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 07:37:21 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 6 -name n2fv_6 -with-ostride 2 -include n2f.h -store-multiple 2 */ 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: * 29 stack variables, 2 constants, and 15 memory accesses cannam@95: */ cannam@95: #include "n2f.h" cannam@95: cannam@95: static void n2fv_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 = 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(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; cannam@95: Tg = VADD(Te, Tf); cannam@95: Ti = VMUL(LDK(KP866025403), VSUB(Tf, Te)); cannam@95: Ta = VADD(T6, T9); cannam@95: Tc = VMUL(LDK(KP866025403), VSUB(T9, T6)); cannam@95: { cannam@95: V Th, Tj, Tb, Tk; cannam@95: Th = VFNMS(LDK(KP500000000), Tg, Td); cannam@95: Tj = VADD(Td, Tg); cannam@95: STM2(&(xo[0]), Tj, ovs, &(xo[0])); cannam@95: Tb = VFNMS(LDK(KP500000000), Ta, T3); cannam@95: Tk = VADD(T3, Ta); cannam@95: STM2(&(xo[6]), Tk, ovs, &(xo[2])); cannam@95: { cannam@95: V Tl, Tm, Tn, To; cannam@95: Tl = VFMAI(Ti, Th); cannam@95: STM2(&(xo[8]), Tl, ovs, &(xo[0])); cannam@95: Tm = VFNMSI(Ti, Th); cannam@95: STM2(&(xo[4]), Tm, ovs, &(xo[0])); cannam@95: STN2(&(xo[4]), Tm, Tk, ovs); cannam@95: Tn = VFMAI(Tc, Tb); cannam@95: STM2(&(xo[2]), Tn, ovs, &(xo[2])); cannam@95: STN2(&(xo[0]), Tj, Tn, ovs); cannam@95: To = VFNMSI(Tc, Tb); cannam@95: STM2(&(xo[10]), To, ovs, &(xo[2])); cannam@95: STN2(&(xo[8]), Tl, To, ovs); 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 = { 6, XSIMD_STRING("n2fv_6"), {12, 2, 6, 0}, &GENUS, 0, 2, 0, 0 }; cannam@95: cannam@95: void XSIMD(codelet_n2fv_6) (planner *p) { cannam@95: X(kdft_register) (p, n2fv_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 -n 6 -name n2fv_6 -with-ostride 2 -include n2f.h -store-multiple 2 */ 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: * 25 stack variables, 2 constants, and 15 memory accesses cannam@95: */ cannam@95: #include "n2f.h" cannam@95: cannam@95: static void n2fv_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(KP866025403, +0.866025403784438646763723170752936183471402627); cannam@95: DVK(KP500000000, +0.500000000000000000000000000000000000000000000); 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(12, is), MAKE_VOLATILE_STRIDE(12, os)) { cannam@95: V T3, Td, T6, Te, T9, Tf, Ta, Tg, T1, T2, Tj, Tk; cannam@95: T1 = LD(&(xi[0]), ivs, &(xi[0])); cannam@95: T2 = LD(&(xi[WS(is, 3)]), ivs, &(xi[WS(is, 1)])); cannam@95: T3 = VSUB(T1, T2); cannam@95: Td = VADD(T1, T2); cannam@95: { cannam@95: V T4, T5, T7, T8; 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: T6 = VSUB(T4, T5); cannam@95: Te = VADD(T4, T5); 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: T9 = VSUB(T7, T8); cannam@95: Tf = VADD(T7, T8); cannam@95: } cannam@95: Ta = VADD(T6, T9); cannam@95: Tg = VADD(Te, Tf); cannam@95: Tj = VADD(T3, Ta); cannam@95: STM2(&(xo[6]), Tj, ovs, &(xo[2])); cannam@95: Tk = VADD(Td, Tg); cannam@95: STM2(&(xo[0]), Tk, ovs, &(xo[0])); cannam@95: { cannam@95: V Tl, Tb, Tc, Tm; cannam@95: Tb = VFNMS(LDK(KP500000000), Ta, T3); cannam@95: Tc = VBYI(VMUL(LDK(KP866025403), VSUB(T9, T6))); cannam@95: Tl = VSUB(Tb, Tc); cannam@95: STM2(&(xo[10]), Tl, ovs, &(xo[2])); cannam@95: Tm = VADD(Tb, Tc); cannam@95: STM2(&(xo[2]), Tm, ovs, &(xo[2])); cannam@95: STN2(&(xo[0]), Tk, Tm, ovs); cannam@95: { cannam@95: V Th, Ti, Tn, To; cannam@95: Th = VFNMS(LDK(KP500000000), Tg, Td); cannam@95: Ti = VBYI(VMUL(LDK(KP866025403), VSUB(Tf, Te))); cannam@95: Tn = VSUB(Th, Ti); cannam@95: STM2(&(xo[4]), Tn, ovs, &(xo[0])); cannam@95: STN2(&(xo[4]), Tn, Tj, ovs); cannam@95: To = VADD(Th, Ti); cannam@95: STM2(&(xo[8]), To, ovs, &(xo[0])); cannam@95: STN2(&(xo[8]), To, Tl, ovs); 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("n2fv_6"), {16, 2, 2, 0}, &GENUS, 0, 2, 0, 0 }; cannam@95: cannam@95: void XSIMD(codelet_n2fv_6) (planner *p) { cannam@95: X(kdft_register) (p, n2fv_6, &desc); cannam@95: } cannam@95: cannam@95: #endif /* HAVE_FMA */