cannam@167: /* cannam@167: * Copyright (c) 2003, 2007-14 Matteo Frigo cannam@167: * Copyright (c) 2003, 2007-14 Massachusetts Institute of Technology cannam@167: * cannam@167: * This program is free software; you can redistribute it and/or modify cannam@167: * it under the terms of the GNU General Public License as published by cannam@167: * the Free Software Foundation; either version 2 of the License, or cannam@167: * (at your option) any later version. cannam@167: * cannam@167: * This program is distributed in the hope that it will be useful, cannam@167: * but WITHOUT ANY WARRANTY; without even the implied warranty of cannam@167: * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the cannam@167: * GNU General Public License for more details. cannam@167: * cannam@167: * You should have received a copy of the GNU General Public License cannam@167: * along with this program; if not, write to the Free Software cannam@167: * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA cannam@167: * cannam@167: */ cannam@167: cannam@167: /* This file was automatically generated --- DO NOT EDIT */ cannam@167: /* Generated on Thu May 24 08:05:19 EDT 2018 */ cannam@167: cannam@167: #include "dft/codelet-dft.h" cannam@167: cannam@167: #if defined(ARCH_PREFERS_FMA) || defined(ISA_EXTENSION_PREFERS_FMA) cannam@167: cannam@167: /* Generated by: ../../../genfft/gen_notw.native -fma -simd -compact -variables 4 -pipeline-latency 8 -n 8 -name n2sv_8 -with-ostride 1 -include dft/simd/n2s.h -store-multiple 4 */ cannam@167: cannam@167: /* cannam@167: * This function contains 52 FP additions, 8 FP multiplications, cannam@167: * (or, 44 additions, 0 multiplications, 8 fused multiply/add), cannam@167: * 34 stack variables, 1 constants, and 36 memory accesses cannam@167: */ cannam@167: #include "dft/simd/n2s.h" cannam@167: cannam@167: static void n2sv_8(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs) cannam@167: { cannam@167: DVK(KP707106781, +0.707106781186547524400844362104849039284835938); cannam@167: { cannam@167: INT i; cannam@167: for (i = v; i > 0; i = i - (2 * VL), ri = ri + ((2 * VL) * ivs), ii = ii + ((2 * VL) * ivs), ro = ro + ((2 * VL) * ovs), io = io + ((2 * VL) * ovs), MAKE_VOLATILE_STRIDE(32, is), MAKE_VOLATILE_STRIDE(32, os)) { cannam@167: V T3, Tn, Ti, TC, T6, TB, Tl, To, Td, TN, Tz, TH, Ta, TM, Tu; cannam@167: V TG; cannam@167: { cannam@167: V T1, T2, Tj, Tk; cannam@167: T1 = LD(&(ri[0]), ivs, &(ri[0])); cannam@167: T2 = LD(&(ri[WS(is, 4)]), ivs, &(ri[0])); cannam@167: T3 = VADD(T1, T2); cannam@167: Tn = VSUB(T1, T2); cannam@167: { cannam@167: V Tg, Th, T4, T5; cannam@167: Tg = LD(&(ii[0]), ivs, &(ii[0])); cannam@167: Th = LD(&(ii[WS(is, 4)]), ivs, &(ii[0])); cannam@167: Ti = VADD(Tg, Th); cannam@167: TC = VSUB(Tg, Th); cannam@167: T4 = LD(&(ri[WS(is, 2)]), ivs, &(ri[0])); cannam@167: T5 = LD(&(ri[WS(is, 6)]), ivs, &(ri[0])); cannam@167: T6 = VADD(T4, T5); cannam@167: TB = VSUB(T4, T5); cannam@167: } cannam@167: Tj = LD(&(ii[WS(is, 2)]), ivs, &(ii[0])); cannam@167: Tk = LD(&(ii[WS(is, 6)]), ivs, &(ii[0])); cannam@167: Tl = VADD(Tj, Tk); cannam@167: To = VSUB(Tj, Tk); cannam@167: { cannam@167: V Tb, Tc, Tv, Tw, Tx, Ty; cannam@167: Tb = LD(&(ri[WS(is, 7)]), ivs, &(ri[WS(is, 1)])); cannam@167: Tc = LD(&(ri[WS(is, 3)]), ivs, &(ri[WS(is, 1)])); cannam@167: Tv = VSUB(Tb, Tc); cannam@167: Tw = LD(&(ii[WS(is, 7)]), ivs, &(ii[WS(is, 1)])); cannam@167: Tx = LD(&(ii[WS(is, 3)]), ivs, &(ii[WS(is, 1)])); cannam@167: Ty = VSUB(Tw, Tx); cannam@167: Td = VADD(Tb, Tc); cannam@167: TN = VADD(Tw, Tx); cannam@167: Tz = VSUB(Tv, Ty); cannam@167: TH = VADD(Tv, Ty); cannam@167: } cannam@167: { cannam@167: V T8, T9, Tq, Tr, Ts, Tt; cannam@167: T8 = LD(&(ri[WS(is, 1)]), ivs, &(ri[WS(is, 1)])); cannam@167: T9 = LD(&(ri[WS(is, 5)]), ivs, &(ri[WS(is, 1)])); cannam@167: Tq = VSUB(T8, T9); cannam@167: Tr = LD(&(ii[WS(is, 1)]), ivs, &(ii[WS(is, 1)])); cannam@167: Ts = LD(&(ii[WS(is, 5)]), ivs, &(ii[WS(is, 1)])); cannam@167: Tt = VSUB(Tr, Ts); cannam@167: Ta = VADD(T8, T9); cannam@167: TM = VADD(Tr, Ts); cannam@167: Tu = VADD(Tq, Tt); cannam@167: TG = VSUB(Tt, Tq); cannam@167: } cannam@167: } cannam@167: { cannam@167: V TR, TS, TT, TU, TV, TW, TX, TY; cannam@167: { cannam@167: V T7, Te, TP, TQ; cannam@167: T7 = VADD(T3, T6); cannam@167: Te = VADD(Ta, Td); cannam@167: TR = VSUB(T7, Te); cannam@167: STM4(&(ro[4]), TR, ovs, &(ro[0])); cannam@167: TS = VADD(T7, Te); cannam@167: STM4(&(ro[0]), TS, ovs, &(ro[0])); cannam@167: TP = VADD(Ti, Tl); cannam@167: TQ = VADD(TM, TN); cannam@167: TT = VSUB(TP, TQ); cannam@167: STM4(&(io[4]), TT, ovs, &(io[0])); cannam@167: TU = VADD(TP, TQ); cannam@167: STM4(&(io[0]), TU, ovs, &(io[0])); cannam@167: } cannam@167: { cannam@167: V Tf, Tm, TL, TO; cannam@167: Tf = VSUB(Td, Ta); cannam@167: Tm = VSUB(Ti, Tl); cannam@167: TV = VADD(Tf, Tm); cannam@167: STM4(&(io[2]), TV, ovs, &(io[0])); cannam@167: TW = VSUB(Tm, Tf); cannam@167: STM4(&(io[6]), TW, ovs, &(io[0])); cannam@167: TL = VSUB(T3, T6); cannam@167: TO = VSUB(TM, TN); cannam@167: TX = VSUB(TL, TO); cannam@167: STM4(&(ro[6]), TX, ovs, &(ro[0])); cannam@167: TY = VADD(TL, TO); cannam@167: STM4(&(ro[2]), TY, ovs, &(ro[0])); cannam@167: } cannam@167: { cannam@167: V TZ, T10, T11, T12; cannam@167: { cannam@167: V Tp, TA, TJ, TK; cannam@167: Tp = VADD(Tn, To); cannam@167: TA = VADD(Tu, Tz); cannam@167: TZ = VFNMS(LDK(KP707106781), TA, Tp); cannam@167: STM4(&(ro[5]), TZ, ovs, &(ro[1])); cannam@167: T10 = VFMA(LDK(KP707106781), TA, Tp); cannam@167: STM4(&(ro[1]), T10, ovs, &(ro[1])); cannam@167: TJ = VSUB(TC, TB); cannam@167: TK = VADD(TG, TH); cannam@167: T11 = VFNMS(LDK(KP707106781), TK, TJ); cannam@167: STM4(&(io[5]), T11, ovs, &(io[1])); cannam@167: T12 = VFMA(LDK(KP707106781), TK, TJ); cannam@167: STM4(&(io[1]), T12, ovs, &(io[1])); cannam@167: } cannam@167: { cannam@167: V TD, TE, T13, T14; cannam@167: TD = VADD(TB, TC); cannam@167: TE = VSUB(Tz, Tu); cannam@167: T13 = VFNMS(LDK(KP707106781), TE, TD); cannam@167: STM4(&(io[7]), T13, ovs, &(io[1])); cannam@167: STN4(&(io[4]), TT, T11, TW, T13, ovs); cannam@167: T14 = VFMA(LDK(KP707106781), TE, TD); cannam@167: STM4(&(io[3]), T14, ovs, &(io[1])); cannam@167: STN4(&(io[0]), TU, T12, TV, T14, ovs); cannam@167: } cannam@167: { cannam@167: V TF, TI, T15, T16; cannam@167: TF = VSUB(Tn, To); cannam@167: TI = VSUB(TG, TH); cannam@167: T15 = VFNMS(LDK(KP707106781), TI, TF); cannam@167: STM4(&(ro[7]), T15, ovs, &(ro[1])); cannam@167: STN4(&(ro[4]), TR, TZ, TX, T15, ovs); cannam@167: T16 = VFMA(LDK(KP707106781), TI, TF); cannam@167: STM4(&(ro[3]), T16, ovs, &(ro[1])); cannam@167: STN4(&(ro[0]), TS, T10, TY, T16, ovs); cannam@167: } cannam@167: } cannam@167: } cannam@167: } cannam@167: } cannam@167: VLEAVE(); cannam@167: } cannam@167: cannam@167: static const kdft_desc desc = { 8, XSIMD_STRING("n2sv_8"), {44, 0, 8, 0}, &GENUS, 0, 1, 0, 0 }; cannam@167: cannam@167: void XSIMD(codelet_n2sv_8) (planner *p) { cannam@167: X(kdft_register) (p, n2sv_8, &desc); cannam@167: } cannam@167: cannam@167: #else cannam@167: cannam@167: /* Generated by: ../../../genfft/gen_notw.native -simd -compact -variables 4 -pipeline-latency 8 -n 8 -name n2sv_8 -with-ostride 1 -include dft/simd/n2s.h -store-multiple 4 */ cannam@167: cannam@167: /* cannam@167: * This function contains 52 FP additions, 4 FP multiplications, cannam@167: * (or, 52 additions, 4 multiplications, 0 fused multiply/add), cannam@167: * 34 stack variables, 1 constants, and 36 memory accesses cannam@167: */ cannam@167: #include "dft/simd/n2s.h" cannam@167: cannam@167: static void n2sv_8(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs) cannam@167: { cannam@167: DVK(KP707106781, +0.707106781186547524400844362104849039284835938); cannam@167: { cannam@167: INT i; cannam@167: for (i = v; i > 0; i = i - (2 * VL), ri = ri + ((2 * VL) * ivs), ii = ii + ((2 * VL) * ivs), ro = ro + ((2 * VL) * ovs), io = io + ((2 * VL) * ovs), MAKE_VOLATILE_STRIDE(32, is), MAKE_VOLATILE_STRIDE(32, os)) { cannam@167: V T3, Tn, Ti, TC, T6, TB, Tl, To, Td, TN, Tz, TH, Ta, TM, Tu; cannam@167: V TG; cannam@167: { cannam@167: V T1, T2, Tj, Tk; cannam@167: T1 = LD(&(ri[0]), ivs, &(ri[0])); cannam@167: T2 = LD(&(ri[WS(is, 4)]), ivs, &(ri[0])); cannam@167: T3 = VADD(T1, T2); cannam@167: Tn = VSUB(T1, T2); cannam@167: { cannam@167: V Tg, Th, T4, T5; cannam@167: Tg = LD(&(ii[0]), ivs, &(ii[0])); cannam@167: Th = LD(&(ii[WS(is, 4)]), ivs, &(ii[0])); cannam@167: Ti = VADD(Tg, Th); cannam@167: TC = VSUB(Tg, Th); cannam@167: T4 = LD(&(ri[WS(is, 2)]), ivs, &(ri[0])); cannam@167: T5 = LD(&(ri[WS(is, 6)]), ivs, &(ri[0])); cannam@167: T6 = VADD(T4, T5); cannam@167: TB = VSUB(T4, T5); cannam@167: } cannam@167: Tj = LD(&(ii[WS(is, 2)]), ivs, &(ii[0])); cannam@167: Tk = LD(&(ii[WS(is, 6)]), ivs, &(ii[0])); cannam@167: Tl = VADD(Tj, Tk); cannam@167: To = VSUB(Tj, Tk); cannam@167: { cannam@167: V Tb, Tc, Tv, Tw, Tx, Ty; cannam@167: Tb = LD(&(ri[WS(is, 7)]), ivs, &(ri[WS(is, 1)])); cannam@167: Tc = LD(&(ri[WS(is, 3)]), ivs, &(ri[WS(is, 1)])); cannam@167: Tv = VSUB(Tb, Tc); cannam@167: Tw = LD(&(ii[WS(is, 7)]), ivs, &(ii[WS(is, 1)])); cannam@167: Tx = LD(&(ii[WS(is, 3)]), ivs, &(ii[WS(is, 1)])); cannam@167: Ty = VSUB(Tw, Tx); cannam@167: Td = VADD(Tb, Tc); cannam@167: TN = VADD(Tw, Tx); cannam@167: Tz = VSUB(Tv, Ty); cannam@167: TH = VADD(Tv, Ty); cannam@167: } cannam@167: { cannam@167: V T8, T9, Tq, Tr, Ts, Tt; cannam@167: T8 = LD(&(ri[WS(is, 1)]), ivs, &(ri[WS(is, 1)])); cannam@167: T9 = LD(&(ri[WS(is, 5)]), ivs, &(ri[WS(is, 1)])); cannam@167: Tq = VSUB(T8, T9); cannam@167: Tr = LD(&(ii[WS(is, 1)]), ivs, &(ii[WS(is, 1)])); cannam@167: Ts = LD(&(ii[WS(is, 5)]), ivs, &(ii[WS(is, 1)])); cannam@167: Tt = VSUB(Tr, Ts); cannam@167: Ta = VADD(T8, T9); cannam@167: TM = VADD(Tr, Ts); cannam@167: Tu = VADD(Tq, Tt); cannam@167: TG = VSUB(Tt, Tq); cannam@167: } cannam@167: } cannam@167: { cannam@167: V TR, TS, TT, TU, TV, TW, TX, TY; cannam@167: { cannam@167: V T7, Te, TP, TQ; cannam@167: T7 = VADD(T3, T6); cannam@167: Te = VADD(Ta, Td); cannam@167: TR = VSUB(T7, Te); cannam@167: STM4(&(ro[4]), TR, ovs, &(ro[0])); cannam@167: TS = VADD(T7, Te); cannam@167: STM4(&(ro[0]), TS, ovs, &(ro[0])); cannam@167: TP = VADD(Ti, Tl); cannam@167: TQ = VADD(TM, TN); cannam@167: TT = VSUB(TP, TQ); cannam@167: STM4(&(io[4]), TT, ovs, &(io[0])); cannam@167: TU = VADD(TP, TQ); cannam@167: STM4(&(io[0]), TU, ovs, &(io[0])); cannam@167: } cannam@167: { cannam@167: V Tf, Tm, TL, TO; cannam@167: Tf = VSUB(Td, Ta); cannam@167: Tm = VSUB(Ti, Tl); cannam@167: TV = VADD(Tf, Tm); cannam@167: STM4(&(io[2]), TV, ovs, &(io[0])); cannam@167: TW = VSUB(Tm, Tf); cannam@167: STM4(&(io[6]), TW, ovs, &(io[0])); cannam@167: TL = VSUB(T3, T6); cannam@167: TO = VSUB(TM, TN); cannam@167: TX = VSUB(TL, TO); cannam@167: STM4(&(ro[6]), TX, ovs, &(ro[0])); cannam@167: TY = VADD(TL, TO); cannam@167: STM4(&(ro[2]), TY, ovs, &(ro[0])); cannam@167: } cannam@167: { cannam@167: V TZ, T10, T11, T12; cannam@167: { cannam@167: V Tp, TA, TJ, TK; cannam@167: Tp = VADD(Tn, To); cannam@167: TA = VMUL(LDK(KP707106781), VADD(Tu, Tz)); cannam@167: TZ = VSUB(Tp, TA); cannam@167: STM4(&(ro[5]), TZ, ovs, &(ro[1])); cannam@167: T10 = VADD(Tp, TA); cannam@167: STM4(&(ro[1]), T10, ovs, &(ro[1])); cannam@167: TJ = VSUB(TC, TB); cannam@167: TK = VMUL(LDK(KP707106781), VADD(TG, TH)); cannam@167: T11 = VSUB(TJ, TK); cannam@167: STM4(&(io[5]), T11, ovs, &(io[1])); cannam@167: T12 = VADD(TJ, TK); cannam@167: STM4(&(io[1]), T12, ovs, &(io[1])); cannam@167: } cannam@167: { cannam@167: V TD, TE, T13, T14; cannam@167: TD = VADD(TB, TC); cannam@167: TE = VMUL(LDK(KP707106781), VSUB(Tz, Tu)); cannam@167: T13 = VSUB(TD, TE); cannam@167: STM4(&(io[7]), T13, ovs, &(io[1])); cannam@167: STN4(&(io[4]), TT, T11, TW, T13, ovs); cannam@167: T14 = VADD(TD, TE); cannam@167: STM4(&(io[3]), T14, ovs, &(io[1])); cannam@167: STN4(&(io[0]), TU, T12, TV, T14, ovs); cannam@167: } cannam@167: { cannam@167: V TF, TI, T15, T16; cannam@167: TF = VSUB(Tn, To); cannam@167: TI = VMUL(LDK(KP707106781), VSUB(TG, TH)); cannam@167: T15 = VSUB(TF, TI); cannam@167: STM4(&(ro[7]), T15, ovs, &(ro[1])); cannam@167: STN4(&(ro[4]), TR, TZ, TX, T15, ovs); cannam@167: T16 = VADD(TF, TI); cannam@167: STM4(&(ro[3]), T16, ovs, &(ro[1])); cannam@167: STN4(&(ro[0]), TS, T10, TY, T16, ovs); cannam@167: } cannam@167: } cannam@167: } cannam@167: } cannam@167: } cannam@167: VLEAVE(); cannam@167: } cannam@167: cannam@167: static const kdft_desc desc = { 8, XSIMD_STRING("n2sv_8"), {52, 4, 0, 0}, &GENUS, 0, 1, 0, 0 }; cannam@167: cannam@167: void XSIMD(codelet_n2sv_8) (planner *p) { cannam@167: X(kdft_register) (p, n2sv_8, &desc); cannam@167: } cannam@167: cannam@167: #endif