gcc/testsuite/gcc.target/powerpc/pr80718.c

   1 /* { dg-do compile { target { powerpc*-*-* && lp64 } } } */
   2 /* { dg-skip-if "" { powerpc*-*-darwin* } } */
   3 /* { dg-require-effective-target powerpc_p8vector_ok } */
   4 /* { dg-skip-if "do not override -mcpu" { powerpc*-*-* } { "-mcpu=*" } { "-mcpu=power8" } } */
   5 /* { dg-options "-mcpu=power8 -O3 -ffast-math" } */
   6
   7 /* Taken from the Spec 2006 milc brenchmark.  Ultimately, GCC wants to generate
   8    a DF splat from offsettable memory.  The register allocator decided it was
   9    better to do the load in the GPR registers and do a move direct, rather than
  10    doing a load in the VSX register sets.  */
  11
  12 typedef struct
  13 {
  14   double real;
  15   double imag;
  16 } complex;
  17
  18 typedef struct
  19 {
  20   double real;
  21   double imag;
  22 } double_complex;
  23
  24 complex cmplx (double x, double y);
  25 complex cadd (complex * a, complex * b);
  26 complex cmul (complex * a, complex * b);
  27 complex csub (complex * a, complex * b);
  28 complex cdiv (complex * a, complex * b);
  29 complex conjg (complex * a);
  30 complex ce_itheta (double theta);
  31
  32 double_complex dcmplx (double x, double y);
  33 double_complex dcadd (double_complex * a, double_complex * b);
  34 double_complex dcmul (double_complex * a, double_complex * b);
  35 double_complex dcsub (double_complex * a, double_complex * b);
  36 double_complex dcdiv (double_complex * a, double_complex * b);
  37 double_complex dconjg (double_complex * a);
  38 double_complex dcexp (double_complex * a);
  39 double_complex dclog (double_complex * a);
  40 double_complex dcsqrt (double_complex * z);
  41 double_complex dce_itheta (double theta);
  42
  43 typedef struct
  44 {
  45   unsigned long r0, r1, r2, r3, r4, r5, r6;
  46   unsigned long multiplier, addend, ic_state;
  47   double scale;
  48 } double_prn;
  49
  50 double myrand (double_prn * prn_pt);
  51
  52 typedef struct
  53 {
  54   complex e[3][3];
  55 } su3_matrix;
  56
  57 typedef struct
  58 {
  59   complex c[3];
  60 } su3_vector;
  61
  62 typedef struct
  63 {
  64   complex m01, m02, m12;
  65   double m00im, m11im, m22im;
  66   double space;
  67 } anti_hermitmat;
  68
  69 typedef struct
  70 {
  71   complex e[2][2];
  72 } su2_matrix;
  73 typedef struct
  74 {
  75   su3_vector d[4];
  76 } wilson_vector;
  77 typedef struct
  78 {
  79   su3_vector h[2];
  80 } half_wilson_vector;
  81 typedef struct
  82 {
  83   wilson_vector c[3];
  84 } color_wilson_vector;
  85 typedef struct
  86 {
  87   wilson_vector d[4];
  88 } spin_wilson_vector;
  89 typedef struct
  90 {
  91   color_wilson_vector d[4];
  92 } wilson_matrix;
  93 typedef struct
  94 {
  95   spin_wilson_vector c[3];
  96 } wilson_propagator;
  97
  98 void mult_su3_nn (su3_matrix * a, su3_matrix * b, su3_matrix * c);
  99 void mult_su3_na (su3_matrix * a, su3_matrix * b, su3_matrix * c);
 100 void mult_su3_an (su3_matrix * a, su3_matrix * b, su3_matrix * c);
 101 double realtrace_su3 (su3_matrix * a, su3_matrix * b);
 102 complex trace_su3 (su3_matrix * a);
 103 complex complextrace_su3 (su3_matrix * a, su3_matrix * b);
 104 complex det_su3 (su3_matrix * a);
 105 void add_su3_matrix (su3_matrix * a, su3_matrix * b, su3_matrix * c);
 106 void sub_su3_matrix (su3_matrix * a, su3_matrix * b, su3_matrix * c);
 107 void scalar_mult_su3_matrix (su3_matrix * src, double scalar,
 108                              su3_matrix * dest);
 109 void scalar_mult_add_su3_matrix (su3_matrix * src1, su3_matrix * src2,
 110                                  double scalar, su3_matrix * dest);
 111 void scalar_mult_sub_su3_matrix (su3_matrix * src1, su3_matrix * src2,
 112                                  double scalar, su3_matrix * dest);
 113 void c_scalar_mult_su3mat (su3_matrix * src, complex * scalar,
 114                            su3_matrix * dest);
 115 void c_scalar_mult_add_su3mat (su3_matrix * src1, su3_matrix * src2,
 116                                complex * scalar, su3_matrix * dest);
 117 void c_scalar_mult_sub_su3mat (su3_matrix * src1, su3_matrix * src2,
 118                                complex * scalar, su3_matrix * dest);
 119 void su3_adjoint (su3_matrix * a, su3_matrix * b);
 120 void make_anti_hermitian (su3_matrix * m3, anti_hermitmat * ah3);
 121 void random_anti_hermitian (anti_hermitmat * mat_antihermit,
 122                             double_prn * prn_pt);
 123 void uncompress_anti_hermitian (anti_hermitmat * mat_anti, su3_matrix * mat);
 124 void compress_anti_hermitian (su3_matrix * mat, anti_hermitmat * mat_anti);
 125 void clear_su3mat (su3_matrix * dest);
 126 void su3mat_copy (su3_matrix * a, su3_matrix * b);
 127 void dumpmat (su3_matrix * m);
 128
 129 void su3_projector (su3_vector * a, su3_vector * b, su3_matrix * c);
 130 complex su3_dot (su3_vector * a, su3_vector * b);
 131 double su3_rdot (su3_vector * a, su3_vector * b);
 132 double magsq_su3vec (su3_vector * a);
 133 void su3vec_copy (su3_vector * a, su3_vector * b);
 134 void dumpvec (su3_vector * v);
 135 void clearvec (su3_vector * v);
 136
 137 void mult_su3_mat_vec (su3_matrix * a, su3_vector * b, su3_vector * c);
 138 void mult_su3_mat_vec_sum (su3_matrix * a, su3_vector * b, su3_vector * c);
 139 void mult_su3_mat_vec_sum_4dir (su3_matrix * a, su3_vector * b0,
 140                                 su3_vector * b1, su3_vector * b2,
 141                                 su3_vector * b3, su3_vector * c);
 142 void mult_su3_mat_vec_nsum (su3_matrix * a, su3_vector * b, su3_vector * c);
 143 void mult_adj_su3_mat_vec (su3_matrix * a, su3_vector * b, su3_vector * c);
 144 void mult_adj_su3_mat_vec_4dir (su3_matrix * a, su3_vector * b,
 145                                 su3_vector * c);
 146 void mult_adj_su3_mat_4vec (su3_matrix * mat, su3_vector * src,
 147                             su3_vector * dest0, su3_vector * dest1,
 148                             su3_vector * dest2, su3_vector * dest3);
 149 void mult_adj_su3_mat_vec_sum (su3_matrix * a, su3_vector * b,
 150                                su3_vector * c);
 151 void mult_adj_su3_mat_vec_nsum (su3_matrix * a, su3_vector * b,
 152                                 su3_vector * c);
 153
 154 void add_su3_vector (su3_vector * a, su3_vector * b, su3_vector * c);
 155 void sub_su3_vector (su3_vector * a, su3_vector * b, su3_vector * c);
 156 void sub_four_su3_vecs (su3_vector * a, su3_vector * b1, su3_vector * b2,
 157                         su3_vector * b3, su3_vector * b4);
 158
 159 void scalar_mult_su3_vector (su3_vector * src, double scalar,
 160                              su3_vector * dest);
 161 void scalar_mult_add_su3_vector (su3_vector * src1, su3_vector * src2,
 162                                  double scalar, su3_vector * dest);
 163 void scalar_mult_sum_su3_vector (su3_vector * src1, su3_vector * src2,
 164                                  double scalar);
 165 void scalar_mult_sub_su3_vector (su3_vector * src1, su3_vector * src2,
 166                                  double scalar, su3_vector * dest);
 167 void scalar_mult_wvec (wilson_vector * src, double s, wilson_vector * dest);
 168 void scalar_mult_hwvec (half_wilson_vector * src, double s,
 169                         half_wilson_vector * dest);
 170 void scalar_mult_add_wvec (wilson_vector * src1, wilson_vector * src2,
 171                            double scalar, wilson_vector * dest);
 172 void scalar_mult_addtm_wvec (wilson_vector * src1, wilson_vector * src2,
 173                              double scalar, wilson_vector * dest);
 174 void c_scalar_mult_wvec (wilson_vector * src1, complex * phase,
 175                          wilson_vector * dest);
 176 void c_scalar_mult_add_wvec (wilson_vector * src1, wilson_vector * src2,
 177                              complex * phase, wilson_vector * dest);
 178 void c_scalar_mult_add_wvec2 (wilson_vector * src1, wilson_vector * src2,
 179                               complex s, wilson_vector * dest);
 180 void c_scalar_mult_su3vec (su3_vector * src, complex * phase,
 181                            su3_vector * dest);
 182 void c_scalar_mult_add_su3vec (su3_vector * v1, complex * phase,
 183                                su3_vector * v2);
 184 void c_scalar_mult_sub_su3vec (su3_vector * v1, complex * phase,
 185                                su3_vector * v2);
 186
 187 void left_su2_hit_n (su2_matrix * u, int p, int q, su3_matrix * link);
 188 void right_su2_hit_a (su2_matrix * u, int p, int q, su3_matrix * link);
 189 void dumpsu2 (su2_matrix * u);
 190 void mult_su2_mat_vec_elem_n (su2_matrix * u, complex * x0, complex * x1);
 191 void mult_su2_mat_vec_elem_a (su2_matrix * u, complex * x0, complex * x1);
 192
 193 void mult_mat_wilson_vec (su3_matrix * mat, wilson_vector * src,
 194                           wilson_vector * dest);
 195 void mult_su3_mat_hwvec (su3_matrix * mat, half_wilson_vector * src,
 196                          half_wilson_vector * dest);
 197 void mult_adj_mat_wilson_vec (su3_matrix * mat, wilson_vector * src,
 198                               wilson_vector * dest);
 199 void mult_adj_su3_mat_hwvec (su3_matrix * mat, half_wilson_vector * src,
 200                              half_wilson_vector * dest);
 201
 202 void add_wilson_vector (wilson_vector * src1, wilson_vector * src2,
 203                         wilson_vector * dest);
 204 void sub_wilson_vector (wilson_vector * src1, wilson_vector * src2,
 205                         wilson_vector * dest);
 206 double magsq_wvec (wilson_vector * src);
 207 complex wvec_dot (wilson_vector * src1, wilson_vector * src2);
 208 complex wvec2_dot (wilson_vector * src1, wilson_vector * src2);
 209 double wvec_rdot (wilson_vector * a, wilson_vector * b);
 210
 211 void wp_shrink (wilson_vector * src, half_wilson_vector * dest,
 212                 int dir, int sign);
 213 void wp_shrink_4dir (wilson_vector * a, half_wilson_vector * b1,
 214                      half_wilson_vector * b2, half_wilson_vector * b3,
 215                      half_wilson_vector * b4, int sign);
 216 void wp_grow (half_wilson_vector * src, wilson_vector * dest,
 217               int dir, int sign);
 218 void wp_grow_add (half_wilson_vector * src, wilson_vector * dest,
 219                   int dir, int sign);
 220 void grow_add_four_wvecs (wilson_vector * a, half_wilson_vector * b1,
 221                           half_wilson_vector * b2, half_wilson_vector * b3,
 222                           half_wilson_vector * b4, int sign, int sum);
 223 void mult_by_gamma (wilson_vector * src, wilson_vector * dest, int dir);
 224 void mult_by_gamma_left (wilson_matrix * src, wilson_matrix * dest, int dir);
 225 void mult_by_gamma_right (wilson_matrix * src, wilson_matrix * dest, int dir);
 226 void mult_swv_by_gamma_l (spin_wilson_vector * src, spin_wilson_vector * dest,
 227                           int dir);
 228 void mult_swv_by_gamma_r (spin_wilson_vector * src, spin_wilson_vector * dest,
 229                           int dir);
 230 void su3_projector_w (wilson_vector * a, wilson_vector * b, su3_matrix * c);
 231 void clear_wvec (wilson_vector * dest);
 232 void copy_wvec (wilson_vector * src, wilson_vector * dest);
 233 void dump_wilson_vec (wilson_vector * src);
 234
 235 double gaussian_rand_no (double_prn * prn_pt);
 236 typedef int int32type;
 237 typedef unsigned int u_int32type;
 238 void byterevn (int32type w[], int n);
 239
 240 void
 241 mult_adj_su3_mat_vec (su3_matrix * a, su3_vector * b, su3_vector * c)
 242 {
 243   int i;
 244   register double t, ar, ai, br, bi, cr, ci;
 245   for (i = 0; i < 3; i++)
 246     {
 247       ar = a->e[0][i].real;
 248       ai = a->e[0][i].imag;
 249
 250       br = b->c[0].real;
 251       bi = b->c[0].imag;
 252
 253       cr = ar * br;
 254       t = ai * bi;
 255       cr += t;
 256
 257       ci = ar * bi;
 258       t = ai * br;
 259       ci -= t;
 260
 261       ar = a->e[1][i].real;
 262       ai = a->e[1][i].imag;
 263
 264       br = b->c[1].real;
 265       bi = b->c[1].imag;
 266
 267       t = ar * br;
 268       cr += t;
 269       t = ai * bi;
 270       cr += t;
 271
 272       t = ar * bi;
 273       ci += t;
 274       t = ai * br;
 275       ci -= t;
 276
 277       ar = a->e[2][i].real;
 278       ai = a->e[2][i].imag;
 279
 280       br = b->c[2].real;
 281       bi = b->c[2].imag;
 282
 283       t = ar * br;
 284       cr += t;
 285       t = ai * bi;
 286       cr += t;
 287
 288       t = ar * bi;
 289       ci += t;
 290       t = ai * br;
 291       ci -= t;
 292
 293       c->c[i].real = cr;
 294       c->c[i].imag = ci;
 295     }
 296 }
 297
 298 /* { dg-final { scan-assembler-not "mtvsrd" } } */