Hi Bohan,
> When the iteration count is unknown, actually compare the costs of
> different VLA modes instead of blindly choosing the bigger LMUL as long
> as there's no unexpected spill. This can prevent some cases when
> -mrvv-max-lmul=dynamic emits worse code than -mrvv-max-lmul=m8, and
Could you give an example for this?
> - else
> - {
> - if (dump_enabled_p ())
> - dump_printf_loc (MSG_NOTE, vect_location,
> - "Keep current LMUL loop because"
> - " it is unknown NITERS\n");
> - return false;
> - }
I think this is OK (and I guess this should have been similar to the
check below in the first place).
> }
> }
> /* If NITERS is unknown, we should not use VLS modes to vectorize
> @@ -1277,8 +1269,8 @@ costs::better_main_loop_than_p (const vector_costs
> *uncast_other) const
> and partial vectors (VLAmodes) on loop epilogue which is very
> inefficient. Instead, we should apply partial vectors (VLAmodes)
> on loop body without an epilogue on unknown NITERS loop. */
> - else if (!LOOP_VINFO_NITERS_KNOWN_P (this_loop_vinfo)
> - && m_cost_type == VLS_VECTOR_COST)
> + if (!LOOP_VINFO_NITERS_KNOWN_P (this_loop_vinfo)
> + && m_cost_type == VLS_VECTOR_COST)
> return false;
This is partially ok but should additionally use
!LOOP_VINFO_USING_PARTIAL_VECTORS_P. We do support VLS partial
vectorization.
> +/* { dg-final { scan-assembler {e8,m1} } } */
> +/* { dg-final { scan-assembler-times {csrr} 0 } } */
scan-assembler-not
--
Regards
Robin
