| 1 |
/* { dg-do compile { target { powerpc*-*-* && lp64 } } } */ |
| 2 |
/* { dg-skip-if "" { powerpc*-*-darwin* } } */ |
| 3 |
/* { dg-options "-mdejagnu-cpu=power8 -mvsx -O3 -ffast-math" } */ |
| 4 |
/* { dg-require-effective-target powerpc_vsx } */ |
| 5 |
|
| 6 |
/* Taken from the Spec 2006 milc brenchmark. Ultimately, GCC wants to generate |
| 7 |
a DF splat from offsettable memory. The register allocator decided it was |
| 8 |
better to do the load in the GPR registers and do a move direct, rather than |
| 9 |
doing a load in the VSX register sets. */ |
| 10 |
|
| 11 |
typedef struct |
| 12 |
{ |
| 13 |
double real; |
| 14 |
double imag; |
| 15 |
} complex; |
| 16 |
|
| 17 |
typedef struct |
| 18 |
{ |
| 19 |
double real; |
| 20 |
double imag; |
| 21 |
} double_complex; |
| 22 |
|
| 23 |
complex cmplx (double x, double y); |
| 24 |
complex cadd (complex * a, complex * b); |
| 25 |
complex cmul (complex * a, complex * b); |
| 26 |
complex csub (complex * a, complex * b); |
| 27 |
complex cdiv (complex * a, complex * b); |
| 28 |
complex conjg (complex * a); |
| 29 |
complex ce_itheta (double theta); |
| 30 |
|
| 31 |
double_complex dcmplx (double x, double y); |
| 32 |
double_complex dcadd (double_complex * a, double_complex * b); |
| 33 |
double_complex dcmul (double_complex * a, double_complex * b); |
| 34 |
double_complex dcsub (double_complex * a, double_complex * b); |
| 35 |
double_complex dcdiv (double_complex * a, double_complex * b); |
| 36 |
double_complex dconjg (double_complex * a); |
| 37 |
double_complex dcexp (double_complex * a); |
| 38 |
double_complex dclog (double_complex * a); |
| 39 |
double_complex dcsqrt (double_complex * z); |
| 40 |
double_complex dce_itheta (double theta); |
| 41 |
|
| 42 |
typedef struct |
| 43 |
{ |
| 44 |
unsigned long r0, r1, r2, r3, r4, r5, r6; |
| 45 |
unsigned long multiplier, addend, ic_state; |
| 46 |
double scale; |
| 47 |
} double_prn; |
| 48 |
|
| 49 |
double myrand (double_prn * prn_pt); |
| 50 |
|
| 51 |
typedef struct |
| 52 |
{ |
| 53 |
complex e[3][3]; |
| 54 |
} su3_matrix; |
| 55 |
|
| 56 |
typedef struct |
| 57 |
{ |
| 58 |
complex c[3]; |
| 59 |
} su3_vector; |
| 60 |
|
| 61 |
typedef struct |
| 62 |
{ |
| 63 |
complex m01, m02, m12; |
| 64 |
double m00im, m11im, m22im; |
| 65 |
double space; |
| 66 |
} anti_hermitmat; |
| 67 |
|
| 68 |
typedef struct |
| 69 |
{ |
| 70 |
complex e[2][2]; |
| 71 |
} su2_matrix; |
| 72 |
typedef struct |
| 73 |
{ |
| 74 |
su3_vector d[4]; |
| 75 |
} wilson_vector; |
| 76 |
typedef struct |
| 77 |
{ |
| 78 |
su3_vector h[2]; |
| 79 |
} half_wilson_vector; |
| 80 |
typedef struct |
| 81 |
{ |
| 82 |
wilson_vector c[3]; |
| 83 |
} color_wilson_vector; |
| 84 |
typedef struct |
| 85 |
{ |
| 86 |
wilson_vector d[4]; |
| 87 |
} spin_wilson_vector; |
| 88 |
typedef struct |
| 89 |
{ |
| 90 |
color_wilson_vector d[4]; |
| 91 |
} wilson_matrix; |
| 92 |
typedef struct |
| 93 |
{ |
| 94 |
spin_wilson_vector c[3]; |
| 95 |
} wilson_propagator; |
| 96 |
|
| 97 |
void mult_su3_nn (su3_matrix * a, su3_matrix * b, su3_matrix * c); |
| 98 |
void mult_su3_na (su3_matrix * a, su3_matrix * b, su3_matrix * c); |
| 99 |
void mult_su3_an (su3_matrix * a, su3_matrix * b, su3_matrix * c); |
| 100 |
double realtrace_su3 (su3_matrix * a, su3_matrix * b); |
| 101 |
complex trace_su3 (su3_matrix * a); |
| 102 |
complex complextrace_su3 (su3_matrix * a, su3_matrix * b); |
| 103 |
complex det_su3 (su3_matrix * a); |
| 104 |
void add_su3_matrix (su3_matrix * a, su3_matrix * b, su3_matrix * c); |
| 105 |
void sub_su3_matrix (su3_matrix * a, su3_matrix * b, su3_matrix * c); |
| 106 |
void scalar_mult_su3_matrix (su3_matrix * src, double scalar, |
| 107 |
su3_matrix * dest); |
| 108 |
void scalar_mult_add_su3_matrix (su3_matrix * src1, su3_matrix * src2, |
| 109 |
double scalar, su3_matrix * dest); |
| 110 |
void scalar_mult_sub_su3_matrix (su3_matrix * src1, su3_matrix * src2, |
| 111 |
double scalar, su3_matrix * dest); |
| 112 |
void c_scalar_mult_su3mat (su3_matrix * src, complex * scalar, |
| 113 |
su3_matrix * dest); |
| 114 |
void c_scalar_mult_add_su3mat (su3_matrix * src1, su3_matrix * src2, |
| 115 |
complex * scalar, su3_matrix * dest); |
| 116 |
void c_scalar_mult_sub_su3mat (su3_matrix * src1, su3_matrix * src2, |
| 117 |
complex * scalar, su3_matrix * dest); |
| 118 |
void su3_adjoint (su3_matrix * a, su3_matrix * b); |
| 119 |
void make_anti_hermitian (su3_matrix * m3, anti_hermitmat * ah3); |
| 120 |
void random_anti_hermitian (anti_hermitmat * mat_antihermit, |
| 121 |
double_prn * prn_pt); |
| 122 |
void uncompress_anti_hermitian (anti_hermitmat * mat_anti, su3_matrix * mat); |
| 123 |
void compress_anti_hermitian (su3_matrix * mat, anti_hermitmat * mat_anti); |
| 124 |
void clear_su3mat (su3_matrix * dest); |
| 125 |
void su3mat_copy (su3_matrix * a, su3_matrix * b); |
| 126 |
void dumpmat (su3_matrix * m); |
| 127 |
|
| 128 |
void su3_projector (su3_vector * a, su3_vector * b, su3_matrix * c); |
| 129 |
complex su3_dot (su3_vector * a, su3_vector * b); |
| 130 |
double su3_rdot (su3_vector * a, su3_vector * b); |
| 131 |
double magsq_su3vec (su3_vector * a); |
| 132 |
void su3vec_copy (su3_vector * a, su3_vector * b); |
| 133 |
void dumpvec (su3_vector * v); |
| 134 |
void clearvec (su3_vector * v); |
| 135 |
|
| 136 |
void mult_su3_mat_vec (su3_matrix * a, su3_vector * b, su3_vector * c); |
| 137 |
void mult_su3_mat_vec_sum (su3_matrix * a, su3_vector * b, su3_vector * c); |
| 138 |
void mult_su3_mat_vec_sum_4dir (su3_matrix * a, su3_vector * b0, |
| 139 |
su3_vector * b1, su3_vector * b2, |
| 140 |
su3_vector * b3, su3_vector * c); |
| 141 |
void mult_su3_mat_vec_nsum (su3_matrix * a, su3_vector * b, su3_vector * c); |
| 142 |
void mult_adj_su3_mat_vec (su3_matrix * a, su3_vector * b, su3_vector * c); |
| 143 |
void mult_adj_su3_mat_vec_4dir (su3_matrix * a, su3_vector * b, |
| 144 |
su3_vector * c); |
| 145 |
void mult_adj_su3_mat_4vec (su3_matrix * mat, su3_vector * src, |
| 146 |
su3_vector * dest0, su3_vector * dest1, |
| 147 |
su3_vector * dest2, su3_vector * dest3); |
| 148 |
void mult_adj_su3_mat_vec_sum (su3_matrix * a, su3_vector * b, |
| 149 |
su3_vector * c); |
| 150 |
void mult_adj_su3_mat_vec_nsum (su3_matrix * a, su3_vector * b, |
| 151 |
su3_vector * c); |
| 152 |
|
| 153 |
void add_su3_vector (su3_vector * a, su3_vector * b, su3_vector * c); |
| 154 |
void sub_su3_vector (su3_vector * a, su3_vector * b, su3_vector * c); |
| 155 |
void sub_four_su3_vecs (su3_vector * a, su3_vector * b1, su3_vector * b2, |
| 156 |
su3_vector * b3, su3_vector * b4); |
| 157 |
|
| 158 |
void scalar_mult_su3_vector (su3_vector * src, double scalar, |
| 159 |
su3_vector * dest); |
| 160 |
void scalar_mult_add_su3_vector (su3_vector * src1, su3_vector * src2, |
| 161 |
double scalar, su3_vector * dest); |
| 162 |
void scalar_mult_sum_su3_vector (su3_vector * src1, su3_vector * src2, |
| 163 |
double scalar); |
| 164 |
void scalar_mult_sub_su3_vector (su3_vector * src1, su3_vector * src2, |
| 165 |
double scalar, su3_vector * dest); |
| 166 |
void scalar_mult_wvec (wilson_vector * src, double s, wilson_vector * dest); |
| 167 |
void scalar_mult_hwvec (half_wilson_vector * src, double s, |
| 168 |
half_wilson_vector * dest); |
| 169 |
void scalar_mult_add_wvec (wilson_vector * src1, wilson_vector * src2, |
| 170 |
double scalar, wilson_vector * dest); |
| 171 |
void scalar_mult_addtm_wvec (wilson_vector * src1, wilson_vector * src2, |
| 172 |
double scalar, wilson_vector * dest); |
| 173 |
void c_scalar_mult_wvec (wilson_vector * src1, complex * phase, |
| 174 |
wilson_vector * dest); |
| 175 |
void c_scalar_mult_add_wvec (wilson_vector * src1, wilson_vector * src2, |
| 176 |
complex * phase, wilson_vector * dest); |
| 177 |
void c_scalar_mult_add_wvec2 (wilson_vector * src1, wilson_vector * src2, |
| 178 |
complex s, wilson_vector * dest); |
| 179 |
void c_scalar_mult_su3vec (su3_vector * src, complex * phase, |
| 180 |
su3_vector * dest); |
| 181 |
void c_scalar_mult_add_su3vec (su3_vector * v1, complex * phase, |
| 182 |
su3_vector * v2); |
| 183 |
void c_scalar_mult_sub_su3vec (su3_vector * v1, complex * phase, |
| 184 |
su3_vector * v2); |
| 185 |
|
| 186 |
void left_su2_hit_n (su2_matrix * u, int p, int q, su3_matrix * link); |
| 187 |
void right_su2_hit_a (su2_matrix * u, int p, int q, su3_matrix * link); |
| 188 |
void dumpsu2 (su2_matrix * u); |
| 189 |
void mult_su2_mat_vec_elem_n (su2_matrix * u, complex * x0, complex * x1); |
| 190 |
void mult_su2_mat_vec_elem_a (su2_matrix * u, complex * x0, complex * x1); |
| 191 |
|
| 192 |
void mult_mat_wilson_vec (su3_matrix * mat, wilson_vector * src, |
| 193 |
wilson_vector * dest); |
| 194 |
void mult_su3_mat_hwvec (su3_matrix * mat, half_wilson_vector * src, |
| 195 |
half_wilson_vector * dest); |
| 196 |
void mult_adj_mat_wilson_vec (su3_matrix * mat, wilson_vector * src, |
| 197 |
wilson_vector * dest); |
| 198 |
void mult_adj_su3_mat_hwvec (su3_matrix * mat, half_wilson_vector * src, |
| 199 |
half_wilson_vector * dest); |
| 200 |
|
| 201 |
void add_wilson_vector (wilson_vector * src1, wilson_vector * src2, |
| 202 |
wilson_vector * dest); |
| 203 |
void sub_wilson_vector (wilson_vector * src1, wilson_vector * src2, |
| 204 |
wilson_vector * dest); |
| 205 |
double magsq_wvec (wilson_vector * src); |
| 206 |
complex wvec_dot (wilson_vector * src1, wilson_vector * src2); |
| 207 |
complex wvec2_dot (wilson_vector * src1, wilson_vector * src2); |
| 208 |
double wvec_rdot (wilson_vector * a, wilson_vector * b); |
| 209 |
|
| 210 |
void wp_shrink (wilson_vector * src, half_wilson_vector * dest, |
| 211 |
int dir, int sign); |
| 212 |
void wp_shrink_4dir (wilson_vector * a, half_wilson_vector * b1, |
| 213 |
half_wilson_vector * b2, half_wilson_vector * b3, |
| 214 |
half_wilson_vector * b4, int sign); |
| 215 |
void wp_grow (half_wilson_vector * src, wilson_vector * dest, |
| 216 |
int dir, int sign); |
| 217 |
void wp_grow_add (half_wilson_vector * src, wilson_vector * dest, |
| 218 |
int dir, int sign); |
| 219 |
void grow_add_four_wvecs (wilson_vector * a, half_wilson_vector * b1, |
| 220 |
half_wilson_vector * b2, half_wilson_vector * b3, |
| 221 |
half_wilson_vector * b4, int sign, int sum); |
| 222 |
void mult_by_gamma (wilson_vector * src, wilson_vector * dest, int dir); |
| 223 |
void mult_by_gamma_left (wilson_matrix * src, wilson_matrix * dest, int dir); |
| 224 |
void mult_by_gamma_right (wilson_matrix * src, wilson_matrix * dest, int dir); |
| 225 |
void mult_swv_by_gamma_l (spin_wilson_vector * src, spin_wilson_vector * dest, |
| 226 |
int dir); |
| 227 |
void mult_swv_by_gamma_r (spin_wilson_vector * src, spin_wilson_vector * dest, |
| 228 |
int dir); |
| 229 |
void su3_projector_w (wilson_vector * a, wilson_vector * b, su3_matrix * c); |
| 230 |
void clear_wvec (wilson_vector * dest); |
| 231 |
void copy_wvec (wilson_vector * src, wilson_vector * dest); |
| 232 |
void dump_wilson_vec (wilson_vector * src); |
| 233 |
|
| 234 |
double gaussian_rand_no (double_prn * prn_pt); |
| 235 |
typedef int int32type; |
| 236 |
typedef unsigned int u_int32type; |
| 237 |
void byterevn (int32type w[], int n); |
| 238 |
|
| 239 |
void |
| 240 |
mult_adj_su3_mat_vec (su3_matrix * a, su3_vector * b, su3_vector * c) |
| 241 |
{ |
| 242 |
int i; |
| 243 |
register double t, ar, ai, br, bi, cr, ci; |
| 244 |
for (i = 0; i < 3; i++) |
| 245 |
{ |
| 246 |
ar = a->e[0][i].real; |
| 247 |
ai = a->e[0][i].imag; |
| 248 |
|
| 249 |
br = b->c[0].real; |
| 250 |
bi = b->c[0].imag; |
| 251 |
|
| 252 |
cr = ar * br; |
| 253 |
t = ai * bi; |
| 254 |
cr += t; |
| 255 |
|
| 256 |
ci = ar * bi; |
| 257 |
t = ai * br; |
| 258 |
ci -= t; |
| 259 |
|
| 260 |
ar = a->e[1][i].real; |
| 261 |
ai = a->e[1][i].imag; |
| 262 |
|
| 263 |
br = b->c[1].real; |
| 264 |
bi = b->c[1].imag; |
| 265 |
|
| 266 |
t = ar * br; |
| 267 |
cr += t; |
| 268 |
t = ai * bi; |
| 269 |
cr += t; |
| 270 |
|
| 271 |
t = ar * bi; |
| 272 |
ci += t; |
| 273 |
t = ai * br; |
| 274 |
ci -= t; |
| 275 |
|
| 276 |
ar = a->e[2][i].real; |
| 277 |
ai = a->e[2][i].imag; |
| 278 |
|
| 279 |
br = b->c[2].real; |
| 280 |
bi = b->c[2].imag; |
| 281 |
|
| 282 |
t = ar * br; |
| 283 |
cr += t; |
| 284 |
t = ai * bi; |
| 285 |
cr += t; |
| 286 |
|
| 287 |
t = ar * bi; |
| 288 |
ci += t; |
| 289 |
t = ai * br; |
| 290 |
ci -= t; |
| 291 |
|
| 292 |
c->c[i].real = cr; |
| 293 |
c->c[i].imag = ci; |
| 294 |
} |
| 295 |
} |
| 296 |
|
| 297 |
/* { dg-final { scan-assembler-not "mtvsrd" } } */ |