123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357 |
- #include "arm_math.h"
- extern void arm_radix4_butterfly_q15(
- q15_t * pSrc,
- uint32_t fftLen,
- q15_t * pCoef,
- uint32_t twidCoefModifier);
- extern void arm_radix4_butterfly_inverse_q15(
- q15_t * pSrc,
- uint32_t fftLen,
- q15_t * pCoef,
- uint32_t twidCoefModifier);
- extern void arm_bitreversal_16(
- uint16_t * pSrc,
- const uint16_t bitRevLen,
- const uint16_t * pBitRevTable);
-
- void arm_cfft_radix4by2_q15(
- q15_t * pSrc,
- uint32_t fftLen,
- const q15_t * pCoef);
-
- void arm_cfft_radix4by2_inverse_q15(
- q15_t * pSrc,
- uint32_t fftLen,
- const q15_t * pCoef);
- void arm_cfft_q15(
- const arm_cfft_instance_q15 * S,
- q15_t * p1,
- uint8_t ifftFlag,
- uint8_t bitReverseFlag)
- {
- uint32_t L = S->fftLen;
- if(ifftFlag == 1u)
- {
- switch (L)
- {
- case 16:
- case 64:
- case 256:
- case 1024:
- case 4096:
- arm_radix4_butterfly_inverse_q15 ( p1, L, (q15_t*)S->pTwiddle, 1 );
- break;
-
- case 32:
- case 128:
- case 512:
- case 2048:
- arm_cfft_radix4by2_inverse_q15 ( p1, L, S->pTwiddle );
- break;
- }
- }
- else
- {
- switch (L)
- {
- case 16:
- case 64:
- case 256:
- case 1024:
- case 4096:
- arm_radix4_butterfly_q15 ( p1, L, (q15_t*)S->pTwiddle, 1 );
- break;
-
- case 32:
- case 128:
- case 512:
- case 2048:
- arm_cfft_radix4by2_q15 ( p1, L, S->pTwiddle );
- break;
- }
- }
-
- if( bitReverseFlag )
- arm_bitreversal_16((uint16_t*)p1,S->bitRevLength,S->pBitRevTable);
- }
- void arm_cfft_radix4by2_q15(
- q15_t * pSrc,
- uint32_t fftLen,
- const q15_t * pCoef)
- {
- uint32_t i;
- uint32_t n2;
- q15_t p0, p1, p2, p3;
- #ifndef ARM_MATH_CM0_FAMILY
- q31_t T, S, R;
- q31_t coeff, out1, out2;
- const q15_t *pC = pCoef;
- q15_t *pSi = pSrc;
- q15_t *pSl = pSrc + fftLen;
- #else
- uint32_t ia, l;
- q15_t xt, yt, cosVal, sinVal;
- #endif
-
- n2 = fftLen >> 1;
- #ifndef ARM_MATH_CM0_FAMILY
- for (i = n2; i > 0; i--)
- {
- coeff = _SIMD32_OFFSET(pC);
- pC += 2;
- T = _SIMD32_OFFSET(pSi);
- T = __SHADD16(T, 0);
- S = _SIMD32_OFFSET(pSl);
- S = __SHADD16(S, 0);
- R = __QSUB16(T, S);
- _SIMD32_OFFSET(pSi) = __SHADD16(T, S);
- pSi += 2;
- #ifndef ARM_MATH_BIG_ENDIAN
- out1 = __SMUAD(coeff, R) >> 16;
- out2 = __SMUSDX(coeff, R);
- #else
- out1 = __SMUSDX(R, coeff) >> 16u;
- out2 = __SMUAD(coeff, R);
- #endif
- _SIMD32_OFFSET(pSl) =
- (q31_t) ((out2) & 0xFFFF0000) | (out1 & 0x0000FFFF);
- pSl += 2;
- }
-
- #else
-
- ia = 0;
- for (i = 0; i < n2; i++)
- {
- cosVal = pCoef[ia * 2];
- sinVal = pCoef[(ia * 2) + 1];
- ia++;
-
- l = i + n2;
-
- xt = (pSrc[2 * i] >> 1u) - (pSrc[2 * l] >> 1u);
- pSrc[2 * i] = ((pSrc[2 * i] >> 1u) + (pSrc[2 * l] >> 1u)) >> 1u;
-
- yt = (pSrc[2 * i + 1] >> 1u) - (pSrc[2 * l + 1] >> 1u);
- pSrc[2 * i + 1] =
- ((pSrc[2 * l + 1] >> 1u) + (pSrc[2 * i + 1] >> 1u)) >> 1u;
- pSrc[2u * l] = (((int16_t) (((q31_t) xt * cosVal) >> 16)) +
- ((int16_t) (((q31_t) yt * sinVal) >> 16)));
- pSrc[2u * l + 1u] = (((int16_t) (((q31_t) yt * cosVal) >> 16)) -
- ((int16_t) (((q31_t) xt * sinVal) >> 16)));
- }
-
- #endif
-
-
- arm_radix4_butterfly_q15( pSrc, n2, (q15_t*)pCoef, 2u);
-
- arm_radix4_butterfly_q15( pSrc + fftLen, n2, (q15_t*)pCoef, 2u);
-
- for (i = 0; i < fftLen >> 1; i++)
- {
- p0 = pSrc[4*i+0];
- p1 = pSrc[4*i+1];
- p2 = pSrc[4*i+2];
- p3 = pSrc[4*i+3];
-
- p0 <<= 1;
- p1 <<= 1;
- p2 <<= 1;
- p3 <<= 1;
-
- pSrc[4*i+0] = p0;
- pSrc[4*i+1] = p1;
- pSrc[4*i+2] = p2;
- pSrc[4*i+3] = p3;
- }
- }
- void arm_cfft_radix4by2_inverse_q15(
- q15_t * pSrc,
- uint32_t fftLen,
- const q15_t * pCoef)
- {
- uint32_t i;
- uint32_t n2;
- q15_t p0, p1, p2, p3;
- #ifndef ARM_MATH_CM0_FAMILY
- q31_t T, S, R;
- q31_t coeff, out1, out2;
- const q15_t *pC = pCoef;
- q15_t *pSi = pSrc;
- q15_t *pSl = pSrc + fftLen;
- #else
- uint32_t ia, l;
- q15_t xt, yt, cosVal, sinVal;
- #endif
-
- n2 = fftLen >> 1;
- #ifndef ARM_MATH_CM0_FAMILY
- for (i = n2; i > 0; i--)
- {
- coeff = _SIMD32_OFFSET(pC);
- pC += 2;
- T = _SIMD32_OFFSET(pSi);
- T = __SHADD16(T, 0);
- S = _SIMD32_OFFSET(pSl);
- S = __SHADD16(S, 0);
- R = __QSUB16(T, S);
- _SIMD32_OFFSET(pSi) = __SHADD16(T, S);
- pSi += 2;
- #ifndef ARM_MATH_BIG_ENDIAN
- out1 = __SMUSD(coeff, R) >> 16;
- out2 = __SMUADX(coeff, R);
- #else
- out1 = __SMUADX(R, coeff) >> 16u;
- out2 = __SMUSD(__QSUB(0, coeff), R);
- #endif
- _SIMD32_OFFSET(pSl) =
- (q31_t) ((out2) & 0xFFFF0000) | (out1 & 0x0000FFFF);
- pSl += 2;
- }
-
- #else
- ia = 0;
- for (i = 0; i < n2; i++)
- {
- cosVal = pCoef[ia * 2];
- sinVal = pCoef[(ia * 2) + 1];
- ia++;
-
- l = i + n2;
- xt = (pSrc[2 * i] >> 1u) - (pSrc[2 * l] >> 1u);
- pSrc[2 * i] = ((pSrc[2 * i] >> 1u) + (pSrc[2 * l] >> 1u)) >> 1u;
-
- yt = (pSrc[2 * i + 1] >> 1u) - (pSrc[2 * l + 1] >> 1u);
- pSrc[2 * i + 1] =
- ((pSrc[2 * l + 1] >> 1u) + (pSrc[2 * i + 1] >> 1u)) >> 1u;
-
- pSrc[2u * l] = (((int16_t) (((q31_t) xt * cosVal) >> 16)) -
- ((int16_t) (((q31_t) yt * sinVal) >> 16)));
-
- pSrc[2u * l + 1u] = (((int16_t) (((q31_t) yt * cosVal) >> 16)) +
- ((int16_t) (((q31_t) xt * sinVal) >> 16)));
- }
-
- #endif
-
- arm_radix4_butterfly_inverse_q15( pSrc, n2, (q15_t*)pCoef, 2u);
-
- arm_radix4_butterfly_inverse_q15( pSrc + fftLen, n2, (q15_t*)pCoef, 2u);
-
- for (i = 0; i < fftLen >> 1; i++)
- {
- p0 = pSrc[4*i+0];
- p1 = pSrc[4*i+1];
- p2 = pSrc[4*i+2];
- p3 = pSrc[4*i+3];
-
- p0 <<= 1;
- p1 <<= 1;
- p2 <<= 1;
- p3 <<= 1;
-
- pSrc[4*i+0] = p0;
- pSrc[4*i+1] = p1;
- pSrc[4*i+2] = p2;
- pSrc[4*i+3] = p3;
- }
- }
|