cannam@95: /* cannam@95: * Copyright (c) 2003, 2007-11 Matteo Frigo cannam@95: * Copyright (c) 2003, 2007-11 Massachusetts Institute of Technology cannam@95: * cannam@95: * This program is free software; you can redistribute it and/or modify cannam@95: * it under the terms of the GNU General Public License as published by cannam@95: * the Free Software Foundation; either version 2 of the License, or cannam@95: * (at your option) any later version. cannam@95: * cannam@95: * This program is distributed in the hope that it will be useful, cannam@95: * but WITHOUT ANY WARRANTY; without even the implied warranty of cannam@95: * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the cannam@95: * GNU General Public License for more details. cannam@95: * cannam@95: * You should have received a copy of the GNU General Public License cannam@95: * along with this program; if not, write to the Free Software cannam@95: * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA cannam@95: * cannam@95: */ cannam@95: cannam@95: /* This file was automatically generated --- DO NOT EDIT */ cannam@95: /* Generated on Sun Nov 25 07:36:52 EST 2012 */ cannam@95: cannam@95: #include "codelet-dft.h" cannam@95: cannam@95: #ifdef HAVE_FMA cannam@95: cannam@95: /* Generated by: ../../../genfft/gen_notw_c.native -fma -reorder-insns -schedule-for-pipeline -simd -compact -variables 4 -pipeline-latency 8 -n 9 -name n1fv_9 -include n1f.h */ cannam@95: cannam@95: /* cannam@95: * This function contains 46 FP additions, 38 FP multiplications, cannam@95: * (or, 12 additions, 4 multiplications, 34 fused multiply/add), cannam@95: * 68 stack variables, 19 constants, and 18 memory accesses cannam@95: */ cannam@95: #include "n1f.h" cannam@95: cannam@95: static void n1fv_9(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs) cannam@95: { cannam@95: DVK(KP939692620, +0.939692620785908384054109277324731469936208134); cannam@95: DVK(KP826351822, +0.826351822333069651148283373230685203999624323); cannam@95: DVK(KP879385241, +0.879385241571816768108218554649462939872416269); cannam@95: DVK(KP984807753, +0.984807753012208059366743024589523013670643252); cannam@95: DVK(KP666666666, +0.666666666666666666666666666666666666666666667); cannam@95: DVK(KP852868531, +0.852868531952443209628250963940074071936020296); cannam@95: DVK(KP907603734, +0.907603734547952313649323976213898122064543220); cannam@95: DVK(KP420276625, +0.420276625461206169731530603237061658838781920); cannam@95: DVK(KP673648177, +0.673648177666930348851716626769314796000375677); cannam@95: DVK(KP898197570, +0.898197570222573798468955502359086394667167570); cannam@95: DVK(KP347296355, +0.347296355333860697703433253538629592000751354); cannam@95: DVK(KP866025403, +0.866025403784438646763723170752936183471402627); cannam@95: DVK(KP439692620, +0.439692620785908384054109277324731469936208134); cannam@95: DVK(KP203604859, +0.203604859554852403062088995281827210665664861); cannam@95: DVK(KP152703644, +0.152703644666139302296566746461370407999248646); cannam@95: DVK(KP586256827, +0.586256827714544512072145703099641959914944179); cannam@95: DVK(KP968908795, +0.968908795874236621082202410917456709164223497); cannam@95: DVK(KP726681596, +0.726681596905677465811651808188092531873167623); cannam@95: DVK(KP500000000, +0.500000000000000000000000000000000000000000000); cannam@95: { cannam@95: INT i; cannam@95: const R *xi; cannam@95: R *xo; cannam@95: xi = ri; cannam@95: xo = ro; cannam@95: for (i = v; i > 0; i = i - VL, xi = xi + (VL * ivs), xo = xo + (VL * ovs), MAKE_VOLATILE_STRIDE(18, is), MAKE_VOLATILE_STRIDE(18, os)) { cannam@95: V T1, T2, T3, T6, Tb, T7, T8, Tc, Td, Tv, T4; cannam@95: T1 = LD(&(xi[0]), ivs, &(xi[0])); cannam@95: T2 = LD(&(xi[WS(is, 3)]), ivs, &(xi[WS(is, 1)])); cannam@95: T3 = LD(&(xi[WS(is, 6)]), ivs, &(xi[0])); cannam@95: T6 = LD(&(xi[WS(is, 1)]), ivs, &(xi[WS(is, 1)])); cannam@95: Tb = LD(&(xi[WS(is, 2)]), ivs, &(xi[0])); cannam@95: T7 = LD(&(xi[WS(is, 4)]), ivs, &(xi[0])); cannam@95: T8 = LD(&(xi[WS(is, 7)]), ivs, &(xi[WS(is, 1)])); cannam@95: Tc = LD(&(xi[WS(is, 5)]), ivs, &(xi[WS(is, 1)])); cannam@95: Td = LD(&(xi[WS(is, 8)]), ivs, &(xi[0])); cannam@95: Tv = VSUB(T3, T2); cannam@95: T4 = VADD(T2, T3); cannam@95: { cannam@95: V Tl, T9, Tm, Te, Tj, T5; cannam@95: Tl = VSUB(T7, T8); cannam@95: T9 = VADD(T7, T8); cannam@95: Tm = VSUB(Td, Tc); cannam@95: Te = VADD(Tc, Td); cannam@95: Tj = VFNMS(LDK(KP500000000), T4, T1); cannam@95: T5 = VADD(T1, T4); cannam@95: { cannam@95: V Tn, Ta, Tk, Tf; cannam@95: Tn = VFNMS(LDK(KP500000000), T9, T6); cannam@95: Ta = VADD(T6, T9); cannam@95: Tk = VFNMS(LDK(KP500000000), Te, Tb); cannam@95: Tf = VADD(Tb, Te); cannam@95: { cannam@95: V Ty, TC, To, TB, Tx, Ts, Tg, Ti; cannam@95: Ty = VFNMS(LDK(KP726681596), Tl, Tn); cannam@95: TC = VFMA(LDK(KP968908795), Tn, Tl); cannam@95: To = VFNMS(LDK(KP586256827), Tn, Tm); cannam@95: TB = VFNMS(LDK(KP152703644), Tm, Tk); cannam@95: Tx = VFMA(LDK(KP203604859), Tk, Tm); cannam@95: Ts = VFNMS(LDK(KP439692620), Tl, Tk); cannam@95: Tg = VADD(Ta, Tf); cannam@95: Ti = VMUL(LDK(KP866025403), VSUB(Tf, Ta)); cannam@95: { cannam@95: V Tz, TI, TF, TD, Tt, Th, Tq, Tp; cannam@95: Tp = VFNMS(LDK(KP347296355), To, Tl); cannam@95: Tz = VFMA(LDK(KP898197570), Ty, Tx); cannam@95: TI = VFNMS(LDK(KP898197570), Ty, Tx); cannam@95: TF = VFNMS(LDK(KP673648177), TC, TB); cannam@95: TD = VFMA(LDK(KP673648177), TC, TB); cannam@95: Tt = VFNMS(LDK(KP420276625), Ts, Tm); cannam@95: ST(&(xo[0]), VADD(T5, Tg), ovs, &(xo[0])); cannam@95: Th = VFNMS(LDK(KP500000000), Tg, T5); cannam@95: Tq = VFNMS(LDK(KP907603734), Tp, Tk); cannam@95: { cannam@95: V TA, TJ, TE, TG, Tu, Tr, TK, TH, Tw; cannam@95: TA = VFMA(LDK(KP852868531), Tz, Tj); cannam@95: TJ = VFMA(LDK(KP666666666), TD, TI); cannam@95: TE = VMUL(LDK(KP984807753), VFNMS(LDK(KP879385241), Tv, TD)); cannam@95: TG = VFNMS(LDK(KP500000000), Tz, TF); cannam@95: Tu = VFNMS(LDK(KP826351822), Tt, Tn); cannam@95: ST(&(xo[WS(os, 6)]), VFNMSI(Ti, Th), ovs, &(xo[0])); cannam@95: ST(&(xo[WS(os, 3)]), VFMAI(Ti, Th), ovs, &(xo[WS(os, 1)])); cannam@95: Tr = VFNMS(LDK(KP939692620), Tq, Tj); cannam@95: TK = VMUL(LDK(KP866025403), VFMA(LDK(KP852868531), TJ, Tv)); cannam@95: ST(&(xo[WS(os, 8)]), VFMAI(TE, TA), ovs, &(xo[0])); cannam@95: ST(&(xo[WS(os, 1)]), VFNMSI(TE, TA), ovs, &(xo[WS(os, 1)])); cannam@95: TH = VFMA(LDK(KP852868531), TG, Tj); cannam@95: Tw = VMUL(LDK(KP984807753), VFMA(LDK(KP879385241), Tv, Tu)); cannam@95: ST(&(xo[WS(os, 4)]), VFMAI(TK, TH), ovs, &(xo[0])); cannam@95: ST(&(xo[WS(os, 5)]), VFNMSI(TK, TH), ovs, &(xo[WS(os, 1)])); cannam@95: ST(&(xo[WS(os, 7)]), VFMAI(Tw, Tr), ovs, &(xo[WS(os, 1)])); cannam@95: ST(&(xo[WS(os, 2)]), VFNMSI(Tw, Tr), ovs, &(xo[0])); cannam@95: } cannam@95: } cannam@95: } cannam@95: } cannam@95: } cannam@95: } cannam@95: } cannam@95: VLEAVE(); cannam@95: } cannam@95: cannam@95: static const kdft_desc desc = { 9, XSIMD_STRING("n1fv_9"), {12, 4, 34, 0}, &GENUS, 0, 0, 0, 0 }; cannam@95: cannam@95: void XSIMD(codelet_n1fv_9) (planner *p) { cannam@95: X(kdft_register) (p, n1fv_9, &desc); cannam@95: } cannam@95: cannam@95: #else /* HAVE_FMA */ cannam@95: cannam@95: /* Generated by: ../../../genfft/gen_notw_c.native -simd -compact -variables 4 -pipeline-latency 8 -n 9 -name n1fv_9 -include n1f.h */ cannam@95: cannam@95: /* cannam@95: * This function contains 46 FP additions, 26 FP multiplications, cannam@95: * (or, 30 additions, 10 multiplications, 16 fused multiply/add), cannam@95: * 41 stack variables, 14 constants, and 18 memory accesses cannam@95: */ cannam@95: #include "n1f.h" cannam@95: cannam@95: static void n1fv_9(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs) cannam@95: { cannam@95: DVK(KP342020143, +0.342020143325668733044099614682259580763083368); cannam@95: DVK(KP813797681, +0.813797681349373692844693217248393223289101568); cannam@95: DVK(KP939692620, +0.939692620785908384054109277324731469936208134); cannam@95: DVK(KP296198132, +0.296198132726023843175338011893050938967728390); cannam@95: DVK(KP642787609, +0.642787609686539326322643409907263432907559884); cannam@95: DVK(KP663413948, +0.663413948168938396205421319635891297216863310); cannam@95: DVK(KP556670399, +0.556670399226419366452912952047023132968291906); cannam@95: DVK(KP766044443, +0.766044443118978035202392650555416673935832457); cannam@95: DVK(KP984807753, +0.984807753012208059366743024589523013670643252); cannam@95: DVK(KP150383733, +0.150383733180435296639271897612501926072238258); cannam@95: DVK(KP852868531, +0.852868531952443209628250963940074071936020296); cannam@95: DVK(KP173648177, +0.173648177666930348851716626769314796000375677); cannam@95: DVK(KP500000000, +0.500000000000000000000000000000000000000000000); cannam@95: DVK(KP866025403, +0.866025403784438646763723170752936183471402627); cannam@95: { cannam@95: INT i; cannam@95: const R *xi; cannam@95: R *xo; cannam@95: xi = ri; cannam@95: xo = ro; cannam@95: for (i = v; i > 0; i = i - VL, xi = xi + (VL * ivs), xo = xo + (VL * ovs), MAKE_VOLATILE_STRIDE(18, is), MAKE_VOLATILE_STRIDE(18, os)) { cannam@95: V T5, Ts, Tj, To, Tf, Tn, Tp, Tu, Tl, Ta, Tk, Tm, Tt; cannam@95: { cannam@95: V T1, T2, T3, T4; cannam@95: T1 = LD(&(xi[0]), ivs, &(xi[0])); cannam@95: T2 = LD(&(xi[WS(is, 3)]), ivs, &(xi[WS(is, 1)])); cannam@95: T3 = LD(&(xi[WS(is, 6)]), ivs, &(xi[0])); cannam@95: T4 = VADD(T2, T3); cannam@95: T5 = VADD(T1, T4); cannam@95: Ts = VMUL(LDK(KP866025403), VSUB(T3, T2)); cannam@95: Tj = VFNMS(LDK(KP500000000), T4, T1); cannam@95: } cannam@95: { cannam@95: V Tb, Te, Tc, Td; cannam@95: Tb = LD(&(xi[WS(is, 2)]), ivs, &(xi[0])); cannam@95: Tc = LD(&(xi[WS(is, 5)]), ivs, &(xi[WS(is, 1)])); cannam@95: Td = LD(&(xi[WS(is, 8)]), ivs, &(xi[0])); cannam@95: Te = VADD(Tc, Td); cannam@95: To = VSUB(Td, Tc); cannam@95: Tf = VADD(Tb, Te); cannam@95: Tn = VFNMS(LDK(KP500000000), Te, Tb); cannam@95: Tp = VFMA(LDK(KP173648177), Tn, VMUL(LDK(KP852868531), To)); cannam@95: Tu = VFNMS(LDK(KP984807753), Tn, VMUL(LDK(KP150383733), To)); cannam@95: } cannam@95: { cannam@95: V T6, T9, T7, T8; cannam@95: T6 = LD(&(xi[WS(is, 1)]), ivs, &(xi[WS(is, 1)])); cannam@95: T7 = LD(&(xi[WS(is, 4)]), ivs, &(xi[0])); cannam@95: T8 = LD(&(xi[WS(is, 7)]), ivs, &(xi[WS(is, 1)])); cannam@95: T9 = VADD(T7, T8); cannam@95: Tl = VSUB(T8, T7); cannam@95: Ta = VADD(T6, T9); cannam@95: Tk = VFNMS(LDK(KP500000000), T9, T6); cannam@95: Tm = VFMA(LDK(KP766044443), Tk, VMUL(LDK(KP556670399), Tl)); cannam@95: Tt = VFNMS(LDK(KP642787609), Tk, VMUL(LDK(KP663413948), Tl)); cannam@95: } cannam@95: { cannam@95: V Ti, Tg, Th, Tz, TA; cannam@95: Ti = VBYI(VMUL(LDK(KP866025403), VSUB(Tf, Ta))); cannam@95: Tg = VADD(Ta, Tf); cannam@95: Th = VFNMS(LDK(KP500000000), Tg, T5); cannam@95: ST(&(xo[0]), VADD(T5, Tg), ovs, &(xo[0])); cannam@95: ST(&(xo[WS(os, 3)]), VADD(Th, Ti), ovs, &(xo[WS(os, 1)])); cannam@95: ST(&(xo[WS(os, 6)]), VSUB(Th, Ti), ovs, &(xo[0])); cannam@95: Tz = VFMA(LDK(KP173648177), Tk, VFNMS(LDK(KP296198132), To, VFNMS(LDK(KP939692620), Tn, VFNMS(LDK(KP852868531), Tl, Tj)))); cannam@95: TA = VBYI(VSUB(VFNMS(LDK(KP342020143), Tn, VFNMS(LDK(KP150383733), Tl, VFNMS(LDK(KP984807753), Tk, VMUL(LDK(KP813797681), To)))), Ts)); cannam@95: ST(&(xo[WS(os, 7)]), VSUB(Tz, TA), ovs, &(xo[WS(os, 1)])); cannam@95: ST(&(xo[WS(os, 2)]), VADD(Tz, TA), ovs, &(xo[0])); cannam@95: { cannam@95: V Tr, Tx, Tw, Ty, Tq, Tv; cannam@95: Tq = VADD(Tm, Tp); cannam@95: Tr = VADD(Tj, Tq); cannam@95: Tx = VFMA(LDK(KP866025403), VSUB(Tt, Tu), VFNMS(LDK(KP500000000), Tq, Tj)); cannam@95: Tv = VADD(Tt, Tu); cannam@95: Tw = VBYI(VADD(Ts, Tv)); cannam@95: Ty = VBYI(VADD(Ts, VFNMS(LDK(KP500000000), Tv, VMUL(LDK(KP866025403), VSUB(Tp, Tm))))); cannam@95: ST(&(xo[WS(os, 8)]), VSUB(Tr, Tw), ovs, &(xo[0])); cannam@95: ST(&(xo[WS(os, 4)]), VADD(Tx, Ty), ovs, &(xo[0])); cannam@95: ST(&(xo[WS(os, 1)]), VADD(Tw, Tr), ovs, &(xo[WS(os, 1)])); cannam@95: ST(&(xo[WS(os, 5)]), VSUB(Tx, Ty), ovs, &(xo[WS(os, 1)])); cannam@95: } cannam@95: } cannam@95: } cannam@95: } cannam@95: VLEAVE(); cannam@95: } cannam@95: cannam@95: static const kdft_desc desc = { 9, XSIMD_STRING("n1fv_9"), {30, 10, 16, 0}, &GENUS, 0, 0, 0, 0 }; cannam@95: cannam@95: void XSIMD(codelet_n1fv_9) (planner *p) { cannam@95: X(kdft_register) (p, n1fv_9, &desc); cannam@95: } cannam@95: cannam@95: #endif /* HAVE_FMA */