On 2013-10-10 14:07 , tsaund...@mozilla.com wrote:
This makes the implementation of stack vectors simpler and easier to use. This works by making the size of the on stack storage a template argument, so the size is embedded in the type. This allows you to implicitly convert a stack_vec<T, N> to a vec<T, va_heap> *, and it will just work. Because there's no need to support stack vectors in unions we can make them be a more normal c++ class with a constructor and destructor that are nontrivial.
Thanks. This looks much simpler, indeed. The patch is fine to commit. Just a couple of observations/questions:
@@ -638,7 +638,7 @@ propagate_threaded_block_debug_into (basic_block dest, basic_block src) i++; }- vec<tree, va_stack> fewvars = vNULL;+ stack_vec<tree, alloc_count> fewvars; pointer_set_t *vars = NULL;
Hm, what will happen now if alloc_count == 0? If I'm following the logic, this is tied to the presence of the 'vars' local, so it seems that we are fine. Otherwise, the quick_push operation will run into trouble.
/* If we're already starting with 3/4 of alloc_count, go for a@@ -646,8 +646,6 @@ propagate_threaded_block_debug_into (basic_block dest, basic_block src) VEC. */ if (i * 4 > alloc_count * 3) vars = pointer_set_create (); - else if (alloc_count) - vec_stack_alloc (tree, fewvars, alloc_count);/* Now go through the initial debug stmts in DEST again, this timeactually inserting in VARS or FEWVARS. Don't bother checking for diff --git a/gcc/tree-vect-loop-manip.c b/gcc/tree-vect-loop-manip.c index 574446a..4a14607 100644 --- a/gcc/tree-vect-loop-manip.c +++ b/gcc/tree-vect-loop-manip.c @@ -107,7 +107,7 @@ typedef struct with a PHI DEF that would soon become non-dominant, and when we got to the suitable one, it wouldn't have anything to substitute any more. */ -static vec<adjust_info, va_stack> adjust_vec; +static vec<adjust_info, va_heap> adjust_vec;
A file global was declared as a stack vector? Sigh. Diego.
