gcc/testsuite/gcc.target/aarch64/sve/vcond_8.c

   1 /* { dg-do compile } */
   2 /* { dg-options "-O2 -ftree-vectorize -fno-trapping-math -ffinite-math-only" } */
   3
   4 #include <stdint.h>
   5
   6 #define add(A, B) ((A) + (B))
   7 #define sub(A, B) ((A) - (B))
   8 #define mul(A, B) ((A) * (B))
   9 #define div(A, B) ((A) / (B))
  10 #define max(A, B) ((A) > (B) ? (A) : (B))
  11 #define min(A, B) ((A) < (B) ? (A) : (B))
  12 #define and(A, B) ((A) & (B))
  13 #define ior(A, B) ((A) | (B))
  14 #define xor(A, B) ((A) ^ (B))
  15
  16 #define DEF_LOOP(TYPE, CMPTYPE, OP)                             \
  17   void __attribute__((noipa))                                   \
  18   f_##OP##_##TYPE (TYPE *restrict dest, CMPTYPE *restrict cond, \
  19                    CMPTYPE limit, TYPE *restrict src,           \
  20                    TYPE val, unsigned int n)                    \
  21   {                                                             \
  22     for (unsigned int i = 0; i < n; ++i)                        \
  23       {                                                         \
  24         TYPE truev = OP (src[i], val);                          \
  25         dest[i] = cond[i] < limit ? truev : src[i];             \
  26       }                                                         \
  27   }
  28
  29 #define FOR_EACH_INT_TYPE(T, TYPE) \
  30   T (TYPE, TYPE, add) \
  31   T (TYPE, TYPE, sub) \
  32   T (TYPE, TYPE, mul) \
  33   T (TYPE, TYPE, max) \
  34   T (TYPE, TYPE, min) \
  35   T (TYPE, TYPE, and) \
  36   T (TYPE, TYPE, ior) \
  37   T (TYPE, TYPE, xor)
  38
  39 #define FOR_EACH_FP_TYPE(T, TYPE, CMPTYPE, SUFFIX) \
  40   T (TYPE, CMPTYPE, add) \
  41   T (TYPE, CMPTYPE, sub) \
  42   T (TYPE, CMPTYPE, mul) \
  43   T (TYPE, CMPTYPE, div) \
  44   T (TYPE, CMPTYPE, __builtin_fmax##SUFFIX) \
  45   T (TYPE, CMPTYPE, __builtin_fmin##SUFFIX)
  46
  47 #define FOR_EACH_LOOP(T) \
  48   FOR_EACH_INT_TYPE (T, int8_t) \
  49   FOR_EACH_INT_TYPE (T, int16_t) \
  50   FOR_EACH_INT_TYPE (T, int32_t) \
  51   FOR_EACH_INT_TYPE (T, int64_t) \
  52   FOR_EACH_INT_TYPE (T, uint8_t) \
  53   FOR_EACH_INT_TYPE (T, uint16_t) \
  54   FOR_EACH_INT_TYPE (T, uint32_t) \
  55   FOR_EACH_INT_TYPE (T, uint64_t) \
  56   FOR_EACH_FP_TYPE (T, _Float16, uint16_t, f16) \
  57   FOR_EACH_FP_TYPE (T, float, float, f32) \
  58   FOR_EACH_FP_TYPE (T, double, double, f64)
  59
  60 FOR_EACH_LOOP (DEF_LOOP)
  61
  62 /* { dg-final { scan-assembler-not {\tsel\t} } } */
  63 /* { dg-final { scan-assembler-not {\tmov\tz[0-9]+\.., z[0-9]+} } } */
  64
  65 /* { dg-final { scan-assembler-times {\tadd\tz[0-9]+\.b, p[0-7]/m,} 2 } } */
  66 /* { dg-final { scan-assembler-times {\tadd\tz[0-9]+\.h, p[0-7]/m,} 2 } } */
  67 /* { dg-final { scan-assembler-times {\tadd\tz[0-9]+\.s, p[0-7]/m,} 2 } } */
  68 /* { dg-final { scan-assembler-times {\tadd\tz[0-9]+\.d, p[0-7]/m,} 2 } } */
  69
  70 /* { dg-final { scan-assembler-times {\tsub\tz[0-9]+\.b, p[0-7]/m,} 2 } } */
  71 /* { dg-final { scan-assembler-times {\tsub\tz[0-9]+\.h, p[0-7]/m,} 2 } } */
  72 /* { dg-final { scan-assembler-times {\tsub\tz[0-9]+\.s, p[0-7]/m,} 2 } } */
  73 /* { dg-final { scan-assembler-times {\tsub\tz[0-9]+\.d, p[0-7]/m,} 2 } } */
  74
  75 /* { dg-final { scan-assembler-times {\tmul\tz[0-9]+\.b, p[0-7]/m,} 2 } } */
  76 /* { dg-final { scan-assembler-times {\tmul\tz[0-9]+\.h, p[0-7]/m,} 2 } } */
  77 /* { dg-final { scan-assembler-times {\tmul\tz[0-9]+\.s, p[0-7]/m,} 2 } } */
  78 /* { dg-final { scan-assembler-times {\tmul\tz[0-9]+\.d, p[0-7]/m,} 2 } } */
  79
  80 /* { dg-final { scan-assembler-times {\tsmax\tz[0-9]+\.b, p[0-7]/m,} 1 } } */
  81 /* { dg-final { scan-assembler-times {\tsmax\tz[0-9]+\.h, p[0-7]/m,} 1 } } */
  82 /* { dg-final { scan-assembler-times {\tsmax\tz[0-9]+\.s, p[0-7]/m,} 1 } } */
  83 /* { dg-final { scan-assembler-times {\tsmax\tz[0-9]+\.d, p[0-7]/m,} 1 } } */
  84
  85 /* { dg-final { scan-assembler-times {\tsmin\tz[0-9]+\.b, p[0-7]/m,} 1 } } */
  86 /* { dg-final { scan-assembler-times {\tsmin\tz[0-9]+\.h, p[0-7]/m,} 1 } } */
  87 /* { dg-final { scan-assembler-times {\tsmin\tz[0-9]+\.s, p[0-7]/m,} 1 } } */
  88 /* { dg-final { scan-assembler-times {\tsmin\tz[0-9]+\.d, p[0-7]/m,} 1 } } */
  89
  90 /* { dg-final { scan-assembler-times {\tumax\tz[0-9]+\.b, p[0-7]/m,} 1 } } */
  91 /* { dg-final { scan-assembler-times {\tumax\tz[0-9]+\.h, p[0-7]/m,} 1 } } */
  92 /* { dg-final { scan-assembler-times {\tumax\tz[0-9]+\.s, p[0-7]/m,} 1 } } */
  93 /* { dg-final { scan-assembler-times {\tumax\tz[0-9]+\.d, p[0-7]/m,} 1 } } */
  94
  95 /* { dg-final { scan-assembler-times {\tumin\tz[0-9]+\.b, p[0-7]/m,} 1 } } */
  96 /* { dg-final { scan-assembler-times {\tumin\tz[0-9]+\.h, p[0-7]/m,} 1 } } */
  97 /* { dg-final { scan-assembler-times {\tumin\tz[0-9]+\.s, p[0-7]/m,} 1 } } */
  98 /* { dg-final { scan-assembler-times {\tumin\tz[0-9]+\.d, p[0-7]/m,} 1 } } */
  99
 100 /* { dg-final { scan-assembler-times {\tand\tz[0-9]+\.b, p[0-7]/m,} 2 } } */
 101 /* { dg-final { scan-assembler-times {\tand\tz[0-9]+\.h, p[0-7]/m,} 2 } } */
 102 /* { dg-final { scan-assembler-times {\tand\tz[0-9]+\.s, p[0-7]/m,} 2 } } */
 103 /* { dg-final { scan-assembler-times {\tand\tz[0-9]+\.d, p[0-7]/m,} 2 } } */
 104
 105 /* { dg-final { scan-assembler-times {\torr\tz[0-9]+\.b, p[0-7]/m,} 2 } } */
 106 /* { dg-final { scan-assembler-times {\torr\tz[0-9]+\.h, p[0-7]/m,} 2 } } */
 107 /* { dg-final { scan-assembler-times {\torr\tz[0-9]+\.s, p[0-7]/m,} 2 } } */
 108 /* { dg-final { scan-assembler-times {\torr\tz[0-9]+\.d, p[0-7]/m,} 2 } } */
 109
 110 /* { dg-final { scan-assembler-times {\teor\tz[0-9]+\.b, p[0-7]/m,} 2 } } */
 111 /* { dg-final { scan-assembler-times {\teor\tz[0-9]+\.h, p[0-7]/m,} 2 } } */
 112 /* { dg-final { scan-assembler-times {\teor\tz[0-9]+\.s, p[0-7]/m,} 2 } } */
 113 /* { dg-final { scan-assembler-times {\teor\tz[0-9]+\.d, p[0-7]/m,} 2 } } */
 114
 115 /* { dg-final { scan-assembler-times {\tfadd\tz[0-9]+\.h, p[0-7]/m,} 1 } } */
 116 /* { dg-final { scan-assembler-times {\tfadd\tz[0-9]+\.s, p[0-7]/m,} 1 } } */
 117 /* { dg-final { scan-assembler-times {\tfadd\tz[0-9]+\.d, p[0-7]/m,} 1 } } */
 118
 119 /* { dg-final { scan-assembler-times {\tfsub\tz[0-9]+\.h, p[0-7]/m,} 1 } } */
 120 /* { dg-final { scan-assembler-times {\tfsub\tz[0-9]+\.s, p[0-7]/m,} 1 } } */
 121 /* { dg-final { scan-assembler-times {\tfsub\tz[0-9]+\.d, p[0-7]/m,} 1 } } */
 122
 123 /* { dg-final { scan-assembler-times {\tfmul\tz[0-9]+\.h, p[0-7]/m,} 1 } } */
 124 /* { dg-final { scan-assembler-times {\tfmul\tz[0-9]+\.s, p[0-7]/m,} 1 } } */
 125 /* { dg-final { scan-assembler-times {\tfmul\tz[0-9]+\.d, p[0-7]/m,} 1 } } */
 126
 127 /* { dg-final { scan-assembler-times {\tfdiv\tz[0-9]+\.h, p[0-7]/m,} 1 } } */
 128 /* { dg-final { scan-assembler-times {\tfdiv\tz[0-9]+\.s, p[0-7]/m,} 1 } } */
 129 /* { dg-final { scan-assembler-times {\tfdiv\tz[0-9]+\.d, p[0-7]/m,} 1 } } */
 130
 131 /* { dg-final { scan-assembler-times {\tfmaxnm\tz[0-9]+\.h, p[0-7]/m,} 1 } } */
 132 /* { dg-final { scan-assembler-times {\tfmaxnm\tz[0-9]+\.s, p[0-7]/m,} 1 } } */
 133 /* { dg-final { scan-assembler-times {\tfmaxnm\tz[0-9]+\.d, p[0-7]/m,} 1 } } */
 134
 135 /* { dg-final { scan-assembler-times {\tfminnm\tz[0-9]+\.h, p[0-7]/m,} 1 } } */
 136 /* { dg-final { scan-assembler-times {\tfminnm\tz[0-9]+\.s, p[0-7]/m,} 1 } } */
 137 /* { dg-final { scan-assembler-times {\tfminnm\tz[0-9]+\.d, p[0-7]/m,} 1 } } */