
+
while (blkCnt > 0U)
{
*pOut++ = (q7_t) __SSAT(*pScr2++ >> 7, 8);
- /* Decrement the blockSize loop counter */
+ /* Decrement loop counter */
blkCnt--;
}
-#endif /* #if defined (ARM_MATH_DSP) */
-
}
/**
- * @} end of FIR_Sparse group
+ @} end of FIR_Sparse group
*/