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 07:41:36 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_r2cb.native -fma -reorder-insns -schedule-for-pipeline -compact -variables 4 -pipeline-latency 4 -sign 1 -n 15 -name r2cbIII_15 -dft-III -include r2cbIII.h */ Chris@10: Chris@10: /* Chris@10: * This function contains 64 FP additions, 43 FP multiplications, Chris@10: * (or, 21 additions, 0 multiplications, 43 fused multiply/add), Chris@10: * 48 stack variables, 9 constants, and 30 memory accesses Chris@10: */ Chris@10: #include "r2cbIII.h" Chris@10: Chris@10: static void r2cbIII_15(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(KP951056516, +0.951056516295153572116439333379382143405698634); Chris@10: DK(KP1_732050807, +1.732050807568877293527446341505872366942805254); Chris@10: DK(KP559016994, +0.559016994374947424102293417182819058860154590); Chris@10: DK(KP250000000, +0.250000000000000000000000000000000000000000000); Chris@10: DK(KP1_902113032, +1.902113032590307144232878666758764286811397268); Chris@10: DK(KP1_118033988, +1.118033988749894848204586834365638117720309180); Chris@10: DK(KP2_000000000, +2.000000000000000000000000000000000000000000000); Chris@10: DK(KP500000000, +0.500000000000000000000000000000000000000000000); Chris@10: DK(KP618033988, +0.618033988749894848204586834365638117720309180); Chris@10: { Chris@10: INT i; Chris@10: for (i = v; i > 0; i = i - 1, R0 = R0 + ovs, R1 = R1 + ovs, Cr = Cr + ivs, Ci = Ci + ivs, MAKE_VOLATILE_STRIDE(60, rs), MAKE_VOLATILE_STRIDE(60, csr), MAKE_VOLATILE_STRIDE(60, csi)) { Chris@10: E TX, Tv, To, TW, Tl, Tx, Ty, Tw; Chris@10: { Chris@10: E TA, Tk, T6, T5, Tz, Th, TI, Tp, Tu, TK, TR, Tn, Td, Tq; Chris@10: { Chris@10: E T1, T2, T3, Ti, Tj; Chris@10: Ti = Ci[WS(csi, 4)]; Chris@10: Tj = Ci[WS(csi, 1)]; Chris@10: T1 = Cr[WS(csr, 7)]; Chris@10: T2 = Cr[WS(csr, 4)]; Chris@10: T3 = Cr[WS(csr, 1)]; Chris@10: TA = FNMS(KP618033988, Ti, Tj); Chris@10: Tk = FMA(KP618033988, Tj, Ti); Chris@10: { Chris@10: E T7, TP, Tc, T8; Chris@10: T6 = Cr[WS(csr, 2)]; Chris@10: { Chris@10: E T4, Tg, Ta, Tb, Tf; Chris@10: T4 = T2 + T3; Chris@10: Tg = T2 - T3; Chris@10: Ta = Cr[WS(csr, 3)]; Chris@10: Tb = Cr[WS(csr, 6)]; Chris@10: T7 = Cr[0]; Chris@10: Tf = FNMS(KP500000000, T4, T1); Chris@10: T5 = FMA(KP2_000000000, T4, T1); Chris@10: TP = Ta - Tb; Chris@10: Tc = Ta + Tb; Chris@10: Tz = FNMS(KP1_118033988, Tg, Tf); Chris@10: Th = FMA(KP1_118033988, Tg, Tf); Chris@10: T8 = Cr[WS(csr, 5)]; Chris@10: } Chris@10: TI = Ci[WS(csi, 2)]; Chris@10: { Chris@10: E Ts, Tt, TQ, T9; Chris@10: Ts = Ci[WS(csi, 3)]; Chris@10: Tt = Ci[WS(csi, 6)]; Chris@10: TQ = T7 - T8; Chris@10: T9 = T7 + T8; Chris@10: Tp = Ci[0]; Chris@10: Tu = Ts - Tt; Chris@10: TK = Ts + Tt; Chris@10: TX = FMA(KP618033988, TP, TQ); Chris@10: TR = FNMS(KP618033988, TQ, TP); Chris@10: Tn = T9 - Tc; Chris@10: Td = T9 + Tc; Chris@10: Tq = Ci[WS(csi, 5)]; Chris@10: } Chris@10: } Chris@10: } Chris@10: { Chris@10: E TB, TF, TO, TG, TE; Chris@10: { Chris@10: E Tm, T11, TN, TD, TM, T12, TC; Chris@10: TB = FNMS(KP1_902113032, TA, Tz); Chris@10: TF = FMA(KP1_902113032, TA, Tz); Chris@10: { Chris@10: E Te, Tr, TJ, TL; Chris@10: Tm = FNMS(KP250000000, Td, T6); Chris@10: Te = T6 + Td; Chris@10: Tr = Tp + Tq; Chris@10: TJ = Tq - Tp; Chris@10: R0[0] = FMA(KP2_000000000, Te, T5); Chris@10: T11 = Te - T5; Chris@10: TN = TJ + TK; Chris@10: TL = TJ - TK; Chris@10: Tv = FMA(KP618033988, Tu, Tr); Chris@10: TD = FNMS(KP618033988, Tr, Tu); Chris@10: TM = FNMS(KP250000000, TL, TI); Chris@10: T12 = TL + TI; Chris@10: } Chris@10: TC = FNMS(KP559016994, Tn, Tm); Chris@10: To = FMA(KP559016994, Tn, Tm); Chris@10: R1[WS(rs, 2)] = FMA(KP1_732050807, T12, T11); Chris@10: R0[WS(rs, 5)] = FMS(KP1_732050807, T12, T11); Chris@10: TW = FMA(KP559016994, TN, TM); Chris@10: TO = FNMS(KP559016994, TN, TM); Chris@10: TG = FNMS(KP951056516, TD, TC); Chris@10: TE = FMA(KP951056516, TD, TC); Chris@10: } Chris@10: Tl = FNMS(KP1_902113032, Tk, Th); Chris@10: Tx = FMA(KP1_902113032, Tk, Th); Chris@10: { Chris@10: E TS, TU, TT, TH; Chris@10: TS = FMA(KP951056516, TR, TO); Chris@10: TU = FNMS(KP951056516, TR, TO); Chris@10: TT = TF - TG; Chris@10: R1[WS(rs, 1)] = -(FMA(KP2_000000000, TG, TF)); Chris@10: TH = TB - TE; Chris@10: R0[WS(rs, 6)] = FMA(KP2_000000000, TE, TB); Chris@10: R1[WS(rs, 6)] = -(FMA(KP1_732050807, TU, TT)); Chris@10: R0[WS(rs, 4)] = FNMS(KP1_732050807, TU, TT); Chris@10: R1[WS(rs, 3)] = -(FMA(KP1_732050807, TS, TH)); Chris@10: R0[WS(rs, 1)] = FNMS(KP1_732050807, TS, TH); Chris@10: } Chris@10: } Chris@10: } Chris@10: Ty = FNMS(KP951056516, Tv, To); Chris@10: Tw = FMA(KP951056516, Tv, To); Chris@10: { Chris@10: E T10, TY, TV, TZ; Chris@10: T10 = FMA(KP951056516, TX, TW); Chris@10: TY = FNMS(KP951056516, TX, TW); Chris@10: TV = Ty - Tx; Chris@10: R0[WS(rs, 3)] = FMA(KP2_000000000, Ty, Tx); Chris@10: TZ = Tl - Tw; Chris@10: R1[WS(rs, 4)] = -(FMA(KP2_000000000, Tw, Tl)); Chris@10: R1[WS(rs, 5)] = FMA(KP1_732050807, TY, TV); Chris@10: R1[0] = FNMS(KP1_732050807, TY, TV); Chris@10: R0[WS(rs, 2)] = FMA(KP1_732050807, T10, TZ); Chris@10: R0[WS(rs, 7)] = FNMS(KP1_732050807, T10, TZ); Chris@10: } Chris@10: } Chris@10: } Chris@10: } Chris@10: Chris@10: static const kr2c_desc desc = { 15, "r2cbIII_15", {21, 0, 43, 0}, &GENUS }; Chris@10: Chris@10: void X(codelet_r2cbIII_15) (planner *p) { Chris@10: X(kr2c_register) (p, r2cbIII_15, &desc); Chris@10: } Chris@10: Chris@10: #else /* HAVE_FMA */ Chris@10: Chris@10: /* Generated by: ../../../genfft/gen_r2cb.native -compact -variables 4 -pipeline-latency 4 -sign 1 -n 15 -name r2cbIII_15 -dft-III -include r2cbIII.h */ Chris@10: Chris@10: /* Chris@10: * This function contains 64 FP additions, 26 FP multiplications, Chris@10: * (or, 49 additions, 11 multiplications, 15 fused multiply/add), Chris@10: * 47 stack variables, 14 constants, and 30 memory accesses Chris@10: */ Chris@10: #include "r2cbIII.h" Chris@10: Chris@10: static void r2cbIII_15(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(KP1_732050807, +1.732050807568877293527446341505872366942805254); Chris@10: DK(KP433012701, +0.433012701892219323381861585376468091735701313); Chris@10: DK(KP968245836, +0.968245836551854221294816349945599902708230426); Chris@10: DK(KP587785252, +0.587785252292473129168705954639072768597652438); Chris@10: DK(KP951056516, +0.951056516295153572116439333379382143405698634); Chris@10: DK(KP250000000, +0.250000000000000000000000000000000000000000000); Chris@10: DK(KP1_647278207, +1.647278207092663851754840078556380006059321028); Chris@10: DK(KP1_018073920, +1.018073920910254366901961726787815297021466329); Chris@10: DK(KP559016994, +0.559016994374947424102293417182819058860154590); Chris@10: DK(KP500000000, +0.500000000000000000000000000000000000000000000); Chris@10: DK(KP2_000000000, +2.000000000000000000000000000000000000000000000); Chris@10: DK(KP1_118033988, +1.118033988749894848204586834365638117720309180); Chris@10: DK(KP1_175570504, +1.175570504584946258337411909278145537195304875); Chris@10: DK(KP1_902113032, +1.902113032590307144232878666758764286811397268); Chris@10: { Chris@10: INT i; Chris@10: for (i = v; i > 0; i = i - 1, R0 = R0 + ovs, R1 = R1 + ovs, Cr = Cr + ivs, Ci = Ci + ivs, MAKE_VOLATILE_STRIDE(60, rs), MAKE_VOLATILE_STRIDE(60, csr), MAKE_VOLATILE_STRIDE(60, csi)) { Chris@10: E Tv, TD, T5, Ts, TC, T6, Tf, TW, TK, Td, Tg, TP, To, TN, TA; Chris@10: E TO, TQ, Tt, Tu, T12, Te, T11; Chris@10: Tt = Ci[WS(csi, 4)]; Chris@10: Tu = Ci[WS(csi, 1)]; Chris@10: Tv = FMA(KP1_902113032, Tt, KP1_175570504 * Tu); Chris@10: TD = FNMS(KP1_175570504, Tt, KP1_902113032 * Tu); Chris@10: { Chris@10: E T1, T4, Tq, T2, T3, Tr; Chris@10: T1 = Cr[WS(csr, 7)]; Chris@10: T2 = Cr[WS(csr, 4)]; Chris@10: T3 = Cr[WS(csr, 1)]; Chris@10: T4 = T2 + T3; Chris@10: Tq = KP1_118033988 * (T2 - T3); Chris@10: T5 = FMA(KP2_000000000, T4, T1); Chris@10: Tr = FNMS(KP500000000, T4, T1); Chris@10: Ts = Tq + Tr; Chris@10: TC = Tr - Tq; Chris@10: } Chris@10: { Chris@10: E Tc, TJ, T9, TI; Chris@10: T6 = Cr[WS(csr, 2)]; Chris@10: { Chris@10: E Ta, Tb, T7, T8; Chris@10: Ta = Cr[WS(csr, 3)]; Chris@10: Tb = Cr[WS(csr, 6)]; Chris@10: Tc = Ta + Tb; Chris@10: TJ = Ta - Tb; Chris@10: T7 = Cr[0]; Chris@10: T8 = Cr[WS(csr, 5)]; Chris@10: T9 = T7 + T8; Chris@10: TI = T7 - T8; Chris@10: } Chris@10: Tf = KP559016994 * (T9 - Tc); Chris@10: TW = FNMS(KP1_647278207, TJ, KP1_018073920 * TI); Chris@10: TK = FMA(KP1_647278207, TI, KP1_018073920 * TJ); Chris@10: Td = T9 + Tc; Chris@10: Tg = FNMS(KP250000000, Td, T6); Chris@10: } Chris@10: { Chris@10: E Tn, TM, Tk, TL; Chris@10: TP = Ci[WS(csi, 2)]; Chris@10: { Chris@10: E Tl, Tm, Ti, Tj; Chris@10: Tl = Ci[WS(csi, 3)]; Chris@10: Tm = Ci[WS(csi, 6)]; Chris@10: Tn = Tl - Tm; Chris@10: TM = Tl + Tm; Chris@10: Ti = Ci[0]; Chris@10: Tj = Ci[WS(csi, 5)]; Chris@10: Tk = Ti + Tj; Chris@10: TL = Ti - Tj; Chris@10: } Chris@10: To = FMA(KP951056516, Tk, KP587785252 * Tn); Chris@10: TN = KP968245836 * (TL - TM); Chris@10: TA = FNMS(KP587785252, Tk, KP951056516 * Tn); Chris@10: TO = TL + TM; Chris@10: TQ = FMA(KP433012701, TO, KP1_732050807 * TP); Chris@10: } Chris@10: T12 = KP1_732050807 * (TP - TO); Chris@10: Te = T6 + Td; Chris@10: T11 = Te - T5; Chris@10: R0[0] = FMA(KP2_000000000, Te, T5); Chris@10: R0[WS(rs, 5)] = T12 - T11; Chris@10: R1[WS(rs, 2)] = T11 + T12; Chris@10: { Chris@10: E TE, TG, TB, TF, TY, T10, Tz, TX, TV, TZ; Chris@10: TE = TC - TD; Chris@10: TG = TC + TD; Chris@10: Tz = Tg - Tf; Chris@10: TB = Tz + TA; Chris@10: TF = TA - Tz; Chris@10: TX = TN + TQ; Chris@10: TY = TW - TX; Chris@10: T10 = TW + TX; Chris@10: R0[WS(rs, 6)] = FMA(KP2_000000000, TB, TE); Chris@10: R1[WS(rs, 1)] = FMS(KP2_000000000, TF, TG); Chris@10: TV = TE - TB; Chris@10: R0[WS(rs, 1)] = TV + TY; Chris@10: R1[WS(rs, 3)] = TY - TV; Chris@10: TZ = TF + TG; Chris@10: R0[WS(rs, 4)] = TZ - T10; Chris@10: R1[WS(rs, 6)] = -(TZ + T10); Chris@10: } Chris@10: { Chris@10: E Tw, Ty, Tp, Tx, TS, TU, Th, TR, TH, TT; Chris@10: Tw = Ts - Tv; Chris@10: Ty = Ts + Tv; Chris@10: Th = Tf + Tg; Chris@10: Tp = Th + To; Chris@10: Tx = Th - To; Chris@10: TR = TN - TQ; Chris@10: TS = TK + TR; Chris@10: TU = TR - TK; Chris@10: R1[WS(rs, 4)] = -(FMA(KP2_000000000, Tp, Tw)); Chris@10: R0[WS(rs, 3)] = FMA(KP2_000000000, Tx, Ty); Chris@10: TH = Tx - Ty; Chris@10: R1[WS(rs, 5)] = TH - TS; Chris@10: R1[0] = TH + TS; Chris@10: TT = Tw - Tp; Chris@10: R0[WS(rs, 2)] = TT - TU; Chris@10: R0[WS(rs, 7)] = TT + TU; Chris@10: } Chris@10: } Chris@10: } Chris@10: } Chris@10: Chris@10: static const kr2c_desc desc = { 15, "r2cbIII_15", {49, 11, 15, 0}, &GENUS }; Chris@10: Chris@10: void X(codelet_r2cbIII_15) (planner *p) { Chris@10: X(kr2c_register) (p, r2cbIII_15, &desc); Chris@10: } Chris@10: Chris@10: #endif /* HAVE_FMA */