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 07:38:54 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_twiddle_c.native -fma -reorder-insns -schedule-for-pipeline -simd -compact -variables 4 -pipeline-latency 8 -twiddle-log3 -precompute-twiddles -no-generate-bytw -n 5 -name t3fv_5 -include t3f.h */ Chris@10: Chris@10: /* Chris@10: * This function contains 22 FP additions, 23 FP multiplications, Chris@10: * (or, 13 additions, 14 multiplications, 9 fused multiply/add), Chris@10: * 30 stack variables, 4 constants, and 10 memory accesses Chris@10: */ Chris@10: #include "t3f.h" Chris@10: Chris@10: static void t3fv_5(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms) 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 m; Chris@10: R *x; Chris@10: x = ri; Chris@10: for (m = mb, W = W + (mb * ((TWVL / VL) * 4)); m < me; m = m + VL, x = x + (VL * ms), W = W + (TWVL * 4), MAKE_VOLATILE_STRIDE(5, rs)) { Chris@10: V T2, T5, T1, T3, Td, T7, Tb; Chris@10: T2 = LDW(&(W[0])); Chris@10: T5 = LDW(&(W[TWVL * 2])); Chris@10: T1 = LD(&(x[0]), ms, &(x[0])); Chris@10: T3 = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)])); Chris@10: Td = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)])); Chris@10: T7 = LD(&(x[WS(rs, 4)]), ms, &(x[0])); Chris@10: Tb = LD(&(x[WS(rs, 2)]), ms, &(x[0])); Chris@10: { Chris@10: V Ta, T6, T4, Te, Tc, T8; Chris@10: Ta = VZMULJ(T2, T5); Chris@10: T6 = VZMUL(T2, T5); Chris@10: T4 = VZMULJ(T2, T3); Chris@10: Te = VZMULJ(T5, Td); Chris@10: Tc = VZMULJ(Ta, Tb); Chris@10: T8 = VZMULJ(T6, T7); Chris@10: { Chris@10: V Tf, Tl, T9, Tk; Chris@10: Tf = VADD(Tc, Te); Chris@10: Tl = VSUB(Tc, Te); Chris@10: T9 = VADD(T4, T8); Chris@10: Tk = VSUB(T4, T8); Chris@10: { Chris@10: V Ti, Tg, To, Tm, Th, Tn, Tj; Chris@10: Ti = VSUB(T9, Tf); Chris@10: Tg = VADD(T9, Tf); Chris@10: To = VMUL(LDK(KP951056516), VFNMS(LDK(KP618033988), Tk, Tl)); Chris@10: Tm = VMUL(LDK(KP951056516), VFMA(LDK(KP618033988), Tl, Tk)); Chris@10: Th = VFNMS(LDK(KP250000000), Tg, T1); Chris@10: ST(&(x[0]), VADD(T1, Tg), ms, &(x[0])); Chris@10: Tn = VFNMS(LDK(KP559016994), Ti, Th); Chris@10: Tj = VFMA(LDK(KP559016994), Ti, Th); Chris@10: ST(&(x[WS(rs, 2)]), VFMAI(To, Tn), ms, &(x[0])); Chris@10: ST(&(x[WS(rs, 3)]), VFNMSI(To, Tn), ms, &(x[WS(rs, 1)])); Chris@10: ST(&(x[WS(rs, 4)]), VFMAI(Tm, Tj), ms, &(x[0])); Chris@10: ST(&(x[WS(rs, 1)]), VFNMSI(Tm, Tj), ms, &(x[WS(rs, 1)])); Chris@10: } Chris@10: } Chris@10: } Chris@10: } Chris@10: } Chris@10: VLEAVE(); Chris@10: } Chris@10: Chris@10: static const tw_instr twinstr[] = { Chris@10: VTW(0, 1), Chris@10: VTW(0, 3), Chris@10: {TW_NEXT, VL, 0} Chris@10: }; Chris@10: Chris@10: static const ct_desc desc = { 5, XSIMD_STRING("t3fv_5"), twinstr, &GENUS, {13, 14, 9, 0}, 0, 0, 0 }; Chris@10: Chris@10: void XSIMD(codelet_t3fv_5) (planner *p) { Chris@10: X(kdft_dit_register) (p, t3fv_5, &desc); Chris@10: } Chris@10: #else /* HAVE_FMA */ Chris@10: Chris@10: /* Generated by: ../../../genfft/gen_twiddle_c.native -simd -compact -variables 4 -pipeline-latency 8 -twiddle-log3 -precompute-twiddles -no-generate-bytw -n 5 -name t3fv_5 -include t3f.h */ Chris@10: Chris@10: /* Chris@10: * This function contains 22 FP additions, 18 FP multiplications, Chris@10: * (or, 19 additions, 15 multiplications, 3 fused multiply/add), Chris@10: * 24 stack variables, 4 constants, and 10 memory accesses Chris@10: */ Chris@10: #include "t3f.h" Chris@10: Chris@10: static void t3fv_5(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms) Chris@10: { Chris@10: DVK(KP250000000, +0.250000000000000000000000000000000000000000000); Chris@10: DVK(KP559016994, +0.559016994374947424102293417182819058860154590); Chris@10: DVK(KP587785252, +0.587785252292473129168705954639072768597652438); Chris@10: DVK(KP951056516, +0.951056516295153572116439333379382143405698634); Chris@10: { Chris@10: INT m; Chris@10: R *x; Chris@10: x = ri; Chris@10: for (m = mb, W = W + (mb * ((TWVL / VL) * 4)); m < me; m = m + VL, x = x + (VL * ms), W = W + (TWVL * 4), MAKE_VOLATILE_STRIDE(5, rs)) { Chris@10: V T1, T4, T5, T9; Chris@10: T1 = LDW(&(W[0])); Chris@10: T4 = LDW(&(W[TWVL * 2])); Chris@10: T5 = VZMUL(T1, T4); Chris@10: T9 = VZMULJ(T1, T4); Chris@10: { Chris@10: V Tg, Tk, Tl, T8, Te, Th; Chris@10: Tg = LD(&(x[0]), ms, &(x[0])); Chris@10: { Chris@10: V T3, Td, T7, Tb; Chris@10: { Chris@10: V T2, Tc, T6, Ta; Chris@10: T2 = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)])); Chris@10: T3 = VZMULJ(T1, T2); Chris@10: Tc = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)])); Chris@10: Td = VZMULJ(T4, Tc); Chris@10: T6 = LD(&(x[WS(rs, 4)]), ms, &(x[0])); Chris@10: T7 = VZMULJ(T5, T6); Chris@10: Ta = LD(&(x[WS(rs, 2)]), ms, &(x[0])); Chris@10: Tb = VZMULJ(T9, Ta); Chris@10: } Chris@10: Tk = VSUB(T3, T7); Chris@10: Tl = VSUB(Tb, Td); Chris@10: T8 = VADD(T3, T7); Chris@10: Te = VADD(Tb, Td); Chris@10: Th = VADD(T8, Te); Chris@10: } Chris@10: ST(&(x[0]), VADD(Tg, Th), ms, &(x[0])); Chris@10: { Chris@10: V Tm, Tn, Tj, To, Tf, Ti; Chris@10: Tm = VBYI(VFMA(LDK(KP951056516), Tk, VMUL(LDK(KP587785252), Tl))); Chris@10: Tn = VBYI(VFNMS(LDK(KP587785252), Tk, VMUL(LDK(KP951056516), Tl))); Chris@10: Tf = VMUL(LDK(KP559016994), VSUB(T8, Te)); Chris@10: Ti = VFNMS(LDK(KP250000000), Th, Tg); Chris@10: Tj = VADD(Tf, Ti); Chris@10: To = VSUB(Ti, Tf); Chris@10: ST(&(x[WS(rs, 1)]), VSUB(Tj, Tm), ms, &(x[WS(rs, 1)])); Chris@10: ST(&(x[WS(rs, 3)]), VSUB(To, Tn), ms, &(x[WS(rs, 1)])); Chris@10: ST(&(x[WS(rs, 4)]), VADD(Tm, Tj), ms, &(x[0])); Chris@10: ST(&(x[WS(rs, 2)]), VADD(Tn, To), ms, &(x[0])); Chris@10: } Chris@10: } Chris@10: } Chris@10: } Chris@10: VLEAVE(); Chris@10: } Chris@10: Chris@10: static const tw_instr twinstr[] = { Chris@10: VTW(0, 1), Chris@10: VTW(0, 3), Chris@10: {TW_NEXT, VL, 0} Chris@10: }; Chris@10: Chris@10: static const ct_desc desc = { 5, XSIMD_STRING("t3fv_5"), twinstr, &GENUS, {19, 15, 3, 0}, 0, 0, 0 }; Chris@10: Chris@10: void XSIMD(codelet_t3fv_5) (planner *p) { Chris@10: X(kdft_dit_register) (p, t3fv_5, &desc); Chris@10: } Chris@10: #endif /* HAVE_FMA */