Chris@10: /* Chris@10: * Copyright (c) 2003, 2007-11 Matteo Frigo Chris@10: * Copyright (c) 2003, 2007-11 Massachusetts Institute of Technology Chris@10: * Chris@10: * This program is free software; you can redistribute it and/or modify Chris@10: * it under the terms of the GNU General Public License as published by Chris@10: * the Free Software Foundation; either version 2 of the License, or Chris@10: * (at your option) any later version. Chris@10: * Chris@10: * This program is distributed in the hope that it will be useful, Chris@10: * but WITHOUT ANY WARRANTY; without even the implied warranty of Chris@10: * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the Chris@10: * GNU General Public License for more details. Chris@10: * Chris@10: * You should have received a copy of the GNU General Public License Chris@10: * along with this program; if not, write to the Free Software Chris@10: * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA Chris@10: * Chris@10: */ Chris@10: Chris@10: /* This file was automatically generated --- DO NOT EDIT */ Chris@10: /* Generated on Sun Nov 25 07:37:00 EST 2012 */ Chris@10: Chris@10: #include "codelet-dft.h" Chris@10: Chris@10: #ifdef HAVE_FMA Chris@10: Chris@10: /* 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 */ Chris@10: Chris@10: /* Chris@10: * This function contains 70 FP additions, 60 FP multiplications, Chris@10: * (or, 15 additions, 5 multiplications, 55 fused multiply/add), Chris@10: * 67 stack variables, 11 constants, and 22 memory accesses Chris@10: */ Chris@10: #include "n1b.h" Chris@10: Chris@10: static void n1bv_11(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs) Chris@10: { Chris@10: DVK(KP959492973, +0.959492973614497389890368057066327699062454848); Chris@10: DVK(KP876768831, +0.876768831002589333891339807079336796764054852); Chris@10: DVK(KP918985947, +0.918985947228994779780736114132655398124909697); Chris@10: DVK(KP989821441, +0.989821441880932732376092037776718787376519372); Chris@10: DVK(KP778434453, +0.778434453334651800608337670740821884709317477); Chris@10: DVK(KP830830026, +0.830830026003772851058548298459246407048009821); Chris@10: DVK(KP372785597, +0.372785597771792209609773152906148328659002598); Chris@10: DVK(KP634356270, +0.634356270682424498893150776899916060542806975); Chris@10: DVK(KP715370323, +0.715370323453429719112414662767260662417897278); Chris@10: DVK(KP342584725, +0.342584725681637509502641509861112333758894680); Chris@10: DVK(KP521108558, +0.521108558113202722944698153526659300680427422); Chris@10: { Chris@10: INT i; Chris@10: const R *xi; Chris@10: R *xo; Chris@10: xi = ii; Chris@10: xo = io; Chris@10: 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)) { Chris@10: V T1, Tb, T4, Tq, Tg, Tm, T7, Tp, Ta, To, Tc, T11; Chris@10: T1 = LD(&(xi[0]), ivs, &(xi[0])); Chris@10: { Chris@10: V T2, T3, Te, Tf; Chris@10: T2 = LD(&(xi[WS(is, 1)]), ivs, &(xi[WS(is, 1)])); Chris@10: T3 = LD(&(xi[WS(is, 10)]), ivs, &(xi[0])); Chris@10: Te = LD(&(xi[WS(is, 5)]), ivs, &(xi[WS(is, 1)])); Chris@10: Tf = LD(&(xi[WS(is, 6)]), ivs, &(xi[0])); Chris@10: { Chris@10: V T5, T6, T8, T9; Chris@10: T5 = LD(&(xi[WS(is, 2)]), ivs, &(xi[0])); Chris@10: T6 = LD(&(xi[WS(is, 9)]), ivs, &(xi[WS(is, 1)])); Chris@10: T8 = LD(&(xi[WS(is, 3)]), ivs, &(xi[WS(is, 1)])); Chris@10: T9 = LD(&(xi[WS(is, 8)]), ivs, &(xi[0])); Chris@10: Tb = LD(&(xi[WS(is, 4)]), ivs, &(xi[0])); Chris@10: T4 = VADD(T2, T3); Chris@10: Tq = VSUB(T2, T3); Chris@10: Tg = VADD(Te, Tf); Chris@10: Tm = VSUB(Te, Tf); Chris@10: T7 = VADD(T5, T6); Chris@10: Tp = VSUB(T5, T6); Chris@10: Ta = VADD(T8, T9); Chris@10: To = VSUB(T8, T9); Chris@10: Tc = LD(&(xi[WS(is, 7)]), ivs, &(xi[WS(is, 1)])); Chris@10: } Chris@10: } Chris@10: T11 = VFMA(LDK(KP521108558), Tm, Tq); Chris@10: { Chris@10: V TA, TS, TE, TW, Td, Tn, Ts, Tw, Tr, Tv, TT, TF; Chris@10: Tr = VFNMS(LDK(KP521108558), Tq, Tp); Chris@10: Tv = VFNMS(LDK(KP342584725), T7, Tg); Chris@10: TA = VFMA(LDK(KP715370323), To, Tq); Chris@10: TS = VFMA(LDK(KP521108558), To, Tm); Chris@10: TE = VFNMS(LDK(KP342584725), T4, Ta); Chris@10: TW = VFNMS(LDK(KP342584725), Ta, T7); Chris@10: Td = VADD(Tb, Tc); Chris@10: Tn = VSUB(Tb, Tc); Chris@10: Ts = VFNMS(LDK(KP715370323), Tr, To); Chris@10: Tw = VFNMS(LDK(KP634356270), Tv, T4); Chris@10: TT = VFNMS(LDK(KP715370323), TS, Tp); Chris@10: TF = VFNMS(LDK(KP634356270), TE, Tg); Chris@10: { Chris@10: V Tu, TV, TD, TL, T14, TP, TZ, Tj, Tz, TI, TB, TJ, TM; Chris@10: TB = VFMA(LDK(KP372785597), Tn, TA); Chris@10: TJ = VFNMS(LDK(KP521108558), Tp, Tn); Chris@10: { Chris@10: V T12, TN, TX, Th; Chris@10: T12 = VFMA(LDK(KP715370323), T11, Tn); Chris@10: ST(&(xo[0]), VADD(Tg, VADD(Td, VADD(Ta, VADD(T7, VADD(T4, T1))))), ovs, &(xo[0])); Chris@10: TN = VFNMS(LDK(KP342584725), Td, T4); Chris@10: TX = VFNMS(LDK(KP634356270), TW, Td); Chris@10: Th = VFNMS(LDK(KP342584725), Tg, Td); Chris@10: { Chris@10: V Tt, Tx, TU, TG; Chris@10: Tt = VFNMS(LDK(KP830830026), Ts, Tn); Chris@10: Tx = VFNMS(LDK(KP778434453), Tw, Ta); Chris@10: TU = VFMA(LDK(KP830830026), TT, Tq); Chris@10: TG = VFNMS(LDK(KP778434453), TF, Td); Chris@10: { Chris@10: V TC, TK, T13, TO; Chris@10: TC = VFNMS(LDK(KP830830026), TB, Tm); Chris@10: TK = VFMA(LDK(KP715370323), TJ, Tm); Chris@10: T13 = VFMA(LDK(KP830830026), T12, Tp); Chris@10: TO = VFNMS(LDK(KP634356270), TN, T7); Chris@10: { Chris@10: V TY, Ti, Ty, TH; Chris@10: TY = VFNMS(LDK(KP778434453), TX, T4); Chris@10: Ti = VFNMS(LDK(KP634356270), Th, Ta); Chris@10: Tu = VMUL(LDK(KP989821441), VFNMS(LDK(KP918985947), Tt, Tm)); Chris@10: Ty = VFNMS(LDK(KP876768831), Tx, Td); Chris@10: TV = VMUL(LDK(KP989821441), VFNMS(LDK(KP918985947), TU, Tn)); Chris@10: TH = VFNMS(LDK(KP876768831), TG, T7); Chris@10: TD = VMUL(LDK(KP989821441), VFMA(LDK(KP918985947), TC, Tp)); Chris@10: TL = VFNMS(LDK(KP830830026), TK, To); Chris@10: T14 = VMUL(LDK(KP989821441), VFMA(LDK(KP918985947), T13, To)); Chris@10: TP = VFNMS(LDK(KP778434453), TO, Tg); Chris@10: TZ = VFNMS(LDK(KP876768831), TY, Tg); Chris@10: Tj = VFNMS(LDK(KP778434453), Ti, T7); Chris@10: Tz = VFNMS(LDK(KP959492973), Ty, T1); Chris@10: TI = VFNMS(LDK(KP959492973), TH, T1); Chris@10: } Chris@10: } Chris@10: } Chris@10: } Chris@10: TM = VMUL(LDK(KP989821441), VFNMS(LDK(KP918985947), TL, Tq)); Chris@10: { Chris@10: V TQ, T10, Tk, TR, Tl; Chris@10: TQ = VFNMS(LDK(KP876768831), TP, Ta); Chris@10: T10 = VFNMS(LDK(KP959492973), TZ, T1); Chris@10: Tk = VFNMS(LDK(KP876768831), Tj, T4); Chris@10: ST(&(xo[WS(os, 7)]), VFMAI(TD, Tz), ovs, &(xo[WS(os, 1)])); Chris@10: ST(&(xo[WS(os, 4)]), VFNMSI(TD, Tz), ovs, &(xo[0])); Chris@10: ST(&(xo[WS(os, 8)]), VFNMSI(TM, TI), ovs, &(xo[0])); Chris@10: ST(&(xo[WS(os, 3)]), VFMAI(TM, TI), ovs, &(xo[WS(os, 1)])); Chris@10: TR = VFNMS(LDK(KP959492973), TQ, T1); Chris@10: ST(&(xo[WS(os, 10)]), VFNMSI(T14, T10), ovs, &(xo[0])); Chris@10: ST(&(xo[WS(os, 1)]), VFMAI(T14, T10), ovs, &(xo[WS(os, 1)])); Chris@10: Tl = VFNMS(LDK(KP959492973), Tk, T1); Chris@10: ST(&(xo[WS(os, 9)]), VFMAI(TV, TR), ovs, &(xo[WS(os, 1)])); Chris@10: ST(&(xo[WS(os, 2)]), VFNMSI(TV, TR), ovs, &(xo[0])); Chris@10: ST(&(xo[WS(os, 6)]), VFNMSI(Tu, Tl), ovs, &(xo[0])); Chris@10: ST(&(xo[WS(os, 5)]), VFMAI(Tu, Tl), ovs, &(xo[WS(os, 1)])); Chris@10: } Chris@10: } Chris@10: } Chris@10: } Chris@10: } Chris@10: VLEAVE(); Chris@10: } Chris@10: Chris@10: static const kdft_desc desc = { 11, XSIMD_STRING("n1bv_11"), {15, 5, 55, 0}, &GENUS, 0, 0, 0, 0 }; Chris@10: Chris@10: void XSIMD(codelet_n1bv_11) (planner *p) { Chris@10: X(kdft_register) (p, n1bv_11, &desc); Chris@10: } Chris@10: Chris@10: #else /* HAVE_FMA */ Chris@10: Chris@10: /* Generated by: ../../../genfft/gen_notw_c.native -simd -compact -variables 4 -pipeline-latency 8 -sign 1 -n 11 -name n1bv_11 -include n1b.h */ Chris@10: Chris@10: /* Chris@10: * This function contains 70 FP additions, 50 FP multiplications, Chris@10: * (or, 30 additions, 10 multiplications, 40 fused multiply/add), Chris@10: * 32 stack variables, 10 constants, and 22 memory accesses Chris@10: */ Chris@10: #include "n1b.h" Chris@10: Chris@10: static void n1bv_11(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs) Chris@10: { Chris@10: DVK(KP959492973, +0.959492973614497389890368057066327699062454848); Chris@10: DVK(KP654860733, +0.654860733945285064056925072466293553183791199); Chris@10: DVK(KP142314838, +0.142314838273285140443792668616369668791051361); Chris@10: DVK(KP415415013, +0.415415013001886425529274149229623203524004910); Chris@10: DVK(KP841253532, +0.841253532831181168861811648919367717513292498); Chris@10: DVK(KP540640817, +0.540640817455597582107635954318691695431770608); Chris@10: DVK(KP909631995, +0.909631995354518371411715383079028460060241051); Chris@10: DVK(KP989821441, +0.989821441880932732376092037776718787376519372); Chris@10: DVK(KP755749574, +0.755749574354258283774035843972344420179717445); Chris@10: DVK(KP281732556, +0.281732556841429697711417915346616899035777899); Chris@10: { Chris@10: INT i; Chris@10: const R *xi; Chris@10: R *xo; Chris@10: xi = ii; Chris@10: xo = io; Chris@10: 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)) { Chris@10: V Th, T3, Tm, Tf, Ti, Tc, Tj, T9, Tk, T6, Tl, Ta, Tb, Ts, Tt; Chris@10: Th = LD(&(xi[0]), ivs, &(xi[0])); Chris@10: { Chris@10: V T1, T2, Td, Te; Chris@10: T1 = LD(&(xi[WS(is, 1)]), ivs, &(xi[WS(is, 1)])); Chris@10: T2 = LD(&(xi[WS(is, 10)]), ivs, &(xi[0])); Chris@10: T3 = VSUB(T1, T2); Chris@10: Tm = VADD(T1, T2); Chris@10: Td = LD(&(xi[WS(is, 2)]), ivs, &(xi[0])); Chris@10: Te = LD(&(xi[WS(is, 9)]), ivs, &(xi[WS(is, 1)])); Chris@10: Tf = VSUB(Td, Te); Chris@10: Ti = VADD(Td, Te); Chris@10: } Chris@10: Ta = LD(&(xi[WS(is, 4)]), ivs, &(xi[0])); Chris@10: Tb = LD(&(xi[WS(is, 7)]), ivs, &(xi[WS(is, 1)])); Chris@10: Tc = VSUB(Ta, Tb); Chris@10: Tj = VADD(Ta, Tb); Chris@10: { Chris@10: V T7, T8, T4, T5; Chris@10: T7 = LD(&(xi[WS(is, 5)]), ivs, &(xi[WS(is, 1)])); Chris@10: T8 = LD(&(xi[WS(is, 6)]), ivs, &(xi[0])); Chris@10: T9 = VSUB(T7, T8); Chris@10: Tk = VADD(T7, T8); Chris@10: T4 = LD(&(xi[WS(is, 3)]), ivs, &(xi[WS(is, 1)])); Chris@10: T5 = LD(&(xi[WS(is, 8)]), ivs, &(xi[0])); Chris@10: T6 = VSUB(T4, T5); Chris@10: Tl = VADD(T4, T5); Chris@10: } Chris@10: ST(&(xo[0]), VADD(Th, VADD(Tm, VADD(Ti, VADD(Tl, VADD(Tj, Tk))))), ovs, &(xo[0])); Chris@10: { Chris@10: V Tg, Tn, Tu, Tv; Chris@10: Tg = VBYI(VFMA(LDK(KP281732556), T3, VFMA(LDK(KP755749574), T6, VFNMS(LDK(KP909631995), Tc, VFNMS(LDK(KP540640817), Tf, VMUL(LDK(KP989821441), T9)))))); Chris@10: Tn = VFMA(LDK(KP841253532), Ti, VFMA(LDK(KP415415013), Tj, VFNMS(LDK(KP142314838), Tk, VFNMS(LDK(KP654860733), Tl, VFNMS(LDK(KP959492973), Tm, Th))))); Chris@10: ST(&(xo[WS(os, 5)]), VADD(Tg, Tn), ovs, &(xo[WS(os, 1)])); Chris@10: ST(&(xo[WS(os, 6)]), VSUB(Tn, Tg), ovs, &(xo[0])); Chris@10: Tu = VBYI(VFMA(LDK(KP755749574), T3, VFMA(LDK(KP540640817), T6, VFNMS(LDK(KP909631995), T9, VFNMS(LDK(KP989821441), Tf, VMUL(LDK(KP281732556), Tc)))))); Chris@10: Tv = VFMA(LDK(KP841253532), Tl, VFMA(LDK(KP415415013), Tk, VFNMS(LDK(KP959492973), Tj, VFNMS(LDK(KP142314838), Ti, VFNMS(LDK(KP654860733), Tm, Th))))); Chris@10: ST(&(xo[WS(os, 4)]), VADD(Tu, Tv), ovs, &(xo[0])); Chris@10: ST(&(xo[WS(os, 7)]), VSUB(Tv, Tu), ovs, &(xo[WS(os, 1)])); Chris@10: } Chris@10: Ts = VBYI(VFMA(LDK(KP909631995), T3, VFNMS(LDK(KP540640817), T9, VFNMS(LDK(KP989821441), Tc, VFNMS(LDK(KP281732556), T6, VMUL(LDK(KP755749574), Tf)))))); Chris@10: Tt = VFMA(LDK(KP415415013), Tm, VFMA(LDK(KP841253532), Tk, VFNMS(LDK(KP142314838), Tj, VFNMS(LDK(KP959492973), Tl, VFNMS(LDK(KP654860733), Ti, Th))))); Chris@10: ST(&(xo[WS(os, 2)]), VADD(Ts, Tt), ovs, &(xo[0])); Chris@10: ST(&(xo[WS(os, 9)]), VSUB(Tt, Ts), ovs, &(xo[WS(os, 1)])); Chris@10: { Chris@10: V Tq, Tr, To, Tp; Chris@10: Tq = VBYI(VFMA(LDK(KP540640817), T3, VFMA(LDK(KP909631995), Tf, VFMA(LDK(KP989821441), T6, VFMA(LDK(KP755749574), Tc, VMUL(LDK(KP281732556), T9)))))); Chris@10: Tr = VFMA(LDK(KP841253532), Tm, VFMA(LDK(KP415415013), Ti, VFNMS(LDK(KP959492973), Tk, VFNMS(LDK(KP654860733), Tj, VFNMS(LDK(KP142314838), Tl, Th))))); Chris@10: ST(&(xo[WS(os, 1)]), VADD(Tq, Tr), ovs, &(xo[WS(os, 1)])); Chris@10: ST(&(xo[WS(os, 10)]), VSUB(Tr, Tq), ovs, &(xo[0])); Chris@10: To = VBYI(VFMA(LDK(KP989821441), T3, VFMA(LDK(KP540640817), Tc, VFNMS(LDK(KP909631995), T6, VFNMS(LDK(KP281732556), Tf, VMUL(LDK(KP755749574), T9)))))); Chris@10: Tp = VFMA(LDK(KP415415013), Tl, VFMA(LDK(KP841253532), Tj, VFNMS(LDK(KP654860733), Tk, VFNMS(LDK(KP959492973), Ti, VFNMS(LDK(KP142314838), Tm, Th))))); Chris@10: ST(&(xo[WS(os, 3)]), VADD(To, Tp), ovs, &(xo[WS(os, 1)])); Chris@10: ST(&(xo[WS(os, 8)]), VSUB(Tp, To), ovs, &(xo[0])); Chris@10: } Chris@10: } Chris@10: } Chris@10: VLEAVE(); Chris@10: } Chris@10: Chris@10: static const kdft_desc desc = { 11, XSIMD_STRING("n1bv_11"), {30, 10, 40, 0}, &GENUS, 0, 0, 0, 0 }; Chris@10: Chris@10: void XSIMD(codelet_n1bv_11) (planner *p) { Chris@10: X(kdft_register) (p, n1bv_11, &desc); Chris@10: } Chris@10: Chris@10: #endif /* HAVE_FMA */