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 16:50:42 EDT 2016 */ cannam@127: cannam@127: #include "codelet-rdft.h" cannam@127: cannam@127: #ifdef HAVE_FMA cannam@127: cannam@127: /* 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 */ cannam@127: cannam@127: /* cannam@127: * This function contains 64 FP additions, 43 FP multiplications, cannam@127: * (or, 21 additions, 0 multiplications, 43 fused multiply/add), cannam@127: * 48 stack variables, 9 constants, and 30 memory accesses cannam@127: */ cannam@127: #include "r2cbIII.h" cannam@127: cannam@127: static void r2cbIII_15(R *R0, R *R1, R *Cr, R *Ci, stride rs, stride csr, stride csi, INT v, INT ivs, INT ovs) cannam@127: { cannam@127: DK(KP951056516, +0.951056516295153572116439333379382143405698634); cannam@127: DK(KP1_732050807, +1.732050807568877293527446341505872366942805254); cannam@127: DK(KP559016994, +0.559016994374947424102293417182819058860154590); cannam@127: DK(KP250000000, +0.250000000000000000000000000000000000000000000); cannam@127: DK(KP1_902113032, +1.902113032590307144232878666758764286811397268); cannam@127: DK(KP1_118033988, +1.118033988749894848204586834365638117720309180); cannam@127: DK(KP2_000000000, +2.000000000000000000000000000000000000000000000); cannam@127: DK(KP500000000, +0.500000000000000000000000000000000000000000000); cannam@127: DK(KP618033988, +0.618033988749894848204586834365638117720309180); cannam@127: { cannam@127: INT i; cannam@127: 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)) { cannam@127: E TX, Tv, To, TW, Tl, Tx, Ty, Tw; cannam@127: { cannam@127: E TA, Tk, T6, T5, Tz, Th, TI, Tp, Tu, TK, TR, Tn, Td, Tq; cannam@127: { cannam@127: E T1, T2, T3, Ti, Tj; cannam@127: Ti = Ci[WS(csi, 4)]; cannam@127: Tj = Ci[WS(csi, 1)]; cannam@127: T1 = Cr[WS(csr, 7)]; cannam@127: T2 = Cr[WS(csr, 4)]; cannam@127: T3 = Cr[WS(csr, 1)]; cannam@127: TA = FNMS(KP618033988, Ti, Tj); cannam@127: Tk = FMA(KP618033988, Tj, Ti); cannam@127: { cannam@127: E T7, TP, Tc, T8; cannam@127: T6 = Cr[WS(csr, 2)]; cannam@127: { cannam@127: E T4, Tg, Ta, Tb, Tf; cannam@127: T4 = T2 + T3; cannam@127: Tg = T2 - T3; cannam@127: Ta = Cr[WS(csr, 3)]; cannam@127: Tb = Cr[WS(csr, 6)]; cannam@127: T7 = Cr[0]; cannam@127: Tf = FNMS(KP500000000, T4, T1); cannam@127: T5 = FMA(KP2_000000000, T4, T1); cannam@127: TP = Ta - Tb; cannam@127: Tc = Ta + Tb; cannam@127: Tz = FNMS(KP1_118033988, Tg, Tf); cannam@127: Th = FMA(KP1_118033988, Tg, Tf); cannam@127: T8 = Cr[WS(csr, 5)]; cannam@127: } cannam@127: TI = Ci[WS(csi, 2)]; cannam@127: { cannam@127: E Ts, Tt, TQ, T9; cannam@127: Ts = Ci[WS(csi, 3)]; cannam@127: Tt = Ci[WS(csi, 6)]; cannam@127: TQ = T7 - T8; cannam@127: T9 = T7 + T8; cannam@127: Tp = Ci[0]; cannam@127: Tu = Ts - Tt; cannam@127: TK = Ts + Tt; cannam@127: TX = FMA(KP618033988, TP, TQ); cannam@127: TR = FNMS(KP618033988, TQ, TP); cannam@127: Tn = T9 - Tc; cannam@127: Td = T9 + Tc; cannam@127: Tq = Ci[WS(csi, 5)]; cannam@127: } cannam@127: } cannam@127: } cannam@127: { cannam@127: E TB, TF, TO, TG, TE; cannam@127: { cannam@127: E Tm, T11, TN, TD, TM, T12, TC; cannam@127: TB = FNMS(KP1_902113032, TA, Tz); cannam@127: TF = FMA(KP1_902113032, TA, Tz); cannam@127: { cannam@127: E Te, Tr, TJ, TL; cannam@127: Tm = FNMS(KP250000000, Td, T6); cannam@127: Te = T6 + Td; cannam@127: Tr = Tp + Tq; cannam@127: TJ = Tq - Tp; cannam@127: R0[0] = FMA(KP2_000000000, Te, T5); cannam@127: T11 = Te - T5; cannam@127: TN = TJ + TK; cannam@127: TL = TJ - TK; cannam@127: Tv = FMA(KP618033988, Tu, Tr); cannam@127: TD = FNMS(KP618033988, Tr, Tu); cannam@127: TM = FNMS(KP250000000, TL, TI); cannam@127: T12 = TL + TI; cannam@127: } cannam@127: TC = FNMS(KP559016994, Tn, Tm); cannam@127: To = FMA(KP559016994, Tn, Tm); cannam@127: R1[WS(rs, 2)] = FMA(KP1_732050807, T12, T11); cannam@127: R0[WS(rs, 5)] = FMS(KP1_732050807, T12, T11); cannam@127: TW = FMA(KP559016994, TN, TM); cannam@127: TO = FNMS(KP559016994, TN, TM); cannam@127: TG = FNMS(KP951056516, TD, TC); cannam@127: TE = FMA(KP951056516, TD, TC); cannam@127: } cannam@127: Tl = FNMS(KP1_902113032, Tk, Th); cannam@127: Tx = FMA(KP1_902113032, Tk, Th); cannam@127: { cannam@127: E TS, TU, TT, TH; cannam@127: TS = FMA(KP951056516, TR, TO); cannam@127: TU = FNMS(KP951056516, TR, TO); cannam@127: TT = TF - TG; cannam@127: R1[WS(rs, 1)] = -(FMA(KP2_000000000, TG, TF)); cannam@127: TH = TB - TE; cannam@127: R0[WS(rs, 6)] = FMA(KP2_000000000, TE, TB); cannam@127: R1[WS(rs, 6)] = -(FMA(KP1_732050807, TU, TT)); cannam@127: R0[WS(rs, 4)] = FNMS(KP1_732050807, TU, TT); cannam@127: R1[WS(rs, 3)] = -(FMA(KP1_732050807, TS, TH)); cannam@127: R0[WS(rs, 1)] = FNMS(KP1_732050807, TS, TH); cannam@127: } cannam@127: } cannam@127: } cannam@127: Ty = FNMS(KP951056516, Tv, To); cannam@127: Tw = FMA(KP951056516, Tv, To); cannam@127: { cannam@127: E T10, TY, TV, TZ; cannam@127: T10 = FMA(KP951056516, TX, TW); cannam@127: TY = FNMS(KP951056516, TX, TW); cannam@127: TV = Ty - Tx; cannam@127: R0[WS(rs, 3)] = FMA(KP2_000000000, Ty, Tx); cannam@127: TZ = Tl - Tw; cannam@127: R1[WS(rs, 4)] = -(FMA(KP2_000000000, Tw, Tl)); cannam@127: R1[WS(rs, 5)] = FMA(KP1_732050807, TY, TV); cannam@127: R1[0] = FNMS(KP1_732050807, TY, TV); cannam@127: R0[WS(rs, 2)] = FMA(KP1_732050807, T10, TZ); cannam@127: R0[WS(rs, 7)] = FNMS(KP1_732050807, T10, TZ); cannam@127: } cannam@127: } cannam@127: } cannam@127: } cannam@127: cannam@127: static const kr2c_desc desc = { 15, "r2cbIII_15", {21, 0, 43, 0}, &GENUS }; cannam@127: cannam@127: void X(codelet_r2cbIII_15) (planner *p) { cannam@127: X(kr2c_register) (p, r2cbIII_15, &desc); cannam@127: } cannam@127: cannam@127: #else /* HAVE_FMA */ cannam@127: cannam@127: /* Generated by: ../../../genfft/gen_r2cb.native -compact -variables 4 -pipeline-latency 4 -sign 1 -n 15 -name r2cbIII_15 -dft-III -include r2cbIII.h */ cannam@127: cannam@127: /* cannam@127: * This function contains 64 FP additions, 26 FP multiplications, cannam@127: * (or, 49 additions, 11 multiplications, 15 fused multiply/add), cannam@127: * 47 stack variables, 14 constants, and 30 memory accesses cannam@127: */ cannam@127: #include "r2cbIII.h" cannam@127: cannam@127: static void r2cbIII_15(R *R0, R *R1, R *Cr, R *Ci, stride rs, stride csr, stride csi, INT v, INT ivs, INT ovs) cannam@127: { cannam@127: DK(KP1_732050807, +1.732050807568877293527446341505872366942805254); cannam@127: DK(KP433012701, +0.433012701892219323381861585376468091735701313); cannam@127: DK(KP968245836, +0.968245836551854221294816349945599902708230426); cannam@127: DK(KP587785252, +0.587785252292473129168705954639072768597652438); cannam@127: DK(KP951056516, +0.951056516295153572116439333379382143405698634); cannam@127: DK(KP250000000, +0.250000000000000000000000000000000000000000000); cannam@127: DK(KP1_647278207, +1.647278207092663851754840078556380006059321028); cannam@127: DK(KP1_018073920, +1.018073920910254366901961726787815297021466329); cannam@127: DK(KP559016994, +0.559016994374947424102293417182819058860154590); cannam@127: DK(KP500000000, +0.500000000000000000000000000000000000000000000); cannam@127: DK(KP2_000000000, +2.000000000000000000000000000000000000000000000); cannam@127: DK(KP1_118033988, +1.118033988749894848204586834365638117720309180); cannam@127: DK(KP1_175570504, +1.175570504584946258337411909278145537195304875); cannam@127: DK(KP1_902113032, +1.902113032590307144232878666758764286811397268); cannam@127: { cannam@127: INT i; cannam@127: 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)) { cannam@127: E Tv, TD, T5, Ts, TC, T6, Tf, TW, TK, Td, Tg, TP, To, TN, TA; cannam@127: E TO, TQ, Tt, Tu, T12, Te, T11; cannam@127: Tt = Ci[WS(csi, 4)]; cannam@127: Tu = Ci[WS(csi, 1)]; cannam@127: Tv = FMA(KP1_902113032, Tt, KP1_175570504 * Tu); cannam@127: TD = FNMS(KP1_175570504, Tt, KP1_902113032 * Tu); cannam@127: { cannam@127: E T1, T4, Tq, T2, T3, Tr; cannam@127: T1 = Cr[WS(csr, 7)]; cannam@127: T2 = Cr[WS(csr, 4)]; cannam@127: T3 = Cr[WS(csr, 1)]; cannam@127: T4 = T2 + T3; cannam@127: Tq = KP1_118033988 * (T2 - T3); cannam@127: T5 = FMA(KP2_000000000, T4, T1); cannam@127: Tr = FNMS(KP500000000, T4, T1); cannam@127: Ts = Tq + Tr; cannam@127: TC = Tr - Tq; cannam@127: } cannam@127: { cannam@127: E Tc, TJ, T9, TI; cannam@127: T6 = Cr[WS(csr, 2)]; cannam@127: { cannam@127: E Ta, Tb, T7, T8; cannam@127: Ta = Cr[WS(csr, 3)]; cannam@127: Tb = Cr[WS(csr, 6)]; cannam@127: Tc = Ta + Tb; cannam@127: TJ = Ta - Tb; cannam@127: T7 = Cr[0]; cannam@127: T8 = Cr[WS(csr, 5)]; cannam@127: T9 = T7 + T8; cannam@127: TI = T7 - T8; cannam@127: } cannam@127: Tf = KP559016994 * (T9 - Tc); cannam@127: TW = FNMS(KP1_647278207, TJ, KP1_018073920 * TI); cannam@127: TK = FMA(KP1_647278207, TI, KP1_018073920 * TJ); cannam@127: Td = T9 + Tc; cannam@127: Tg = FNMS(KP250000000, Td, T6); cannam@127: } cannam@127: { cannam@127: E Tn, TM, Tk, TL; cannam@127: TP = Ci[WS(csi, 2)]; cannam@127: { cannam@127: E Tl, Tm, Ti, Tj; cannam@127: Tl = Ci[WS(csi, 3)]; cannam@127: Tm = Ci[WS(csi, 6)]; cannam@127: Tn = Tl - Tm; cannam@127: TM = Tl + Tm; cannam@127: Ti = Ci[0]; cannam@127: Tj = Ci[WS(csi, 5)]; cannam@127: Tk = Ti + Tj; cannam@127: TL = Ti - Tj; cannam@127: } cannam@127: To = FMA(KP951056516, Tk, KP587785252 * Tn); cannam@127: TN = KP968245836 * (TL - TM); cannam@127: TA = FNMS(KP587785252, Tk, KP951056516 * Tn); cannam@127: TO = TL + TM; cannam@127: TQ = FMA(KP433012701, TO, KP1_732050807 * TP); cannam@127: } cannam@127: T12 = KP1_732050807 * (TP - TO); cannam@127: Te = T6 + Td; cannam@127: T11 = Te - T5; cannam@127: R0[0] = FMA(KP2_000000000, Te, T5); cannam@127: R0[WS(rs, 5)] = T12 - T11; cannam@127: R1[WS(rs, 2)] = T11 + T12; cannam@127: { cannam@127: E TE, TG, TB, TF, TY, T10, Tz, TX, TV, TZ; cannam@127: TE = TC - TD; cannam@127: TG = TC + TD; cannam@127: Tz = Tg - Tf; cannam@127: TB = Tz + TA; cannam@127: TF = TA - Tz; cannam@127: TX = TN + TQ; cannam@127: TY = TW - TX; cannam@127: T10 = TW + TX; cannam@127: R0[WS(rs, 6)] = FMA(KP2_000000000, TB, TE); cannam@127: R1[WS(rs, 1)] = FMS(KP2_000000000, TF, TG); cannam@127: TV = TE - TB; cannam@127: R0[WS(rs, 1)] = TV + TY; cannam@127: R1[WS(rs, 3)] = TY - TV; cannam@127: TZ = TF + TG; cannam@127: R0[WS(rs, 4)] = TZ - T10; cannam@127: R1[WS(rs, 6)] = -(TZ + T10); cannam@127: } cannam@127: { cannam@127: E Tw, Ty, Tp, Tx, TS, TU, Th, TR, TH, TT; cannam@127: Tw = Ts - Tv; cannam@127: Ty = Ts + Tv; cannam@127: Th = Tf + Tg; cannam@127: Tp = Th + To; cannam@127: Tx = Th - To; cannam@127: TR = TN - TQ; cannam@127: TS = TK + TR; cannam@127: TU = TR - TK; cannam@127: R1[WS(rs, 4)] = -(FMA(KP2_000000000, Tp, Tw)); cannam@127: R0[WS(rs, 3)] = FMA(KP2_000000000, Tx, Ty); cannam@127: TH = Tx - Ty; cannam@127: R1[WS(rs, 5)] = TH - TS; cannam@127: R1[0] = TH + TS; cannam@127: TT = Tw - Tp; cannam@127: R0[WS(rs, 2)] = TT - TU; cannam@127: R0[WS(rs, 7)] = TT + TU; cannam@127: } cannam@127: } cannam@127: } cannam@127: } cannam@127: cannam@127: static const kr2c_desc desc = { 15, "r2cbIII_15", {49, 11, 15, 0}, &GENUS }; cannam@127: cannam@127: void X(codelet_r2cbIII_15) (planner *p) { cannam@127: X(kr2c_register) (p, r2cbIII_15, &desc); cannam@127: } cannam@127: cannam@127: #endif /* HAVE_FMA */