PR tree-optimization/30730

	PR tree-optimization/26900
	* tree-ssa-loop-niter.c: Include gmp.h.
	(bounds): New type.
	(mpz_set_double_int, get_type_bounds, mpz_to_double_int,
	split_to_var_and_offset, determine_value_range,
	bound_difference_of_offsetted_base, refine_bounds_using_guard,
	bound_difference, bounds_add, bounds_negate,
	number_of_iterations_ne_max, dump_affine_iv): New functions.
	(number_of_iterations_ne, number_of_iterations_lt_to_ne,
	assert_loop_rolls_lt, assert_loop_rolls_le): Use bounds on the
	difference of initial and final value of control iv to validate
	results.
	(number_of_iterations_cond): Add loop parameter.  Determine bounds
	on the difference of the extremes of the control iv.  Add dumps.
	(expand_simple_operations): Handle phi nodes.
	(simplify_using_initial_conditions): Do not record used conditions.
	(number_of_iterations_exit): Pass loop to number_of_iterations_cond.
	Do not set additional_info.
	(implies_nonnegative_p, implies_ge_p): Removed.
	(derive_constant_upper_bound): Do not use parameter `additional'.
	(record_estimate): Parameter `additional' removed.  Parameter
	`i_bound' added.  Do not call derive_constant_upper_bound.
	(record_nonwrapping_iv): Use derive_constant_upper_bound to
	bound the number of iterations estimate.
	(estimate_numbers_of_iterations_loop): Pass the estimate from
	the number of iterations analysis to record_estimate.
	* tree.h (multiple_of_p): Declare.
	* tree-scalar-evolution.c (expression_expensive_p): Removed.
	(scev_const_prop): Do not check expression_expensive_p.
	* fold-const.c (multiple_of_p): Exported.
	* double-int.c (double_int_mask): Exported.
	* double-int.h (double_int_mask): Declare.
	* tree-flow.h (struct tree_niter_desc): Removed additional_info
	field.  Added max field.

	* gcc.dg/tree-ssa/loop-26.c: New test.



git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/trunk@122896 138bc75d-0d04-0410-961f-82ee72b054a4

1 files changed, 7 insertions, 12 deletions

diff --git a/gcc/tree-scalar-evolution.c b/gcc/tree-scalar-evolution.c
index f1914c3d51e..846d27414d5 100644
--- a/gcc/tree-scalar-evolution.c
+++ b/gcc/tree-scalar-evolution.c
@@ -2864,14 +2864,6 @@ scev_finalize (void)
   BITMAP_FREE (already_instantiated);
 }
 
-/* Returns true if EXPR looks expensive.  */
-
-static bool
-expression_expensive_p (tree expr)
-{
-  return force_expr_to_var_cost (expr) >= target_spill_cost;
-}
-
 /* Replace ssa names for that scev can prove they are constant by the
    appropriate constants.  Also perform final value replacement in loops,
    in case the replacement expressions are cheap.
@@ -2958,10 +2950,13 @@ scev_const_prop (void)
 	continue;
 
       niter = number_of_latch_executions (loop);
-      if (niter == chrec_dont_know
-	  /* If computing the number of iterations is expensive, it may be
-	     better not to introduce computations involving it.  */
-	  || expression_expensive_p (niter))
+      /* We used to check here whether the computation of NITER is expensive,
+	 and avoided final value elimination if that is the case.  The problem
+	 is that it is hard to evaluate whether the expression is too
+	 expensive, as we do not know what optimization opportunities the
+	 the elimination of the final value may reveal.  Therefore, we now
+	 eliminate the final values of induction variables unconditionally.  */
+      if (niter == chrec_dont_know)
 	continue;
 
       /* Ensure that it is possible to insert new statements somewhere.  */