On 12/8/22 15:01, Jason Merrill via Gcc-patches wrote:
After the previous patches, I noticed that we were putting the array of
strings into .rodata, but then memcpying it into an automatic array, which
is pointless; we should be able to use it directly.
C++ doesn't allow us to do this for the backing array of an
initializer_list, but should be able to do it for the implementation detail
array we use to construct the backing array.
This doesn't happen automatically because TREE_ADDRESSABLE is set, and
gimplify_init_constructor uses that to decide whether to promote a variable
to static. Ideally this could use escape analysis to recognize that the
address, though taken, never leaves the function; that should allow
promotion when we're only using the address for indexing within the
function, as in initlist-opt2.C.
But in initlist-opt1.C, we're passing the array address to another function,
so it definitely escapes; it's only safe in this case because it's calling a
standard library function that we know only uses it for indexing. So, a
flag seems needed. I first thought to put the flag on the TARGET_EXPR, but
the VAR_DECL seems more appropriate.
Bikeshedding, or other approaches, welcome.
PR c++/105838
gcc/ChangeLog:
* tree.h (DECL_NOT_OBSERVABLE): New.
* tree-core.h (struct tree_decl_common): Mention it.
* gimplify.cc (gimplify_init_constructor): Check it.
gcc/cp/ChangeLog:
* call.cc (maybe_init_list_as_array): Set DECL_NOT_OBSERVABLE.
gcc/testsuite/ChangeLog:
* g++.dg/tree-ssa/initlist-opt1.C: Check for static array.
* g++.dg/tree-ssa/initlist-opt2.C: Likewise.
Presumably the DECL in question isn't actually compiler generated and
thus DECL_ARTIFICIAL is not appropriate here. Assuming that's the case,
should we be setting DECL_NON_OBSERVABLE on VAR_DECLs with
DECL_ARTIFICIAL set? I don't think that needs to be done for this patch
to move forward though.
There haven't been any objections, so OK.
Jeff
