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 07:37:22 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 10 -name n2fv_10 -with-ostride 2 -include n2f.h -store-multiple 2 */ cannam@95: cannam@95: /* cannam@95: * This function contains 42 FP additions, 22 FP multiplications, cannam@95: * (or, 24 additions, 4 multiplications, 18 fused multiply/add), cannam@95: * 53 stack variables, 4 constants, and 25 memory accesses cannam@95: */ cannam@95: #include "n2f.h" cannam@95: cannam@95: static void n2fv_10(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(KP559016994, +0.559016994374947424102293417182819058860154590); cannam@95: DVK(KP250000000, +0.250000000000000000000000000000000000000000000); cannam@95: DVK(KP618033988, +0.618033988749894848204586834365638117720309180); cannam@95: DVK(KP951056516, +0.951056516295153572116439333379382143405698634); 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(20, is), MAKE_VOLATILE_STRIDE(20, os)) { cannam@95: V Tb, Tr, T3, Ts, T6, Tw, Tg, Tt, T9, Tc, T1, T2; cannam@95: T1 = LD(&(xi[0]), ivs, &(xi[0])); cannam@95: T2 = LD(&(xi[WS(is, 5)]), ivs, &(xi[WS(is, 1)])); cannam@95: { cannam@95: V T4, T5, Te, Tf, T7, T8; cannam@95: T4 = LD(&(xi[WS(is, 2)]), ivs, &(xi[0])); cannam@95: T5 = LD(&(xi[WS(is, 7)]), ivs, &(xi[WS(is, 1)])); cannam@95: Te = LD(&(xi[WS(is, 6)]), ivs, &(xi[0])); cannam@95: Tf = LD(&(xi[WS(is, 1)]), ivs, &(xi[WS(is, 1)])); cannam@95: T7 = LD(&(xi[WS(is, 8)]), ivs, &(xi[0])); cannam@95: T8 = LD(&(xi[WS(is, 3)]), ivs, &(xi[WS(is, 1)])); cannam@95: Tb = LD(&(xi[WS(is, 4)]), ivs, &(xi[0])); cannam@95: Tr = VADD(T1, T2); cannam@95: T3 = VSUB(T1, T2); cannam@95: Ts = VADD(T4, T5); cannam@95: T6 = VSUB(T4, T5); cannam@95: Tw = VADD(Te, Tf); cannam@95: Tg = VSUB(Te, Tf); cannam@95: Tt = VADD(T7, T8); cannam@95: T9 = VSUB(T7, T8); cannam@95: Tc = LD(&(xi[WS(is, 9)]), ivs, &(xi[WS(is, 1)])); cannam@95: } cannam@95: { cannam@95: V TD, Tu, Tm, Ta, Td, Tv; cannam@95: TD = VSUB(Ts, Tt); cannam@95: Tu = VADD(Ts, Tt); cannam@95: Tm = VSUB(T6, T9); cannam@95: Ta = VADD(T6, T9); cannam@95: Td = VSUB(Tb, Tc); cannam@95: Tv = VADD(Tb, Tc); cannam@95: { cannam@95: V TC, Tx, Tn, Th; cannam@95: TC = VSUB(Tv, Tw); cannam@95: Tx = VADD(Tv, Tw); cannam@95: Tn = VSUB(Td, Tg); cannam@95: Th = VADD(Td, Tg); cannam@95: { cannam@95: V Ty, TA, TE, TG, Ti, Tk, To, Tq; cannam@95: Ty = VADD(Tu, Tx); cannam@95: TA = VSUB(Tu, Tx); cannam@95: TE = VMUL(LDK(KP951056516), VFNMS(LDK(KP618033988), TD, TC)); cannam@95: TG = VMUL(LDK(KP951056516), VFMA(LDK(KP618033988), TC, TD)); cannam@95: Ti = VADD(Ta, Th); cannam@95: Tk = VSUB(Ta, Th); cannam@95: To = VMUL(LDK(KP951056516), VFMA(LDK(KP618033988), Tn, Tm)); cannam@95: Tq = VMUL(LDK(KP951056516), VFNMS(LDK(KP618033988), Tm, Tn)); cannam@95: { cannam@95: V Tz, TH, Tj, TI; cannam@95: Tz = VFNMS(LDK(KP250000000), Ty, Tr); cannam@95: TH = VADD(Tr, Ty); cannam@95: STM2(&(xo[0]), TH, ovs, &(xo[0])); cannam@95: Tj = VFNMS(LDK(KP250000000), Ti, T3); cannam@95: TI = VADD(T3, Ti); cannam@95: STM2(&(xo[10]), TI, ovs, &(xo[2])); cannam@95: { cannam@95: V TB, TF, Tl, Tp; cannam@95: TB = VFNMS(LDK(KP559016994), TA, Tz); cannam@95: TF = VFMA(LDK(KP559016994), TA, Tz); cannam@95: Tl = VFMA(LDK(KP559016994), Tk, Tj); cannam@95: Tp = VFNMS(LDK(KP559016994), Tk, Tj); cannam@95: { cannam@95: V TJ, TK, TL, TM; cannam@95: TJ = VFMAI(TG, TF); cannam@95: STM2(&(xo[8]), TJ, ovs, &(xo[0])); cannam@95: STN2(&(xo[8]), TJ, TI, ovs); cannam@95: TK = VFNMSI(TG, TF); cannam@95: STM2(&(xo[12]), TK, ovs, &(xo[0])); cannam@95: TL = VFNMSI(TE, TB); cannam@95: STM2(&(xo[16]), TL, ovs, &(xo[0])); cannam@95: TM = VFMAI(TE, TB); cannam@95: STM2(&(xo[4]), TM, ovs, &(xo[0])); cannam@95: { cannam@95: V TN, TO, TP, TQ; cannam@95: TN = VFNMSI(Tq, Tp); cannam@95: STM2(&(xo[6]), TN, ovs, &(xo[2])); cannam@95: STN2(&(xo[4]), TM, TN, ovs); cannam@95: TO = VFMAI(Tq, Tp); cannam@95: STM2(&(xo[14]), TO, ovs, &(xo[2])); cannam@95: STN2(&(xo[12]), TK, TO, ovs); cannam@95: TP = VFMAI(To, Tl); cannam@95: STM2(&(xo[18]), TP, ovs, &(xo[2])); cannam@95: STN2(&(xo[16]), TL, TP, ovs); cannam@95: TQ = VFNMSI(To, Tl); cannam@95: STM2(&(xo[2]), TQ, ovs, &(xo[2])); cannam@95: STN2(&(xo[0]), TH, TQ, ovs); cannam@95: } cannam@95: } 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 = { 10, XSIMD_STRING("n2fv_10"), {24, 4, 18, 0}, &GENUS, 0, 2, 0, 0 }; cannam@95: cannam@95: void XSIMD(codelet_n2fv_10) (planner *p) { cannam@95: X(kdft_register) (p, n2fv_10, &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 10 -name n2fv_10 -with-ostride 2 -include n2f.h -store-multiple 2 */ cannam@95: cannam@95: /* cannam@95: * This function contains 42 FP additions, 12 FP multiplications, cannam@95: * (or, 36 additions, 6 multiplications, 6 fused multiply/add), cannam@95: * 36 stack variables, 4 constants, and 25 memory accesses cannam@95: */ cannam@95: #include "n2f.h" cannam@95: cannam@95: static void n2fv_10(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(KP250000000, +0.250000000000000000000000000000000000000000000); cannam@95: DVK(KP559016994, +0.559016994374947424102293417182819058860154590); cannam@95: DVK(KP587785252, +0.587785252292473129168705954639072768597652438); cannam@95: DVK(KP951056516, +0.951056516295153572116439333379382143405698634); 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(20, is), MAKE_VOLATILE_STRIDE(20, os)) { cannam@95: V Ti, Ty, Tm, Tn, Tw, Tt, Tz, TA, TB, T7, Te, Tj, Tg, Th; cannam@95: Tg = LD(&(xi[0]), ivs, &(xi[0])); cannam@95: Th = LD(&(xi[WS(is, 5)]), ivs, &(xi[WS(is, 1)])); cannam@95: Ti = VSUB(Tg, Th); cannam@95: Ty = VADD(Tg, Th); cannam@95: { cannam@95: V T3, Tu, Td, Ts, T6, Tv, Ta, Tr; cannam@95: { cannam@95: V T1, T2, Tb, Tc; cannam@95: T1 = LD(&(xi[WS(is, 2)]), ivs, &(xi[0])); cannam@95: T2 = LD(&(xi[WS(is, 7)]), ivs, &(xi[WS(is, 1)])); cannam@95: T3 = VSUB(T1, T2); cannam@95: Tu = VADD(T1, T2); cannam@95: Tb = LD(&(xi[WS(is, 6)]), ivs, &(xi[0])); cannam@95: Tc = LD(&(xi[WS(is, 1)]), ivs, &(xi[WS(is, 1)])); cannam@95: Td = VSUB(Tb, Tc); cannam@95: Ts = VADD(Tb, Tc); cannam@95: } cannam@95: { cannam@95: V T4, T5, T8, T9; cannam@95: T4 = LD(&(xi[WS(is, 8)]), ivs, &(xi[0])); cannam@95: T5 = LD(&(xi[WS(is, 3)]), ivs, &(xi[WS(is, 1)])); cannam@95: T6 = VSUB(T4, T5); cannam@95: Tv = VADD(T4, T5); cannam@95: T8 = LD(&(xi[WS(is, 4)]), ivs, &(xi[0])); cannam@95: T9 = LD(&(xi[WS(is, 9)]), ivs, &(xi[WS(is, 1)])); cannam@95: Ta = VSUB(T8, T9); cannam@95: Tr = VADD(T8, T9); cannam@95: } cannam@95: Tm = VSUB(T3, T6); cannam@95: Tn = VSUB(Ta, Td); cannam@95: Tw = VSUB(Tu, Tv); cannam@95: Tt = VSUB(Tr, Ts); cannam@95: Tz = VADD(Tu, Tv); cannam@95: TA = VADD(Tr, Ts); cannam@95: TB = VADD(Tz, TA); cannam@95: T7 = VADD(T3, T6); cannam@95: Te = VADD(Ta, Td); cannam@95: Tj = VADD(T7, Te); cannam@95: } cannam@95: { cannam@95: V TH, TI, TK, TL, TM; cannam@95: TH = VADD(Ti, Tj); cannam@95: STM2(&(xo[10]), TH, ovs, &(xo[2])); cannam@95: TI = VADD(Ty, TB); cannam@95: STM2(&(xo[0]), TI, ovs, &(xo[0])); cannam@95: { cannam@95: V To, Tq, Tl, Tp, Tf, Tk, TJ; cannam@95: To = VBYI(VFMA(LDK(KP951056516), Tm, VMUL(LDK(KP587785252), Tn))); cannam@95: Tq = VBYI(VFNMS(LDK(KP587785252), Tm, VMUL(LDK(KP951056516), Tn))); cannam@95: Tf = VMUL(LDK(KP559016994), VSUB(T7, Te)); cannam@95: Tk = VFNMS(LDK(KP250000000), Tj, Ti); cannam@95: Tl = VADD(Tf, Tk); cannam@95: Tp = VSUB(Tk, Tf); cannam@95: TJ = VSUB(Tl, To); cannam@95: STM2(&(xo[2]), TJ, ovs, &(xo[2])); cannam@95: STN2(&(xo[0]), TI, TJ, ovs); cannam@95: TK = VADD(Tq, Tp); cannam@95: STM2(&(xo[14]), TK, ovs, &(xo[2])); cannam@95: TL = VADD(To, Tl); cannam@95: STM2(&(xo[18]), TL, ovs, &(xo[2])); cannam@95: TM = VSUB(Tp, Tq); cannam@95: STM2(&(xo[6]), TM, ovs, &(xo[2])); cannam@95: } cannam@95: { cannam@95: V Tx, TF, TE, TG, TC, TD; cannam@95: Tx = VBYI(VFNMS(LDK(KP587785252), Tw, VMUL(LDK(KP951056516), Tt))); cannam@95: TF = VBYI(VFMA(LDK(KP951056516), Tw, VMUL(LDK(KP587785252), Tt))); cannam@95: TC = VFNMS(LDK(KP250000000), TB, Ty); cannam@95: TD = VMUL(LDK(KP559016994), VSUB(Tz, TA)); cannam@95: TE = VSUB(TC, TD); cannam@95: TG = VADD(TD, TC); cannam@95: { cannam@95: V TN, TO, TP, TQ; cannam@95: TN = VADD(Tx, TE); cannam@95: STM2(&(xo[4]), TN, ovs, &(xo[0])); cannam@95: STN2(&(xo[4]), TN, TM, ovs); cannam@95: TO = VSUB(TG, TF); cannam@95: STM2(&(xo[12]), TO, ovs, &(xo[0])); cannam@95: STN2(&(xo[12]), TO, TK, ovs); cannam@95: TP = VSUB(TE, Tx); cannam@95: STM2(&(xo[16]), TP, ovs, &(xo[0])); cannam@95: STN2(&(xo[16]), TP, TL, ovs); cannam@95: TQ = VADD(TF, TG); cannam@95: STM2(&(xo[8]), TQ, ovs, &(xo[0])); cannam@95: STN2(&(xo[8]), TQ, TH, ovs); cannam@95: } cannam@95: } cannam@95: } cannam@95: } cannam@95: } cannam@95: VLEAVE(); cannam@95: } cannam@95: cannam@95: static const kdft_desc desc = { 10, XSIMD_STRING("n2fv_10"), {36, 6, 6, 0}, &GENUS, 0, 2, 0, 0 }; cannam@95: cannam@95: void XSIMD(codelet_n2fv_10) (planner *p) { cannam@95: X(kdft_register) (p, n2fv_10, &desc); cannam@95: } cannam@95: cannam@95: #endif /* HAVE_FMA */