gcc/testsuite/gcc.target/aarch64/sve/mask_gather_load_7.c

   1 /* { dg-do assemble { target aarch64_asm_sve_ok } } */
   2 /* { dg-options "-O2 -ftree-vectorize -ffast-math --save-temps" } */
   3
   4 #include <stdint.h>
   5
   6 #define TEST_LOOP(DATA_TYPE, CMP_TYPE, INDEX_TYPE)                      \
   7   void                                                                  \
   8   f_##DATA_TYPE##_##CMP_TYPE##_##INDEX_TYPE                             \
   9     (DATA_TYPE *restrict dest, DATA_TYPE *restrict src,                 \
  10      CMP_TYPE *cmp1, CMP_TYPE *cmp2, INDEX_TYPE *indices, int n)        \
  11   {                                                                     \
  12     for (int i = 0; i < n; ++i)                                         \
  13       if (cmp1[i] == cmp2[i])                                           \
  14         dest[i] += src[indices[i]];                                     \
  15   }
  16
  17 #define TEST32(T, DATA_TYPE)                    \
  18   T (DATA_TYPE, int16_t, int32_t)               \
  19   T (DATA_TYPE, uint16_t, int32_t)              \
  20   T (DATA_TYPE, _Float16, int32_t)              \
  21   T (DATA_TYPE, int16_t, uint32_t)              \
  22   T (DATA_TYPE, uint16_t, uint32_t)             \
  23   T (DATA_TYPE, _Float16, uint32_t)
  24
  25 #define TEST64(T, DATA_TYPE)                    \
  26   T (DATA_TYPE, int32_t, int64_t)               \
  27   T (DATA_TYPE, uint32_t, int64_t)              \
  28   T (DATA_TYPE, float, int64_t)                 \
  29   T (DATA_TYPE, int32_t, uint64_t)              \
  30   T (DATA_TYPE, uint32_t, uint64_t)             \
  31   T (DATA_TYPE, float, uint64_t)
  32
  33 #define TEST_ALL(T)                     \
  34   TEST32 (T, int32_t)                   \
  35   TEST32 (T, uint32_t)                  \
  36   TEST32 (T, float)                     \
  37   TEST64 (T, int64_t)                   \
  38   TEST64 (T, uint64_t)                  \
  39   TEST64 (T, double)
  40
  41 TEST_ALL (TEST_LOOP)
  42
  43 /* { dg-final { scan-assembler-times {\tld1h\tz[0-9]+\.h, p[0-7]/z, \[x[0-9]+, x[0-9]+, lsl 1\]\n} 36 } } */
  44 /* { dg-final { scan-assembler-times {\tcmpeq\tp[0-7]\.h, p[0-7]/z, z[0-9]+\.h, z[0-9]+\.h\n} 12 } } */
  45 /* { dg-final { scan-assembler-times {\tfcmeq\tp[0-7]\.h, p[0-7]/z, z[0-9]+\.h, z[0-9]+\.h\n} 6 } } */
  46 /* { dg-final { scan-assembler-times {\tld1w\tz[0-9]+\.s, p[0-7]/z, \[x[0-9]+, z[0-9]+\.s, sxtw 2\]\n} 18 } } */
  47 /* { dg-final { scan-assembler-times {\tld1w\tz[0-9]+\.s, p[0-7]/z, \[x[0-9]+, z[0-9]+\.s, uxtw 2\]\n} 18 } } */
  48
  49 /* Also used for the TEST32 indices.  */
  50 /* { dg-final { scan-assembler-times {\tld1w\tz[0-9]+\.s, p[0-7]/z, \[x[0-9]+, x[0-9]+, lsl 2\]\n} 72 } } */
  51 /* { dg-final { scan-assembler-times {\tcmpeq\tp[0-7]\.s, p[0-7]/z, z[0-9]+\.s, z[0-9]+\.s\n} 12 } } */
  52 /* { dg-final { scan-assembler-times {\tfcmeq\tp[0-7]\.s, p[0-7]/z, z[0-9]+\.s, z[0-9]+\.s\n} 6 } } */
  53 /* { dg-final { scan-assembler-times {\tld1d\tz[0-9]+\.d, p[0-7]/z, \[x[0-9]+, z[0-9]+\.d, lsl 3\]\n} 36 } } */