On 12/16/2016 05:50 AM, Richard Biener wrote:
+ gimple *defstmt = SSA_NAME_DEF_STMT (vuse);
+ tree src2 = NULL_TREE, len2 = NULL_TREE;
+ HOST_WIDE_INT offset, offset2;
+ tree val = integer_zero_node;
+ if (gimple_store_p (defstmt)
+ && gimple_assign_single_p (defstmt)
+ && TREE_CODE (gimple_assign_rhs1 (defstmt)) == CONSTRUCTOR
+ && CONSTRUCTOR_NELTS (gimple_assign_rhs1 (defstmt)) == 0
Should be always true for stores from constructors.
Seems like I should drop similar checks from the in-progress DSE
changes. I thought I saw something in SRA to cover when this wasn't
true as well that could probably be simplified.
jeff
