1  /* { dg-options "-O2 -ftree-vectorize -fno-tree-loop-distribute-patterns -msve-vector-bits=512" } */
       2  
       3  #include <stdint.h>
       4  
       5  #define TEST_LOOP(TYPE1, TYPE2)						\
       6    void									\
       7    f_##TYPE1##_##TYPE2 (TYPE1 *restrict dst1, TYPE1 *restrict src1,	\
       8  		       TYPE2 *restrict dst2, TYPE2 *restrict src2,	\
       9  		       int n)						\
      10    {									\
      11      for (int i = 0; i < n; ++i)						\
      12        {									\
      13  	dst1[i * 2] = src1[i * 2] + 1;					\
      14  	dst1[i * 2 + 1] = src1[i * 2 + 1] + 1;				\
      15  	dst2[i * 2] = 2;						\
      16  	dst2[i * 2 + 1] = 3;						\
      17        }									\
      18    }
      19  
      20  #define TEST_ALL(T) \
      21    T (uint16_t, uint8_t) \
      22    T (uint32_t, uint16_t) \
      23    T (uint32_t, _Float16) \
      24    T (uint64_t, uint32_t) \
      25    T (uint64_t, float)
      26  
      27  TEST_ALL (TEST_LOOP)
      28  
      29  /* { dg-final { scan-assembler-times {\tld1rw\tz[0-9]+\.s,} 1 } } */
      30  /* { dg-final { scan-assembler-times {\tld1rd\tz[0-9]+\.d,} 2 } } */
      31  /* { dg-final { scan-assembler-times {\tld1rqw\tz[0-9]+\.s,} 2 } } */
      32  
      33  /* { dg-final { scan-assembler-times {\tst1b\tz[0-9]+\.h,} 1 } } */
      34  /* { dg-final { scan-assembler-times {\tst1h\tz[0-9]+\.s,} 2 } } */
      35  /* { dg-final { scan-assembler-times {\tst1w\tz[0-9]+\.d,} 2 } } */
      36  
      37  /* { dg-final { scan-assembler-times {\tld1h\tz[0-9]+\.h,} 1 } } */
      38  /* { dg-final { scan-assembler-times {\tst1h\tz[0-9]+\.h,} 1 } } */
      39  /* { dg-final { scan-assembler-times {\tld1w\tz[0-9]+\.s,} 2 } } */
      40  /* { dg-final { scan-assembler-times {\tst1w\tz[0-9]+\.s,} 2 } } */
      41  /* { dg-final { scan-assembler-times {\tld1d\tz[0-9]+\.d,} 2 } } */
      42  /* { dg-final { scan-assembler-times {\tst1d\tz[0-9]+\.d,} 2 } } */