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 07:37:29 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 -sign 1 -n 6 -name n2bv_6 -with-ostride 2 -include n2b.h -store-multiple 2 */ Chris@10: Chris@10: /* Chris@10: * This function contains 18 FP additions, 8 FP multiplications, Chris@10: * (or, 12 additions, 2 multiplications, 6 fused multiply/add), Chris@10: * 29 stack variables, 2 constants, and 15 memory accesses Chris@10: */ Chris@10: #include "n2b.h" Chris@10: Chris@10: static void n2bv_6(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(KP500000000, +0.500000000000000000000000000000000000000000000); Chris@10: DVK(KP866025403, +0.866025403784438646763723170752936183471402627); Chris@10: { Chris@10: INT i; Chris@10: const R *xi; Chris@10: R *xo; Chris@10: xi = ii; Chris@10: xo = io; Chris@10: 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)) { Chris@10: V T1, T2, T4, T5, T7, T8; Chris@10: T1 = LD(&(xi[0]), ivs, &(xi[0])); Chris@10: T2 = LD(&(xi[WS(is, 3)]), ivs, &(xi[WS(is, 1)])); Chris@10: T4 = LD(&(xi[WS(is, 2)]), ivs, &(xi[0])); Chris@10: T5 = LD(&(xi[WS(is, 5)]), ivs, &(xi[WS(is, 1)])); Chris@10: T7 = LD(&(xi[WS(is, 4)]), ivs, &(xi[0])); Chris@10: T8 = LD(&(xi[WS(is, 1)]), ivs, &(xi[WS(is, 1)])); Chris@10: { Chris@10: V T3, Td, T6, Te, T9, Tf; Chris@10: T3 = VSUB(T1, T2); Chris@10: Td = VADD(T1, T2); Chris@10: T6 = VSUB(T4, T5); Chris@10: Te = VADD(T4, T5); Chris@10: T9 = VSUB(T7, T8); Chris@10: Tf = VADD(T7, T8); Chris@10: { Chris@10: V Tg, Ti, Ta, Tc; Chris@10: Tg = VADD(Te, Tf); Chris@10: Ti = VMUL(LDK(KP866025403), VSUB(Te, Tf)); Chris@10: Ta = VADD(T6, T9); Chris@10: Tc = VMUL(LDK(KP866025403), VSUB(T6, T9)); Chris@10: { Chris@10: V Th, Tj, Tb, Tk; Chris@10: Th = VFNMS(LDK(KP500000000), Tg, Td); Chris@10: Tj = VADD(Td, Tg); Chris@10: STM2(&(xo[0]), Tj, ovs, &(xo[0])); Chris@10: Tb = VFNMS(LDK(KP500000000), Ta, T3); Chris@10: Tk = VADD(T3, Ta); Chris@10: STM2(&(xo[6]), Tk, ovs, &(xo[2])); Chris@10: { Chris@10: V Tl, Tm, Tn, To; Chris@10: Tl = VFMAI(Ti, Th); Chris@10: STM2(&(xo[8]), Tl, ovs, &(xo[0])); Chris@10: Tm = VFNMSI(Ti, Th); Chris@10: STM2(&(xo[4]), Tm, ovs, &(xo[0])); Chris@10: STN2(&(xo[4]), Tm, Tk, ovs); Chris@10: Tn = VFNMSI(Tc, Tb); Chris@10: STM2(&(xo[10]), Tn, ovs, &(xo[2])); Chris@10: STN2(&(xo[8]), Tl, Tn, ovs); Chris@10: To = VFMAI(Tc, Tb); Chris@10: STM2(&(xo[2]), To, ovs, &(xo[2])); Chris@10: STN2(&(xo[0]), Tj, To, ovs); Chris@10: } Chris@10: } Chris@10: } Chris@10: } Chris@10: } Chris@10: } Chris@10: VLEAVE(); Chris@10: } Chris@10: Chris@10: static const kdft_desc desc = { 6, XSIMD_STRING("n2bv_6"), {12, 2, 6, 0}, &GENUS, 0, 2, 0, 0 }; Chris@10: Chris@10: void XSIMD(codelet_n2bv_6) (planner *p) { Chris@10: X(kdft_register) (p, n2bv_6, &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 -sign 1 -n 6 -name n2bv_6 -with-ostride 2 -include n2b.h -store-multiple 2 */ Chris@10: Chris@10: /* Chris@10: * This function contains 18 FP additions, 4 FP multiplications, Chris@10: * (or, 16 additions, 2 multiplications, 2 fused multiply/add), Chris@10: * 25 stack variables, 2 constants, and 15 memory accesses Chris@10: */ Chris@10: #include "n2b.h" Chris@10: Chris@10: static void n2bv_6(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(KP500000000, +0.500000000000000000000000000000000000000000000); Chris@10: DVK(KP866025403, +0.866025403784438646763723170752936183471402627); Chris@10: { Chris@10: INT i; Chris@10: const R *xi; Chris@10: R *xo; Chris@10: xi = ii; Chris@10: xo = io; Chris@10: 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)) { Chris@10: V Ta, Td, T3, Te, T6, Tf, Tb, Tg, T8, T9, Tj, Tk; Chris@10: T8 = LD(&(xi[0]), ivs, &(xi[0])); Chris@10: T9 = LD(&(xi[WS(is, 3)]), ivs, &(xi[WS(is, 1)])); Chris@10: Ta = VSUB(T8, T9); Chris@10: Td = VADD(T8, T9); Chris@10: { Chris@10: V T1, T2, T4, T5; Chris@10: T1 = LD(&(xi[WS(is, 2)]), ivs, &(xi[0])); Chris@10: T2 = LD(&(xi[WS(is, 5)]), ivs, &(xi[WS(is, 1)])); Chris@10: T3 = VSUB(T1, T2); Chris@10: Te = VADD(T1, T2); Chris@10: T4 = LD(&(xi[WS(is, 4)]), ivs, &(xi[0])); Chris@10: T5 = LD(&(xi[WS(is, 1)]), ivs, &(xi[WS(is, 1)])); Chris@10: T6 = VSUB(T4, T5); Chris@10: Tf = VADD(T4, T5); Chris@10: } Chris@10: Tb = VADD(T3, T6); Chris@10: Tg = VADD(Te, Tf); Chris@10: Tj = VADD(Ta, Tb); Chris@10: STM2(&(xo[6]), Tj, ovs, &(xo[2])); Chris@10: Tk = VADD(Td, Tg); Chris@10: STM2(&(xo[0]), Tk, ovs, &(xo[0])); Chris@10: { Chris@10: V Tm, T7, Tc, Tl; Chris@10: T7 = VBYI(VMUL(LDK(KP866025403), VSUB(T3, T6))); Chris@10: Tc = VFNMS(LDK(KP500000000), Tb, Ta); Chris@10: Tl = VADD(T7, Tc); Chris@10: STM2(&(xo[2]), Tl, ovs, &(xo[2])); Chris@10: STN2(&(xo[0]), Tk, Tl, ovs); Chris@10: Tm = VSUB(Tc, T7); Chris@10: STM2(&(xo[10]), Tm, ovs, &(xo[2])); Chris@10: { Chris@10: V Th, Ti, Tn, To; Chris@10: Th = VFNMS(LDK(KP500000000), Tg, Td); Chris@10: Ti = VBYI(VMUL(LDK(KP866025403), VSUB(Te, Tf))); Chris@10: Tn = VSUB(Th, Ti); Chris@10: STM2(&(xo[4]), Tn, ovs, &(xo[0])); Chris@10: STN2(&(xo[4]), Tn, Tj, ovs); Chris@10: To = VADD(Ti, Th); Chris@10: STM2(&(xo[8]), To, ovs, &(xo[0])); Chris@10: STN2(&(xo[8]), To, Tm, ovs); Chris@10: } Chris@10: } Chris@10: } Chris@10: } Chris@10: VLEAVE(); Chris@10: } Chris@10: Chris@10: static const kdft_desc desc = { 6, XSIMD_STRING("n2bv_6"), {16, 2, 2, 0}, &GENUS, 0, 2, 0, 0 }; Chris@10: Chris@10: void XSIMD(codelet_n2bv_6) (planner *p) { Chris@10: X(kdft_register) (p, n2bv_6, &desc); Chris@10: } Chris@10: Chris@10: #endif /* HAVE_FMA */