Chris@10: /* Chris@10: * Copyright (c) 2003, 2007-11 Matteo Frigo Chris@10: * Copyright (c) 2003, 2007-11 Massachusetts Institute of Technology Chris@10: * Chris@10: * This program is free software; you can redistribute it and/or modify Chris@10: * it under the terms of the GNU General Public License as published by Chris@10: * the Free Software Foundation; either version 2 of the License, or Chris@10: * (at your option) any later version. Chris@10: * Chris@10: * This program is distributed in the hope that it will be useful, Chris@10: * but WITHOUT ANY WARRANTY; without even the implied warranty of Chris@10: * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the Chris@10: * GNU General Public License for more details. Chris@10: * Chris@10: * You should have received a copy of the GNU General Public License Chris@10: * along with this program; if not, write to the Free Software Chris@10: * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA Chris@10: * Chris@10: */ Chris@10: Chris@10: /* This file was automatically generated --- DO NOT EDIT */ Chris@10: /* Generated on Sun Nov 25 07:36:52 EST 2012 */ Chris@10: Chris@10: #include "codelet-dft.h" Chris@10: Chris@10: #ifdef HAVE_FMA Chris@10: Chris@10: /* Generated by: ../../../genfft/gen_notw_c.native -fma -reorder-insns -schedule-for-pipeline -simd -compact -variables 4 -pipeline-latency 8 -n 5 -name n1fv_5 -include n1f.h */ Chris@10: Chris@10: /* Chris@10: * This function contains 16 FP additions, 11 FP multiplications, Chris@10: * (or, 7 additions, 2 multiplications, 9 fused multiply/add), Chris@10: * 23 stack variables, 4 constants, and 10 memory accesses Chris@10: */ Chris@10: #include "n1f.h" Chris@10: Chris@10: static void n1fv_5(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs) Chris@10: { Chris@10: DVK(KP559016994, +0.559016994374947424102293417182819058860154590); Chris@10: DVK(KP250000000, +0.250000000000000000000000000000000000000000000); Chris@10: DVK(KP618033988, +0.618033988749894848204586834365638117720309180); Chris@10: DVK(KP951056516, +0.951056516295153572116439333379382143405698634); Chris@10: { Chris@10: INT i; Chris@10: const R *xi; Chris@10: R *xo; Chris@10: xi = ri; Chris@10: xo = ro; Chris@10: for (i = v; i > 0; i = i - VL, xi = xi + (VL * ivs), xo = xo + (VL * ovs), MAKE_VOLATILE_STRIDE(10, is), MAKE_VOLATILE_STRIDE(10, os)) { Chris@10: V T1, T2, T3, T5, T6; Chris@10: T1 = LD(&(xi[0]), ivs, &(xi[0])); Chris@10: T2 = LD(&(xi[WS(is, 1)]), ivs, &(xi[WS(is, 1)])); Chris@10: T3 = LD(&(xi[WS(is, 4)]), ivs, &(xi[0])); Chris@10: T5 = LD(&(xi[WS(is, 2)]), ivs, &(xi[0])); Chris@10: T6 = LD(&(xi[WS(is, 3)]), ivs, &(xi[WS(is, 1)])); Chris@10: { Chris@10: V Tc, T4, Td, T7; Chris@10: Tc = VSUB(T2, T3); Chris@10: T4 = VADD(T2, T3); Chris@10: Td = VSUB(T5, T6); Chris@10: T7 = VADD(T5, T6); Chris@10: { Chris@10: V Tg, Te, Ta, T8, T9, Tf, Tb; Chris@10: Tg = VMUL(LDK(KP951056516), VFNMS(LDK(KP618033988), Tc, Td)); Chris@10: Te = VMUL(LDK(KP951056516), VFMA(LDK(KP618033988), Td, Tc)); Chris@10: Ta = VSUB(T4, T7); Chris@10: T8 = VADD(T4, T7); Chris@10: T9 = VFNMS(LDK(KP250000000), T8, T1); Chris@10: ST(&(xo[0]), VADD(T1, T8), ovs, &(xo[0])); Chris@10: Tf = VFNMS(LDK(KP559016994), Ta, T9); Chris@10: Tb = VFMA(LDK(KP559016994), Ta, T9); Chris@10: ST(&(xo[WS(os, 2)]), VFMAI(Tg, Tf), ovs, &(xo[0])); Chris@10: ST(&(xo[WS(os, 3)]), VFNMSI(Tg, Tf), ovs, &(xo[WS(os, 1)])); Chris@10: ST(&(xo[WS(os, 4)]), VFMAI(Te, Tb), ovs, &(xo[0])); Chris@10: ST(&(xo[WS(os, 1)]), VFNMSI(Te, Tb), ovs, &(xo[WS(os, 1)])); Chris@10: } Chris@10: } Chris@10: } Chris@10: } Chris@10: VLEAVE(); Chris@10: } Chris@10: Chris@10: static const kdft_desc desc = { 5, XSIMD_STRING("n1fv_5"), {7, 2, 9, 0}, &GENUS, 0, 0, 0, 0 }; Chris@10: Chris@10: void XSIMD(codelet_n1fv_5) (planner *p) { Chris@10: X(kdft_register) (p, n1fv_5, &desc); Chris@10: } Chris@10: Chris@10: #else /* HAVE_FMA */ Chris@10: Chris@10: /* Generated by: ../../../genfft/gen_notw_c.native -simd -compact -variables 4 -pipeline-latency 8 -n 5 -name n1fv_5 -include n1f.h */ Chris@10: Chris@10: /* Chris@10: * This function contains 16 FP additions, 6 FP multiplications, Chris@10: * (or, 13 additions, 3 multiplications, 3 fused multiply/add), Chris@10: * 18 stack variables, 4 constants, and 10 memory accesses Chris@10: */ Chris@10: #include "n1f.h" Chris@10: Chris@10: static void n1fv_5(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs) Chris@10: { Chris@10: DVK(KP250000000, +0.250000000000000000000000000000000000000000000); Chris@10: DVK(KP587785252, +0.587785252292473129168705954639072768597652438); Chris@10: DVK(KP951056516, +0.951056516295153572116439333379382143405698634); Chris@10: DVK(KP559016994, +0.559016994374947424102293417182819058860154590); Chris@10: { Chris@10: INT i; Chris@10: const R *xi; Chris@10: R *xo; Chris@10: xi = ri; Chris@10: xo = ro; Chris@10: for (i = v; i > 0; i = i - VL, xi = xi + (VL * ivs), xo = xo + (VL * ovs), MAKE_VOLATILE_STRIDE(10, is), MAKE_VOLATILE_STRIDE(10, os)) { Chris@10: V T8, T7, Td, T9, Tc; Chris@10: T8 = LD(&(xi[0]), ivs, &(xi[0])); Chris@10: { Chris@10: V T1, T2, T3, T4, T5, T6; Chris@10: T1 = LD(&(xi[WS(is, 1)]), ivs, &(xi[WS(is, 1)])); Chris@10: T2 = LD(&(xi[WS(is, 4)]), ivs, &(xi[0])); Chris@10: T3 = VADD(T1, T2); Chris@10: T4 = LD(&(xi[WS(is, 2)]), ivs, &(xi[0])); Chris@10: T5 = LD(&(xi[WS(is, 3)]), ivs, &(xi[WS(is, 1)])); Chris@10: T6 = VADD(T4, T5); Chris@10: T7 = VMUL(LDK(KP559016994), VSUB(T3, T6)); Chris@10: Td = VSUB(T4, T5); Chris@10: T9 = VADD(T3, T6); Chris@10: Tc = VSUB(T1, T2); Chris@10: } Chris@10: ST(&(xo[0]), VADD(T8, T9), ovs, &(xo[0])); Chris@10: { Chris@10: V Te, Tf, Tb, Tg, Ta; Chris@10: Te = VBYI(VFMA(LDK(KP951056516), Tc, VMUL(LDK(KP587785252), Td))); Chris@10: Tf = VBYI(VFNMS(LDK(KP587785252), Tc, VMUL(LDK(KP951056516), Td))); Chris@10: Ta = VFNMS(LDK(KP250000000), T9, T8); Chris@10: Tb = VADD(T7, Ta); Chris@10: Tg = VSUB(Ta, T7); Chris@10: ST(&(xo[WS(os, 1)]), VSUB(Tb, Te), ovs, &(xo[WS(os, 1)])); Chris@10: ST(&(xo[WS(os, 3)]), VSUB(Tg, Tf), ovs, &(xo[WS(os, 1)])); Chris@10: ST(&(xo[WS(os, 4)]), VADD(Te, Tb), ovs, &(xo[0])); Chris@10: ST(&(xo[WS(os, 2)]), VADD(Tf, Tg), ovs, &(xo[0])); Chris@10: } Chris@10: } Chris@10: } Chris@10: VLEAVE(); Chris@10: } Chris@10: Chris@10: static const kdft_desc desc = { 5, XSIMD_STRING("n1fv_5"), {13, 3, 3, 0}, &GENUS, 0, 0, 0, 0 }; Chris@10: Chris@10: void XSIMD(codelet_n1fv_5) (planner *p) { Chris@10: X(kdft_register) (p, n1fv_5, &desc); Chris@10: } Chris@10: Chris@10: #endif /* HAVE_FMA */