cannam@127: /* cannam@127: * Copyright (c) 2003, 2007-14 Matteo Frigo cannam@127: * Copyright (c) 2003, 2007-14 Massachusetts Institute of Technology cannam@127: * cannam@127: * This program is free software; you can redistribute it and/or modify cannam@127: * it under the terms of the GNU General Public License as published by cannam@127: * the Free Software Foundation; either version 2 of the License, or cannam@127: * (at your option) any later version. cannam@127: * cannam@127: * This program is distributed in the hope that it will be useful, cannam@127: * but WITHOUT ANY WARRANTY; without even the implied warranty of cannam@127: * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the cannam@127: * GNU General Public License for more details. cannam@127: * cannam@127: * You should have received a copy of the GNU General Public License cannam@127: * along with this program; if not, write to the Free Software cannam@127: * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA cannam@127: * cannam@127: */ cannam@127: cannam@127: /* This file was automatically generated --- DO NOT EDIT */ cannam@127: /* Generated on Sat Jul 30 16:39:06 EDT 2016 */ cannam@127: cannam@127: #include "codelet-dft.h" cannam@127: cannam@127: #ifdef HAVE_FMA cannam@127: cannam@127: /* Generated by: ../../../genfft/gen_notw_c.native -fma -reorder-insns -schedule-for-pipeline -simd -compact -variables 4 -pipeline-latency 8 -sign 1 -n 11 -name n1bv_11 -include n1b.h */ cannam@127: cannam@127: /* cannam@127: * This function contains 70 FP additions, 60 FP multiplications, cannam@127: * (or, 15 additions, 5 multiplications, 55 fused multiply/add), cannam@127: * 67 stack variables, 11 constants, and 22 memory accesses cannam@127: */ cannam@127: #include "n1b.h" cannam@127: cannam@127: static void n1bv_11(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs) cannam@127: { cannam@127: DVK(KP959492973, +0.959492973614497389890368057066327699062454848); cannam@127: DVK(KP876768831, +0.876768831002589333891339807079336796764054852); cannam@127: DVK(KP918985947, +0.918985947228994779780736114132655398124909697); cannam@127: DVK(KP989821441, +0.989821441880932732376092037776718787376519372); cannam@127: DVK(KP778434453, +0.778434453334651800608337670740821884709317477); cannam@127: DVK(KP830830026, +0.830830026003772851058548298459246407048009821); cannam@127: DVK(KP372785597, +0.372785597771792209609773152906148328659002598); cannam@127: DVK(KP634356270, +0.634356270682424498893150776899916060542806975); cannam@127: DVK(KP715370323, +0.715370323453429719112414662767260662417897278); cannam@127: DVK(KP342584725, +0.342584725681637509502641509861112333758894680); cannam@127: DVK(KP521108558, +0.521108558113202722944698153526659300680427422); cannam@127: { cannam@127: INT i; cannam@127: const R *xi; cannam@127: R *xo; cannam@127: xi = ii; cannam@127: xo = io; cannam@127: for (i = v; i > 0; i = i - VL, xi = xi + (VL * ivs), xo = xo + (VL * ovs), MAKE_VOLATILE_STRIDE(22, is), MAKE_VOLATILE_STRIDE(22, os)) { cannam@127: V T1, Tb, T4, Tq, Tg, Tm, T7, Tp, Ta, To, Tc, T11; cannam@127: T1 = LD(&(xi[0]), ivs, &(xi[0])); cannam@127: { cannam@127: V T2, T3, Te, Tf; cannam@127: T2 = LD(&(xi[WS(is, 1)]), ivs, &(xi[WS(is, 1)])); cannam@127: T3 = LD(&(xi[WS(is, 10)]), ivs, &(xi[0])); cannam@127: Te = LD(&(xi[WS(is, 5)]), ivs, &(xi[WS(is, 1)])); cannam@127: Tf = LD(&(xi[WS(is, 6)]), ivs, &(xi[0])); cannam@127: { cannam@127: V T5, T6, T8, T9; cannam@127: T5 = LD(&(xi[WS(is, 2)]), ivs, &(xi[0])); cannam@127: T6 = LD(&(xi[WS(is, 9)]), ivs, &(xi[WS(is, 1)])); cannam@127: T8 = LD(&(xi[WS(is, 3)]), ivs, &(xi[WS(is, 1)])); cannam@127: T9 = LD(&(xi[WS(is, 8)]), ivs, &(xi[0])); cannam@127: Tb = LD(&(xi[WS(is, 4)]), ivs, &(xi[0])); cannam@127: T4 = VADD(T2, T3); cannam@127: Tq = VSUB(T2, T3); cannam@127: Tg = VADD(Te, Tf); cannam@127: Tm = VSUB(Te, Tf); cannam@127: T7 = VADD(T5, T6); cannam@127: Tp = VSUB(T5, T6); cannam@127: Ta = VADD(T8, T9); cannam@127: To = VSUB(T8, T9); cannam@127: Tc = LD(&(xi[WS(is, 7)]), ivs, &(xi[WS(is, 1)])); cannam@127: } cannam@127: } cannam@127: T11 = VFMA(LDK(KP521108558), Tm, Tq); cannam@127: { cannam@127: V TA, TS, TE, TW, Td, Tn, Ts, Tw, Tr, Tv, TT, TF; cannam@127: Tr = VFNMS(LDK(KP521108558), Tq, Tp); cannam@127: Tv = VFNMS(LDK(KP342584725), T7, Tg); cannam@127: TA = VFMA(LDK(KP715370323), To, Tq); cannam@127: TS = VFMA(LDK(KP521108558), To, Tm); cannam@127: TE = VFNMS(LDK(KP342584725), T4, Ta); cannam@127: TW = VFNMS(LDK(KP342584725), Ta, T7); cannam@127: Td = VADD(Tb, Tc); cannam@127: Tn = VSUB(Tb, Tc); cannam@127: Ts = VFNMS(LDK(KP715370323), Tr, To); cannam@127: Tw = VFNMS(LDK(KP634356270), Tv, T4); cannam@127: TT = VFNMS(LDK(KP715370323), TS, Tp); cannam@127: TF = VFNMS(LDK(KP634356270), TE, Tg); cannam@127: { cannam@127: V Tu, TV, TD, TL, T14, TP, TZ, Tj, Tz, TI, TB, TJ, TM; cannam@127: TB = VFMA(LDK(KP372785597), Tn, TA); cannam@127: TJ = VFNMS(LDK(KP521108558), Tp, Tn); cannam@127: { cannam@127: V T12, TN, TX, Th; cannam@127: T12 = VFMA(LDK(KP715370323), T11, Tn); cannam@127: ST(&(xo[0]), VADD(Tg, VADD(Td, VADD(Ta, VADD(T7, VADD(T4, T1))))), ovs, &(xo[0])); cannam@127: TN = VFNMS(LDK(KP342584725), Td, T4); cannam@127: TX = VFNMS(LDK(KP634356270), TW, Td); cannam@127: Th = VFNMS(LDK(KP342584725), Tg, Td); cannam@127: { cannam@127: V Tt, Tx, TU, TG; cannam@127: Tt = VFNMS(LDK(KP830830026), Ts, Tn); cannam@127: Tx = VFNMS(LDK(KP778434453), Tw, Ta); cannam@127: TU = VFMA(LDK(KP830830026), TT, Tq); cannam@127: TG = VFNMS(LDK(KP778434453), TF, Td); cannam@127: { cannam@127: V TC, TK, T13, TO; cannam@127: TC = VFNMS(LDK(KP830830026), TB, Tm); cannam@127: TK = VFMA(LDK(KP715370323), TJ, Tm); cannam@127: T13 = VFMA(LDK(KP830830026), T12, Tp); cannam@127: TO = VFNMS(LDK(KP634356270), TN, T7); cannam@127: { cannam@127: V TY, Ti, Ty, TH; cannam@127: TY = VFNMS(LDK(KP778434453), TX, T4); cannam@127: Ti = VFNMS(LDK(KP634356270), Th, Ta); cannam@127: Tu = VMUL(LDK(KP989821441), VFNMS(LDK(KP918985947), Tt, Tm)); cannam@127: Ty = VFNMS(LDK(KP876768831), Tx, Td); cannam@127: TV = VMUL(LDK(KP989821441), VFNMS(LDK(KP918985947), TU, Tn)); cannam@127: TH = VFNMS(LDK(KP876768831), TG, T7); cannam@127: TD = VMUL(LDK(KP989821441), VFMA(LDK(KP918985947), TC, Tp)); cannam@127: TL = VFNMS(LDK(KP830830026), TK, To); cannam@127: T14 = VMUL(LDK(KP989821441), VFMA(LDK(KP918985947), T13, To)); cannam@127: TP = VFNMS(LDK(KP778434453), TO, Tg); cannam@127: TZ = VFNMS(LDK(KP876768831), TY, Tg); cannam@127: Tj = VFNMS(LDK(KP778434453), Ti, T7); cannam@127: Tz = VFNMS(LDK(KP959492973), Ty, T1); cannam@127: TI = VFNMS(LDK(KP959492973), TH, T1); cannam@127: } cannam@127: } cannam@127: } cannam@127: } cannam@127: TM = VMUL(LDK(KP989821441), VFNMS(LDK(KP918985947), TL, Tq)); cannam@127: { cannam@127: V TQ, T10, Tk, TR, Tl; cannam@127: TQ = VFNMS(LDK(KP876768831), TP, Ta); cannam@127: T10 = VFNMS(LDK(KP959492973), TZ, T1); cannam@127: Tk = VFNMS(LDK(KP876768831), Tj, T4); cannam@127: ST(&(xo[WS(os, 7)]), VFMAI(TD, Tz), ovs, &(xo[WS(os, 1)])); cannam@127: ST(&(xo[WS(os, 4)]), VFNMSI(TD, Tz), ovs, &(xo[0])); cannam@127: ST(&(xo[WS(os, 8)]), VFNMSI(TM, TI), ovs, &(xo[0])); cannam@127: ST(&(xo[WS(os, 3)]), VFMAI(TM, TI), ovs, &(xo[WS(os, 1)])); cannam@127: TR = VFNMS(LDK(KP959492973), TQ, T1); cannam@127: ST(&(xo[WS(os, 10)]), VFNMSI(T14, T10), ovs, &(xo[0])); cannam@127: ST(&(xo[WS(os, 1)]), VFMAI(T14, T10), ovs, &(xo[WS(os, 1)])); cannam@127: Tl = VFNMS(LDK(KP959492973), Tk, T1); cannam@127: ST(&(xo[WS(os, 9)]), VFMAI(TV, TR), ovs, &(xo[WS(os, 1)])); cannam@127: ST(&(xo[WS(os, 2)]), VFNMSI(TV, TR), ovs, &(xo[0])); cannam@127: ST(&(xo[WS(os, 6)]), VFNMSI(Tu, Tl), ovs, &(xo[0])); cannam@127: ST(&(xo[WS(os, 5)]), VFMAI(Tu, Tl), ovs, &(xo[WS(os, 1)])); cannam@127: } cannam@127: } cannam@127: } cannam@127: } cannam@127: } cannam@127: VLEAVE(); cannam@127: } cannam@127: cannam@127: static const kdft_desc desc = { 11, XSIMD_STRING("n1bv_11"), {15, 5, 55, 0}, &GENUS, 0, 0, 0, 0 }; cannam@127: cannam@127: void XSIMD(codelet_n1bv_11) (planner *p) { cannam@127: X(kdft_register) (p, n1bv_11, &desc); cannam@127: } cannam@127: cannam@127: #else /* HAVE_FMA */ cannam@127: cannam@127: /* Generated by: ../../../genfft/gen_notw_c.native -simd -compact -variables 4 -pipeline-latency 8 -sign 1 -n 11 -name n1bv_11 -include n1b.h */ cannam@127: cannam@127: /* cannam@127: * This function contains 70 FP additions, 50 FP multiplications, cannam@127: * (or, 30 additions, 10 multiplications, 40 fused multiply/add), cannam@127: * 32 stack variables, 10 constants, and 22 memory accesses cannam@127: */ cannam@127: #include "n1b.h" cannam@127: cannam@127: static void n1bv_11(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs) cannam@127: { cannam@127: DVK(KP959492973, +0.959492973614497389890368057066327699062454848); cannam@127: DVK(KP654860733, +0.654860733945285064056925072466293553183791199); cannam@127: DVK(KP142314838, +0.142314838273285140443792668616369668791051361); cannam@127: DVK(KP415415013, +0.415415013001886425529274149229623203524004910); cannam@127: DVK(KP841253532, +0.841253532831181168861811648919367717513292498); cannam@127: DVK(KP540640817, +0.540640817455597582107635954318691695431770608); cannam@127: DVK(KP909631995, +0.909631995354518371411715383079028460060241051); cannam@127: DVK(KP989821441, +0.989821441880932732376092037776718787376519372); cannam@127: DVK(KP755749574, +0.755749574354258283774035843972344420179717445); cannam@127: DVK(KP281732556, +0.281732556841429697711417915346616899035777899); cannam@127: { cannam@127: INT i; cannam@127: const R *xi; cannam@127: R *xo; cannam@127: xi = ii; cannam@127: xo = io; cannam@127: for (i = v; i > 0; i = i - VL, xi = xi + (VL * ivs), xo = xo + (VL * ovs), MAKE_VOLATILE_STRIDE(22, is), MAKE_VOLATILE_STRIDE(22, os)) { cannam@127: V Th, T3, Tm, Tf, Ti, Tc, Tj, T9, Tk, T6, Tl, Ta, Tb, Ts, Tt; cannam@127: Th = LD(&(xi[0]), ivs, &(xi[0])); cannam@127: { cannam@127: V T1, T2, Td, Te; cannam@127: T1 = LD(&(xi[WS(is, 1)]), ivs, &(xi[WS(is, 1)])); cannam@127: T2 = LD(&(xi[WS(is, 10)]), ivs, &(xi[0])); cannam@127: T3 = VSUB(T1, T2); cannam@127: Tm = VADD(T1, T2); cannam@127: Td = LD(&(xi[WS(is, 2)]), ivs, &(xi[0])); cannam@127: Te = LD(&(xi[WS(is, 9)]), ivs, &(xi[WS(is, 1)])); cannam@127: Tf = VSUB(Td, Te); cannam@127: Ti = VADD(Td, Te); cannam@127: } cannam@127: Ta = LD(&(xi[WS(is, 4)]), ivs, &(xi[0])); cannam@127: Tb = LD(&(xi[WS(is, 7)]), ivs, &(xi[WS(is, 1)])); cannam@127: Tc = VSUB(Ta, Tb); cannam@127: Tj = VADD(Ta, Tb); cannam@127: { cannam@127: V T7, T8, T4, T5; cannam@127: T7 = LD(&(xi[WS(is, 5)]), ivs, &(xi[WS(is, 1)])); cannam@127: T8 = LD(&(xi[WS(is, 6)]), ivs, &(xi[0])); cannam@127: T9 = VSUB(T7, T8); cannam@127: Tk = VADD(T7, T8); cannam@127: T4 = LD(&(xi[WS(is, 3)]), ivs, &(xi[WS(is, 1)])); cannam@127: T5 = LD(&(xi[WS(is, 8)]), ivs, &(xi[0])); cannam@127: T6 = VSUB(T4, T5); cannam@127: Tl = VADD(T4, T5); cannam@127: } cannam@127: ST(&(xo[0]), VADD(Th, VADD(Tm, VADD(Ti, VADD(Tl, VADD(Tj, Tk))))), ovs, &(xo[0])); cannam@127: { cannam@127: V Tg, Tn, Tu, Tv; cannam@127: Tg = VBYI(VFMA(LDK(KP281732556), T3, VFMA(LDK(KP755749574), T6, VFNMS(LDK(KP909631995), Tc, VFNMS(LDK(KP540640817), Tf, VMUL(LDK(KP989821441), T9)))))); cannam@127: Tn = VFMA(LDK(KP841253532), Ti, VFMA(LDK(KP415415013), Tj, VFNMS(LDK(KP142314838), Tk, VFNMS(LDK(KP654860733), Tl, VFNMS(LDK(KP959492973), Tm, Th))))); cannam@127: ST(&(xo[WS(os, 5)]), VADD(Tg, Tn), ovs, &(xo[WS(os, 1)])); cannam@127: ST(&(xo[WS(os, 6)]), VSUB(Tn, Tg), ovs, &(xo[0])); cannam@127: Tu = VBYI(VFMA(LDK(KP755749574), T3, VFMA(LDK(KP540640817), T6, VFNMS(LDK(KP909631995), T9, VFNMS(LDK(KP989821441), Tf, VMUL(LDK(KP281732556), Tc)))))); cannam@127: Tv = VFMA(LDK(KP841253532), Tl, VFMA(LDK(KP415415013), Tk, VFNMS(LDK(KP959492973), Tj, VFNMS(LDK(KP142314838), Ti, VFNMS(LDK(KP654860733), Tm, Th))))); cannam@127: ST(&(xo[WS(os, 4)]), VADD(Tu, Tv), ovs, &(xo[0])); cannam@127: ST(&(xo[WS(os, 7)]), VSUB(Tv, Tu), ovs, &(xo[WS(os, 1)])); cannam@127: } cannam@127: Ts = VBYI(VFMA(LDK(KP909631995), T3, VFNMS(LDK(KP540640817), T9, VFNMS(LDK(KP989821441), Tc, VFNMS(LDK(KP281732556), T6, VMUL(LDK(KP755749574), Tf)))))); cannam@127: Tt = VFMA(LDK(KP415415013), Tm, VFMA(LDK(KP841253532), Tk, VFNMS(LDK(KP142314838), Tj, VFNMS(LDK(KP959492973), Tl, VFNMS(LDK(KP654860733), Ti, Th))))); cannam@127: ST(&(xo[WS(os, 2)]), VADD(Ts, Tt), ovs, &(xo[0])); cannam@127: ST(&(xo[WS(os, 9)]), VSUB(Tt, Ts), ovs, &(xo[WS(os, 1)])); cannam@127: { cannam@127: V Tq, Tr, To, Tp; cannam@127: Tq = VBYI(VFMA(LDK(KP540640817), T3, VFMA(LDK(KP909631995), Tf, VFMA(LDK(KP989821441), T6, VFMA(LDK(KP755749574), Tc, VMUL(LDK(KP281732556), T9)))))); cannam@127: Tr = VFMA(LDK(KP841253532), Tm, VFMA(LDK(KP415415013), Ti, VFNMS(LDK(KP959492973), Tk, VFNMS(LDK(KP654860733), Tj, VFNMS(LDK(KP142314838), Tl, Th))))); cannam@127: ST(&(xo[WS(os, 1)]), VADD(Tq, Tr), ovs, &(xo[WS(os, 1)])); cannam@127: ST(&(xo[WS(os, 10)]), VSUB(Tr, Tq), ovs, &(xo[0])); cannam@127: To = VBYI(VFMA(LDK(KP989821441), T3, VFMA(LDK(KP540640817), Tc, VFNMS(LDK(KP909631995), T6, VFNMS(LDK(KP281732556), Tf, VMUL(LDK(KP755749574), T9)))))); cannam@127: Tp = VFMA(LDK(KP415415013), Tl, VFMA(LDK(KP841253532), Tj, VFNMS(LDK(KP654860733), Tk, VFNMS(LDK(KP959492973), Ti, VFNMS(LDK(KP142314838), Tm, Th))))); cannam@127: ST(&(xo[WS(os, 3)]), VADD(To, Tp), ovs, &(xo[WS(os, 1)])); cannam@127: ST(&(xo[WS(os, 8)]), VSUB(Tp, To), ovs, &(xo[0])); cannam@127: } cannam@127: } cannam@127: } cannam@127: VLEAVE(); cannam@127: } cannam@127: cannam@127: static const kdft_desc desc = { 11, XSIMD_STRING("n1bv_11"), {30, 10, 40, 0}, &GENUS, 0, 0, 0, 0 }; cannam@127: cannam@127: void XSIMD(codelet_n1bv_11) (planner *p) { cannam@127: X(kdft_register) (p, n1bv_11, &desc); cannam@127: } cannam@127: cannam@127: #endif /* HAVE_FMA */