#include "speech_fft.h"

#include "basic_op.h"

















#ifndef OPT_KHW_BASIC_OP



static const short fft_bitrev60[15] = 

{

	0, 10, 20, 2, 12, 22, 4, 14, 24, 6, 16, 26, 8, 18, 28, 

};



static const short fft_twiddles16000_240[120] = 

{

	32767,	 0, 32588, -3425,

	32051, -6813, 31165, -10126,

	29936, -13328,28379, -16384,

	26510, -19260,24353, -21926,

	21927, -24352,19261, -26509,

	16385, -28378,13329, -29934,

	10127, -31164,6815, -32051,

	3426, -32588, 0,

	-32767,

	-3425, -32588, -6813, -32051,

	-10126, -31165,-13328, -29936,

	-16384, -28379,-19260, -26510,

	-21926, -24353,-24352, -21927,

	-26509, -19261,-28378, -16385,

	-29934, -13329,-31164, -10127,

	-32051, -6815,-32588, -3426,



	-32767, 0,-32588, 3425,

	-32051, 6813,-31165, 10126,

	-29936, 13328,-28379, 16384,

	-26510, 19260,-24353, 21926,

	-21927, 24352,-19261, 26509,

	-16385, 28378,-13329, 29934,

	-10127, 31164,-6815, 32051,

	-3426, 32588,0, 32767,

	3425, 32588,6813, 32051,

	10126, 31165,13328, 29936,

	16384, 28379,19260, 26510,

	21926, 24353,24352, 21927,

	26509, 19261,28378, 16385,

	29934, 13329,31164, 10127,

	32051, 6815,32588, 3426,

};



#ifndef kf_bfly3_opt_hifi3_ZH

static void kf_bfly3(int * Fout, short _fwd)

{

	short i, k; // addit = 0;

	const short m = 10;

	const short m2 = 20;

	const short *tw1, *tw2;



	int scratch_0, scratch_1, scratch_2, scratch_3, scratch_4, scratch_5;



	short fwd = (_fwd<<1)-1;

	short epi3 = -28379;

#ifdef	HW_HIFI3

	long long tmp_val,tmp_val2;

#endif



#ifdef HW_FFT_5

	if(_fwd)

	{



		for(i=0;i<120;i=i+4)

		{

/*

			tmp_val = AE_MUL32X16_H0(Fout[i], 10923);

			Fout[i] = (int)(tmp_val>>15);



			tmp_val = AE_MUL32X16_H0(Fout[i+1], 10923);

			Fout[i+1] = (int)(tmp_val>>15);



			tmp_val = AE_MUL32X16_H0(Fout[i+2], 10923);

			Fout[i+2] = (int)(tmp_val>>15);



			tmp_val = AE_MUL32X16_H0(Fout[i+3], 10923);

			Fout[i+3] = (int)(tmp_val>>15);

*/

			Fout[i] = S_MUL(Fout[i], 10923);

			Fout[i+1] = S_MUL(Fout[i+1], 10923);

			Fout[i+2] = S_MUL(Fout[i+2], 10923);

			Fout[i+3] = S_MUL(Fout[i+3], 10923);





		}



	}

#endif







	for (i = 0; i<4; i++)

	{

		tw1 = tw2 = fft_twiddles16000_240;

		k = 5;

		do {



#ifndef HW_FFT_5

			if (_fwd) {

#ifndef	HW_HIFI3

				Fout[0] = S_MUL(Fout[0], 10923);

				Fout[1] = S_MUL(Fout[1], 10923);

				Fout[m] = S_MUL(Fout[m], 10923);

				Fout[m + 1] = S_MUL(Fout[m + 1], 10923);

				Fout[m2] = S_MUL(Fout[m2], 10923);

				Fout[m2 + 1] = S_MUL(Fout[m2 + 1], 10923);

#else

				tmp_val = AE_MUL32X16_H0(Fout[0], 10923);

				Fout[0] = (int)(tmp_val>>15);



				tmp_val = AE_MUL32X16_H0(Fout[1], 10923);

				Fout[1] = (int)(tmp_val>>15);



				tmp_val = AE_MUL32X16_H0(Fout[m], 10923);

				Fout[m] = (int)(tmp_val>>15);



				tmp_val = AE_MUL32X16_H0(Fout[m+1], 10923);

				Fout[m+1] = (int)(tmp_val>>15);



				tmp_val = AE_MUL32X16_H0(Fout[m2], 10923);

				Fout[m2] = (int)(tmp_val>>15);



				tmp_val = AE_MUL32X16_H0(Fout[m2+1], 10923);

				Fout[m2+1] = (int)(tmp_val>>15);

#endif

			}

#endif



#ifndef	HW_HIFI3

			scratch_2 = SSC_ADD(S_MUL(Fout[m], tw1[0]), (-1)*fwd * S_MUL(Fout[m + 1], tw1[1]));

			scratch_3 = SSC_ADD(S_MUL(Fout[m + 1 - _fwd], tw1[_fwd]), fwd*S_MUL(Fout[m + _fwd], tw1[1 - _fwd]));

			scratch_4 = SSC_ADD(S_MUL(Fout[m2], tw2[0]), (-1)*fwd * S_MUL(Fout[m2 + 1], tw2[1]));

			scratch_5 = SSC_ADD(S_MUL(Fout[m2 + 1 - _fwd], tw2[_fwd]), fwd*S_MUL(Fout[m2 + _fwd], tw2[1 - _fwd]));

#else

			tmp_val = AE_MUL32X16_H0(Fout[m], tw1[0]);

			tmp_val2 = AE_MUL32X16_H0(Fout[m+1], tw1[1]);

		//	scratch_2 = SSC_SUB((int)(tmp_val>>15) , fwd*(int)(tmp_val2>>15) );

			scratch_2 = SSC_SUB((int)(tmp_val>>15) , FFT_MUL_16_32(fwd,(int)(tmp_val2>>15))    );





			tmp_val = AE_MUL32X16_H0(Fout[m + 1 - _fwd], tw1[_fwd]);

			tmp_val2 = AE_MUL32X16_H0(Fout[m + _fwd], tw1[1 - _fwd]);

		//	scratch_3 = SSC_ADD((int)(tmp_val>>15) , fwd*(int)(tmp_val2>>15) );

			scratch_3 = SSC_ADD((int)(tmp_val>>15) , FFT_MUL_16_32(fwd,(int)(tmp_val2>>15)) );



			tmp_val = AE_MUL32X16_H0(Fout[m2], tw2[0]);

			tmp_val2 = AE_MUL32X16_H0(Fout[m2+1], tw2[1]);

	//		scratch_4 = SSC_SUB((int)(tmp_val>>15) , fwd*(int)(tmp_val2>>15) );

			scratch_4 = SSC_SUB((int)(tmp_val>>15) , FFT_MUL_16_32(fwd,(int)(tmp_val2>>15)) );



			tmp_val = AE_MUL32X16_H0(Fout[m2 + 1 - _fwd], tw2[_fwd]);

			tmp_val2 = AE_MUL32X16_H0(Fout[m2 + _fwd], tw2[1 - _fwd]);

	//		scratch_5 = SSC_ADD((int)(tmp_val>>15) , fwd*(int)(tmp_val2>>15) );

			scratch_5 = SSC_ADD((int)(tmp_val>>15) , FFT_MUL_16_32(fwd,(int)(tmp_val2>>15)) );

#endif



			scratch_0 = SSC_SUB(scratch_2, scratch_4); 

			scratch_1 = SSC_ADD(scratch_2, scratch_4); // 6-->1

								   

								   

			scratch_2 = SSC_ADD(scratch_3, scratch_5);	// 7-->2

			scratch_3 = SSC_SUB(scratch_3, scratch_5);   // 1-->3







			tw1 += 8;

			tw2 += 16;

			Fout[m] = Fout[0] - HALF_OF(scratch_1); // 6-->1

			Fout[m + 1] = Fout[1] - HALF_OF(scratch_2); // 7-->2

#ifndef	HW_HIFI3

			scratch_0 = S_MUL(scratch_0, fwd * epi3);

			scratch_3 = S_MUL(scratch_3, fwd * epi3); // 1-->3

#else

			tmp_val = AE_MUL32X16_H0(scratch_0, (SSC_MULT16x16(fwd,epi3)));

			scratch_0 = (int)(tmp_val>>15);

			tmp_val = AE_MUL32X16_H0(scratch_3, (SSC_MULT16x16(fwd,epi3))); // 1-->3

			scratch_3 = (int)(tmp_val>>15);

#endif



			Fout[0] = SSC_ADD(Fout[0], scratch_1); // 6-->1

			Fout[1] = SSC_ADD(Fout[1], scratch_2); // 7-->2

			Fout[m2] = Fout[m] + scratch_3;  // 1-->3

			Fout[m2 + 1] = Fout[m + 1] - scratch_0;

			Fout[m] -= scratch_3; // 1-->3

			Fout[m + 1] += scratch_0;

			Fout = Fout + 2;

		} while (--k);

		Fout = Fout + 20;

	}

}

#else

static void kf_bfly3(int * Fout, short _fwd)

{

	short i, k; // addit = 0;

	const short m = 10;

	const short m2 = 20;

	const short *tw1, *tw2;

	ae_f32x2 tmpL1, tmpL2, tmpL3, tmpL4, flagv;

	ae_f32x2 *pt0, *pt1, *pt2, *pt3, *pt4, *pt5;

	ae_valign align0 ,align1, align2, align3, align4, align5;

	ae_f16x4 epi3v;



	int scratch_0, scratch_1, scratch_2, scratch_3, scratch_4, scratch_5;



	short fwd = (_fwd<<1)-1;

	short epi3 = -28379;

#ifdef	HW_HIFI3

	long long tmp_val,tmp_val2;

#endif



#ifdef HW_FFT_5

	if(_fwd)

	{



		for(i=0;i<120;i=i+4)

		{

			tmp_val = AE_MUL32X16_H0(Fout[i], 10923);  //  ³ªÁß¿¡  AE_MUL32X2_vector¸¦ ÀÌ¿ëÇؼ­ ¹Ù²ÙÀÚ.  AE_SRAI64X2_vector

			Fout[i] = (int)(tmp_val>>15);



			tmp_val = AE_MUL32X16_H0(Fout[i+1], 10923);

			Fout[i+1] = (int)(tmp_val>>15);



			tmp_val = AE_MUL32X16_H0(Fout[i+2], 10923);

			Fout[i+2] = (int)(tmp_val>>15);



			tmp_val = AE_MUL32X16_H0(Fout[i+3], 10923);

			Fout[i+3] = (int)(tmp_val>>15);

		}



	}

#endif





	epi3v = AE_MOVDA16(SSC_MULT16x16(fwd,epi3));

	flagv = AE_MOVDA32X2(1,fwd);

	for (i = 0; i<4; i++)

	{

		ae_f32x2 S01,S23, S45, F01, F23, F45, S12, S03, tmpL, S30;

		ae_f16x4 Tw0,Tw1;



		tw1 = tw2 = fft_twiddles16000_240;

		k = 5;



		pt0 = pt3 =  (ae_f32x2*)&Fout[0];

		pt1 = pt4 = (ae_f32x2*)&Fout[m];

		pt2 = pt5 = (ae_f32x2*)&Fout[m2];

		align0 = AE_LA64_PP(pt0);

		align1 = AE_LA64_PP(pt1);

		align2 = AE_LA64_PP(pt2);

		align3 = AE_ZALIGN64();

		align4 = AE_ZALIGN64();

		align5 = AE_ZALIGN64();





		do {







			AE_LA32X2_IP(F01, align0, pt0);

			AE_LA32X2_IP(F23, align1, pt1);

			AE_LA32X2_IP(F45, align2, pt2);



			Tw0 = AE_MOVDA16X2(tw1[0], tw1[1]*fwd);   //  here

			Tw1 = AE_MOVDA16X2(tw2[0], tw2[1]*fwd);   //







			S23 = AE_MULFC32X16RAS_L(F23, Tw0);

			S45 = AE_MULFC32X16RAS_L(F45, Tw1);



            S03 = AE_SUB32(S23, S45);

            S12 = AE_ADD32(S23, S45);



            tmpL = AE_SRAI32(S12, 1);



			tw1 += 8;

			tw2 += 16;



			F23 = AE_SUB32(F01, tmpL);

			S03 = AE_MULFP32X16X2RAS_L(S03, epi3v);



			S30 = AE_SEL32_LH(S03, S03);



            F01 = AE_ADD32(F01, S12);

            F45 = AE_ADDSUB32(F23, S30);

            F23 = AE_SUBADD32(F23, S30);





            AE_SA32X2_IP(F01, align3, pt3);

            AE_SA32X2_IP(F23, align4, pt4);

            AE_SA32X2_IP(F45, align5, pt5);



//			Fout = Fout + 2;

		} while (--k);

		AE_SA64POS_FP(align3, pt3);

		AE_SA64POS_FP(align4, pt4);

		AE_SA64POS_FP(align5, pt5);

		Fout = Fout + 30;

	}

}



#endif











#ifndef kf_bfly4_opt_hifi3_ZH



static void kf_bfly4(int *Fout, short fwd)

{

	const short *tw1, *tw2, *tw3;

	short flag = 1, shiftbit = 2;



	int scratch_0,scratch_1,scratch_2,scratch_3,scratch_4,scratch_5,scratch_6,scratch_7;



	short m = 30;

	const short m2 = (m << 1);

	const short m3 = 90;

	short j;



#ifdef	HW_HIFI3

	long long tmp_val,tmp_val2;

#endif



	tw3 = tw2 = tw1 = fft_twiddles16000_240;



	j = m >> 1;



	if (0 == fwd)

	{

		flag = -1;

		shiftbit = 0;



	}





	do {



#ifndef HW_HIFI3

		scratch_0 = SHR(SSC_SUB(S_MUL(Fout[m], tw1[0]), flag*S_MUL(Fout[m + 1], tw1[1])), shiftbit);

		scratch_1 = SHR(SSC_ADD(S_MUL(Fout[m + 1], tw1[0]), flag*S_MUL(Fout[m], tw1[1])), shiftbit);

		scratch_2 = SHR(SSC_SUB(S_MUL(Fout[m2], tw2[0]), flag*S_MUL(Fout[m2 + 1], tw2[1])), shiftbit);

		scratch_3 = SHR(SSC_ADD(S_MUL(Fout[m2 + 1], tw2[0]), flag*S_MUL(Fout[m2], tw2[1])), shiftbit);

		scratch_4 = SHR(SSC_SUB(S_MUL(Fout[m3], tw3[0]), flag*S_MUL(Fout[m3 + 1], tw3[1])), shiftbit);



		scratch_5 = SHR(SSC_ADD(S_MUL(Fout[m3 + 1], tw3[0]), flag*S_MUL(Fout[m3], tw3[1])), shiftbit);

#else	

		tmp_val = AE_MUL32X16_H0(Fout[m], tw1[0]);

		tmp_val2 = AE_MUL32X16_H0(Fout[m+1], tw1[1]);

//		scratch_0 =  SHR(SSC_SUB((int)(tmp_val>>15),(flag)*((int)(tmp_val2>>15))),shiftbit);

		scratch_0 =  SHR(SSC_SUB((int)(tmp_val>>15),FFT_MUL_16_32(flag,((int)(tmp_val2>>15)))),shiftbit);





		tmp_val = AE_MUL32X16_H0(Fout[m+1], tw1[0]);

		tmp_val2 = AE_MUL32X16_H0(Fout[m], tw1[1]);

	//	scratch_1 =  SHR(SSC_ADD((int)(tmp_val>>15),(flag)*((int)(tmp_val2>>15))),shiftbit);

		scratch_1 =  SHR(SSC_ADD((int)(tmp_val>>15),FFT_MUL_16_32(flag,((int)(tmp_val2>>15)))),shiftbit);



		tmp_val = AE_MUL32X16_H0(Fout[m2], tw2[0]);

		tmp_val2 = AE_MUL32X16_H0(Fout[m2+1], tw2[1]);

//		scratch_2 =  SHR(SSC_SUB((int)(tmp_val>>15),(flag)*((int)(tmp_val2>>15))),shiftbit);

		scratch_2 =  SHR(SSC_SUB((int)(tmp_val>>15),FFT_MUL_16_32(flag,((int)(tmp_val2>>15)))),shiftbit);



		tmp_val = AE_MUL32X16_H0(Fout[m2+1], tw2[0]);

		tmp_val2 = AE_MUL32X16_H0(Fout[m2], tw2[1]);

//		scratch_3 =  SHR(SSC_ADD((int)(tmp_val>>15),(flag)*((int)(tmp_val2>>15))),shiftbit);

		scratch_3 =  SHR(SSC_ADD((int)(tmp_val>>15),FFT_MUL_16_32(flag,((int)(tmp_val2>>15)))),shiftbit);



		tmp_val = AE_MUL32X16_H0(Fout[m3], tw3[0]);

		tmp_val2 = AE_MUL32X16_H0(Fout[m3+1], tw3[1]);

//		scratch_4 =  SHR(SSC_SUB((int)(tmp_val>>15),(flag)*((int)(tmp_val2>>15))),shiftbit);

		scratch_4 =  SHR(SSC_SUB((int)(tmp_val>>15),FFT_MUL_16_32(flag,((int)(tmp_val2>>15)))),shiftbit);



		tmp_val = AE_MUL32X16_H0(Fout[m3+1], tw3[0]);

		tmp_val2 = AE_MUL32X16_H0(Fout[m3], tw3[1]);

//		scratch_5 =  SHR(SSC_ADD((int)(tmp_val>>15),(flag)*((int)(tmp_val2>>15))),shiftbit);

		scratch_5 =  SHR(SSC_ADD((int)(tmp_val>>15),FFT_MUL_16_32(flag,((int)(tmp_val2>>15)))),shiftbit);

#endif



		Fout[0] = SSC_PSHR(Fout[0], shiftbit);

		Fout[1] = SSC_PSHR(Fout[1], shiftbit);

		scratch_6 = SSC_SUB(Fout[0], scratch_2);  // 10-->6

		scratch_7 = SSC_SUB(Fout[1], scratch_3);  // 11-->7

		Fout[0] = SSC_ADD(Fout[0], scratch_2);

		Fout[1] = SSC_ADD(Fout[1], scratch_3);



		scratch_2 = SSC_ADD(scratch_0, scratch_4);   //6-->2

		scratch_3 = SSC_ADD(scratch_1, scratch_5);   //7-->3

		scratch_0 = SSC_SUB(scratch_0, scratch_4);   //8-->0

		scratch_1 = SSC_SUB(scratch_1, scratch_5);   //9-->1



		Fout[m2] = SSC_SUB(Fout[0], scratch_2); //6-->2

		Fout[m2 + 1] = SSC_SUB(Fout[1], scratch_3); //7-->3

		tw1 += 2;

		tw2 += 4;

		tw3 += 6;

		Fout[0] = SSC_ADD(Fout[0], scratch_2); //6-->2

		Fout[1] = SSC_ADD(Fout[1], scratch_3); //7-->3

#ifndef HW_HIFI3

		Fout[m] = scratch_6 + flag*scratch_1; // 10-->6   //9-->1

		Fout[m + 1] = scratch_7 - flag*scratch_0; // 11-->7  //8-->0

		Fout[m3] = scratch_6 - flag*scratch_1; // 10-->6   //9-->1

		Fout[m3 + 1] = scratch_7 + flag*scratch_0; // 11-->7  //8-->0

#else

		Fout[m] = scratch_6 + FFT_MUL_16_32(flag,scratch_1); // 10-->6   //9-->1

		Fout[m + 1] = scratch_7 - FFT_MUL_16_32(flag,scratch_0); // 11-->7  //8-->0

		Fout[m3] = scratch_6 - FFT_MUL_16_32(flag,scratch_1); // 10-->6   //9-->1

		Fout[m3 + 1] = scratch_7 + FFT_MUL_16_32(flag,scratch_0); // 11-->7  //8-->0

#endif

		Fout = Fout + 2;

	} while (--j);

}

#else

static void kf_bfly4(int *Fout, short fwd)

{

	const short *tw1, *tw2, *tw3;

	short flag = 1, shiftbit = 2;



	int scratch_0,scratch_1,scratch_2,scratch_3,scratch_4,scratch_5,scratch_6,scratch_7;



	short m = 30;

	const short m2 = (m << 1);

	const short m3 = 90;

	short j;





	long long tmp_val,tmp_val2;

	ae_f32x2 *pt0, *pt1, *pt2,*pt3,*pt4,*pt5,*pt6, *pt7;

	ae_f32x2  F01, F23, F45, F67;

	ae_f32x2  S01, S23, S45, S67;

	ae_valign align0 ,align1, align2, align3, align4, align5 , align6, align7;

	ae_f16x4 Tw0,Tw1,Tw2;



	tw3 = tw2 = tw1 = fft_twiddles16000_240;



	j = m >> 1;



	if (0 == fwd)

	{

		flag = -1;

		shiftbit = 0;



	}



	pt0 = pt4 = (ae_f32x2*)&Fout[0];

	pt1 = pt5 = (ae_f32x2*)&Fout[m];

	pt2 = pt6 = (ae_f32x2*)&Fout[m2];

	pt3 = pt7 = (ae_f32x2*)&Fout[m3];

	align0 = AE_LA64_PP(pt0);

	align1 = AE_LA64_PP(pt1);

	align2 = AE_LA64_PP(pt2);

	align3 = AE_LA64_PP(pt3);

	align4 = AE_ZALIGN64();

	align5 = AE_ZALIGN64();

	align6 = AE_ZALIGN64();

	align7 = AE_ZALIGN64();

	do {



		AE_LA32X2_IP(F01, align0, pt0);

		AE_LA32X2_IP(F23, align1, pt1);

		AE_LA32X2_IP(F45, align2, pt2);

		AE_LA32X2_IP(F67, align3, pt3);





		Tw0 = AE_MOVDA16X2(tw1[0], tw1[1]*flag);

		Tw1 = AE_MOVDA16X2(tw2[0], tw2[1]*flag);

		Tw2 = AE_MOVDA16X2(tw3[0], tw3[1]*flag);



		S01 = AE_MULFC32X16RAS_L(F23, Tw0);

		S23 = AE_MULFC32X16RAS_L(F45, Tw1);

		S45 = AE_MULFC32X16RAS_L(F67, Tw2);

		S01 = AE_SRAA32(S01, shiftbit);

		S23 = AE_SRAA32(S23, shiftbit);

		S45 = AE_SRAA32(S45, shiftbit);



		F01 = AE_SRAA32RS(F01, shiftbit);





		S67 = AE_SUB32(F01, S23);

		F01 = AE_ADD32(F01, S23);



		S23 = AE_ADD32(S01, S45);

		S01 = AE_SUB32(S01, S45);



        F45 = AE_SUB32(F01, S23);



		tw1 += 2;

		tw2 += 4;

		tw3 += 6;

		F01 = AE_ADD32(F01, S23);



		S01 = AE_SEL32_LH(S01, S01);

		S01 = AE_MULP32X16X2_L(S01, flag);



		F23 = AE_ADDSUB32(S67, S01);

		F67 = AE_SUBADD32(S67, S01);



        AE_SA32X2_IP(F01, align4, pt4);

        AE_SA32X2_IP(F23, align5, pt5);

        AE_SA32X2_IP(F45, align6, pt6);

        AE_SA32X2_IP(F67, align7, pt7);



	} while (--j);

	AE_SA64POS_FP(align4, pt4);

	AE_SA64POS_FP(align5, pt5);

	AE_SA64POS_FP(align6, pt6);

	AE_SA64POS_FP(align7, pt7);

}

#endif





#ifndef kf_bfly5_opt_hifi3_ZH

static void kf_bfly5(int * Fout, short fwd)

{

	int *Fout0; //,*Fout4;

	short i;

#ifndef HW_FFT_5

	short flag = 1;

#else

	short flag = -1;

#endif

	int scratch_0,scratch_1,scratch_2,scratch_3,scratch_4,scratch_5,scratch_6,scratch_7,scratch_8,scratch_9,scratch_10,scratch_11;



	short ya_0, ya_1, yb_0, yb_1;

#ifdef	HW_HIFI3

	long long tmp_val,tmp_val2;

#endif

	ya_1 = -31164;

	yb_1 = -19261;





	ya_0 = 10127;





	yb_0 = -26509;

#ifndef HW_FFT_5

	if (!fwd)

	{

		Fout0 = Fout;

		flag = -1;

	}

#else

	if(fwd)

	{

		flag = 1;

#if 0

		Fout0 = Fout;

		i=30;

		do

		{

			tmp_val = AE_MUL32X16_H0(*Fout0, 6554);

			*Fout0++ = (int)(tmp_val>>15);

			tmp_val = AE_MUL32X16_H0(*Fout0, 6554);

			*Fout0++ = (int)(tmp_val>>15);

			tmp_val = AE_MUL32X16_H0(*Fout0, 6554);

			*Fout0++ = (int)(tmp_val>>15);

			tmp_val = AE_MUL32X16_H0(*Fout0, 6554);

			*Fout0++ = (int)(tmp_val>>15);

		} while (--i);



#else

		//AE_MULA32X2_vector



		for(i=0;i<120;i=i+4)

		{

			Fout[i] = S_MUL(Fout[i], 6554);

			Fout[i+1] = S_MUL(Fout[i+1], 6554);

			Fout[i+2] = S_MUL(Fout[i+2], 6554);

			Fout[i+3] = S_MUL(Fout[i+3], 6554);

		}

#endif

	}



	Fout0 = Fout;



#endif

	i = 12;

	do {







#ifndef HW_FFT_5

		if (fwd)

		{	

			Fout0 = Fout;

#ifndef HW_HIFI3

			*Fout0++ = S_MUL(*Fout0, 6554);

			*Fout0++ = S_MUL(*Fout0, 6554);

			*Fout0++ = S_MUL(*Fout0, 6554);

			*Fout0++ = S_MUL(*Fout0, 6554);

			*Fout0++ = S_MUL(*Fout0, 6554);

			*Fout0++ = S_MUL(*Fout0, 6554);

			*Fout0++ = S_MUL(*Fout0, 6554);

			*Fout0++ = S_MUL(*Fout0, 6554);

			*Fout0++ = S_MUL(*Fout0, 6554);

			*Fout0 = S_MUL(*Fout0, 6554);

			Fout0 = Fout0 - 9;

#else

			tmp_val = AE_MUL32X16_H0(*Fout0, 6554);

			*Fout0++ = (int)(tmp_val>>15);

		//	*Fout0++ = FFT_SHRI(tmp_val,15);

			tmp_val = AE_MUL32X16_H0(*Fout0, 6554);

			*Fout0++ = (int)(tmp_val>>15);

			tmp_val = AE_MUL32X16_H0(*Fout0, 6554);

			*Fout0++ = (int)(tmp_val>>15);

			tmp_val = AE_MUL32X16_H0(*Fout0, 6554);

			*Fout0++ = (int)(tmp_val>>15);

			tmp_val = AE_MUL32X16_H0(*Fout0, 6554);

			*Fout0++ = (int)(tmp_val>>15);

			tmp_val = AE_MUL32X16_H0(*Fout0, 6554);

			*Fout0++ = (int)(tmp_val>>15);

			tmp_val = AE_MUL32X16_H0(*Fout0, 6554);

			*Fout0++ = (int)(tmp_val>>15);

			tmp_val = AE_MUL32X16_H0(*Fout0, 6554);

			*Fout0++ = (int)(tmp_val>>15);

			tmp_val = AE_MUL32X16_H0(*Fout0, 6554);

			*Fout0++ = (int)(tmp_val>>15);

			tmp_val = AE_MUL32X16_H0(*Fout0, 6554);

			*Fout0 = (int)(tmp_val>>15);



			Fout0 = Fout0 - 9;

#endif

		}

#endif

		scratch_0 = Fout0[0];

		scratch_1 = Fout0[1];



		scratch_2 = SSC_ADD(Fout0[2], Fout0[8]);

		scratch_3 = SSC_ADD(Fout0[3], Fout0[9]);

		scratch_8 = SSC_SUB(Fout0[2], Fout0[8]);

		scratch_9 = SSC_SUB(Fout0[3], Fout0[9]);

		scratch_4 = SSC_ADD(Fout0[4], Fout0[6]);

		scratch_5 = SSC_ADD(Fout0[5], Fout0[7]);

		scratch_6 = SSC_SUB(Fout0[4], Fout0[6]);

		scratch_7 = SSC_SUB(Fout0[5], Fout0[7]);







		*Fout0++ = *Fout0 + scratch_2 + scratch_4;

		*Fout0++ = *Fout0 + scratch_3 + scratch_5;

		

#ifndef HW_HIFI3

		scratch_10 = scratch_0 + S_MUL(scratch_2, ya_0) + S_MUL(scratch_4, yb_0);

		scratch_11 = scratch_1 + S_MUL(scratch_3, ya_0) + S_MUL(scratch_5, yb_0);

		scratch_0 = scratch_0 + S_MUL(scratch_2, yb_0) + S_MUL(scratch_4, ya_0);   //12-->0  //13-->1

		scratch_1 = scratch_1 + S_MUL(scratch_3, yb_0) + S_MUL(scratch_5, ya_0);  

#else



		tmp_val = AE_MUL32X16_H0(scratch_2,ya_0);

		tmp_val2 = AE_MUL32X16_H0(scratch_4,yb_0);

		scratch_10 = scratch_0+(int)(tmp_val>>15) +(int)(tmp_val2>>15);



		tmp_val = AE_MUL32X16_H0(scratch_3,ya_0);

		tmp_val2 = AE_MUL32X16_H0(scratch_5,yb_0);

		scratch_11 = scratch_1+(int)(tmp_val>>15) +(int)(tmp_val2>>15);



		tmp_val = AE_MUL32X16_H0(scratch_2,yb_0);

		tmp_val2 = AE_MUL32X16_H0(scratch_4,ya_0);

		scratch_0 = scratch_0+(int)(tmp_val>>15) +(int)(tmp_val2>>15);



		tmp_val = AE_MUL32X16_H0(scratch_3,yb_0);

		tmp_val2 = AE_MUL32X16_H0(scratch_5,ya_0);

		scratch_1 = scratch_1+(int)(tmp_val>>15) +(int)(tmp_val2>>15);

#endif

		

 



#ifndef HW_HIFI3

		scratch_2 = flag*(S_MUL(scratch_9, ya_1) + S_MUL(scratch_7, yb_1));    //0-->2  // 1->3

		scratch_3 = (-flag)*(S_MUL(scratch_8, ya_1) + S_MUL(scratch_6, yb_1));  



		scratch_4 = flag*(S_MUL(scratch_7, ya_1)-S_MUL(scratch_9, yb_1));  //2-->4  // 3->5

		scratch_5 = flag*(S_MUL(scratch_8, yb_1)-S_MUL(scratch_6, ya_1)); 

#else

		tmp_val = AE_MUL32X16_H0(scratch_9,ya_1);

		tmp_val2 = AE_MUL32X16_H0(scratch_7,yb_1);

	//	scratch_2 = flag*((int)(tmp_val>>15)+(int)(tmp_val2>>15));

		scratch_2 = FFT_MUL_16_32(flag,((int)(tmp_val>>15)+(int)(tmp_val2>>15)));





		tmp_val = AE_MUL32X16_H0(scratch_8,ya_1);

		tmp_val2 = AE_MUL32X16_H0(scratch_6,yb_1);

	//	scratch_3 = (-flag)*((int)(tmp_val>>15)+(int)(tmp_val2>>15));

		scratch_3 = FFT_MUL_16_32((-flag),((int)(tmp_val>>15)+(int)(tmp_val2>>15)));



		tmp_val = AE_MUL32X16_H0(scratch_7,ya_1);

		tmp_val2 = AE_MUL32X16_H0(scratch_9,yb_1);

	//	scratch_4 = flag*((int)(tmp_val>>15)-(int)(tmp_val2>>15));

		scratch_4 = FFT_MUL_16_32(flag,((int)(tmp_val>>15)-(int)(tmp_val2>>15)));

		

		tmp_val = AE_MUL32X16_H0(scratch_8,yb_1);

		tmp_val2 = AE_MUL32X16_H0(scratch_6,ya_1);

	//	scratch_5 = flag*((int)(tmp_val>>15)-(int)(tmp_val2>>15));

		scratch_5 = FFT_MUL_16_32(flag,((int)(tmp_val>>15)-(int)(tmp_val2>>15)));

#endif







		*Fout0++ = SSC_SUB(scratch_10, scratch_2);    // //0-->2  // 1->3

		*Fout0++ = SSC_SUB(scratch_11, scratch_3);  



		*Fout0++ = SSC_ADD(scratch_0, scratch_4); //12-->0  //13-->1  //2-->4  // 3->5

		*Fout0++ = SSC_ADD(scratch_1, scratch_5); 

		*Fout0++ = SSC_SUB(scratch_0, scratch_4);

		*Fout0++ = SSC_SUB(scratch_1, scratch_5);







		*Fout0++ = SSC_ADD(scratch_10, scratch_2);  //0-->2  // 1->3

		*Fout0++ = SSC_ADD(scratch_11, scratch_3); 

#ifndef HW_FFT_5

#ifndef HW_HIFI3

		Fout = Fout + 10*fwd;

#else

		Fout = Fout + SSC_MULT16x16(10,fwd);

#endif

#endif

	}while (--i);

}

#else

static void kf_bfly5(int * Fout, short fwd)

{

	int *Fout0; //,*Fout4;

	short i;



	short flag = -1;



	int scratch_0,scratch_1,scratch_2,scratch_3,scratch_4,scratch_5,scratch_6,scratch_7,scratch_8,scratch_9,scratch_10,scratch_11;



	short ya_0, ya_1, yb_0, yb_1;



	long long tmp_val,tmp_val2;

	ae_valign align1, align2;



	ae_f32x2 tmpL1, tmpL2, tmpL3, tmpL4;

	ae_f32x2 *pt1, *pt2;

	ae_f16x4 tmpS, ya0_b0,yb0_a0,ya1_b1,yb1_a1, flagv;

	ae_f64   tmpLL1, tmpLL2;

	ya_1 = -31164;

	yb_1 = -19261;





	ya_0 = 10127;





	yb_0 = -26509;



	if(fwd)

	{

#if 1

		flag = 1;

		Fout0 = Fout;

		i=30;

		do

		{

			tmp_val = AE_MUL32X16_H0(*Fout0, 6554);    // //  ³ªÁß¿¡  AE_MUL32X2_vector¸¦ ÀÌ¿ëÇؼ­ ¹Ù²ÙÀÚ.  AE_SRAI64X2_vector

			*Fout0++ = (int)(tmp_val>>15);

			tmp_val = AE_MUL32X16_H0(*Fout0, 6554);

			*Fout0++ = (int)(tmp_val>>15);

			tmp_val = AE_MUL32X16_H0(*Fout0, 6554);

			*Fout0++ = (int)(tmp_val>>15);

			tmp_val = AE_MUL32X16_H0(*Fout0, 6554);

			*Fout0++ = (int)(tmp_val>>15);

		} while (--i);



#else

		flag = 1;

		i=30;

		pt1 = (ae_f32x2*)Fout;

		pt2 = (ae_f32x2*)Fout;

		align1 = AE_LA64_PP(pt1);

		align2 = AE_ZALIGN64();

		tmpS = AE_MOVDA16X2(6554, 6554);

		do

		{

			AE_LA32X2_IP(tmpL1, align1, pt1);

			tmpL1 = AE_MULFP32X16X2RAS_L(tmpL1, tmpS);

			AE_SA32X2_IP(tmpL1, align2, pt2);



			AE_LA32X2_IP(tmpL1, align1, pt1);

			tmpL1 = AE_MULFP32X16X2RAS_L(tmpL1, tmpS);

			AE_SA32X2_IP(tmpL1, align2, pt2);



		} while (--i);

		AE_SA64POS_FP(align2, pt2);

#endif





	}







	Fout0 = Fout;



	pt1 = (ae_f32x2*)Fout;

	pt2 = (ae_f32x2*)Fout;

	align1 = AE_LA64_PP(pt1);

	align2 = AE_ZALIGN64();

	i = 12;

	ya0_b0 = AE_MOVDA16X2(ya_0, yb_0);

	yb0_a0 = AE_MOVDA16X2(yb_0, ya_0);

	ya1_b1 = AE_MOVDA16X2(ya_1, yb_1);

	yb1_a1 = AE_MOVDA16X2(yb_1, ya_1);

	flagv =  AE_MOVDA16X2(flag, -flag);

	do {

		ae_f32x2 F01,F23, F45, F67, F89;

		ae_f32x2 S01,S23, S45, S67, S89,S1011, S97, S86;

		ae_f32x2 S24, S35;

		int data1,data2;



		AE_LA32X2_IP(F01, align1, pt1);

		AE_LA32X2_IP(F23, align1, pt1);

		AE_LA32X2_IP(F45, align1, pt1);

		AE_LA32X2_IP(F67, align1, pt1);

		AE_LA32X2_IP(F89, align1, pt1);



//		scratch_0 = Fout0[0];

//		scratch_1 = Fout0[1];



/*		scratch_2 = SSC_ADD(Fout0[2], Fout0[8]);

		scratch_3 = SSC_ADD(Fout0[3], Fout0[9]);

		scratch_8 = SSC_SUB(Fout0[2], Fout0[8]);

		scratch_9 = SSC_SUB(Fout0[3], Fout0[9]);

		scratch_4 = SSC_ADD(Fout0[4], Fout0[6]);

		scratch_5 = SSC_ADD(Fout0[5], Fout0[7]);

		scratch_6 = SSC_SUB(Fout0[4], Fout0[6]);

		scratch_7 = SSC_SUB(Fout0[5], Fout0[7]);

*/



		S01 = AE_MOV32(F01);

		S23 = AE_ADD32(F23, F89);

		S89 = AE_SUB32(F23, F89);

		S45 = AE_ADD32(F45, F67);

		S67 = AE_SUB32(F45, F67);





		F01 = AE_ADD32(F01, S23);

		F01 = AE_ADD32(F01, S45);

		AE_SA32X2_IP(F01, align2, pt2);







		S24 = AE_SEL32_HH(S23, S45);

		S35 = AE_SEL32_LL(S23, S45);





        tmpL1 = AE_MULFP32X16X2RAS_L(S24, ya0_b0);

        tmpL2 = AE_MULFP32X16X2RAS_L(S35, ya0_b0);

        tmpL3 = AE_SEL32_HH(tmpL1, tmpL2);

        tmpL4 = AE_SEL32_LL(tmpL1, tmpL2);

        tmpL3 = AE_ADD32(tmpL3, tmpL4);

        S1011 = AE_ADD32(S01, tmpL3);







        tmpL1 = AE_MULFP32X16X2RAS_L(S24, yb0_a0);

        tmpL2 = AE_MULFP32X16X2RAS_L(S35, yb0_a0);

        tmpL3 = AE_SEL32_HH(tmpL1, tmpL2);

        tmpL4 = AE_SEL32_LL(tmpL1, tmpL2);

        tmpL3 = AE_ADD32(tmpL3, tmpL4);

        S01 = AE_ADD32(S01, tmpL3);







		S97 = AE_SEL32_LL(S89, S67);

		S86 = AE_SEL32_HH(S89, S67);







        tmpL1 = AE_MULFP32X16X2RAS_L(S97, ya1_b1);

        tmpL1 = AE_MULP32X16X2_L(tmpL1, flag);

        tmpL2 = AE_MULFP32X16X2RAS_L(S86, ya1_b1);

        tmpL2 = AE_MULP32X16X2_L(tmpL2, -flag);

        tmpL3 = AE_SEL32_HH(tmpL1, tmpL2);

        tmpL4 = AE_SEL32_LL(tmpL1, tmpL2);

        S23 = AE_ADD32(tmpL3, tmpL4);



        tmpL1 = AE_MULFP32X16X2RAS_L(S97, yb1_a1);

        tmpL1 = AE_MULP32X16X2_L(tmpL1, -flag);

        tmpL2 = AE_MULFP32X16X2RAS_L(S86, yb1_a1);

        tmpL2 = AE_MULP32X16X2_L(tmpL2, flag);

        tmpL3 = AE_SEL32_HH(tmpL1, tmpL2);

        tmpL4 = AE_SEL32_LL(tmpL1, tmpL2);

        S45 = AE_SUB32(tmpL3, tmpL4);







		tmpL1 = AE_SUB32(S1011, S23);

		AE_SA32X2_IP(tmpL1, align2, pt2);



		tmpL1 = AE_ADD32(S01, S45);

		AE_SA32X2_IP(tmpL1, align2, pt2);



		tmpL1 = AE_SUB32(S01, S45);

		AE_SA32X2_IP(tmpL1, align2, pt2);



		tmpL1 = AE_ADD32(S1011, S23);

		AE_SA32X2_IP(tmpL1, align2, pt2);





	}while (--i);

}

#endif







//FILE *fp_check;

void sun_fft(const int *fin,int *fout,short dir)

{

	

	const short *bitrev = fft_bitrev60;

	const int *_fin = fin;

	int *_fout;

	int i;



#ifndef  fft_bitrev_opt_hifi3_ZH

#ifndef HW_FOR_GCC

	i = 15;

	do{

		_fout = fout+(*bitrev++);

		*_fout++ = *_fin++;

		*_fout = *_fin++;

		_fout = _fout+29;



		*_fout++ = *_fin++;

		*_fout = *_fin++;

		_fout = _fout+29;



		*_fout++ = *_fin++;

		*_fout = *_fin++;

		_fout = _fout+29;

		*_fout++ = *_fin++;

		*_fout = *_fin++;

	}while(--i);

#else

	short i;

	for (i=0;i<15;i++)

	{

		fout[bitrev[i]] = fin[(i<<3)];

		fout[bitrev[i]+1] = fin[(i<<3)+1];



		fout[bitrev[i]+30] = fin[(i<<3)+2];

		fout[bitrev[i]+31] = fin[(i<<3)+3];



		fout[bitrev[i]+60] = fin[(i<<3)+4];

		fout[bitrev[i]+61] = fin[(i<<3)+5];



		fout[bitrev[i]+90] = fin[(i<<3)+6];

		fout[bitrev[i]+91] = fin[(i<<3)+7];

	}

#endif

#else

	{

	   ae_f32x2 *pfin, *pfout;

	   pfin = (ae_f32x2*)fin;

	   short i = 15;

	   do{

		   pfout = (ae_f32x2*)(fout+(*bitrev++));



		   *pfout = *pfin++;

		   pfout += 15;



		   *pfout = *pfin++;

		   pfout += 15;



		   *pfout = *pfin++;

		   pfout += 15;



		   *pfout = *pfin++;

		   pfout += 15;



	   }while(--i);

	}

#endif





	kf_bfly5(fout,dir);	

	kf_bfly3(fout,dir);

	kf_bfly4(fout, dir);



}



#endif