1  /* { dg-do compile } */
       2  /* { dg-options "-O2 -ftree-vectorize" } */
       3  
       4  #include <stdint.h>
       5  
       6  #define DEF_SAD(TYPE1, TYPE2)						\
       7  TYPE1 __attribute__ ((noinline, noclone))				\
       8  sum_abs_##TYPE1##_##TYPE2 (TYPE2 *restrict x, TYPE2 *restrict y, int n)	\
       9  {									\
      10    TYPE1 sum = 0;							\
      11    for (int i = 0; i < n; i++)						\
      12      {									\
      13        sum += __builtin_abs (x[i] - y[i]);				\
      14      }									\
      15    return sum;								\
      16  }
      17  
      18  DEF_SAD(int32_t, uint8_t)
      19  DEF_SAD(int32_t, int8_t)
      20  DEF_SAD(int64_t, uint16_t)
      21  DEF_SAD(int64_t, int16_t)
      22  
      23  /* { dg-final { scan-assembler-times {\tuabd\tz[0-9]+\.b, p[0-7]/m, z[0-9]+\.b, z[0-9]+\.b\n} 1 } } */
      24  /* { dg-final { scan-assembler-times {\tsabd\tz[0-9]+\.b, p[0-7]/m, z[0-9]+\.b, z[0-9]+\.b\n} 1 } } */
      25  /* { dg-final { scan-assembler-times {\tudot\tz[0-9]+\.s, z[0-9]+\.b, z[0-9]+\.b\n} 2 } } */
      26  /* { dg-final { scan-assembler-times {\tuabd\tz[0-9]+\.h, p[0-7]/m, z[0-9]+\.h, z[0-9]+\.h\n} 1 } } */
      27  /* { dg-final { scan-assembler-times {\tsabd\tz[0-9]+\.h, p[0-7]/m, z[0-9]+\.h, z[0-9]+\.h\n} 1 } } */
      28  /* { dg-final { scan-assembler-times {\tudot\tz[0-9]+\.d, z[0-9]+\.h, z[0-9]+\.h\n} 2 } } */