diff src/fftw-3.3.3/rdft/simd/common/hc2cbdftv_6.c @ 10:37bf6b4a2645

Add FFTW3

--- /dev/null	Thu Jan 01 00:00:00 1970 +0000
+++ b/src/fftw-3.3.3/rdft/simd/common/hc2cbdftv_6.c	Wed Mar 20 15:35:50 2013 +0000
@@ -0,0 +1,191 @@
+/*
+ * 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 07:42:29 EST 2012 */
+
+#include "codelet-rdft.h"
+
+#ifdef HAVE_FMA
+
+/* Generated by: ../../../genfft/gen_hc2cdft_c.native -fma -reorder-insns -schedule-for-pipeline -simd -compact -variables 4 -pipeline-latency 8 -trivial-stores -variables 32 -no-generate-bytw -n 6 -dif -sign 1 -name hc2cbdftv_6 -include hc2cbv.h */
+
+/*
+ * This function contains 29 FP additions, 24 FP multiplications,
+ * (or, 17 additions, 12 multiplications, 12 fused multiply/add),
+ * 38 stack variables, 2 constants, and 12 memory accesses
+ */
+#include "hc2cbv.h"
+
+static void hc2cbdftv_6(R *Rp, R *Ip, R *Rm, R *Im, const R *W, stride rs, INT mb, INT me, INT ms)
+{
+     DVK(KP500000000, +0.500000000000000000000000000000000000000000000);
+     DVK(KP866025403, +0.866025403784438646763723170752936183471402627);
+     {
+	  INT m;
+	  for (m = mb, W = W + ((mb - 1) * ((TWVL / VL) * 10)); m < me; m = m + VL, Rp = Rp + (VL * ms), Ip = Ip + (VL * ms), Rm = Rm - (VL * ms), Im = Im - (VL * ms), W = W + (TWVL * 10), MAKE_VOLATILE_STRIDE(24, rs)) {
+	       V Tv, Tn, Tr, Te, T4, Tg, Ta, Tf, T7, T1, Td, T2, T3, T8, T9;
+	       V T5, T6, Th, Tj, Tb, Tp, Tx, Ti, Tc, To, Tk, Ts, Tq, Tw, Tm;
+	       V Tl, Tu, Tt, Tz, Ty;
+	       T2 = LD(&(Rp[0]), ms, &(Rp[0]));
+	       T3 = LD(&(Rm[WS(rs, 2)]), -ms, &(Rm[0]));
+	       T8 = LD(&(Rp[WS(rs, 1)]), ms, &(Rp[WS(rs, 1)]));
+	       T9 = LD(&(Rm[WS(rs, 1)]), -ms, &(Rm[WS(rs, 1)]));
+	       T5 = LD(&(Rp[WS(rs, 2)]), ms, &(Rp[0]));
+	       T6 = LD(&(Rm[0]), -ms, &(Rm[0]));
+	       Tv = LDW(&(W[0]));
+	       Tn = LDW(&(W[TWVL * 8]));
+	       Tr = LDW(&(W[TWVL * 6]));
+	       Te = VFMACONJ(T3, T2);
+	       T4 = VFNMSCONJ(T3, T2);
+	       Tg = VFMACONJ(T9, T8);
+	       Ta = VFMSCONJ(T9, T8);
+	       Tf = VFMACONJ(T6, T5);
+	       T7 = VFNMSCONJ(T6, T5);
+	       T1 = LDW(&(W[TWVL * 4]));
+	       Td = LDW(&(W[TWVL * 2]));
+	       Th = VADD(Tf, Tg);
+	       Tj = VMUL(LDK(KP866025403), VSUB(Tf, Tg));
+	       Tb = VADD(T7, Ta);
+	       Tp = VMUL(LDK(KP866025403), VSUB(T7, Ta));
+	       Tx = VADD(Te, Th);
+	       Ti = VFNMS(LDK(KP500000000), Th, Te);
+	       Tc = VZMULI(T1, VADD(T4, Tb));
+	       To = VFNMS(LDK(KP500000000), Tb, T4);
+	       Tk = VZMUL(Td, VFNMSI(Tj, Ti));
+	       Ts = VZMUL(Tr, VFMAI(Tj, Ti));
+	       Tq = VZMULI(Tn, VFNMSI(Tp, To));
+	       Tw = VZMULI(Tv, VFMAI(Tp, To));
+	       Tm = VCONJ(VSUB(Tk, Tc));
+	       Tl = VADD(Tc, Tk);
+	       Tu = VCONJ(VSUB(Ts, Tq));
+	       Tt = VADD(Tq, Ts);
+	       Tz = VCONJ(VSUB(Tx, Tw));
+	       Ty = VADD(Tw, Tx);
+	       ST(&(Rm[WS(rs, 1)]), Tm, -ms, &(Rm[WS(rs, 1)]));
+	       ST(&(Rp[WS(rs, 1)]), Tl, ms, &(Rp[WS(rs, 1)]));
+	       ST(&(Rm[WS(rs, 2)]), Tu, -ms, &(Rm[0]));
+	       ST(&(Rp[WS(rs, 2)]), Tt, ms, &(Rp[0]));
+	       ST(&(Rm[0]), Tz, -ms, &(Rm[0]));
+	       ST(&(Rp[0]), Ty, ms, &(Rp[0]));
+	  }
+     }
+     VLEAVE();
+}
+
+static const tw_instr twinstr[] = {
+     VTW(1, 1),
+     VTW(1, 2),
+     VTW(1, 3),
+     VTW(1, 4),
+     VTW(1, 5),
+     {TW_NEXT, VL, 0}
+};
+
+static const hc2c_desc desc = { 6, XSIMD_STRING("hc2cbdftv_6"), twinstr, &GENUS, {17, 12, 12, 0} };
+
+void XSIMD(codelet_hc2cbdftv_6) (planner *p) {
+     X(khc2c_register) (p, hc2cbdftv_6, &desc, HC2C_VIA_DFT);
+}
+#else				/* HAVE_FMA */
+
+/* Generated by: ../../../genfft/gen_hc2cdft_c.native -simd -compact -variables 4 -pipeline-latency 8 -trivial-stores -variables 32 -no-generate-bytw -n 6 -dif -sign 1 -name hc2cbdftv_6 -include hc2cbv.h */
+
+/*
+ * This function contains 29 FP additions, 14 FP multiplications,
+ * (or, 27 additions, 12 multiplications, 2 fused multiply/add),
+ * 41 stack variables, 2 constants, and 12 memory accesses
+ */
+#include "hc2cbv.h"
+
+static void hc2cbdftv_6(R *Rp, R *Ip, R *Rm, R *Im, const R *W, stride rs, INT mb, INT me, INT ms)
+{
+     DVK(KP500000000, +0.500000000000000000000000000000000000000000000);
+     DVK(KP866025403, +0.866025403784438646763723170752936183471402627);
+     {
+	  INT m;
+	  for (m = mb, W = W + ((mb - 1) * ((TWVL / VL) * 10)); m < me; m = m + VL, Rp = Rp + (VL * ms), Ip = Ip + (VL * ms), Rm = Rm - (VL * ms), Im = Im - (VL * ms), W = W + (TWVL * 10), MAKE_VOLATILE_STRIDE(24, rs)) {
+	       V T5, Th, Te, Ts, Tk, Tm, T2, T4, T3, T6, Tc, T8, Tb, T7, Ta;
+	       V T9, Td, Ti, Tj, TA, Tf, Tn, Tv, Tt, Tz, T1, Tl, Tg, Tu, Tr;
+	       V Tq, Ty, To, Tp, TC, TB, Tx, Tw;
+	       T2 = LD(&(Rp[0]), ms, &(Rp[0]));
+	       T3 = LD(&(Rm[WS(rs, 2)]), -ms, &(Rm[0]));
+	       T4 = VCONJ(T3);
+	       T5 = VSUB(T2, T4);
+	       Th = VADD(T2, T4);
+	       T6 = LD(&(Rp[WS(rs, 2)]), ms, &(Rp[0]));
+	       Tc = LD(&(Rp[WS(rs, 1)]), ms, &(Rp[WS(rs, 1)]));
+	       T7 = LD(&(Rm[0]), -ms, &(Rm[0]));
+	       T8 = VCONJ(T7);
+	       Ta = LD(&(Rm[WS(rs, 1)]), -ms, &(Rm[WS(rs, 1)]));
+	       Tb = VCONJ(Ta);
+	       T9 = VSUB(T6, T8);
+	       Td = VSUB(Tb, Tc);
+	       Te = VADD(T9, Td);
+	       Ts = VBYI(VMUL(LDK(KP866025403), VSUB(T9, Td)));
+	       Ti = VADD(T6, T8);
+	       Tj = VADD(Tb, Tc);
+	       Tk = VADD(Ti, Tj);
+	       Tm = VBYI(VMUL(LDK(KP866025403), VSUB(Ti, Tj)));
+	       TA = VADD(Th, Tk);
+	       T1 = LDW(&(W[TWVL * 4]));
+	       Tf = VZMULI(T1, VADD(T5, Te));
+	       Tl = VFNMS(LDK(KP500000000), Tk, Th);
+	       Tg = LDW(&(W[TWVL * 2]));
+	       Tn = VZMUL(Tg, VSUB(Tl, Tm));
+	       Tu = LDW(&(W[TWVL * 6]));
+	       Tv = VZMUL(Tu, VADD(Tm, Tl));
+	       Tr = VFNMS(LDK(KP500000000), Te, T5);
+	       Tq = LDW(&(W[TWVL * 8]));
+	       Tt = VZMULI(Tq, VSUB(Tr, Ts));
+	       Ty = LDW(&(W[0]));
+	       Tz = VZMULI(Ty, VADD(Ts, Tr));
+	       To = VADD(Tf, Tn);
+	       ST(&(Rp[WS(rs, 1)]), To, ms, &(Rp[WS(rs, 1)]));
+	       Tp = VCONJ(VSUB(Tn, Tf));
+	       ST(&(Rm[WS(rs, 1)]), Tp, -ms, &(Rm[WS(rs, 1)]));
+	       TC = VCONJ(VSUB(TA, Tz));
+	       ST(&(Rm[0]), TC, -ms, &(Rm[0]));
+	       TB = VADD(Tz, TA);
+	       ST(&(Rp[0]), TB, ms, &(Rp[0]));
+	       Tx = VCONJ(VSUB(Tv, Tt));
+	       ST(&(Rm[WS(rs, 2)]), Tx, -ms, &(Rm[0]));
+	       Tw = VADD(Tt, Tv);
+	       ST(&(Rp[WS(rs, 2)]), Tw, ms, &(Rp[0]));
+	  }
+     }
+     VLEAVE();
+}
+
+static const tw_instr twinstr[] = {
+     VTW(1, 1),
+     VTW(1, 2),
+     VTW(1, 3),
+     VTW(1, 4),
+     VTW(1, 5),
+     {TW_NEXT, VL, 0}
+};
+
+static const hc2c_desc desc = { 6, XSIMD_STRING("hc2cbdftv_6"), twinstr, &GENUS, {27, 12, 2, 0} };
+
+void XSIMD(codelet_hc2cbdftv_6) (planner *p) {
+     X(khc2c_register) (p, hc2cbdftv_6, &desc, HC2C_VIA_DFT);
+}
+#endif				/* HAVE_FMA */