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 07:37:47 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.native -fma -reorder-insns -schedule-for-pipeline -simd -compact -variables 4 -pipeline-latency 8 -n 4 -name n2sv_4 -with-ostride 1 -include n2s.h -store-multiple 4 */ cannam@95: cannam@95: /* cannam@95: * This function contains 16 FP additions, 0 FP multiplications, cannam@95: * (or, 16 additions, 0 multiplications, 0 fused multiply/add), cannam@95: * 25 stack variables, 0 constants, and 18 memory accesses cannam@95: */ cannam@95: #include "n2s.h" cannam@95: cannam@95: static void n2sv_4(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs) cannam@95: { cannam@95: { cannam@95: INT i; cannam@95: 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(16, is), MAKE_VOLATILE_STRIDE(16, os)) { cannam@95: V T1, T2, T7, T8, T4, T5, Tc, Td; cannam@95: T1 = LD(&(ri[0]), ivs, &(ri[0])); cannam@95: T2 = LD(&(ri[WS(is, 2)]), ivs, &(ri[0])); cannam@95: T7 = LD(&(ii[0]), ivs, &(ii[0])); cannam@95: T8 = LD(&(ii[WS(is, 2)]), ivs, &(ii[0])); cannam@95: T4 = LD(&(ri[WS(is, 1)]), ivs, &(ri[WS(is, 1)])); cannam@95: T5 = LD(&(ri[WS(is, 3)]), ivs, &(ri[WS(is, 1)])); cannam@95: Tc = LD(&(ii[WS(is, 1)]), ivs, &(ii[WS(is, 1)])); cannam@95: Td = LD(&(ii[WS(is, 3)]), ivs, &(ii[WS(is, 1)])); cannam@95: { cannam@95: V T3, Tb, T9, Tf, T6, Ta, Te, Tg; cannam@95: T3 = VADD(T1, T2); cannam@95: Tb = VSUB(T1, T2); cannam@95: T9 = VSUB(T7, T8); cannam@95: Tf = VADD(T7, T8); cannam@95: T6 = VADD(T4, T5); cannam@95: Ta = VSUB(T4, T5); cannam@95: Te = VSUB(Tc, Td); cannam@95: Tg = VADD(Tc, Td); cannam@95: { cannam@95: V Th, Ti, Tj, Tk; cannam@95: Th = VADD(Ta, T9); cannam@95: STM4(&(io[3]), Th, ovs, &(io[1])); cannam@95: Ti = VSUB(T9, Ta); cannam@95: STM4(&(io[1]), Ti, ovs, &(io[1])); cannam@95: Tj = VADD(T3, T6); cannam@95: STM4(&(ro[0]), Tj, ovs, &(ro[0])); cannam@95: Tk = VSUB(T3, T6); cannam@95: STM4(&(ro[2]), Tk, ovs, &(ro[0])); cannam@95: { cannam@95: V Tl, Tm, Tn, To; cannam@95: Tl = VADD(Tf, Tg); cannam@95: STM4(&(io[0]), Tl, ovs, &(io[0])); cannam@95: Tm = VSUB(Tf, Tg); cannam@95: STM4(&(io[2]), Tm, ovs, &(io[0])); cannam@95: STN4(&(io[0]), Tl, Ti, Tm, Th, ovs); cannam@95: Tn = VSUB(Tb, Te); cannam@95: STM4(&(ro[3]), Tn, ovs, &(ro[1])); cannam@95: To = VADD(Tb, Te); cannam@95: STM4(&(ro[1]), To, ovs, &(ro[1])); cannam@95: STN4(&(ro[0]), Tj, To, Tk, Tn, 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 = { 4, XSIMD_STRING("n2sv_4"), {16, 0, 0, 0}, &GENUS, 0, 1, 0, 0 }; cannam@95: cannam@95: void XSIMD(codelet_n2sv_4) (planner *p) { cannam@95: X(kdft_register) (p, n2sv_4, &desc); cannam@95: } cannam@95: cannam@95: #else /* HAVE_FMA */ cannam@95: cannam@95: /* Generated by: ../../../genfft/gen_notw.native -simd -compact -variables 4 -pipeline-latency 8 -n 4 -name n2sv_4 -with-ostride 1 -include n2s.h -store-multiple 4 */ cannam@95: cannam@95: /* cannam@95: * This function contains 16 FP additions, 0 FP multiplications, cannam@95: * (or, 16 additions, 0 multiplications, 0 fused multiply/add), cannam@95: * 17 stack variables, 0 constants, and 18 memory accesses cannam@95: */ cannam@95: #include "n2s.h" cannam@95: cannam@95: static void n2sv_4(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs) cannam@95: { cannam@95: { cannam@95: INT i; cannam@95: 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(16, is), MAKE_VOLATILE_STRIDE(16, os)) { cannam@95: V T3, Tb, T9, Tf, T6, Ta, Te, Tg; cannam@95: { cannam@95: V T1, T2, T7, T8; cannam@95: T1 = LD(&(ri[0]), ivs, &(ri[0])); cannam@95: T2 = LD(&(ri[WS(is, 2)]), ivs, &(ri[0])); cannam@95: T3 = VADD(T1, T2); cannam@95: Tb = VSUB(T1, T2); cannam@95: T7 = LD(&(ii[0]), ivs, &(ii[0])); cannam@95: T8 = LD(&(ii[WS(is, 2)]), ivs, &(ii[0])); cannam@95: T9 = VSUB(T7, T8); cannam@95: Tf = VADD(T7, T8); cannam@95: } cannam@95: { cannam@95: V T4, T5, Tc, Td; cannam@95: T4 = LD(&(ri[WS(is, 1)]), ivs, &(ri[WS(is, 1)])); cannam@95: T5 = LD(&(ri[WS(is, 3)]), ivs, &(ri[WS(is, 1)])); cannam@95: T6 = VADD(T4, T5); cannam@95: Ta = VSUB(T4, T5); cannam@95: Tc = LD(&(ii[WS(is, 1)]), ivs, &(ii[WS(is, 1)])); cannam@95: Td = LD(&(ii[WS(is, 3)]), ivs, &(ii[WS(is, 1)])); cannam@95: Te = VSUB(Tc, Td); cannam@95: Tg = VADD(Tc, Td); cannam@95: } cannam@95: { cannam@95: V Th, Ti, Tj, Tk; cannam@95: Th = VSUB(T3, T6); cannam@95: STM4(&(ro[2]), Th, ovs, &(ro[0])); cannam@95: Ti = VSUB(Tf, Tg); cannam@95: STM4(&(io[2]), Ti, ovs, &(io[0])); cannam@95: Tj = VADD(T3, T6); cannam@95: STM4(&(ro[0]), Tj, ovs, &(ro[0])); cannam@95: Tk = VADD(Tf, Tg); cannam@95: STM4(&(io[0]), Tk, ovs, &(io[0])); cannam@95: { cannam@95: V Tl, Tm, Tn, To; cannam@95: Tl = VSUB(T9, Ta); cannam@95: STM4(&(io[1]), Tl, ovs, &(io[1])); cannam@95: Tm = VADD(Tb, Te); cannam@95: STM4(&(ro[1]), Tm, ovs, &(ro[1])); cannam@95: Tn = VADD(Ta, T9); cannam@95: STM4(&(io[3]), Tn, ovs, &(io[1])); cannam@95: STN4(&(io[0]), Tk, Tl, Ti, Tn, ovs); cannam@95: To = VSUB(Tb, Te); cannam@95: STM4(&(ro[3]), To, ovs, &(ro[1])); cannam@95: STN4(&(ro[0]), Tj, Tm, Th, To, ovs); cannam@95: } cannam@95: } cannam@95: } cannam@95: } cannam@95: VLEAVE(); cannam@95: } cannam@95: cannam@95: static const kdft_desc desc = { 4, XSIMD_STRING("n2sv_4"), {16, 0, 0, 0}, &GENUS, 0, 1, 0, 0 }; cannam@95: cannam@95: void XSIMD(codelet_n2sv_4) (planner *p) { cannam@95: X(kdft_register) (p, n2sv_4, &desc); cannam@95: } cannam@95: cannam@95: #endif /* HAVE_FMA */