1 /* { dg-require-effective-target scalar_all_fma } */
2 /* { dg-additional-options "-fdump-tree-optimized" } */
3 
4 #include "tree-vect.h"
5 
6 #define N (VECTOR_BITS * 11 / 64 + 3)
7 
8 #define DEF(INV)					\
9   void __attribute__ ((noipa))				\
10   f_##INV (double *restrict a, double *restrict b,	\
11 	   double *restrict c, double *restrict d)	\
12   {							\
13     for (int i = 0; i < N; ++i)				\
14       {							\
15 	double mb = (INV & 1 ? -b[i] : b[i]);		\
16 	double mc = c[i];				\
17 	double md = (INV & 2 ? -d[i] : d[i]);		\
18 	double fma = __builtin_fma (mb, mc, md);	\
19 	double truev = (INV & 4 ? -fma : fma);		\
20 	a[i] = b[i] < 10 ? truev : 10.0;		\
21       }							\
22   }
23 
24 #define TEST(INV)					\
25   {							\
26     f_##INV (a, b, c, d);				\
27     for (int i = 0; i < N; ++i)				\
28       {							\
29 	double mb = (INV & 1 ? -b[i] : b[i]);		\
30 	double mc = c[i];				\
31 	double md = (INV & 2 ? -d[i] : d[i]);		\
32 	double fma = __builtin_fma (mb, mc, md);	\
33 	double truev = (INV & 4 ? -fma : fma);		\
34 	if (a[i] != (i % 17 < 10 ? truev : 10.0))	\
35 	  __builtin_abort ();				\
36 	asm volatile ("" ::: "memory");			\
37       }							\
38   }
39 
40 #define FOR_EACH_INV(T) \
41   T (0) T (1) T (2) T (3) T (4) T (5) T (6) T (7)
42 
FOR_EACH_INV(DEF)43 FOR_EACH_INV (DEF)
44 
45 int
46 main (void)
47 {
48   double a[N], b[N], c[N], d[N];
49   for (int i = 0; i < N; ++i)
50     {
51       b[i] = i % 17;
52       c[i] = i % 9 + 11;
53       d[i] = i % 13 + 14;
54       asm volatile ("" ::: "memory");
55     }
56   FOR_EACH_INV (TEST)
57   return 0;
58 }
59 
60 /* { dg-final { scan-tree-dump-times { = \.COND_FMA } 2 "optimized" { target vect_double_cond_arith } } } */
61 /* { dg-final { scan-tree-dump-times { = \.COND_FMS } 2 "optimized" { target vect_double_cond_arith } } } */
62 /* { dg-final { scan-tree-dump-times { = \.COND_FNMA } 2 "optimized" { target vect_double_cond_arith } } } */
63 /* { dg-final { scan-tree-dump-times { = \.COND_FNMS } 2 "optimized" { target vect_double_cond_arith } } } */
64