gcc/testsuite/gcc.target/aarch64/sve/mask_struct_load_2.c

   1 /* { dg-do compile } */
   2 /* { dg-options "-O2 -ftree-vectorize -ffast-math" } */
   3
   4 #define TEST_LOOP(NAME, OUTTYPE, INTYPE, MASKTYPE)              \
   5   void __attribute__ ((noinline, noclone))                      \
   6   NAME##_3 (OUTTYPE *__restrict dest, INTYPE *__restrict src,   \
   7             MASKTYPE *__restrict cond, int n)                   \
   8   {                                                             \
   9     for (int i = 0; i < n; ++i)                                 \
  10       if (cond[i])                                              \
  11         dest[i] = (src[i * 3]                                   \
  12                    + src[i * 3 + 1]                             \
  13                    + src[i * 3 + 2]);                           \
  14   }
  15
  16 #define TEST2(NAME, OUTTYPE, INTYPE) \
  17   TEST_LOOP (NAME##_i8, OUTTYPE, INTYPE, signed char) \
  18   TEST_LOOP (NAME##_i16, OUTTYPE, INTYPE, unsigned short) \
  19   TEST_LOOP (NAME##_f32, OUTTYPE, INTYPE, float) \
  20   TEST_LOOP (NAME##_f64, OUTTYPE, INTYPE, double)
  21
  22 #define TEST1(NAME, OUTTYPE) \
  23   TEST2 (NAME##_i8, OUTTYPE, signed char) \
  24   TEST2 (NAME##_i16, OUTTYPE, unsigned short) \
  25   TEST2 (NAME##_i32, OUTTYPE, int) \
  26   TEST2 (NAME##_i64, OUTTYPE, unsigned long)
  27
  28 #define TEST(NAME) \
  29   TEST1 (NAME##_i8, signed char) \
  30   TEST1 (NAME##_i16, unsigned short) \
  31   TEST1 (NAME##_i32, int) \
  32   TEST1 (NAME##_i64, unsigned long) \
  33   TEST2 (NAME##_f16_f16, _Float16, _Float16) \
  34   TEST2 (NAME##_f32_f32, float, float) \
  35   TEST2 (NAME##_f64_f64, double, double)
  36
  37 TEST (test)
  38
  39 /*    Mask |  8 16 32 64
  40     -------+------------
  41     Out  8 |  1  1  1  1
  42         16 |  1  1  1  1
  43         32 |  1  1  1  1
  44         64 |  1  1  1  1.  */
  45 /* { dg-final { scan-assembler-times {\tld3b\t.z[0-9]} 16 } } */
  46
  47 /*    Mask |  8 16 32 64
  48     -------+------------
  49     Out  8 |  2  2  2  2
  50         16 |  2  1  1  1 x2 (for _Float16)
  51         32 |  2  1  1  1
  52         64 |  2  1  1  1.  */
  53 /* { dg-final { scan-assembler-times {\tld3h\t.z[0-9]} 28 } } */
  54
  55 /*    Mask |  8 16 32 64
  56     -------+------------
  57     Out  8 |  4  4  4  4
  58         16 |  4  2  2  2
  59         32 |  4  2  1  1 x2 (for float)
  60         64 |  4  2  1  1.  */
  61 /* { dg-final { scan-assembler-times {\tld3w\t.z[0-9]} 50 } } */
  62
  63 /*    Mask |  8 16 32 64
  64     -------+------------
  65     Out  8 |  8  8  8  8
  66         16 |  8  4  4  4
  67         32 |  8  4  2  2
  68         64 |  8  4  2  1 x2 (for double).  */
  69 /* { dg-final { scan-assembler-times {\tld3d\t.z[0-9]} 98 } } */