gcc/testsuite/gcc.target/powerpc/recip-5.c

   1 /* { dg-do compile { target { powerpc*-*-* } } } */
   2 /* { dg-skip-if "" { powerpc*-*-darwin* } } */
   3 /* { dg-require-effective-target powerpc_vsx_ok } */
   4 /* { dg-skip-if "do not override -mcpu" { powerpc*-*-* } { "-mcpu=*" } { "-mcpu=power7" } } */
   5 /* { dg-options "-O3 -ftree-vectorize -mrecip=all -ffast-math -mcpu=power7 -fno-unroll-loops" } */
   6 /* { dg-final { scan-assembler-times "xvredp" 4 } } */
   7 /* { dg-final { scan-assembler-times "xvresp" 5 } } */
   8 /* { dg-final { scan-assembler-times "xsredp\|fre\ " 2 } } */
   9 /* { dg-final { scan-assembler-times "fres\|xsresp" 2 } } */
  10 /* { dg-final { scan-assembler-times "fmuls\|xsmulsp" 2 } } */
  11 /* { dg-final { scan-assembler-times "fnmsubs\|xsnmsub.sp" 2 } } */
  12 /* { dg-final { scan-assembler-times "xsmuldp\|fmul\ " 2 } } */
  13 /* { dg-final { scan-assembler-times "xsnmsub.dp\|fnmsub\ " 4 } } */
  14 /* { dg-final { scan-assembler-times "xvmulsp" 7 } } */
  15 /* { dg-final { scan-assembler-times "xvnmsub.sp" 5 } } */
  16 /* { dg-final { scan-assembler-times "xvmuldp" 6 } } */
  17 /* { dg-final { scan-assembler-times "xvnmsub.dp" 8 } } */
  18
  19 #include <altivec.h>
  20
  21 float f_recip (float a, float b) { return __builtin_recipdivf (a, b); }
  22 double d_recip (double a, double b) { return __builtin_recipdiv (a, b); }
  23
  24 float f_div (float a, float b) { return a / b; }
  25 double d_div (double a, double b) { return a / b; }
  26
  27 #define SIZE 1024
  28
  29 double d_a[SIZE] __attribute__((__aligned__(32)));
  30 double d_b[SIZE] __attribute__((__aligned__(32)));
  31 double d_c[SIZE] __attribute__((__aligned__(32)));
  32
  33 float f_a[SIZE] __attribute__((__aligned__(32)));
  34 float f_b[SIZE] __attribute__((__aligned__(32)));
  35 float f_c[SIZE] __attribute__((__aligned__(32)));
  36
  37 void vec_f_recip (void)
  38 {
  39   int i;
  40
  41   for (i = 0; i < SIZE; i++)
  42     f_a[i] = __builtin_recipdivf (f_b[i], f_c[i]);
  43 }
  44
  45 void vec_d_recip (void)
  46 {
  47   int i;
  48
  49   for (i = 0; i < SIZE; i++)
  50     d_a[i] = __builtin_recipdiv (d_b[i], d_c[i]);
  51 }
  52
  53 void vec_f_div (void)
  54 {
  55   int i;
  56
  57   for (i = 0; i < SIZE; i++)
  58     f_a[i] = f_b[i] / f_c[i];
  59 }
  60
  61 void vec_f_div2 (void)
  62 {
  63   int i;
  64
  65   for (i = 0; i < SIZE; i++)
  66     f_a[i] = f_b[i] / 2.0f;
  67 }
  68
  69 void vec_f_div53 (void)
  70 {
  71   int i;
  72
  73   for (i = 0; i < SIZE; i++)
  74     f_a[i] = f_b[i] / 53.0f;
  75 }
  76
  77 void vec_d_div (void)
  78 {
  79   int i;
  80
  81   for (i = 0; i < SIZE; i++)
  82     d_a[i] = d_b[i] / d_c[i];
  83 }
  84
  85 void vec_d_div2 (void)
  86 {
  87   int i;
  88
  89   for (i = 0; i < SIZE; i++)
  90     d_a[i] = d_b[i] / 2.0;
  91 }
  92
  93 void vec_d_div53 (void)
  94 {
  95   int i;
  96
  97   for (i = 0; i < SIZE; i++)
  98     d_a[i] = d_b[i] / 53.0;
  99 }
 100
 101 vector float v4sf_recip1 (vector float a, vector float b) { return vec_recipdiv (a, b); }
 102 vector float v4sf_recip2 (vector float a, vector float b) { return __builtin_altivec_vrecipdivfp (a, b); }
 103 vector double v2df_recip1 (vector double a, vector double b) { return vec_recipdiv (a, b); }
 104 vector float v4sf_recip3 (vector float a, vector float b) { return __builtin_vsx_xvrecipdivsp (a, b); }
 105 vector double v2df_recip2 (vector double a, vector double b) { return __builtin_vsx_xvrecipdivdp (a, b); }