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 07:39:03 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_twiddle_c.native -fma -reorder-insns -schedule-for-pipeline -simd -compact -variables 4 -pipeline-latency 8 -n 6 -name t1bv_6 -include t1b.h -sign 1 */ cannam@95: cannam@95: /* cannam@95: * This function contains 23 FP additions, 18 FP multiplications, cannam@95: * (or, 17 additions, 12 multiplications, 6 fused multiply/add), cannam@95: * 27 stack variables, 2 constants, and 12 memory accesses cannam@95: */ cannam@95: #include "t1b.h" cannam@95: cannam@95: static void t1bv_6(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms) cannam@95: { cannam@95: DVK(KP500000000, +0.500000000000000000000000000000000000000000000); cannam@95: DVK(KP866025403, +0.866025403784438646763723170752936183471402627); cannam@95: { cannam@95: INT m; cannam@95: R *x; cannam@95: x = ii; cannam@95: for (m = mb, W = W + (mb * ((TWVL / VL) * 10)); m < me; m = m + VL, x = x + (VL * ms), W = W + (TWVL * 10), MAKE_VOLATILE_STRIDE(6, rs)) { cannam@95: V T1, T2, Ta, Tc, T5, T7; cannam@95: T1 = LD(&(x[0]), ms, &(x[0])); cannam@95: T2 = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)])); cannam@95: Ta = LD(&(x[WS(rs, 4)]), ms, &(x[0])); cannam@95: Tc = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)])); cannam@95: T5 = LD(&(x[WS(rs, 2)]), ms, &(x[0])); cannam@95: T7 = LD(&(x[WS(rs, 5)]), ms, &(x[WS(rs, 1)])); cannam@95: { cannam@95: V T3, Tb, Td, T6, T8; cannam@95: T3 = BYTW(&(W[TWVL * 4]), T2); cannam@95: Tb = BYTW(&(W[TWVL * 6]), Ta); cannam@95: Td = BYTW(&(W[0]), Tc); cannam@95: T6 = BYTW(&(W[TWVL * 2]), T5); cannam@95: T8 = BYTW(&(W[TWVL * 8]), T7); cannam@95: { cannam@95: V Ti, T4, Tk, Te, Tj, T9; cannam@95: Ti = VADD(T1, T3); cannam@95: T4 = VSUB(T1, T3); cannam@95: Tk = VADD(Tb, Td); cannam@95: Te = VSUB(Tb, Td); cannam@95: Tj = VADD(T6, T8); cannam@95: T9 = VSUB(T6, T8); cannam@95: { cannam@95: V Tl, Tn, Tf, Th, Tm, Tg; cannam@95: Tl = VADD(Tj, Tk); cannam@95: Tn = VMUL(LDK(KP866025403), VSUB(Tj, Tk)); cannam@95: Tf = VADD(T9, Te); cannam@95: Th = VMUL(LDK(KP866025403), VSUB(T9, Te)); cannam@95: ST(&(x[0]), VADD(Ti, Tl), ms, &(x[0])); cannam@95: Tm = VFNMS(LDK(KP500000000), Tl, Ti); cannam@95: ST(&(x[WS(rs, 3)]), VADD(T4, Tf), ms, &(x[WS(rs, 1)])); cannam@95: Tg = VFNMS(LDK(KP500000000), Tf, T4); cannam@95: ST(&(x[WS(rs, 4)]), VFMAI(Tn, Tm), ms, &(x[0])); cannam@95: ST(&(x[WS(rs, 2)]), VFNMSI(Tn, Tm), ms, &(x[0])); cannam@95: ST(&(x[WS(rs, 5)]), VFNMSI(Th, Tg), ms, &(x[WS(rs, 1)])); cannam@95: ST(&(x[WS(rs, 1)]), VFMAI(Th, Tg), ms, &(x[WS(rs, 1)])); cannam@95: } cannam@95: } cannam@95: } cannam@95: } cannam@95: } cannam@95: VLEAVE(); cannam@95: } cannam@95: cannam@95: static const tw_instr twinstr[] = { cannam@95: VTW(0, 1), cannam@95: VTW(0, 2), cannam@95: VTW(0, 3), cannam@95: VTW(0, 4), cannam@95: VTW(0, 5), cannam@95: {TW_NEXT, VL, 0} cannam@95: }; cannam@95: cannam@95: static const ct_desc desc = { 6, XSIMD_STRING("t1bv_6"), twinstr, &GENUS, {17, 12, 6, 0}, 0, 0, 0 }; cannam@95: cannam@95: void XSIMD(codelet_t1bv_6) (planner *p) { cannam@95: X(kdft_dit_register) (p, t1bv_6, &desc); cannam@95: } cannam@95: #else /* HAVE_FMA */ cannam@95: cannam@95: /* Generated by: ../../../genfft/gen_twiddle_c.native -simd -compact -variables 4 -pipeline-latency 8 -n 6 -name t1bv_6 -include t1b.h -sign 1 */ cannam@95: cannam@95: /* cannam@95: * This function contains 23 FP additions, 14 FP multiplications, cannam@95: * (or, 21 additions, 12 multiplications, 2 fused multiply/add), cannam@95: * 19 stack variables, 2 constants, and 12 memory accesses cannam@95: */ cannam@95: #include "t1b.h" cannam@95: cannam@95: static void t1bv_6(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms) cannam@95: { cannam@95: DVK(KP500000000, +0.500000000000000000000000000000000000000000000); cannam@95: DVK(KP866025403, +0.866025403784438646763723170752936183471402627); cannam@95: { cannam@95: INT m; cannam@95: R *x; cannam@95: x = ii; cannam@95: for (m = mb, W = W + (mb * ((TWVL / VL) * 10)); m < me; m = m + VL, x = x + (VL * ms), W = W + (TWVL * 10), MAKE_VOLATILE_STRIDE(6, rs)) { cannam@95: V Tf, Ti, Ta, Tk, T5, Tj, Tc, Te, Td; cannam@95: Tc = LD(&(x[0]), ms, &(x[0])); cannam@95: Td = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)])); cannam@95: Te = BYTW(&(W[TWVL * 4]), Td); cannam@95: Tf = VSUB(Tc, Te); cannam@95: Ti = VADD(Tc, Te); cannam@95: { cannam@95: V T7, T9, T6, T8; cannam@95: T6 = LD(&(x[WS(rs, 4)]), ms, &(x[0])); cannam@95: T7 = BYTW(&(W[TWVL * 6]), T6); cannam@95: T8 = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)])); cannam@95: T9 = BYTW(&(W[0]), T8); cannam@95: Ta = VSUB(T7, T9); cannam@95: Tk = VADD(T7, T9); cannam@95: } cannam@95: { cannam@95: V T2, T4, T1, T3; cannam@95: T1 = LD(&(x[WS(rs, 2)]), ms, &(x[0])); cannam@95: T2 = BYTW(&(W[TWVL * 2]), T1); cannam@95: T3 = LD(&(x[WS(rs, 5)]), ms, &(x[WS(rs, 1)])); cannam@95: T4 = BYTW(&(W[TWVL * 8]), T3); cannam@95: T5 = VSUB(T2, T4); cannam@95: Tj = VADD(T2, T4); cannam@95: } cannam@95: { cannam@95: V Tb, Tg, Th, Tn, Tl, Tm; cannam@95: Tb = VBYI(VMUL(LDK(KP866025403), VSUB(T5, Ta))); cannam@95: Tg = VADD(T5, Ta); cannam@95: Th = VFNMS(LDK(KP500000000), Tg, Tf); cannam@95: ST(&(x[WS(rs, 1)]), VADD(Tb, Th), ms, &(x[WS(rs, 1)])); cannam@95: ST(&(x[WS(rs, 3)]), VADD(Tf, Tg), ms, &(x[WS(rs, 1)])); cannam@95: ST(&(x[WS(rs, 5)]), VSUB(Th, Tb), ms, &(x[WS(rs, 1)])); cannam@95: Tn = VBYI(VMUL(LDK(KP866025403), VSUB(Tj, Tk))); cannam@95: Tl = VADD(Tj, Tk); cannam@95: Tm = VFNMS(LDK(KP500000000), Tl, Ti); cannam@95: ST(&(x[WS(rs, 2)]), VSUB(Tm, Tn), ms, &(x[0])); cannam@95: ST(&(x[0]), VADD(Ti, Tl), ms, &(x[0])); cannam@95: ST(&(x[WS(rs, 4)]), VADD(Tn, Tm), ms, &(x[0])); cannam@95: } cannam@95: } cannam@95: } cannam@95: VLEAVE(); cannam@95: } cannam@95: cannam@95: static const tw_instr twinstr[] = { cannam@95: VTW(0, 1), cannam@95: VTW(0, 2), cannam@95: VTW(0, 3), cannam@95: VTW(0, 4), cannam@95: VTW(0, 5), cannam@95: {TW_NEXT, VL, 0} cannam@95: }; cannam@95: cannam@95: static const ct_desc desc = { 6, XSIMD_STRING("t1bv_6"), twinstr, &GENUS, {21, 12, 2, 0}, 0, 0, 0 }; cannam@95: cannam@95: void XSIMD(codelet_t1bv_6) (planner *p) { cannam@95: X(kdft_dit_register) (p, t1bv_6, &desc); cannam@95: } cannam@95: #endif /* HAVE_FMA */