~ubuntu-branches/ubuntu/utopic/fftw3/utopic

/* Generated by: ../../../genfft/gen_hc2cdft -fma -reorder-insns -schedule-for-pipeline -compact -variables 4 -pipeline-latency 4 -twiddle-log3 -precompute-twiddles -n 4 -dit -name hc2cfdft2_4 -include hc2cf.h */

/* Generated by: ../../../genfft/gen_hc2cdft.native -fma -reorder-insns -schedule-for-pipeline -compact -variables 4 -pipeline-latency 4 -twiddle-log3 -precompute-twiddles -n 4 -dit -name hc2cfdft2_4 -include hc2cf.h */

* This function contains 32 FP additions, 24 FP multiplications,

static void hc2cfdft2_4(R *Rp, R *Ip, R *Rm, R *Im, const R *W, stride rs, INT mb, INT me, INT ms)

{

DK(KP500000000, +0.500000000000000000000000000000000000000000000);

INT m;

for (m = mb, W = W + ((mb - 1) * 4); m < me; m = m + 1, Rp = Rp + ms, Ip = Ip + ms, Rm = Rm - ms, Im = Im - ms, W = W + 4, MAKE_VOLATILE_STRIDE(rs)) {

E T1, T5, T2, T4;

T1 = W[0];

T5 = W[3];

T2 = W[2];

T4 = W[1];

{

E Tc, T6, Tp, Tj, Tw, Tt, T9, TE, To, TC, Ta, Tr, Tf, Tl, Tm;

{

E Th, Tb, T3, Ti;

Th = Ip[0];

Tb = T1 * T5;

T3 = T1 * T2;

Ti = Im[0];

Tl = Rm[0];

Tc = FNMS(T4, T2, Tb);

T6 = FMA(T4, T5, T3);

Tp = Th + Ti;

Tj = Th - Ti;

Tm = Rp[0];

}

{

E T7, T8, Td, Tn, Te;

T7 = Ip[WS(rs, 1)];

T8 = Im[WS(rs, 1)];

Td = Rp[WS(rs, 1)];

Tw = Tm + Tl;

Tn = Tl - Tm;

Tt = T7 + T8;

T9 = T7 - T8;

Te = Rm[WS(rs, 1)];

TE = T4 * Tn;

To = T1 * Tn;

TC = T2 * Tt;

Ta = T6 * T9;

Tr = Td - Te;

Tf = Td + Te;

}

{

E Tq, Tk, TB, Ty, Tu, TI, TG, TF;

Tq = FNMS(T4, Tp, To);

TF = FMA(T1, Tp, TE);

{

E Tg, Tx, TD, Ts;

Tg = FNMS(Tc, Tf, Ta);

Tx = T6 * Tf;

TD = FNMS(T5, Tr, TC);

Ts = T2 * Tr;

Tk = Tg + Tj;

TB = Tj - Tg;

Ty = FMA(Tc, T9, Tx);

Tu = FMA(T5, Tt, Ts);

TI = TD + TF;

TG = TD - TF;

}

{

E Tz, TH, Tv, TA;

Tz = Tw - Ty;

TH = Tw + Ty;

100

Tv = Tq - Tu;

101

TA = Tu + Tq;

102

Rp[0] = KP500000000 * (TH + TI);

103

Rm[WS(rs, 1)] = KP500000000 * (TH - TI);

104

Rm[0] = KP500000000 * (Tz - TA);

105

Im[WS(rs, 1)] = KP500000000 * (Tv - Tk);

106

Ip[0] = KP500000000 * (Tk + Tv);

107

Im[0] = KP500000000 * (TG - TB);

108

Rp[WS(rs, 1)] = KP500000000 * (Tz + TA);

109

Ip[WS(rs, 1)] = KP500000000 * (TB + TG);

{

INT m;

for (m = mb, W = W + ((mb - 1) * 4); m < me; m = m + 1, Rp = Rp + ms, Ip = Ip + ms, Rm = Rm - ms, Im = Im - ms, W = W + 4, MAKE_VOLATILE_STRIDE(rs)) {

E T1, T5, T2, T4;

T1 = W[0];

T5 = W[3];

T2 = W[2];

T4 = W[1];

{

E Tc, T6, Tp, Tj, Tw, Tt, T9, TE, To, TC, Ta, Tr, Tf, Tl, Tm;

{

E Th, Tb, T3, Ti;

Th = Ip[0];

Tb = T1 * T5;

T3 = T1 * T2;

Ti = Im[0];

Tl = Rm[0];

Tc = FNMS(T4, T2, Tb);

T6 = FMA(T4, T5, T3);

Tp = Th + Ti;

Tj = Th - Ti;

Tm = Rp[0];

}

{

E T7, T8, Td, Tn, Te;

T7 = Ip[WS(rs, 1)];

T8 = Im[WS(rs, 1)];

Td = Rp[WS(rs, 1)];

Tw = Tm + Tl;

Tn = Tl - Tm;

Tt = T7 + T8;

T9 = T7 - T8;

Te = Rm[WS(rs, 1)];

TE = T4 * Tn;

To = T1 * Tn;

TC = T2 * Tt;

Ta = T6 * T9;

Tr = Td - Te;

Tf = Td + Te;

}

{

E Tq, Tk, TB, Ty, Tu, TI, TG, TF;

Tq = FNMS(T4, Tp, To);

TF = FMA(T1, Tp, TE);

{

E Tg, Tx, TD, Ts;

Tg = FNMS(Tc, Tf, Ta);

Tx = T6 * Tf;

TD = FNMS(T5, Tr, TC);

Ts = T2 * Tr;

Tk = Tg + Tj;

TB = Tj - Tg;

Ty = FMA(Tc, T9, Tx);

Tu = FMA(T5, Tt, Ts);

TI = TD + TF;

TG = TD - TF;

}

{

E Tz, TH, Tv, TA;

Tz = Tw - Ty;

100

TH = Tw + Ty;

101

Tv = Tq - Tu;

102

TA = Tu + Tq;

103

Rp[0] = KP500000000 * (TH + TI);

104

Rm[WS(rs, 1)] = KP500000000 * (TH - TI);

105

Rm[0] = KP500000000 * (Tz - TA);

106

Im[WS(rs, 1)] = KP500000000 * (Tv - Tk);

107

Ip[0] = KP500000000 * (Tk + Tv);

108

Im[0] = KP500000000 * (TG - TB);

109

Rp[WS(rs, 1)] = KP500000000 * (Tz + TA);

110

Ip[WS(rs, 1)] = KP500000000 * (TB + TG);

111

}

110

112

}

111

113

}

112

114

}

126

128

}

127

129

#else /* HAVE_FMA */

128

130

129

/* Generated by: ../../../genfft/gen_hc2cdft -compact -variables 4 -pipeline-latency 4 -twiddle-log3 -precompute-twiddles -n 4 -dit -name hc2cfdft2_4 -include hc2cf.h */

131

/* Generated by: ../../../genfft/gen_hc2cdft.native -compact -variables 4 -pipeline-latency 4 -twiddle-log3 -precompute-twiddles -n 4 -dit -name hc2cfdft2_4 -include hc2cf.h */

130

132

131

133

132

134

* This function contains 32 FP additions, 24 FP multiplications,

138

140

static void hc2cfdft2_4(R *Rp, R *Ip, R *Rm, R *Im, const R *W, stride rs, INT mb, INT me, INT ms)

139

141

{

140

142

DK(KP500000000, +0.500000000000000000000000000000000000000000000);

141

INT m;

142

for (m = mb, W = W + ((mb - 1) * 4); m < me; m = m + 1, Rp = Rp + ms, Ip = Ip + ms, Rm = Rm - ms, Im = Im - ms, W = W + 4, MAKE_VOLATILE_STRIDE(rs)) {

143

E T1, T3, T2, T4, T5, T9;

144

T1 = W[0];

145

T3 = W[1];

146

T2 = W[2];

147

T4 = W[3];

148

T5 = FMA(T1, T2, T3 * T4);

149

T9 = FNMS(T3, T2, T1 * T4);

150

{

151

E Tg, Tr, Tm, Tx, Td, Tw, Tp, Ts;

152

{

153

E Te, Tf, Tl, Ti, Tj, Tk;

154

Te = Ip[0];

155

Tf = Im[0];

156

Tl = Te + Tf;

157

Ti = Rm[0];

158

Tj = Rp[0];

159

Tk = Ti - Tj;

160

Tg = Te - Tf;

161

Tr = Tj + Ti;

162

Tm = FNMS(T3, Tl, T1 * Tk);

163

Tx = FMA(T3, Tk, T1 * Tl);

164

}

165

{

166

E T8, To, Tc, Tn;

167

{

168

E T6, T7, Ta, Tb;

169

T6 = Ip[WS(rs, 1)];

170

T7 = Im[WS(rs, 1)];

171

T8 = T6 - T7;

172

To = T6 + T7;

173

Ta = Rp[WS(rs, 1)];

174

Tb = Rm[WS(rs, 1)];

175

Tc = Ta + Tb;

176

Tn = Ta - Tb;

177

}

178

Td = FNMS(T9, Tc, T5 * T8);

179

Tw = FNMS(T4, Tn, T2 * To);

180

Tp = FMA(T2, Tn, T4 * To);

181

Ts = FMA(T5, Tc, T9 * T8);

182

}

183

{

184

E Th, Tq, Tz, TA;

185

Th = Td + Tg;

186

Tq = Tm - Tp;

187

Ip[0] = KP500000000 * (Th + Tq);

188

Im[WS(rs, 1)] = KP500000000 * (Tq - Th);

189

Tz = Tr + Ts;

190

TA = Tw + Tx;

191

Rm[WS(rs, 1)] = KP500000000 * (Tz - TA);

192

Rp[0] = KP500000000 * (Tz + TA);

193

}

194

{

195

E Tt, Tu, Tv, Ty;

196

Tt = Tr - Ts;

197

Tu = Tp + Tm;

198

Rm[0] = KP500000000 * (Tt - Tu);

199

Rp[WS(rs, 1)] = KP500000000 * (Tt + Tu);

200

Tv = Tg - Td;

201

Ty = Tw - Tx;

202

Ip[WS(rs, 1)] = KP500000000 * (Tv + Ty);

203

Im[0] = KP500000000 * (Ty - Tv);

143

{

144

INT m;

145

for (m = mb, W = W + ((mb - 1) * 4); m < me; m = m + 1, Rp = Rp + ms, Ip = Ip + ms, Rm = Rm - ms, Im = Im - ms, W = W + 4, MAKE_VOLATILE_STRIDE(rs)) {

146

E T1, T3, T2, T4, T5, T9;

147

T1 = W[0];

148

T3 = W[1];

149

T2 = W[2];

150

T4 = W[3];

151

T5 = FMA(T1, T2, T3 * T4);

152

T9 = FNMS(T3, T2, T1 * T4);

153

{

154

E Tg, Tr, Tm, Tx, Td, Tw, Tp, Ts;

155

{

156

E Te, Tf, Tl, Ti, Tj, Tk;

157

Te = Ip[0];

158

Tf = Im[0];

159

Tl = Te + Tf;

160

Ti = Rm[0];

161

Tj = Rp[0];

162

Tk = Ti - Tj;

163

Tg = Te - Tf;

164

Tr = Tj + Ti;

165

Tm = FNMS(T3, Tl, T1 * Tk);

166

Tx = FMA(T3, Tk, T1 * Tl);

167

}

168

{

169

E T8, To, Tc, Tn;

170

{

171

E T6, T7, Ta, Tb;

172

T6 = Ip[WS(rs, 1)];

173

T7 = Im[WS(rs, 1)];

174

T8 = T6 - T7;

175

To = T6 + T7;

176

Ta = Rp[WS(rs, 1)];

177

Tb = Rm[WS(rs, 1)];

178

Tc = Ta + Tb;

179

Tn = Ta - Tb;

180

}

181

Td = FNMS(T9, Tc, T5 * T8);

182

Tw = FNMS(T4, Tn, T2 * To);

183

Tp = FMA(T2, Tn, T4 * To);

184

Ts = FMA(T5, Tc, T9 * T8);

185

}

186

{

187

E Th, Tq, Tz, TA;

188

Th = Td + Tg;

189

Tq = Tm - Tp;

190

Ip[0] = KP500000000 * (Th + Tq);

191

Im[WS(rs, 1)] = KP500000000 * (Tq - Th);

192

Tz = Tr + Ts;

193

TA = Tw + Tx;

194

Rm[WS(rs, 1)] = KP500000000 * (Tz - TA);

195

Rp[0] = KP500000000 * (Tz + TA);

196

}

197

{

198

E Tt, Tu, Tv, Ty;

199

Tt = Tr - Ts;

200

Tu = Tp + Tm;

201

Rm[0] = KP500000000 * (Tt - Tu);

202

Rp[WS(rs, 1)] = KP500000000 * (Tt + Tu);

203

Tv = Tg - Td;

204

Ty = Tw - Tx;

205

Ip[WS(rs, 1)] = KP500000000 * (Tv + Ty);

206

Im[0] = KP500000000 * (Ty - Tv);

207

}

204

208

}

205

209

}

206

210

}

Older »