diff src/fftw-3.3.3/dft/scalar/codelets/t1_6.c @ 95:89f5e221ed7b

Add FFTW3

--- /dev/null	Thu Jan 01 00:00:00 1970 +0000
+++ b/src/fftw-3.3.3/dft/scalar/codelets/t1_6.c	Wed Mar 20 15:35:50 2013 +0000
@@ -0,0 +1,290 @@
+/*
+ * Copyright (c) 2003, 2007-11 Matteo Frigo
+ * Copyright (c) 2003, 2007-11 Massachusetts Institute of Technology
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA
+ *
+ */
+
+/* This file was automatically generated --- DO NOT EDIT */
+/* Generated on Sun Nov 25 07:35:48 EST 2012 */
+
+#include "codelet-dft.h"
+
+#ifdef HAVE_FMA
+
+/* Generated by: ../../../genfft/gen_twiddle.native -fma -reorder-insns -schedule-for-pipeline -compact -variables 4 -pipeline-latency 4 -n 6 -name t1_6 -include t.h */
+
+/*
+ * This function contains 46 FP additions, 32 FP multiplications,
+ * (or, 24 additions, 10 multiplications, 22 fused multiply/add),
+ * 47 stack variables, 2 constants, and 24 memory accesses
+ */
+#include "t.h"
+
+static void t1_6(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
+{
+     DK(KP866025403, +0.866025403784438646763723170752936183471402627);
+     DK(KP500000000, +0.500000000000000000000000000000000000000000000);
+     {
+	  INT m;
+	  for (m = mb, W = W + (mb * 10); m < me; m = m + 1, ri = ri + ms, ii = ii + ms, W = W + 10, MAKE_VOLATILE_STRIDE(12, rs)) {
+	       E TY, TU, T10, TZ;
+	       {
+		    E T1, TX, TW, T7, Tn, Tq, TJ, TR, TB, Tl, To, TK, Tt, Tw, Ts;
+		    E Tp, Tv;
+		    T1 = ri[0];
+		    TX = ii[0];
+		    {
+			 E T3, T6, T2, T5;
+			 T3 = ri[WS(rs, 3)];
+			 T6 = ii[WS(rs, 3)];
+			 T2 = W[4];
+			 T5 = W[5];
+			 {
+			      E Ta, Td, Tg, TF, Tb, Tj, Tf, Tc, Ti, TV, T4, T9;
+			      Ta = ri[WS(rs, 2)];
+			      Td = ii[WS(rs, 2)];
+			      TV = T2 * T6;
+			      T4 = T2 * T3;
+			      T9 = W[2];
+			      Tg = ri[WS(rs, 5)];
+			      TW = FNMS(T5, T3, TV);
+			      T7 = FMA(T5, T6, T4);
+			      TF = T9 * Td;
+			      Tb = T9 * Ta;
+			      Tj = ii[WS(rs, 5)];
+			      Tf = W[8];
+			      Tc = W[3];
+			      Ti = W[9];
+			      {
+				   E TG, Te, TI, Tk, TH, Th, Tm;
+				   Tn = ri[WS(rs, 4)];
+				   TH = Tf * Tj;
+				   Th = Tf * Tg;
+				   TG = FNMS(Tc, Ta, TF);
+				   Te = FMA(Tc, Td, Tb);
+				   TI = FNMS(Ti, Tg, TH);
+				   Tk = FMA(Ti, Tj, Th);
+				   Tq = ii[WS(rs, 4)];
+				   Tm = W[6];
+				   TJ = TG - TI;
+				   TR = TG + TI;
+				   TB = Te + Tk;
+				   Tl = Te - Tk;
+				   To = Tm * Tn;
+				   TK = Tm * Tq;
+			      }
+			      Tt = ri[WS(rs, 1)];
+			      Tw = ii[WS(rs, 1)];
+			      Ts = W[0];
+			      Tp = W[7];
+			      Tv = W[1];
+			 }
+		    }
+		    {
+			 E TA, T8, TL, Tr, TN, Tx, T11, TM, Tu;
+			 TA = T1 + T7;
+			 T8 = T1 - T7;
+			 TM = Ts * Tw;
+			 Tu = Ts * Tt;
+			 TL = FNMS(Tp, Tn, TK);
+			 Tr = FMA(Tp, Tq, To);
+			 TN = FNMS(Tv, Tt, TM);
+			 Tx = FMA(Tv, Tw, Tu);
+			 T11 = TX - TW;
+			 TY = TW + TX;
+			 {
+			      E TP, TT, TD, TE, TQ, Tz, T14, T13;
+			      {
+				   E TO, TS, TC, Ty, T12;
+				   TO = TL - TN;
+				   TS = TL + TN;
+				   TC = Tr + Tx;
+				   Ty = Tr - Tx;
+				   T12 = TJ + TO;
+				   TP = TJ - TO;
+				   TT = TR - TS;
+				   TU = TR + TS;
+				   Tz = Tl + Ty;
+				   T14 = Ty - Tl;
+				   ii[WS(rs, 3)] = T12 + T11;
+				   T13 = FNMS(KP500000000, T12, T11);
+				   T10 = TC - TB;
+				   TD = TB + TC;
+			      }
+			      ri[WS(rs, 3)] = T8 + Tz;
+			      TE = FNMS(KP500000000, Tz, T8);
+			      ii[WS(rs, 5)] = FNMS(KP866025403, T14, T13);
+			      ii[WS(rs, 1)] = FMA(KP866025403, T14, T13);
+			      TQ = FNMS(KP500000000, TD, TA);
+			      ri[WS(rs, 5)] = FNMS(KP866025403, TP, TE);
+			      ri[WS(rs, 1)] = FMA(KP866025403, TP, TE);
+			      ri[0] = TA + TD;
+			      ri[WS(rs, 4)] = FMA(KP866025403, TT, TQ);
+			      ri[WS(rs, 2)] = FNMS(KP866025403, TT, TQ);
+			 }
+		    }
+	       }
+	       ii[0] = TU + TY;
+	       TZ = FNMS(KP500000000, TU, TY);
+	       ii[WS(rs, 2)] = FNMS(KP866025403, T10, TZ);
+	       ii[WS(rs, 4)] = FMA(KP866025403, T10, TZ);
+	  }
+     }
+}
+
+static const tw_instr twinstr[] = {
+     {TW_FULL, 0, 6},
+     {TW_NEXT, 1, 0}
+};
+
+static const ct_desc desc = { 6, "t1_6", twinstr, &GENUS, {24, 10, 22, 0}, 0, 0, 0 };
+
+void X(codelet_t1_6) (planner *p) {
+     X(kdft_dit_register) (p, t1_6, &desc);
+}
+#else				/* HAVE_FMA */
+
+/* Generated by: ../../../genfft/gen_twiddle.native -compact -variables 4 -pipeline-latency 4 -n 6 -name t1_6 -include t.h */
+
+/*
+ * This function contains 46 FP additions, 28 FP multiplications,
+ * (or, 32 additions, 14 multiplications, 14 fused multiply/add),
+ * 23 stack variables, 2 constants, and 24 memory accesses
+ */
+#include "t.h"
+
+static void t1_6(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
+{
+     DK(KP500000000, +0.500000000000000000000000000000000000000000000);
+     DK(KP866025403, +0.866025403784438646763723170752936183471402627);
+     {
+	  INT m;
+	  for (m = mb, W = W + (mb * 10); m < me; m = m + 1, ri = ri + ms, ii = ii + ms, W = W + 10, MAKE_VOLATILE_STRIDE(12, rs)) {
+	       E T7, TS, Tv, TO, Tt, TJ, Tx, TF, Ti, TI, Tw, TC;
+	       {
+		    E T1, TN, T6, TM;
+		    T1 = ri[0];
+		    TN = ii[0];
+		    {
+			 E T3, T5, T2, T4;
+			 T3 = ri[WS(rs, 3)];
+			 T5 = ii[WS(rs, 3)];
+			 T2 = W[4];
+			 T4 = W[5];
+			 T6 = FMA(T2, T3, T4 * T5);
+			 TM = FNMS(T4, T3, T2 * T5);
+		    }
+		    T7 = T1 - T6;
+		    TS = TN - TM;
+		    Tv = T1 + T6;
+		    TO = TM + TN;
+	       }
+	       {
+		    E Tn, TD, Ts, TE;
+		    {
+			 E Tk, Tm, Tj, Tl;
+			 Tk = ri[WS(rs, 4)];
+			 Tm = ii[WS(rs, 4)];
+			 Tj = W[6];
+			 Tl = W[7];
+			 Tn = FMA(Tj, Tk, Tl * Tm);
+			 TD = FNMS(Tl, Tk, Tj * Tm);
+		    }
+		    {
+			 E Tp, Tr, To, Tq;
+			 Tp = ri[WS(rs, 1)];
+			 Tr = ii[WS(rs, 1)];
+			 To = W[0];
+			 Tq = W[1];
+			 Ts = FMA(To, Tp, Tq * Tr);
+			 TE = FNMS(Tq, Tp, To * Tr);
+		    }
+		    Tt = Tn - Ts;
+		    TJ = TD + TE;
+		    Tx = Tn + Ts;
+		    TF = TD - TE;
+	       }
+	       {
+		    E Tc, TA, Th, TB;
+		    {
+			 E T9, Tb, T8, Ta;
+			 T9 = ri[WS(rs, 2)];
+			 Tb = ii[WS(rs, 2)];
+			 T8 = W[2];
+			 Ta = W[3];
+			 Tc = FMA(T8, T9, Ta * Tb);
+			 TA = FNMS(Ta, T9, T8 * Tb);
+		    }
+		    {
+			 E Te, Tg, Td, Tf;
+			 Te = ri[WS(rs, 5)];
+			 Tg = ii[WS(rs, 5)];
+			 Td = W[8];
+			 Tf = W[9];
+			 Th = FMA(Td, Te, Tf * Tg);
+			 TB = FNMS(Tf, Te, Td * Tg);
+		    }
+		    Ti = Tc - Th;
+		    TI = TA + TB;
+		    Tw = Tc + Th;
+		    TC = TA - TB;
+	       }
+	       {
+		    E TG, Tu, Tz, TR, TT, TU;
+		    TG = KP866025403 * (TC - TF);
+		    Tu = Ti + Tt;
+		    Tz = FNMS(KP500000000, Tu, T7);
+		    ri[WS(rs, 3)] = T7 + Tu;
+		    ri[WS(rs, 1)] = Tz + TG;
+		    ri[WS(rs, 5)] = Tz - TG;
+		    TR = KP866025403 * (Tt - Ti);
+		    TT = TC + TF;
+		    TU = FNMS(KP500000000, TT, TS);
+		    ii[WS(rs, 1)] = TR + TU;
+		    ii[WS(rs, 3)] = TT + TS;
+		    ii[WS(rs, 5)] = TU - TR;
+	       }
+	       {
+		    E TK, Ty, TH, TQ, TL, TP;
+		    TK = KP866025403 * (TI - TJ);
+		    Ty = Tw + Tx;
+		    TH = FNMS(KP500000000, Ty, Tv);
+		    ri[0] = Tv + Ty;
+		    ri[WS(rs, 4)] = TH + TK;
+		    ri[WS(rs, 2)] = TH - TK;
+		    TQ = KP866025403 * (Tx - Tw);
+		    TL = TI + TJ;
+		    TP = FNMS(KP500000000, TL, TO);
+		    ii[0] = TL + TO;
+		    ii[WS(rs, 4)] = TQ + TP;
+		    ii[WS(rs, 2)] = TP - TQ;
+	       }
+	  }
+     }
+}
+
+static const tw_instr twinstr[] = {
+     {TW_FULL, 0, 6},
+     {TW_NEXT, 1, 0}
+};
+
+static const ct_desc desc = { 6, "t1_6", twinstr, &GENUS, {32, 14, 14, 0}, 0, 0, 0 };
+
+void X(codelet_t1_6) (planner *p) {
+     X(kdft_dit_register) (p, t1_6, &desc);
+}
+#endif				/* HAVE_FMA */