On Mon, 2025-09-22 at 13:27 +0200, Richard Biener wrote:
> 
> /* { dg-additional-options "-mcpu=power8" { target ppc64*-*-* } } */
> 
> > +/* { dg-require-effective-target vect_int} */
> > +/* { dg-require-effective-target vect_condition} */
> > +/* { dg-require-effective-target vect_shift} */
> > +
> > +#define TEST_FN(OP, CONST, NAME) void __GIMPLE
> > (ssa,guessed_local(10737416)) \
> > +NAME (int * a) \
> > +{ \
> > +  int i; \
> > +  long unsigned int _1; \
> > +  long unsigned int _2; \
> > +  int * _3; \
> > +  int _4; \
> > +  int _5; \
> > +  unsigned int _12; \
> > +  unsigned int _13; \
> > + \
> > +  __BB(2,guessed_local(10737416)): \
> > +  goto __BB3(precise(134217728)); \
> > + \
> > +  __BB(3,loop_header(1),guessed_local(1063004408)): \
> > +  i_14 = __PHI (__BB5: i_11, __BB2: 0); \
> > +  _13 = __PHI (__BB5: _12, __BB2: 512u); \
> > +  _1 = (long unsigned int) i_14; \
> > +  _2 = _1 * 4ul; \
> > +  _3 = a_9(D) + _2; \
> > +  _4 = __MEM <int> (_3); \
> > +  _5 = _4 OP CONST; \
> > +  __MEM <int> (_3) = _5; \
> > +  i_11 = i_14 + 2; \
> > +  _12 = _13 - 1u; \
> > +  if (_12 != 0u) \
> > +    goto __BB5(guessed(132861994)); \
> > +  else \
> > +    goto __BB4(guessed(1355734)); \
> > + \
> > +  __BB(5,guessed_local(1052266995)): \
> > +  goto __BB3(precise(134217728)); \
> > + \
> > +  __BB(4,guessed_local(10737416)): \
> > +  return; \
> > +} \
> > +
> > +TEST_FN(%, 2, trunc_mod_2)
> > +TEST_FN(__FLOOR_MOD, 2, floor_mod_2)
> > +TEST_FN(__FLOOR_DIV, 2, floor_div_2)
> > +
> > +TEST_FN(%, 4, trunc_mod_pow2)
> > +TEST_FN(__FLOOR_MOD, 4, floor_mod_pow2)
> > +TEST_FN(__FLOOR_DIV, 4, floor_div_pow2)
> > +
> > +TEST_FN(%, 5, trunc_mod)
> > +TEST_FN(__FLOOR_MOD, 5, floor_mod)
> > +TEST_FN(__FLOOR_DIV, 5, floor_div)
> 
> Can you make this a runtime testcase as well?  You can add a plain C
> main() exercising the GIMPLE FN above.
> 

Thanks for the review comments!
Would it be ok to update the same test case by just adding a main
function? Or create a new test file with same functions but calling
main?

Best regards,
Avinash Jayakar

> 

Reply via email to