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 07:39:45 EST 2012 */ Chris@10: Chris@10: #include "codelet-rdft.h" Chris@10: Chris@10: #ifdef HAVE_FMA Chris@10: Chris@10: /* Generated by: ../../../genfft/gen_r2cf.native -fma -reorder-insns -schedule-for-pipeline -compact -variables 4 -pipeline-latency 4 -n 9 -name r2cf_9 -include r2cf.h */ Chris@10: Chris@10: /* Chris@10: * This function contains 38 FP additions, 30 FP multiplications, Chris@10: * (or, 12 additions, 4 multiplications, 26 fused multiply/add), Chris@10: * 57 stack variables, 18 constants, and 18 memory accesses Chris@10: */ Chris@10: #include "r2cf.h" Chris@10: Chris@10: static void r2cf_9(R *R0, R *R1, R *Cr, R *Ci, stride rs, stride csr, stride csi, INT v, INT ivs, INT ovs) Chris@10: { Chris@10: DK(KP907603734, +0.907603734547952313649323976213898122064543220); Chris@10: DK(KP852868531, +0.852868531952443209628250963940074071936020296); Chris@10: DK(KP347296355, +0.347296355333860697703433253538629592000751354); Chris@10: DK(KP666666666, +0.666666666666666666666666666666666666666666667); Chris@10: DK(KP879385241, +0.879385241571816768108218554649462939872416269); Chris@10: DK(KP984807753, +0.984807753012208059366743024589523013670643252); Chris@10: DK(KP673648177, +0.673648177666930348851716626769314796000375677); Chris@10: DK(KP898197570, +0.898197570222573798468955502359086394667167570); Chris@10: DK(KP939692620, +0.939692620785908384054109277324731469936208134); Chris@10: DK(KP866025403, +0.866025403784438646763723170752936183471402627); Chris@10: DK(KP203604859, +0.203604859554852403062088995281827210665664861); Chris@10: DK(KP152703644, +0.152703644666139302296566746461370407999248646); Chris@10: DK(KP394930843, +0.394930843634698457567117349190734585290304520); Chris@10: DK(KP968908795, +0.968908795874236621082202410917456709164223497); Chris@10: DK(KP726681596, +0.726681596905677465811651808188092531873167623); Chris@10: DK(KP586256827, +0.586256827714544512072145703099641959914944179); Chris@10: DK(KP184792530, +0.184792530904095372701352047572203755870913560); Chris@10: DK(KP500000000, +0.500000000000000000000000000000000000000000000); Chris@10: { Chris@10: INT i; Chris@10: for (i = v; i > 0; i = i - 1, R0 = R0 + ivs, R1 = R1 + ivs, Cr = Cr + ovs, Ci = Ci + ovs, MAKE_VOLATILE_STRIDE(36, rs), MAKE_VOLATILE_STRIDE(36, csr), MAKE_VOLATILE_STRIDE(36, csi)) { Chris@10: E Tp, Tz, Tw, Ts, TA; Chris@10: { Chris@10: E T1, T6, Tb, T7, T4, To, T8, Tc, Td, T2, T3; Chris@10: T1 = R0[0]; Chris@10: T2 = R1[WS(rs, 1)]; Chris@10: T3 = R0[WS(rs, 3)]; Chris@10: T6 = R1[0]; Chris@10: Tb = R0[WS(rs, 1)]; Chris@10: T7 = R0[WS(rs, 2)]; Chris@10: T4 = T2 + T3; Chris@10: To = T3 - T2; Chris@10: T8 = R1[WS(rs, 3)]; Chris@10: Tc = R1[WS(rs, 2)]; Chris@10: Td = R0[WS(rs, 4)]; Chris@10: { Chris@10: E T5, T9, Tk, Te, Ti; Chris@10: T5 = T1 + T4; Chris@10: Tp = FNMS(KP500000000, T4, T1); Chris@10: T9 = T7 + T8; Chris@10: Tk = T7 - T8; Chris@10: Te = Tc + Td; Chris@10: Ti = Td - Tc; Chris@10: { Chris@10: E Tl, Ta, Tu, Tf, Th; Chris@10: Tl = FMS(KP500000000, T9, T6); Chris@10: Ta = T6 + T9; Chris@10: Tu = FMA(KP184792530, Tk, Ti); Chris@10: Tf = Tb + Te; Chris@10: Th = FNMS(KP500000000, Te, Tb); Chris@10: { Chris@10: E Tq, Ty, Tm, Tt; Chris@10: Tq = FMA(KP586256827, Tl, Ti); Chris@10: Ty = FMA(KP726681596, Tk, Tl); Chris@10: Tm = FNMS(KP968908795, Tl, Tk); Chris@10: Tt = FMA(KP394930843, Th, To); Chris@10: { Chris@10: E Tj, Tx, Tg, Tv; Chris@10: Tj = FNMS(KP152703644, Ti, Th); Chris@10: Tx = FMA(KP203604859, Th, Ti); Chris@10: Tg = Ta + Tf; Chris@10: Ci[WS(csi, 3)] = KP866025403 * (Tf - Ta); Chris@10: Tv = FNMS(KP939692620, Tu, Tt); Chris@10: { Chris@10: E TB, Tn, TC, Tr; Chris@10: TB = FMA(KP898197570, Ty, Tx); Chris@10: Tz = FNMS(KP898197570, Ty, Tx); Chris@10: Tw = FNMS(KP673648177, Tm, Tj); Chris@10: Tn = FMA(KP673648177, Tm, Tj); Chris@10: Cr[0] = T5 + Tg; Chris@10: Cr[WS(csr, 3)] = FNMS(KP500000000, Tg, T5); Chris@10: Ci[WS(csi, 2)] = KP984807753 * (FNMS(KP879385241, Tv, Tl)); Chris@10: Ci[WS(csi, 1)] = -(KP984807753 * (FNMS(KP879385241, To, Tn))); Chris@10: TC = FMA(KP666666666, Tn, TB); Chris@10: Tr = FNMS(KP347296355, Tq, Tk); Chris@10: Ci[WS(csi, 4)] = KP866025403 * (FMA(KP852868531, TC, To)); Chris@10: Ts = FNMS(KP907603734, Tr, Th); Chris@10: } Chris@10: } Chris@10: } Chris@10: } Chris@10: } Chris@10: } Chris@10: Cr[WS(csr, 1)] = FMA(KP852868531, Tz, Tp); Chris@10: TA = FNMS(KP500000000, Tz, Tw); Chris@10: Cr[WS(csr, 2)] = FNMS(KP939692620, Ts, Tp); Chris@10: Cr[WS(csr, 4)] = FMA(KP852868531, TA, Tp); Chris@10: } Chris@10: } Chris@10: } Chris@10: Chris@10: static const kr2c_desc desc = { 9, "r2cf_9", {12, 4, 26, 0}, &GENUS }; Chris@10: Chris@10: void X(codelet_r2cf_9) (planner *p) { Chris@10: X(kr2c_register) (p, r2cf_9, &desc); Chris@10: } Chris@10: Chris@10: #else /* HAVE_FMA */ Chris@10: Chris@10: /* Generated by: ../../../genfft/gen_r2cf.native -compact -variables 4 -pipeline-latency 4 -n 9 -name r2cf_9 -include r2cf.h */ Chris@10: Chris@10: /* Chris@10: * This function contains 38 FP additions, 26 FP multiplications, Chris@10: * (or, 21 additions, 9 multiplications, 17 fused multiply/add), Chris@10: * 36 stack variables, 14 constants, and 18 memory accesses Chris@10: */ Chris@10: #include "r2cf.h" Chris@10: Chris@10: static void r2cf_9(R *R0, R *R1, R *Cr, R *Ci, stride rs, stride csr, stride csi, INT v, INT ivs, INT ovs) Chris@10: { Chris@10: DK(KP939692620, +0.939692620785908384054109277324731469936208134); Chris@10: DK(KP296198132, +0.296198132726023843175338011893050938967728390); Chris@10: DK(KP342020143, +0.342020143325668733044099614682259580763083368); Chris@10: DK(KP813797681, +0.813797681349373692844693217248393223289101568); Chris@10: DK(KP984807753, +0.984807753012208059366743024589523013670643252); Chris@10: DK(KP150383733, +0.150383733180435296639271897612501926072238258); Chris@10: DK(KP642787609, +0.642787609686539326322643409907263432907559884); Chris@10: DK(KP663413948, +0.663413948168938396205421319635891297216863310); Chris@10: DK(KP852868531, +0.852868531952443209628250963940074071936020296); Chris@10: DK(KP173648177, +0.173648177666930348851716626769314796000375677); Chris@10: DK(KP556670399, +0.556670399226419366452912952047023132968291906); Chris@10: DK(KP766044443, +0.766044443118978035202392650555416673935832457); Chris@10: DK(KP866025403, +0.866025403784438646763723170752936183471402627); Chris@10: DK(KP500000000, +0.500000000000000000000000000000000000000000000); Chris@10: { Chris@10: INT i; Chris@10: for (i = v; i > 0; i = i - 1, R0 = R0 + ivs, R1 = R1 + ivs, Cr = Cr + ovs, Ci = Ci + ovs, MAKE_VOLATILE_STRIDE(36, rs), MAKE_VOLATILE_STRIDE(36, csr), MAKE_VOLATILE_STRIDE(36, csi)) { Chris@10: E T1, T4, Tr, Ta, Tl, Ti, Tf, Tk, Tj, T2, T3, T5, Tg; Chris@10: T1 = R0[0]; Chris@10: T2 = R1[WS(rs, 1)]; Chris@10: T3 = R0[WS(rs, 3)]; Chris@10: T4 = T2 + T3; Chris@10: Tr = T3 - T2; Chris@10: { Chris@10: E T6, T7, T8, T9; Chris@10: T6 = R1[0]; Chris@10: T7 = R0[WS(rs, 2)]; Chris@10: T8 = R1[WS(rs, 3)]; Chris@10: T9 = T7 + T8; Chris@10: Ta = T6 + T9; Chris@10: Tl = T8 - T7; Chris@10: Ti = FNMS(KP500000000, T9, T6); Chris@10: } Chris@10: { Chris@10: E Tb, Tc, Td, Te; Chris@10: Tb = R0[WS(rs, 1)]; Chris@10: Tc = R1[WS(rs, 2)]; Chris@10: Td = R0[WS(rs, 4)]; Chris@10: Te = Tc + Td; Chris@10: Tf = Tb + Te; Chris@10: Tk = FNMS(KP500000000, Te, Tb); Chris@10: Tj = Td - Tc; Chris@10: } Chris@10: Ci[WS(csi, 3)] = KP866025403 * (Tf - Ta); Chris@10: T5 = T1 + T4; Chris@10: Tg = Ta + Tf; Chris@10: Cr[WS(csr, 3)] = FNMS(KP500000000, Tg, T5); Chris@10: Cr[0] = T5 + Tg; Chris@10: { Chris@10: E Tt, Th, Tm, Tn, To, Tp, Tq, Ts; Chris@10: Tt = KP866025403 * Tr; Chris@10: Th = FNMS(KP500000000, T4, T1); Chris@10: Tm = FMA(KP766044443, Ti, KP556670399 * Tl); Chris@10: Tn = FMA(KP173648177, Tk, KP852868531 * Tj); Chris@10: To = Tm + Tn; Chris@10: Tp = FNMS(KP642787609, Ti, KP663413948 * Tl); Chris@10: Tq = FNMS(KP984807753, Tk, KP150383733 * Tj); Chris@10: Ts = Tp + Tq; Chris@10: Cr[WS(csr, 1)] = Th + To; Chris@10: Ci[WS(csi, 1)] = Tt + Ts; Chris@10: Cr[WS(csr, 4)] = FMA(KP866025403, Tp - Tq, Th) - (KP500000000 * To); Chris@10: Ci[WS(csi, 4)] = FNMS(KP500000000, Ts, KP866025403 * (Tr + (Tn - Tm))); Chris@10: Ci[WS(csi, 2)] = FNMS(KP342020143, Tk, KP813797681 * Tj) + FNMA(KP150383733, Tl, KP984807753 * Ti) - Tt; Chris@10: Cr[WS(csr, 2)] = FMA(KP173648177, Ti, Th) + FNMA(KP296198132, Tj, KP939692620 * Tk) - (KP852868531 * Tl); Chris@10: } Chris@10: } Chris@10: } Chris@10: } Chris@10: Chris@10: static const kr2c_desc desc = { 9, "r2cf_9", {21, 9, 17, 0}, &GENUS }; Chris@10: Chris@10: void X(codelet_r2cf_9) (planner *p) { Chris@10: X(kr2c_register) (p, r2cf_9, &desc); Chris@10: } Chris@10: Chris@10: #endif /* HAVE_FMA */