Chris@82: /* Chris@82: * Copyright (c) 2003, 2007-14 Matteo Frigo Chris@82: * Copyright (c) 2003, 2007-14 Massachusetts Institute of Technology Chris@82: * Chris@82: * This program is free software; you can redistribute it and/or modify Chris@82: * it under the terms of the GNU General Public License as published by Chris@82: * the Free Software Foundation; either version 2 of the License, or Chris@82: * (at your option) any later version. Chris@82: * Chris@82: * This program is distributed in the hope that it will be useful, Chris@82: * but WITHOUT ANY WARRANTY; without even the implied warranty of Chris@82: * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the Chris@82: * GNU General Public License for more details. Chris@82: * Chris@82: * You should have received a copy of the GNU General Public License Chris@82: * along with this program; if not, write to the Free Software Chris@82: * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA Chris@82: * Chris@82: */ Chris@82: Chris@82: /* This file was automatically generated --- DO NOT EDIT */ Chris@82: /* Generated on Thu May 24 08:07:44 EDT 2018 */ Chris@82: Chris@82: #include "rdft/codelet-rdft.h" Chris@82: Chris@82: #if defined(ARCH_PREFERS_FMA) || defined(ISA_EXTENSION_PREFERS_FMA) Chris@82: Chris@82: /* Generated by: ../../../genfft/gen_r2cb.native -fma -compact -variables 4 -pipeline-latency 4 -sign 1 -n 15 -name r2cbIII_15 -dft-III -include rdft/scalar/r2cbIII.h */ Chris@82: Chris@82: /* Chris@82: * This function contains 64 FP additions, 43 FP multiplications, Chris@82: * (or, 21 additions, 0 multiplications, 43 fused multiply/add), Chris@82: * 42 stack variables, 9 constants, and 30 memory accesses Chris@82: */ Chris@82: #include "rdft/scalar/r2cbIII.h" Chris@82: Chris@82: 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@82: { Chris@82: DK(KP951056516, +0.951056516295153572116439333379382143405698634); Chris@82: DK(KP559016994, +0.559016994374947424102293417182819058860154590); Chris@82: DK(KP1_902113032, +1.902113032590307144232878666758764286811397268); Chris@82: DK(KP1_732050807, +1.732050807568877293527446341505872366942805254); Chris@82: DK(KP250000000, +0.250000000000000000000000000000000000000000000); Chris@82: DK(KP1_118033988, +1.118033988749894848204586834365638117720309180); Chris@82: DK(KP500000000, +0.500000000000000000000000000000000000000000000); Chris@82: DK(KP2_000000000, +2.000000000000000000000000000000000000000000000); Chris@82: DK(KP618033988, +0.618033988749894848204586834365638117720309180); Chris@82: { Chris@82: INT i; Chris@82: 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@82: E Tk, TA, T5, Th, Tz, T6, Tn, TX, TR, Td, Tm, TI, Tv, TN, TD; Chris@82: E TL, TM, Ti, Tj, T12, Te, T11; Chris@82: Ti = Ci[WS(csi, 4)]; Chris@82: Tj = Ci[WS(csi, 1)]; Chris@82: Tk = FMA(KP618033988, Tj, Ti); Chris@82: TA = FNMS(KP618033988, Ti, Tj); Chris@82: { Chris@82: E T1, T4, Tg, T2, T3, Tf; Chris@82: T1 = Cr[WS(csr, 7)]; Chris@82: T2 = Cr[WS(csr, 4)]; Chris@82: T3 = Cr[WS(csr, 1)]; Chris@82: T4 = T2 + T3; Chris@82: Tg = T2 - T3; Chris@82: T5 = FMA(KP2_000000000, T4, T1); Chris@82: Tf = FNMS(KP500000000, T4, T1); Chris@82: Th = FMA(KP1_118033988, Tg, Tf); Chris@82: Tz = FNMS(KP1_118033988, Tg, Tf); Chris@82: } Chris@82: { Chris@82: E Tc, TP, T9, TQ; Chris@82: T6 = Cr[WS(csr, 2)]; Chris@82: { Chris@82: E Ta, Tb, T7, T8; Chris@82: Ta = Cr[WS(csr, 3)]; Chris@82: Tb = Cr[WS(csr, 6)]; Chris@82: Tc = Ta + Tb; Chris@82: TP = Ta - Tb; Chris@82: T7 = Cr[0]; Chris@82: T8 = Cr[WS(csr, 5)]; Chris@82: T9 = T7 + T8; Chris@82: TQ = T7 - T8; Chris@82: } Chris@82: Tn = T9 - Tc; Chris@82: TX = FMA(KP618033988, TP, TQ); Chris@82: TR = FNMS(KP618033988, TQ, TP); Chris@82: Td = T9 + Tc; Chris@82: Tm = FNMS(KP250000000, Td, T6); Chris@82: } Chris@82: { Chris@82: E Tu, TK, Tr, TJ; Chris@82: TI = Ci[WS(csi, 2)]; Chris@82: { Chris@82: E Ts, Tt, Tp, Tq; Chris@82: Ts = Ci[WS(csi, 3)]; Chris@82: Tt = Ci[WS(csi, 6)]; Chris@82: Tu = Ts - Tt; Chris@82: TK = Ts + Tt; Chris@82: Tp = Ci[0]; Chris@82: Tq = Ci[WS(csi, 5)]; Chris@82: Tr = Tp + Tq; Chris@82: TJ = Tq - Tp; Chris@82: } Chris@82: Tv = FMA(KP618033988, Tu, Tr); Chris@82: TN = TJ + TK; Chris@82: TD = FNMS(KP618033988, Tr, Tu); Chris@82: TL = TJ - TK; Chris@82: TM = FNMS(KP250000000, TL, TI); Chris@82: } Chris@82: T12 = TL + TI; Chris@82: Te = T6 + Td; Chris@82: T11 = Te - T5; Chris@82: R0[0] = FMA(KP2_000000000, Te, T5); Chris@82: R0[WS(rs, 5)] = FMS(KP1_732050807, T12, T11); Chris@82: R1[WS(rs, 2)] = FMA(KP1_732050807, T12, T11); Chris@82: { Chris@82: E TB, TF, TE, TG, TS, TU, TC, TO, TH, TT; Chris@82: TB = FNMS(KP1_902113032, TA, Tz); Chris@82: TF = FMA(KP1_902113032, TA, Tz); Chris@82: TC = FNMS(KP559016994, Tn, Tm); Chris@82: TE = FMA(KP951056516, TD, TC); Chris@82: TG = FNMS(KP951056516, TD, TC); Chris@82: TO = FNMS(KP559016994, TN, TM); Chris@82: TS = FMA(KP951056516, TR, TO); Chris@82: TU = FNMS(KP951056516, TR, TO); Chris@82: R0[WS(rs, 6)] = FMA(KP2_000000000, TE, TB); Chris@82: R1[WS(rs, 1)] = -(FMA(KP2_000000000, TG, TF)); Chris@82: TH = TB - TE; Chris@82: R0[WS(rs, 1)] = FNMS(KP1_732050807, TS, TH); Chris@82: R1[WS(rs, 3)] = -(FMA(KP1_732050807, TS, TH)); Chris@82: TT = TF - TG; Chris@82: R0[WS(rs, 4)] = FNMS(KP1_732050807, TU, TT); Chris@82: R1[WS(rs, 6)] = -(FMA(KP1_732050807, TU, TT)); Chris@82: } Chris@82: { Chris@82: E Tl, Tx, Tw, Ty, TY, T10, To, TW, TV, TZ; Chris@82: Tl = FNMS(KP1_902113032, Tk, Th); Chris@82: Tx = FMA(KP1_902113032, Tk, Th); Chris@82: To = FMA(KP559016994, Tn, Tm); Chris@82: Tw = FMA(KP951056516, Tv, To); Chris@82: Ty = FNMS(KP951056516, Tv, To); Chris@82: TW = FMA(KP559016994, TN, TM); Chris@82: TY = FNMS(KP951056516, TX, TW); Chris@82: T10 = FMA(KP951056516, TX, TW); Chris@82: R1[WS(rs, 4)] = -(FMA(KP2_000000000, Tw, Tl)); Chris@82: R0[WS(rs, 3)] = FMA(KP2_000000000, Ty, Tx); Chris@82: TV = Ty - Tx; Chris@82: R1[0] = FNMS(KP1_732050807, TY, TV); Chris@82: R1[WS(rs, 5)] = FMA(KP1_732050807, TY, TV); Chris@82: TZ = Tl - Tw; Chris@82: R0[WS(rs, 7)] = FNMS(KP1_732050807, T10, TZ); Chris@82: R0[WS(rs, 2)] = FMA(KP1_732050807, T10, TZ); Chris@82: } Chris@82: } Chris@82: } Chris@82: } Chris@82: Chris@82: static const kr2c_desc desc = { 15, "r2cbIII_15", {21, 0, 43, 0}, &GENUS }; Chris@82: Chris@82: void X(codelet_r2cbIII_15) (planner *p) { Chris@82: X(kr2c_register) (p, r2cbIII_15, &desc); Chris@82: } Chris@82: Chris@82: #else Chris@82: Chris@82: /* Generated by: ../../../genfft/gen_r2cb.native -compact -variables 4 -pipeline-latency 4 -sign 1 -n 15 -name r2cbIII_15 -dft-III -include rdft/scalar/r2cbIII.h */ Chris@82: Chris@82: /* Chris@82: * This function contains 64 FP additions, 26 FP multiplications, Chris@82: * (or, 49 additions, 11 multiplications, 15 fused multiply/add), Chris@82: * 47 stack variables, 14 constants, and 30 memory accesses Chris@82: */ Chris@82: #include "rdft/scalar/r2cbIII.h" Chris@82: Chris@82: 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@82: { Chris@82: DK(KP1_732050807, +1.732050807568877293527446341505872366942805254); Chris@82: DK(KP433012701, +0.433012701892219323381861585376468091735701313); Chris@82: DK(KP968245836, +0.968245836551854221294816349945599902708230426); Chris@82: DK(KP587785252, +0.587785252292473129168705954639072768597652438); Chris@82: DK(KP951056516, +0.951056516295153572116439333379382143405698634); Chris@82: DK(KP250000000, +0.250000000000000000000000000000000000000000000); Chris@82: DK(KP1_647278207, +1.647278207092663851754840078556380006059321028); Chris@82: DK(KP1_018073920, +1.018073920910254366901961726787815297021466329); Chris@82: DK(KP559016994, +0.559016994374947424102293417182819058860154590); Chris@82: DK(KP500000000, +0.500000000000000000000000000000000000000000000); Chris@82: DK(KP2_000000000, +2.000000000000000000000000000000000000000000000); Chris@82: DK(KP1_118033988, +1.118033988749894848204586834365638117720309180); Chris@82: DK(KP1_175570504, +1.175570504584946258337411909278145537195304875); Chris@82: DK(KP1_902113032, +1.902113032590307144232878666758764286811397268); Chris@82: { Chris@82: INT i; Chris@82: 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@82: E Tv, TD, T5, Ts, TC, T6, Tf, TW, TK, Td, Tg, TP, To, TN, TA; Chris@82: E TO, TQ, Tt, Tu, T12, Te, T11; Chris@82: Tt = Ci[WS(csi, 4)]; Chris@82: Tu = Ci[WS(csi, 1)]; Chris@82: Tv = FMA(KP1_902113032, Tt, KP1_175570504 * Tu); Chris@82: TD = FNMS(KP1_175570504, Tt, KP1_902113032 * Tu); Chris@82: { Chris@82: E T1, T4, Tq, T2, T3, Tr; Chris@82: T1 = Cr[WS(csr, 7)]; Chris@82: T2 = Cr[WS(csr, 4)]; Chris@82: T3 = Cr[WS(csr, 1)]; Chris@82: T4 = T2 + T3; Chris@82: Tq = KP1_118033988 * (T2 - T3); Chris@82: T5 = FMA(KP2_000000000, T4, T1); Chris@82: Tr = FNMS(KP500000000, T4, T1); Chris@82: Ts = Tq + Tr; Chris@82: TC = Tr - Tq; Chris@82: } Chris@82: { Chris@82: E Tc, TJ, T9, TI; Chris@82: T6 = Cr[WS(csr, 2)]; Chris@82: { Chris@82: E Ta, Tb, T7, T8; Chris@82: Ta = Cr[WS(csr, 3)]; Chris@82: Tb = Cr[WS(csr, 6)]; Chris@82: Tc = Ta + Tb; Chris@82: TJ = Ta - Tb; Chris@82: T7 = Cr[0]; Chris@82: T8 = Cr[WS(csr, 5)]; Chris@82: T9 = T7 + T8; Chris@82: TI = T7 - T8; Chris@82: } Chris@82: Tf = KP559016994 * (T9 - Tc); Chris@82: TW = FNMS(KP1_647278207, TJ, KP1_018073920 * TI); Chris@82: TK = FMA(KP1_647278207, TI, KP1_018073920 * TJ); Chris@82: Td = T9 + Tc; Chris@82: Tg = FNMS(KP250000000, Td, T6); Chris@82: } Chris@82: { Chris@82: E Tn, TM, Tk, TL; Chris@82: TP = Ci[WS(csi, 2)]; Chris@82: { Chris@82: E Tl, Tm, Ti, Tj; Chris@82: Tl = Ci[WS(csi, 3)]; Chris@82: Tm = Ci[WS(csi, 6)]; Chris@82: Tn = Tl - Tm; Chris@82: TM = Tl + Tm; Chris@82: Ti = Ci[0]; Chris@82: Tj = Ci[WS(csi, 5)]; Chris@82: Tk = Ti + Tj; Chris@82: TL = Ti - Tj; Chris@82: } Chris@82: To = FMA(KP951056516, Tk, KP587785252 * Tn); Chris@82: TN = KP968245836 * (TL - TM); Chris@82: TA = FNMS(KP587785252, Tk, KP951056516 * Tn); Chris@82: TO = TL + TM; Chris@82: TQ = FMA(KP433012701, TO, KP1_732050807 * TP); Chris@82: } Chris@82: T12 = KP1_732050807 * (TP - TO); Chris@82: Te = T6 + Td; Chris@82: T11 = Te - T5; Chris@82: R0[0] = FMA(KP2_000000000, Te, T5); Chris@82: R0[WS(rs, 5)] = T12 - T11; Chris@82: R1[WS(rs, 2)] = T11 + T12; Chris@82: { Chris@82: E TE, TG, TB, TF, TY, T10, Tz, TX, TV, TZ; Chris@82: TE = TC - TD; Chris@82: TG = TC + TD; Chris@82: Tz = Tg - Tf; Chris@82: TB = Tz + TA; Chris@82: TF = TA - Tz; Chris@82: TX = TN + TQ; Chris@82: TY = TW - TX; Chris@82: T10 = TW + TX; Chris@82: R0[WS(rs, 6)] = FMA(KP2_000000000, TB, TE); Chris@82: R1[WS(rs, 1)] = FMS(KP2_000000000, TF, TG); Chris@82: TV = TE - TB; Chris@82: R0[WS(rs, 1)] = TV + TY; Chris@82: R1[WS(rs, 3)] = TY - TV; Chris@82: TZ = TF + TG; Chris@82: R0[WS(rs, 4)] = TZ - T10; Chris@82: R1[WS(rs, 6)] = -(TZ + T10); Chris@82: } Chris@82: { Chris@82: E Tw, Ty, Tp, Tx, TS, TU, Th, TR, TH, TT; Chris@82: Tw = Ts - Tv; Chris@82: Ty = Ts + Tv; Chris@82: Th = Tf + Tg; Chris@82: Tp = Th + To; Chris@82: Tx = Th - To; Chris@82: TR = TN - TQ; Chris@82: TS = TK + TR; Chris@82: TU = TR - TK; Chris@82: R1[WS(rs, 4)] = -(FMA(KP2_000000000, Tp, Tw)); Chris@82: R0[WS(rs, 3)] = FMA(KP2_000000000, Tx, Ty); Chris@82: TH = Tx - Ty; Chris@82: R1[WS(rs, 5)] = TH - TS; Chris@82: R1[0] = TH + TS; Chris@82: TT = Tw - Tp; Chris@82: R0[WS(rs, 2)] = TT - TU; Chris@82: R0[WS(rs, 7)] = TT + TU; Chris@82: } Chris@82: } Chris@82: } Chris@82: } Chris@82: Chris@82: static const kr2c_desc desc = { 15, "r2cbIII_15", {49, 11, 15, 0}, &GENUS }; Chris@82: Chris@82: void X(codelet_r2cbIII_15) (planner *p) { Chris@82: X(kr2c_register) (p, r2cbIII_15, &desc); Chris@82: } Chris@82: Chris@82: #endif