diff options
Diffstat (limited to 'Drivers/CMSIS/DSP/Source/ComplexMathFunctions/arm_cmplx_mult_real_f32.c')
| -rw-r--r-- | Drivers/CMSIS/DSP/Source/ComplexMathFunctions/arm_cmplx_mult_real_f32.c | 230 |
1 files changed, 93 insertions, 137 deletions
diff --git a/Drivers/CMSIS/DSP/Source/ComplexMathFunctions/arm_cmplx_mult_real_f32.c b/Drivers/CMSIS/DSP/Source/ComplexMathFunctions/arm_cmplx_mult_real_f32.c index 8c7ca313b..9651999e3 100644 --- a/Drivers/CMSIS/DSP/Source/ComplexMathFunctions/arm_cmplx_mult_real_f32.c +++ b/Drivers/CMSIS/DSP/Source/ComplexMathFunctions/arm_cmplx_mult_real_f32.c @@ -3,13 +3,13 @@ * Title: arm_cmplx_mult_real_f32.c * Description: Floating-point complex by real multiplication * - * $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 * @@ -29,185 +29,141 @@ #include "arm_math.h" /** - * @ingroup groupCmplxMath + @ingroup groupCmplxMath */ /** - * @defgroup CmplxByRealMult Complex-by-Real Multiplication - * - * Multiplies a complex vector by a real vector and generates a complex result. - * The data in the complex arrays is stored in an interleaved fashion - * (real, imag, real, imag, ...). - * The parameter <code>numSamples</code> represents the number of complex - * samples processed. The complex arrays have a total of <code>2*numSamples</code> - * real values while the real array has a total of <code>numSamples</code> - * real values. - * - * The underlying algorithm is used: - * - * <pre> - * for(n=0; n<numSamples; n++) { - * pCmplxDst[(2*n)+0] = pSrcCmplx[(2*n)+0] * pSrcReal[n]; - * pCmplxDst[(2*n)+1] = pSrcCmplx[(2*n)+1] * pSrcReal[n]; - * } - * </pre> - * - * There are separate functions for floating-point, Q15, and Q31 data types. + @defgroup CmplxByRealMult Complex-by-Real Multiplication + + Multiplies a complex vector by a real vector and generates a complex result. + The data in the complex arrays is stored in an interleaved fashion + (real, imag, real, imag, ...). + The parameter <code>numSamples</code> represents the number of complex + samples processed. The complex arrays have a total of <code>2*numSamples</code> + real values while the real array has a total of <code>numSamples</code> + real values. + + The underlying algorithm is used: + + <pre> + for (n = 0; n < numSamples; n++) { + pCmplxDst[(2*n)+0] = pSrcCmplx[(2*n)+0] * pSrcReal[n]; + pCmplxDst[(2*n)+1] = pSrcCmplx[(2*n)+1] * pSrcReal[n]; + } + </pre> + + There are separate functions for floating-point, Q15, and Q31 data types. */ /** - * @addtogroup CmplxByRealMult - * @{ + @addtogroup CmplxByRealMult + @{ */ - /** - * @brief Floating-point complex-by-real multiplication - * @param[in] *pSrcCmplx points to the complex input vector - * @param[in] *pSrcReal points to the real input vector - * @param[out] *pCmplxDst points to the complex output vector - * @param[in] numSamples number of samples in each vector - * @return none. + @brief Floating-point complex-by-real multiplication. + @param[in] pSrcCmplx points to complex input vector + @param[in] pSrcReal points to real input vector + @param[out] pCmplxDst points to complex output vector + @param[in] numSamples number of samples in each vector + @return none */ void arm_cmplx_mult_real_f32( - float32_t * pSrcCmplx, - float32_t * pSrcReal, - float32_t * pCmplxDst, - uint32_t numSamples) + const float32_t * pSrcCmplx, + const float32_t * pSrcReal, + float32_t * pCmplxDst, + uint32_t numSamples) { - float32_t in; /* Temporary variable to store input value */ - uint32_t blkCnt; /* loop counters */ - -#if defined (ARM_MATH_DSP) - - /* Run the below code for Cortex-M4 and Cortex-M3 */ - float32_t inA1, inA2, inA3, inA4; /* Temporary variables to hold input data */ - float32_t inA5, inA6, inA7, inA8; /* Temporary variables to hold input data */ - float32_t inB1, inB2, inB3, inB4; /* Temporary variables to hold input data */ - float32_t out1, out2, out3, out4; /* Temporary variables to hold output data */ - float32_t out5, out6, out7, out8; /* Temporary variables to hold output data */ - - /* loop Unrolling */ - blkCnt = numSamples >> 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 (blkCnt > 0U) - { - /* C[2 * i] = A[2 * i] * B[i]. */ - /* C[2 * i + 1] = A[2 * i + 1] * B[i]. */ - /* read input from complex input buffer */ - inA1 = pSrcCmplx[0]; - inA2 = pSrcCmplx[1]; - /* read input from real input buffer */ - inB1 = pSrcReal[0]; - - /* read input from complex input buffer */ - inA3 = pSrcCmplx[2]; - - /* multiply complex buffer real input with real buffer input */ - out1 = inA1 * inB1; - - /* read input from complex input buffer */ - inA4 = pSrcCmplx[3]; + uint32_t blkCnt; /* Loop counter */ + float32_t in; /* Temporary variable */ - /* multiply complex buffer imaginary input with real buffer input */ - out2 = inA2 * inB1; +#if defined(ARM_MATH_NEON) + float32x4_t r; + float32x4x2_t ab,outCplx; - /* read input from real input buffer */ - inB2 = pSrcReal[1]; - /* read input from complex input buffer */ - inA5 = pSrcCmplx[4]; + /* Compute 4 outputs at a time */ + blkCnt = numSamples >> 2U; - /* multiply complex buffer real input with real buffer input */ - out3 = inA3 * inB2; + while (blkCnt > 0U) + { + ab = vld2q_f32(pSrcCmplx); // load & separate real/imag pSrcA (de-interleave 2) + r = vld1q_f32(pSrcReal); // load & separate real/imag pSrcB - /* read input from complex input buffer */ - inA6 = pSrcCmplx[5]; - /* read input from real input buffer */ - inB3 = pSrcReal[2]; + /* Increment pointers */ + pSrcCmplx += 8; + pSrcReal += 4; - /* multiply complex buffer imaginary input with real buffer input */ - out4 = inA4 * inB2; + outCplx.val[0] = vmulq_f32(ab.val[0], r); + outCplx.val[1] = vmulq_f32(ab.val[1], r); - /* read input from complex input buffer */ - inA7 = pSrcCmplx[6]; + vst2q_f32(pCmplxDst, outCplx); + pCmplxDst += 8; - /* multiply complex buffer real input with real buffer input */ - out5 = inA5 * inB3; + blkCnt--; + } - /* read input from complex input buffer */ - inA8 = pSrcCmplx[7]; - - /* multiply complex buffer imaginary input with real buffer input */ - out6 = inA6 * inB3; - - /* read input from real input buffer */ - inB4 = pSrcReal[3]; - - /* store result to destination bufer */ - pCmplxDst[0] = out1; - - /* multiply complex buffer real input with real buffer input */ - out7 = inA7 * inB4; - - /* store result to destination bufer */ - pCmplxDst[1] = out2; - - /* multiply complex buffer imaginary input with real buffer input */ - out8 = inA8 * inB4; + /* Tail */ + blkCnt = numSamples & 3; +#else +#if defined (ARM_MATH_LOOPUNROLL) - /* store result to destination bufer */ - pCmplxDst[2] = out3; - pCmplxDst[3] = out4; - pCmplxDst[4] = out5; + /* Loop unrolling: Compute 4 outputs at a time */ + blkCnt = numSamples >> 2U; - /* incremnet complex input buffer by 8 to process next samples */ - pSrcCmplx += 8U; + while (blkCnt > 0U) + { + /* C[2 * i ] = A[2 * i ] * B[i]. */ + /* C[2 * i + 1] = A[2 * i + 1] * B[i]. */ - /* store result to destination bufer */ - pCmplxDst[5] = out6; + in = *pSrcReal++; + /* store result in destination buffer. */ + *pCmplxDst++ = *pSrcCmplx++ * in; + *pCmplxDst++ = *pSrcCmplx++ * in; - /* increment real input buffer by 4 to process next samples */ - pSrcReal += 4U; + in = *pSrcReal++; + *pCmplxDst++ = *pSrcCmplx++ * in; + *pCmplxDst++ = *pSrcCmplx++ * in; - /* store result to destination bufer */ - pCmplxDst[6] = out7; - pCmplxDst[7] = out8; + in = *pSrcReal++; + *pCmplxDst++ = *pSrcCmplx++ * in; + *pCmplxDst++ = *pSrcCmplx++ * in; - /* increment destination buffer by 8 to process next sampels */ - pCmplxDst += 8U; + in = *pSrcReal++; + *pCmplxDst++ = *pSrcCmplx++* in; + *pCmplxDst++ = *pSrcCmplx++ * in; - /* Decrement the numSamples loop counter */ + /* Decrement loop counter */ blkCnt--; } - /* If the numSamples is not a multiple of 4, compute any remaining output samples here. - ** No loop unrolling is used. */ + /* Loop unrolling: Compute remaining outputs */ blkCnt = numSamples % 0x4U; #else - /* Run the below code for Cortex-M0 */ + /* Initialize blkCnt with number of samples */ blkCnt = numSamples; -#endif /* #if defined (ARM_MATH_DSP) */ +#endif /* #if defined (ARM_MATH_LOOPUNROLL) */ +#endif /* #if defined(ARM_MATH_NEON) */ while (blkCnt > 0U) { - /* C[2 * i] = A[2 * i] * B[i]. */ - /* C[2 * i + 1] = A[2 * i + 1] * B[i]. */ + /* C[2 * i ] = A[2 * i ] * B[i]. */ + /* C[2 * i + 1] = A[2 * i + 1] * B[i]. */ + in = *pSrcReal++; - /* store the result in the destination buffer. */ - *pCmplxDst++ = (*pSrcCmplx++) * (in); - *pCmplxDst++ = (*pSrcCmplx++) * (in); + /* store result in destination buffer. */ + *pCmplxDst++ = *pSrcCmplx++ * in; + *pCmplxDst++ = *pSrcCmplx++ * in; - /* Decrement the numSamples loop counter */ + /* Decrement loop counter */ blkCnt--; } + } /** - * @} end of CmplxByRealMult group + @} end of CmplxByRealMult group */ |