Re: [PATCH] c++: coroutines - Overlap variables in frame [PR105989]

[Iain Sandoe <iain@sandoe.co.uk>]

Hi Michal,

> On 12 Jul 2022, at 14:35, Michal Jankovič via Gcc-patches <gcc-patches@gcc.gnu.org> wrote:
> 
> Currently, coroutine frames store all variables of a coroutine separately,
> even if their lifetime does not overlap (they are in distinct scopes). This
> patch implements overlapping distinct variable scopes in the coroutine frame,
> by storing the frame fields in nested unions of structs. This lowers the size
> of the frame for larger coroutines significantly, and makes them more usable
> on systems with limited memory.

not a review (I will try to take a look at the weekend).

but … this is one of the two main optimisations on my TODO - so cool for doing it.

(the other related optimisation is to eliminate frame entries for scopes without any suspend
 points - which has the potential to save even more space for code with sparse use of co_xxxx)

Iain

> Bootstrapped and regression tested on x86_64-pc-linux-gnu; new test fails
> before the patch and succeeds after with no regressions.
> 
> 	PR c++/105989
> 
> gcc/cp/ChangeLog:
> 
> 	* coroutines.cc (struct local_var_info): Add field_access_path.
> 	(build_local_var_frame_access_expr): New.
> 	(transform_local_var_uses): Use build_local_var_frame_access_expr.
> 	(coro_make_frame_entry_id): New.
> 	(coro_make_frame_entry): Delegate to coro_make_frame_entry_id.
> 	(struct local_vars_frame_data): Add orig, field_access_path.
> 	(register_local_var_uses): Generate new frame layout. Create access
> 	paths to vars.
> 	(morph_fn_to_coro): Set new fields in local_vars_frame_data. 
> 
> gcc/testsuite/ChangeLog:
> 
> 	* g++.dg/coroutines/pr105989.C: New test.
> 
> <pr105989.patch>