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 07:36:58 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 -sign 1 -n 6 -name n1bv_6 -include n1b.h */ cannam@95: cannam@95: /* cannam@95: * This function contains 18 FP additions, 8 FP multiplications, cannam@95: * (or, 12 additions, 2 multiplications, 6 fused multiply/add), cannam@95: * 23 stack variables, 2 constants, and 12 memory accesses cannam@95: */ cannam@95: #include "n1b.h" cannam@95: cannam@95: static void n1bv_6(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(KP500000000, +0.500000000000000000000000000000000000000000000); cannam@95: DVK(KP866025403, +0.866025403784438646763723170752936183471402627); cannam@95: { cannam@95: INT i; cannam@95: const R *xi; cannam@95: R *xo; cannam@95: xi = ii; cannam@95: xo = io; cannam@95: for (i = v; i > 0; i = i - VL, xi = xi + (VL * ivs), xo = xo + (VL * ovs), MAKE_VOLATILE_STRIDE(12, is), MAKE_VOLATILE_STRIDE(12, os)) { cannam@95: V T1, T2, T4, T5, T7, T8; cannam@95: T1 = LD(&(xi[0]), ivs, &(xi[0])); cannam@95: T2 = LD(&(xi[WS(is, 3)]), ivs, &(xi[WS(is, 1)])); cannam@95: T4 = LD(&(xi[WS(is, 2)]), ivs, &(xi[0])); cannam@95: T5 = LD(&(xi[WS(is, 5)]), ivs, &(xi[WS(is, 1)])); cannam@95: T7 = LD(&(xi[WS(is, 4)]), ivs, &(xi[0])); cannam@95: T8 = LD(&(xi[WS(is, 1)]), ivs, &(xi[WS(is, 1)])); cannam@95: { cannam@95: V T3, Td, T6, Te, T9, Tf; cannam@95: T3 = VSUB(T1, T2); cannam@95: Td = VADD(T1, T2); cannam@95: T6 = VSUB(T4, T5); cannam@95: Te = VADD(T4, T5); cannam@95: T9 = VSUB(T7, T8); cannam@95: Tf = VADD(T7, T8); cannam@95: { cannam@95: V Tg, Ti, Ta, Tc, Th, Tb; cannam@95: Tg = VADD(Te, Tf); cannam@95: Ti = VMUL(LDK(KP866025403), VSUB(Te, Tf)); cannam@95: Ta = VADD(T6, T9); cannam@95: Tc = VMUL(LDK(KP866025403), VSUB(T6, T9)); cannam@95: Th = VFNMS(LDK(KP500000000), Tg, Td); cannam@95: ST(&(xo[0]), VADD(Td, Tg), ovs, &(xo[0])); cannam@95: Tb = VFNMS(LDK(KP500000000), Ta, T3); cannam@95: ST(&(xo[WS(os, 3)]), VADD(T3, Ta), ovs, &(xo[WS(os, 1)])); cannam@95: ST(&(xo[WS(os, 4)]), VFMAI(Ti, Th), ovs, &(xo[0])); cannam@95: ST(&(xo[WS(os, 2)]), VFNMSI(Ti, Th), ovs, &(xo[0])); cannam@95: ST(&(xo[WS(os, 5)]), VFNMSI(Tc, Tb), ovs, &(xo[WS(os, 1)])); cannam@95: ST(&(xo[WS(os, 1)]), VFMAI(Tc, Tb), ovs, &(xo[WS(os, 1)])); cannam@95: } cannam@95: } cannam@95: } cannam@95: } cannam@95: VLEAVE(); cannam@95: } cannam@95: cannam@95: static const kdft_desc desc = { 6, XSIMD_STRING("n1bv_6"), {12, 2, 6, 0}, &GENUS, 0, 0, 0, 0 }; cannam@95: cannam@95: void XSIMD(codelet_n1bv_6) (planner *p) { cannam@95: X(kdft_register) (p, n1bv_6, &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 -sign 1 -n 6 -name n1bv_6 -include n1b.h */ cannam@95: cannam@95: /* cannam@95: * This function contains 18 FP additions, 4 FP multiplications, cannam@95: * (or, 16 additions, 2 multiplications, 2 fused multiply/add), cannam@95: * 19 stack variables, 2 constants, and 12 memory accesses cannam@95: */ cannam@95: #include "n1b.h" cannam@95: cannam@95: static void n1bv_6(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(KP500000000, +0.500000000000000000000000000000000000000000000); cannam@95: DVK(KP866025403, +0.866025403784438646763723170752936183471402627); cannam@95: { cannam@95: INT i; cannam@95: const R *xi; cannam@95: R *xo; cannam@95: xi = ii; cannam@95: xo = io; cannam@95: for (i = v; i > 0; i = i - VL, xi = xi + (VL * ivs), xo = xo + (VL * ovs), MAKE_VOLATILE_STRIDE(12, is), MAKE_VOLATILE_STRIDE(12, os)) { cannam@95: V Ta, Td, T3, Te, T6, Tf, Tb, Tg, T8, T9; cannam@95: T8 = LD(&(xi[0]), ivs, &(xi[0])); cannam@95: T9 = LD(&(xi[WS(is, 3)]), ivs, &(xi[WS(is, 1)])); cannam@95: Ta = VSUB(T8, T9); cannam@95: Td = VADD(T8, T9); cannam@95: { cannam@95: V T1, T2, T4, T5; cannam@95: T1 = LD(&(xi[WS(is, 2)]), ivs, &(xi[0])); cannam@95: T2 = LD(&(xi[WS(is, 5)]), ivs, &(xi[WS(is, 1)])); cannam@95: T3 = VSUB(T1, T2); cannam@95: Te = VADD(T1, T2); cannam@95: T4 = LD(&(xi[WS(is, 4)]), ivs, &(xi[0])); cannam@95: T5 = LD(&(xi[WS(is, 1)]), ivs, &(xi[WS(is, 1)])); cannam@95: T6 = VSUB(T4, T5); cannam@95: Tf = VADD(T4, T5); cannam@95: } cannam@95: Tb = VADD(T3, T6); cannam@95: Tg = VADD(Te, Tf); cannam@95: ST(&(xo[WS(os, 3)]), VADD(Ta, Tb), ovs, &(xo[WS(os, 1)])); cannam@95: ST(&(xo[0]), VADD(Td, Tg), ovs, &(xo[0])); cannam@95: { cannam@95: V T7, Tc, Th, Ti; cannam@95: T7 = VBYI(VMUL(LDK(KP866025403), VSUB(T3, T6))); cannam@95: Tc = VFNMS(LDK(KP500000000), Tb, Ta); cannam@95: ST(&(xo[WS(os, 1)]), VADD(T7, Tc), ovs, &(xo[WS(os, 1)])); cannam@95: ST(&(xo[WS(os, 5)]), VSUB(Tc, T7), ovs, &(xo[WS(os, 1)])); cannam@95: Th = VFNMS(LDK(KP500000000), Tg, Td); cannam@95: Ti = VBYI(VMUL(LDK(KP866025403), VSUB(Te, Tf))); cannam@95: ST(&(xo[WS(os, 2)]), VSUB(Th, Ti), ovs, &(xo[0])); cannam@95: ST(&(xo[WS(os, 4)]), VADD(Ti, Th), ovs, &(xo[0])); cannam@95: } cannam@95: } cannam@95: } cannam@95: VLEAVE(); cannam@95: } cannam@95: cannam@95: static const kdft_desc desc = { 6, XSIMD_STRING("n1bv_6"), {16, 2, 2, 0}, &GENUS, 0, 0, 0, 0 }; cannam@95: cannam@95: void XSIMD(codelet_n1bv_6) (planner *p) { cannam@95: X(kdft_register) (p, n1bv_6, &desc); cannam@95: } cannam@95: cannam@95: #endif /* HAVE_FMA */