[Bug tree-optimization/84037] [8 Regression] Speed regression of polyhedron benchmark since r256644

[hubicka at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=84037

--- Comment #17 from Jan Hubicka <hubicka at gcc dot gnu.org> ---
We already have
/* This function adjusts the unroll factor based on
   the hardware capabilities. For ex, bdver3 has
   a loop buffer which makes unrolling of smaller
   loops less important. This function decides the
   unroll factor using number of memory references
   (value 32 is used) as a heuristic. */

static unsigned
ix86_loop_unroll_adjust (unsigned nunroll, struct loop *loop)

which triggers with TARGET_ADJUST_UNROLL
/* X86_TUNE_ADJUST_UNROLL: This enables adjusting the unroll factor based       
   on hardware capabilities. Bdver3 hardware has a loop buffer which makes      
   unrolling small loop less important. For, such architectures we adjust       
   the unroll factor so that the unrolled loop fits the loop buffer.  */        
DEF_TUNE (X86_TUNE_ADJUST_UNROLL, "adjust_unroll_factor", m_BDVER3 | m_BDVER4)  

so perhaps what you propose can be done by making this one more general?