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 16:40:34 EDT 2016 */ cannam@127: cannam@127: #include "codelet-dft.h" cannam@127: cannam@127: #ifdef HAVE_FMA cannam@127: cannam@127: /* Generated by: ../../../genfft/gen_notw_c.native -fma -reorder-insns -schedule-for-pipeline -simd -compact -variables 4 -pipeline-latency 8 -sign 1 -n 8 -name n2bv_8 -with-ostride 2 -include n2b.h -store-multiple 2 */ cannam@127: cannam@127: /* cannam@127: * This function contains 26 FP additions, 10 FP multiplications, cannam@127: * (or, 16 additions, 0 multiplications, 10 fused multiply/add), cannam@127: * 38 stack variables, 1 constants, and 20 memory accesses cannam@127: */ cannam@127: #include "n2b.h" cannam@127: cannam@127: static void n2bv_8(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs) cannam@127: { cannam@127: DVK(KP707106781, +0.707106781186547524400844362104849039284835938); cannam@127: { cannam@127: INT i; cannam@127: const R *xi; cannam@127: R *xo; cannam@127: xi = ii; cannam@127: xo = io; cannam@127: for (i = v; i > 0; i = i - VL, xi = xi + (VL * ivs), xo = xo + (VL * ovs), MAKE_VOLATILE_STRIDE(16, is), MAKE_VOLATILE_STRIDE(16, os)) { cannam@127: V T1, T2, Tc, Td, T4, T5, T7, T8; cannam@127: T1 = LD(&(xi[0]), ivs, &(xi[0])); cannam@127: T2 = LD(&(xi[WS(is, 4)]), ivs, &(xi[0])); cannam@127: Tc = LD(&(xi[WS(is, 2)]), ivs, &(xi[0])); cannam@127: Td = LD(&(xi[WS(is, 6)]), ivs, &(xi[0])); cannam@127: T4 = LD(&(xi[WS(is, 1)]), ivs, &(xi[WS(is, 1)])); cannam@127: T5 = LD(&(xi[WS(is, 5)]), ivs, &(xi[WS(is, 1)])); cannam@127: T7 = LD(&(xi[WS(is, 7)]), ivs, &(xi[WS(is, 1)])); cannam@127: T8 = LD(&(xi[WS(is, 3)]), ivs, &(xi[WS(is, 1)])); cannam@127: { cannam@127: V T3, Tj, Te, Tk, T6, Tm, T9, Tn, Tp, Tl; cannam@127: T3 = VSUB(T1, T2); cannam@127: Tj = VADD(T1, T2); cannam@127: Te = VSUB(Tc, Td); cannam@127: Tk = VADD(Tc, Td); cannam@127: T6 = VSUB(T4, T5); cannam@127: Tm = VADD(T4, T5); cannam@127: T9 = VSUB(T7, T8); cannam@127: Tn = VADD(T7, T8); cannam@127: Tp = VADD(Tj, Tk); cannam@127: Tl = VSUB(Tj, Tk); cannam@127: { cannam@127: V Tq, To, Ta, Tf; cannam@127: Tq = VADD(Tm, Tn); cannam@127: To = VSUB(Tm, Tn); cannam@127: Ta = VADD(T6, T9); cannam@127: Tf = VSUB(T6, T9); cannam@127: { cannam@127: V Tr, Ts, Tt, Tu, Tg, Ti, Tb, Th; cannam@127: Tr = VFMAI(To, Tl); cannam@127: STM2(&(xo[4]), Tr, ovs, &(xo[0])); cannam@127: Ts = VFNMSI(To, Tl); cannam@127: STM2(&(xo[12]), Ts, ovs, &(xo[0])); cannam@127: Tt = VADD(Tp, Tq); cannam@127: STM2(&(xo[0]), Tt, ovs, &(xo[0])); cannam@127: Tu = VSUB(Tp, Tq); cannam@127: STM2(&(xo[8]), Tu, ovs, &(xo[0])); cannam@127: Tg = VFNMS(LDK(KP707106781), Tf, Te); cannam@127: Ti = VFMA(LDK(KP707106781), Tf, Te); cannam@127: Tb = VFNMS(LDK(KP707106781), Ta, T3); cannam@127: Th = VFMA(LDK(KP707106781), Ta, T3); cannam@127: { cannam@127: V Tv, Tw, Tx, Ty; cannam@127: Tv = VFNMSI(Ti, Th); cannam@127: STM2(&(xo[14]), Tv, ovs, &(xo[2])); cannam@127: STN2(&(xo[12]), Ts, Tv, ovs); cannam@127: Tw = VFMAI(Ti, Th); cannam@127: STM2(&(xo[2]), Tw, ovs, &(xo[2])); cannam@127: STN2(&(xo[0]), Tt, Tw, ovs); cannam@127: Tx = VFMAI(Tg, Tb); cannam@127: STM2(&(xo[10]), Tx, ovs, &(xo[2])); cannam@127: STN2(&(xo[8]), Tu, Tx, ovs); cannam@127: Ty = VFNMSI(Tg, Tb); cannam@127: STM2(&(xo[6]), Ty, ovs, &(xo[2])); cannam@127: STN2(&(xo[4]), Tr, Ty, ovs); cannam@127: } cannam@127: } cannam@127: } cannam@127: } cannam@127: } cannam@127: } cannam@127: VLEAVE(); cannam@127: } cannam@127: cannam@127: static const kdft_desc desc = { 8, XSIMD_STRING("n2bv_8"), {16, 0, 10, 0}, &GENUS, 0, 2, 0, 0 }; cannam@127: cannam@127: void XSIMD(codelet_n2bv_8) (planner *p) { cannam@127: X(kdft_register) (p, n2bv_8, &desc); cannam@127: } cannam@127: cannam@127: #else /* HAVE_FMA */ cannam@127: cannam@127: /* Generated by: ../../../genfft/gen_notw_c.native -simd -compact -variables 4 -pipeline-latency 8 -sign 1 -n 8 -name n2bv_8 -with-ostride 2 -include n2b.h -store-multiple 2 */ cannam@127: cannam@127: /* cannam@127: * This function contains 26 FP additions, 2 FP multiplications, cannam@127: * (or, 26 additions, 2 multiplications, 0 fused multiply/add), cannam@127: * 24 stack variables, 1 constants, and 20 memory accesses cannam@127: */ cannam@127: #include "n2b.h" cannam@127: cannam@127: static void n2bv_8(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs) cannam@127: { cannam@127: DVK(KP707106781, +0.707106781186547524400844362104849039284835938); cannam@127: { cannam@127: INT i; cannam@127: const R *xi; cannam@127: R *xo; cannam@127: xi = ii; cannam@127: xo = io; cannam@127: for (i = v; i > 0; i = i - VL, xi = xi + (VL * ivs), xo = xo + (VL * ovs), MAKE_VOLATILE_STRIDE(16, is), MAKE_VOLATILE_STRIDE(16, os)) { cannam@127: V Ta, Tk, Te, Tj, T7, Tn, Tf, Tm, Tr, Tu; cannam@127: { cannam@127: V T8, T9, Tc, Td; cannam@127: T8 = LD(&(xi[WS(is, 2)]), ivs, &(xi[0])); cannam@127: T9 = LD(&(xi[WS(is, 6)]), ivs, &(xi[0])); cannam@127: Ta = VSUB(T8, T9); cannam@127: Tk = VADD(T8, T9); cannam@127: Tc = LD(&(xi[0]), ivs, &(xi[0])); cannam@127: Td = LD(&(xi[WS(is, 4)]), ivs, &(xi[0])); cannam@127: Te = VSUB(Tc, Td); cannam@127: Tj = VADD(Tc, Td); cannam@127: { cannam@127: V T1, T2, T3, T4, T5, T6; cannam@127: T1 = LD(&(xi[WS(is, 1)]), ivs, &(xi[WS(is, 1)])); cannam@127: T2 = LD(&(xi[WS(is, 5)]), ivs, &(xi[WS(is, 1)])); cannam@127: T3 = VSUB(T1, T2); cannam@127: T4 = LD(&(xi[WS(is, 7)]), ivs, &(xi[WS(is, 1)])); cannam@127: T5 = LD(&(xi[WS(is, 3)]), ivs, &(xi[WS(is, 1)])); cannam@127: T6 = VSUB(T4, T5); cannam@127: T7 = VMUL(LDK(KP707106781), VSUB(T3, T6)); cannam@127: Tn = VADD(T4, T5); cannam@127: Tf = VMUL(LDK(KP707106781), VADD(T3, T6)); cannam@127: Tm = VADD(T1, T2); cannam@127: } cannam@127: } cannam@127: { cannam@127: V Ts, Tb, Tg, Tp, Tq, Tt; cannam@127: Tb = VBYI(VSUB(T7, Ta)); cannam@127: Tg = VSUB(Te, Tf); cannam@127: Tr = VADD(Tb, Tg); cannam@127: STM2(&(xo[6]), Tr, ovs, &(xo[2])); cannam@127: Ts = VSUB(Tg, Tb); cannam@127: STM2(&(xo[10]), Ts, ovs, &(xo[2])); cannam@127: Tp = VADD(Tj, Tk); cannam@127: Tq = VADD(Tm, Tn); cannam@127: Tt = VSUB(Tp, Tq); cannam@127: STM2(&(xo[8]), Tt, ovs, &(xo[0])); cannam@127: STN2(&(xo[8]), Tt, Ts, ovs); cannam@127: Tu = VADD(Tp, Tq); cannam@127: STM2(&(xo[0]), Tu, ovs, &(xo[0])); cannam@127: } cannam@127: { cannam@127: V Tw, Th, Ti, Tv; cannam@127: Th = VBYI(VADD(Ta, T7)); cannam@127: Ti = VADD(Te, Tf); cannam@127: Tv = VADD(Th, Ti); cannam@127: STM2(&(xo[2]), Tv, ovs, &(xo[2])); cannam@127: STN2(&(xo[0]), Tu, Tv, ovs); cannam@127: Tw = VSUB(Ti, Th); cannam@127: STM2(&(xo[14]), Tw, ovs, &(xo[2])); cannam@127: { cannam@127: V Tl, To, Tx, Ty; cannam@127: Tl = VSUB(Tj, Tk); cannam@127: To = VBYI(VSUB(Tm, Tn)); cannam@127: Tx = VSUB(Tl, To); cannam@127: STM2(&(xo[12]), Tx, ovs, &(xo[0])); cannam@127: STN2(&(xo[12]), Tx, Tw, ovs); cannam@127: Ty = VADD(Tl, To); cannam@127: STM2(&(xo[4]), Ty, ovs, &(xo[0])); cannam@127: STN2(&(xo[4]), Ty, Tr, ovs); cannam@127: } cannam@127: } cannam@127: } cannam@127: } cannam@127: VLEAVE(); cannam@127: } cannam@127: cannam@127: static const kdft_desc desc = { 8, XSIMD_STRING("n2bv_8"), {26, 2, 0, 0}, &GENUS, 0, 2, 0, 0 }; cannam@127: cannam@127: void XSIMD(codelet_n2bv_8) (planner *p) { cannam@127: X(kdft_register) (p, n2bv_8, &desc); cannam@127: } cannam@127: cannam@127: #endif /* HAVE_FMA */