Mercurial > hg > sv-dependency-builds

diff src/fftw-3.3.8/dft/simd/common/n2bv_14.c @ 82:d0c2a83c1364

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
Add FFTW 3.3.8 source, and a Linux build
author Chris Cannam
date Tue, 19 Nov 2019 14:52:55 +0000
parents
children
line wrap: on
line diff
--- /dev/null	Thu Jan 01 00:00:00 1970 +0000
+++ b/src/fftw-3.3.8/dft/simd/common/n2bv_14.c	Tue Nov 19 14:52:55 2019 +0000
@@ -0,0 +1,365 @@
+/*
+ * Copyright (c) 2003, 2007-14 Matteo Frigo
+ * Copyright (c) 2003, 2007-14 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 Thu May 24 08:05:11 EDT 2018 */
+
+#include "dft/codelet-dft.h"
+
+#if defined(ARCH_PREFERS_FMA) || defined(ISA_EXTENSION_PREFERS_FMA)
+
+/* Generated by: ../../../genfft/gen_notw_c.native -fma -simd -compact -variables 4 -pipeline-latency 8 -sign 1 -n 14 -name n2bv_14 -with-ostride 2 -include dft/simd/n2b.h -store-multiple 2 */
+
+/*
+ * This function contains 74 FP additions, 48 FP multiplications,
+ * (or, 32 additions, 6 multiplications, 42 fused multiply/add),
+ * 51 stack variables, 6 constants, and 35 memory accesses
+ */
+#include "dft/simd/n2b.h"
+
+static void n2bv_14(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs)
+{
+     DVK(KP801937735, +0.801937735804838252472204639014890102331838324);
+     DVK(KP974927912, +0.974927912181823607018131682993931217232785801);
+     DVK(KP554958132, +0.554958132087371191422194871006410481067288862);
+     DVK(KP900968867, +0.900968867902419126236102319507445051165919162);
+     DVK(KP692021471, +0.692021471630095869627814897002069140197260599);
+     DVK(KP356895867, +0.356895867892209443894399510021300583399127187);
+     {
+	  INT i;
+	  const R *xi;
+	  R *xo;
+	  xi = ii;
+	  xo = io;
+	  for (i = v; i > 0; i = i - VL, xi = xi + (VL * ivs), xo = xo + (VL * ovs), MAKE_VOLATILE_STRIDE(28, is), MAKE_VOLATILE_STRIDE(28, os)) {
+	       V T3, TH, Ts, TV, TW, Tt, Tu, TU, Ta, To, Th, Tp, TC, Tx, TK;
+	       V TQ, TN, TR, T14, TZ, T1, T2;
+	       T1 = LD(&(xi[0]), ivs, &(xi[0]));
+	       T2 = LD(&(xi[WS(is, 7)]), ivs, &(xi[WS(is, 1)]));
+	       T3 = VSUB(T1, T2);
+	       TH = VADD(T1, T2);
+	       {
+		    V T6, TI, T9, TJ, Tn, TP, Tk, TO, Tg, TM, Td, TL;
+		    {
+			 V T4, T5, Ti, Tj;
+			 T4 = LD(&(xi[WS(is, 2)]), ivs, &(xi[0]));
+			 T5 = LD(&(xi[WS(is, 9)]), ivs, &(xi[WS(is, 1)]));
+			 T6 = VSUB(T4, T5);
+			 TI = VADD(T4, T5);
+			 {
+			      V T7, T8, Tl, Tm;
+			      T7 = LD(&(xi[WS(is, 12)]), ivs, &(xi[0]));
+			      T8 = LD(&(xi[WS(is, 5)]), ivs, &(xi[WS(is, 1)]));
+			      T9 = VSUB(T7, T8);
+			      TJ = VADD(T7, T8);
+			      Tl = LD(&(xi[WS(is, 8)]), ivs, &(xi[0]));
+			      Tm = LD(&(xi[WS(is, 1)]), ivs, &(xi[WS(is, 1)]));
+			      Tn = VSUB(Tl, Tm);
+			      TP = VADD(Tl, Tm);
+			 }
+			 Ti = LD(&(xi[WS(is, 6)]), ivs, &(xi[0]));
+			 Tj = LD(&(xi[WS(is, 13)]), ivs, &(xi[WS(is, 1)]));
+			 Tk = VSUB(Ti, Tj);
+			 TO = VADD(Ti, Tj);
+			 {
+			      V Te, Tf, Tb, Tc;
+			      Te = LD(&(xi[WS(is, 10)]), ivs, &(xi[0]));
+			      Tf = LD(&(xi[WS(is, 3)]), ivs, &(xi[WS(is, 1)]));
+			      Tg = VSUB(Te, Tf);
+			      TM = VADD(Te, Tf);
+			      Tb = LD(&(xi[WS(is, 4)]), ivs, &(xi[0]));
+			      Tc = LD(&(xi[WS(is, 11)]), ivs, &(xi[WS(is, 1)]));
+			      Td = VSUB(Tb, Tc);
+			      TL = VADD(Tb, Tc);
+			 }
+		    }
+		    Ts = VSUB(Tk, Tn);
+		    TV = VSUB(TP, TO);
+		    TW = VSUB(TM, TL);
+		    Tt = VSUB(Td, Tg);
+		    Tu = VSUB(T6, T9);
+		    TU = VSUB(TI, TJ);
+		    Ta = VADD(T6, T9);
+		    To = VADD(Tk, Tn);
+		    Th = VADD(Td, Tg);
+		    Tp = VFNMS(LDK(KP356895867), To, Th);
+		    TC = VFNMS(LDK(KP356895867), Th, Ta);
+		    Tx = VFNMS(LDK(KP356895867), Ta, To);
+		    TK = VADD(TI, TJ);
+		    TQ = VADD(TO, TP);
+		    TN = VADD(TL, TM);
+		    TR = VFNMS(LDK(KP356895867), TK, TQ);
+		    T14 = VFNMS(LDK(KP356895867), TQ, TN);
+		    TZ = VFNMS(LDK(KP356895867), TN, TK);
+	       }
+	       {
+		    V T19, T1a, T1b, T1e, T1c, T1g, T1h;
+		    T19 = VADD(T3, VADD(Ta, VADD(Th, To)));
+		    STM2(&(xo[14]), T19, ovs, &(xo[2]));
+		    T1a = VADD(TH, VADD(TK, VADD(TN, TQ)));
+		    STM2(&(xo[0]), T1a, ovs, &(xo[0]));
+		    {
+			 V Tr, Tw, Tq, Tv;
+			 Tq = VFNMS(LDK(KP692021471), Tp, Ta);
+			 Tr = VFNMS(LDK(KP900968867), Tq, T3);
+			 Tv = VFNMS(LDK(KP554958132), Tu, Tt);
+			 Tw = VMUL(LDK(KP974927912), VFNMS(LDK(KP801937735), Tv, Ts));
+			 T1b = VFMAI(Tw, Tr);
+			 STM2(&(xo[6]), T1b, ovs, &(xo[2]));
+			 T1c = VFNMSI(Tw, Tr);
+			 STM2(&(xo[22]), T1c, ovs, &(xo[2]));
+		    }
+		    {
+			 V T16, T18, T15, T17, T1d;
+			 T15 = VFNMS(LDK(KP692021471), T14, TK);
+			 T16 = VFNMS(LDK(KP900968867), T15, TH);
+			 T17 = VFMA(LDK(KP554958132), TU, TW);
+			 T18 = VMUL(LDK(KP974927912), VFNMS(LDK(KP801937735), T17, TV));
+			 T1d = VFNMSI(T18, T16);
+			 STM2(&(xo[20]), T1d, ovs, &(xo[0]));
+			 STN2(&(xo[20]), T1d, T1c, ovs);
+			 T1e = VFMAI(T18, T16);
+			 STM2(&(xo[8]), T1e, ovs, &(xo[0]));
+		    }
+		    {
+			 V Tz, TB, Ty, TA, T1f;
+			 Ty = VFNMS(LDK(KP692021471), Tx, Th);
+			 Tz = VFNMS(LDK(KP900968867), Ty, T3);
+			 TA = VFMA(LDK(KP554958132), Tt, Ts);
+			 TB = VMUL(LDK(KP974927912), VFNMS(LDK(KP801937735), TA, Tu));
+			 T1f = VFNMSI(TB, Tz);
+			 STM2(&(xo[10]), T1f, ovs, &(xo[2]));
+			 STN2(&(xo[8]), T1e, T1f, ovs);
+			 T1g = VFMAI(TB, Tz);
+			 STM2(&(xo[18]), T1g, ovs, &(xo[2]));
+		    }
+		    {
+			 V TT, TY, TS, TX, T1i;
+			 TS = VFNMS(LDK(KP692021471), TR, TN);
+			 TT = VFNMS(LDK(KP900968867), TS, TH);
+			 TX = VFMA(LDK(KP554958132), TW, TV);
+			 TY = VMUL(LDK(KP974927912), VFMA(LDK(KP801937735), TX, TU));
+			 T1h = VFNMSI(TY, TT);
+			 STM2(&(xo[24]), T1h, ovs, &(xo[0]));
+			 T1i = VFMAI(TY, TT);
+			 STM2(&(xo[4]), T1i, ovs, &(xo[0]));
+			 STN2(&(xo[4]), T1i, T1b, ovs);
+		    }
+		    {
+			 V T11, T13, T10, T12, T1j, T1k;
+			 T10 = VFNMS(LDK(KP692021471), TZ, TQ);
+			 T11 = VFNMS(LDK(KP900968867), T10, TH);
+			 T12 = VFNMS(LDK(KP554958132), TV, TU);
+			 T13 = VMUL(LDK(KP974927912), VFNMS(LDK(KP801937735), T12, TW));
+			 T1j = VFNMSI(T13, T11);
+			 STM2(&(xo[16]), T1j, ovs, &(xo[0]));
+			 STN2(&(xo[16]), T1j, T1g, ovs);
+			 T1k = VFMAI(T13, T11);
+			 STM2(&(xo[12]), T1k, ovs, &(xo[0]));
+			 STN2(&(xo[12]), T1k, T19, ovs);
+		    }
+		    {
+			 V TE, TG, TD, TF, T1l, T1m;
+			 TD = VFNMS(LDK(KP692021471), TC, To);
+			 TE = VFNMS(LDK(KP900968867), TD, T3);
+			 TF = VFMA(LDK(KP554958132), Ts, Tu);
+			 TG = VMUL(LDK(KP974927912), VFMA(LDK(KP801937735), TF, Tt));
+			 T1l = VFMAI(TG, TE);
+			 STM2(&(xo[2]), T1l, ovs, &(xo[2]));
+			 STN2(&(xo[0]), T1a, T1l, ovs);
+			 T1m = VFNMSI(TG, TE);
+			 STM2(&(xo[26]), T1m, ovs, &(xo[2]));
+			 STN2(&(xo[24]), T1h, T1m, ovs);
+		    }
+	       }
+	  }
+     }
+     VLEAVE();
+}
+
+static const kdft_desc desc = { 14, XSIMD_STRING("n2bv_14"), {32, 6, 42, 0}, &GENUS, 0, 2, 0, 0 };
+
+void XSIMD(codelet_n2bv_14) (planner *p) {
+     X(kdft_register) (p, n2bv_14, &desc);
+}
+
+#else
+
+/* Generated by: ../../../genfft/gen_notw_c.native -simd -compact -variables 4 -pipeline-latency 8 -sign 1 -n 14 -name n2bv_14 -with-ostride 2 -include dft/simd/n2b.h -store-multiple 2 */
+
+/*
+ * This function contains 74 FP additions, 36 FP multiplications,
+ * (or, 50 additions, 12 multiplications, 24 fused multiply/add),
+ * 41 stack variables, 6 constants, and 35 memory accesses
+ */
+#include "dft/simd/n2b.h"
+
+static void n2bv_14(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs)
+{
+     DVK(KP900968867, +0.900968867902419126236102319507445051165919162);
+     DVK(KP222520933, +0.222520933956314404288902564496794759466355569);
+     DVK(KP623489801, +0.623489801858733530525004884004239810632274731);
+     DVK(KP781831482, +0.781831482468029808708444526674057750232334519);
+     DVK(KP974927912, +0.974927912181823607018131682993931217232785801);
+     DVK(KP433883739, +0.433883739117558120475768332848358754609990728);
+     {
+	  INT i;
+	  const R *xi;
+	  R *xo;
+	  xi = ii;
+	  xo = io;
+	  for (i = v; i > 0; i = i - VL, xi = xi + (VL * ivs), xo = xo + (VL * ovs), MAKE_VOLATILE_STRIDE(28, is), MAKE_VOLATILE_STRIDE(28, os)) {
+	       V Tp, Ty, Tl, TL, Tq, TE, T7, TJ, Ts, TB, Te, TK, Tr, TH, Tn;
+	       V To;
+	       Tn = LD(&(xi[0]), ivs, &(xi[0]));
+	       To = LD(&(xi[WS(is, 7)]), ivs, &(xi[WS(is, 1)]));
+	       Tp = VSUB(Tn, To);
+	       Ty = VADD(Tn, To);
+	       {
+		    V Th, TC, Tk, TD;
+		    {
+			 V Tf, Tg, Ti, Tj;
+			 Tf = LD(&(xi[WS(is, 4)]), ivs, &(xi[0]));
+			 Tg = LD(&(xi[WS(is, 11)]), ivs, &(xi[WS(is, 1)]));
+			 Th = VSUB(Tf, Tg);
+			 TC = VADD(Tf, Tg);
+			 Ti = LD(&(xi[WS(is, 10)]), ivs, &(xi[0]));
+			 Tj = LD(&(xi[WS(is, 3)]), ivs, &(xi[WS(is, 1)]));
+			 Tk = VSUB(Ti, Tj);
+			 TD = VADD(Ti, Tj);
+		    }
+		    Tl = VSUB(Th, Tk);
+		    TL = VSUB(TD, TC);
+		    Tq = VADD(Th, Tk);
+		    TE = VADD(TC, TD);
+	       }
+	       {
+		    V T3, Tz, T6, TA;
+		    {
+			 V T1, T2, T4, T5;
+			 T1 = LD(&(xi[WS(is, 2)]), ivs, &(xi[0]));
+			 T2 = LD(&(xi[WS(is, 9)]), ivs, &(xi[WS(is, 1)]));
+			 T3 = VSUB(T1, T2);
+			 Tz = VADD(T1, T2);
+			 T4 = LD(&(xi[WS(is, 12)]), ivs, &(xi[0]));
+			 T5 = LD(&(xi[WS(is, 5)]), ivs, &(xi[WS(is, 1)]));
+			 T6 = VSUB(T4, T5);
+			 TA = VADD(T4, T5);
+		    }
+		    T7 = VSUB(T3, T6);
+		    TJ = VSUB(Tz, TA);
+		    Ts = VADD(T3, T6);
+		    TB = VADD(Tz, TA);
+	       }
+	       {
+		    V Ta, TF, Td, TG;
+		    {
+			 V T8, T9, Tb, Tc;
+			 T8 = LD(&(xi[WS(is, 6)]), ivs, &(xi[0]));
+			 T9 = LD(&(xi[WS(is, 13)]), ivs, &(xi[WS(is, 1)]));
+			 Ta = VSUB(T8, T9);
+			 TF = VADD(T8, T9);
+			 Tb = LD(&(xi[WS(is, 8)]), ivs, &(xi[0]));
+			 Tc = LD(&(xi[WS(is, 1)]), ivs, &(xi[WS(is, 1)]));
+			 Td = VSUB(Tb, Tc);
+			 TG = VADD(Tb, Tc);
+		    }
+		    Te = VSUB(Ta, Td);
+		    TK = VSUB(TG, TF);
+		    Tr = VADD(Ta, Td);
+		    TH = VADD(TF, TG);
+	       }
+	       {
+		    V TR, TS, TU, TV;
+		    TR = VADD(Tp, VADD(Ts, VADD(Tq, Tr)));
+		    STM2(&(xo[14]), TR, ovs, &(xo[2]));
+		    TS = VADD(Ty, VADD(TB, VADD(TE, TH)));
+		    STM2(&(xo[0]), TS, ovs, &(xo[0]));
+		    {
+			 V TT, Tm, Tt, TQ, TP, TW;
+			 Tm = VBYI(VFMA(LDK(KP433883739), T7, VFNMS(LDK(KP781831482), Tl, VMUL(LDK(KP974927912), Te))));
+			 Tt = VFMA(LDK(KP623489801), Tq, VFNMS(LDK(KP222520933), Tr, VFNMS(LDK(KP900968867), Ts, Tp)));
+			 TT = VADD(Tm, Tt);
+			 STM2(&(xo[6]), TT, ovs, &(xo[2]));
+			 TU = VSUB(Tt, Tm);
+			 STM2(&(xo[22]), TU, ovs, &(xo[2]));
+			 TQ = VBYI(VFMA(LDK(KP974927912), TJ, VFMA(LDK(KP433883739), TL, VMUL(LDK(KP781831482), TK))));
+			 TP = VFMA(LDK(KP623489801), TH, VFNMS(LDK(KP900968867), TE, VFNMS(LDK(KP222520933), TB, Ty)));
+			 TV = VSUB(TP, TQ);
+			 STM2(&(xo[24]), TV, ovs, &(xo[0]));
+			 TW = VADD(TP, TQ);
+			 STM2(&(xo[4]), TW, ovs, &(xo[0]));
+			 STN2(&(xo[4]), TW, TT, ovs);
+		    }
+		    {
+			 V T10, TM, TI, TZ;
+			 {
+			      V Tu, Tv, TX, TY;
+			      Tu = VBYI(VFMA(LDK(KP781831482), T7, VFMA(LDK(KP974927912), Tl, VMUL(LDK(KP433883739), Te))));
+			      Tv = VFMA(LDK(KP623489801), Ts, VFNMS(LDK(KP900968867), Tr, VFNMS(LDK(KP222520933), Tq, Tp)));
+			      TX = VADD(Tu, Tv);
+			      STM2(&(xo[2]), TX, ovs, &(xo[2]));
+			      STN2(&(xo[0]), TS, TX, ovs);
+			      TY = VSUB(Tv, Tu);
+			      STM2(&(xo[26]), TY, ovs, &(xo[2]));
+			      STN2(&(xo[24]), TV, TY, ovs);
+			 }
+			 TM = VBYI(VFNMS(LDK(KP433883739), TK, VFNMS(LDK(KP974927912), TL, VMUL(LDK(KP781831482), TJ))));
+			 TI = VFMA(LDK(KP623489801), TB, VFNMS(LDK(KP900968867), TH, VFNMS(LDK(KP222520933), TE, Ty)));
+			 TZ = VSUB(TI, TM);
+			 STM2(&(xo[12]), TZ, ovs, &(xo[0]));
+			 STN2(&(xo[12]), TZ, TR, ovs);
+			 T10 = VADD(TI, TM);
+			 STM2(&(xo[16]), T10, ovs, &(xo[0]));
+			 {
+			      V T11, TO, TN, T12;
+			      TO = VBYI(VFMA(LDK(KP433883739), TJ, VFNMS(LDK(KP974927912), TK, VMUL(LDK(KP781831482), TL))));
+			      TN = VFMA(LDK(KP623489801), TE, VFNMS(LDK(KP222520933), TH, VFNMS(LDK(KP900968867), TB, Ty)));
+			      T11 = VSUB(TN, TO);
+			      STM2(&(xo[8]), T11, ovs, &(xo[0]));
+			      T12 = VADD(TN, TO);
+			      STM2(&(xo[20]), T12, ovs, &(xo[0]));
+			      STN2(&(xo[20]), T12, TU, ovs);
+			      {
+				   V Tx, Tw, T13, T14;
+				   Tx = VBYI(VFNMS(LDK(KP781831482), Te, VFNMS(LDK(KP433883739), Tl, VMUL(LDK(KP974927912), T7))));
+				   Tw = VFMA(LDK(KP623489801), Tr, VFNMS(LDK(KP900968867), Tq, VFNMS(LDK(KP222520933), Ts, Tp)));
+				   T13 = VSUB(Tw, Tx);
+				   STM2(&(xo[10]), T13, ovs, &(xo[2]));
+				   STN2(&(xo[8]), T11, T13, ovs);
+				   T14 = VADD(Tx, Tw);
+				   STM2(&(xo[18]), T14, ovs, &(xo[2]));
+				   STN2(&(xo[16]), T10, T14, ovs);
+			      }
+			 }
+		    }
+	       }
+	  }
+     }
+     VLEAVE();
+}
+
+static const kdft_desc desc = { 14, XSIMD_STRING("n2bv_14"), {50, 12, 24, 0}, &GENUS, 0, 2, 0, 0 };
+
+void XSIMD(codelet_n2bv_14) (planner *p) {
+     X(kdft_register) (p, n2bv_14, &desc);
+}
+
+#endif