On 3/18/25 08:36, Paul-Antoine Arras wrote:
This patch partially enables use of the OpenMP interop construct by adding
middle end support, mostly in the omplower pass, and in the target-independent
part of the libgomp runtime. It follows up on previous patches for C, C++ and
Fortran front ends support. The full interop feature requires another patch to
enable foreign runtime support in libgomp plugins.
gcc/ChangeLog:
* builtin-types.def
(BT_FN_VOID_INT_INT_PTR_PTR_PTR_INT_PTR_INT_PTR_UINT_PTR): New.
* gimple-low.cc (lower_stmt): Handle GIMPLE_OMP_INTEROP.
* gimple-pretty-print.cc (dump_gimple_omp_interop): New function.
(pp_gimple_stmt_1): Handle GIMPLE_OMP_INTEROP.
* gimple.cc (gimple_build_omp_interop): New function.
(gimple_copy): Handle GIMPLE_OMP_INTEROP.
* gimple.def (GIMPLE_OMP_INTEROP): Define.
* gimple.h (gimple_build_omp_interop): Declare.
(gimple_omp_interop_clauses): New function.
(gimple_omp_interop_clauses_ptr): Likewise.
(gimple_omp_interop_set_clauses): Likewise.
(gimple_return_set_retval): Handle GIMPLE_OMP_INTEROP.
* gimplify.cc (gimplify_scan_omp_clauses): Handle OMP_CLAUSE_INIT,
OMP_CLAUSE_USE and OMP_CLAUSE_DESTROY.
(gimplify_omp_interop): New function.
(gimplify_expr): Replace sorry with call to gimplify_omp_interop.
* omp-builtins.def (BUILT_IN_GOMP_INTEROP): Define.
* omp-low.cc (scan_sharing_clauses): Handle OMP_CLAUSE_INIT,
OMP_CLAUSE_USE and OMP_CLAUSE_DESTROY.
(scan_omp_1_stmt): Handle GIMPLE_OMP_INTEROP.
(lower_omp_interop_action_clauses): New function.
(lower_omp_interop): Likewise.
(lower_omp_1): Handle GIMPLE_OMP_INTEROP.
gcc/c/ChangeLog:
* c-parser.cc (c_parser_omp_clause_destroy): Make addressable.
(c_parser_omp_clause_init): Make addressable.
gcc/cp/ChangeLog:
* parser.cc (cp_parser_omp_clause_init): Make addressable.
gcc/fortran/ChangeLog:
* trans-openmp.cc (gfc_trans_omp_clauses): Make OMP_CLAUSE_DESTROY and
OMP_CLAUSE_INIT addressable.
* types.def (BT_FN_VOID_INT_INT_PTR_PTR_PTR_INT_PTR_INT_PTR_UINT_PTR):
New.
include/ChangeLog:
* gomp-constants.h (GOMP_DEVICE_DEFAULT_OMP_61): Define.
(GOMP_INTEROP_FLAG_TARGET): Define.
(GOMP_INTEROP_FLAG_TARGETSYNC): Define.
libgomp/ChangeLog:
* icv-device.c (omp_set_default_device): Check
GOMP_DEVICE_DEFAULT_OMP_61.
* libgomp-plugin.h (struct interop_obj_t): New.
(enum gomp_interop_flag): New.
(GOMP_OFFLOAD_interop): Declare.
(GOMP_OFFLOAD_get_interop_int): Declare.
(GOMP_OFFLOAD_get_interop_ptr): Declare.
(GOMP_OFFLOAD_get_interop_str): Declare.
(GOMP_OFFLOAD_get_interop_type_desc): Declare.
* libgomp.h (_LIBGOMP_OMP_LOCK_DEFINED): Define.
(struct gomp_device_descr): Add interop_func, get_interop_int_func,
get_interop_ptr_func, get_interop_str_func, get_interop_type_desc_func.
* libgomp.map: Add GOMP_interop.
* libgomp_g.h (GOMP_interop): Declare.
* target.c (resolve_device): Handle GOMP_DEVICE_DEFAULT_OMP_61.
(omp_get_interop_int): Replace stub with actual implementation.
(omp_get_interop_ptr): Likewise.
(omp_get_interop_str): Likewise.
(omp_get_interop_type_desc): Likewise.
(struct interop_data_t): Define.
(gomp_interop_internal): New function.
(GOMP_interop): Likewise.
(gomp_load_plugin_for_device): Load symbols for get_interop_int,
get_interop_ptr, get_interop_str and get_interop_type_desc.
* testsuite/libgomp.c-c++-common/interop-1.c: New test.
gcc/testsuite/ChangeLog:
* c-c++-common/gomp/interop-1.c: Remove dg-prune-output "sorry".
* c-c++-common/gomp/interop-2.c: Likewise.
* c-c++-common/gomp/interop-3.c: Likewise.
* c-c++-common/gomp/interop-4.c: Remove dg-message "not supported".
* g++.dg/gomp/interop-5.C: Likewise.
* gfortran.dg/gomp/interop-4.f90: Likewise.
* c-c++-common/gomp/interop-5.c: New test.
* gfortran.dg/gomp/interop-5.f90: New test.
I have a general complaint, which applies most particularly to this
piece in libgomp/target.c:
+
+void
+GOMP_interop (int device_num, int n_init, void *init,
+ const void *target_targetsync, const void *prefer_type,
+ int n_use, void *use, int n_destroy, void *destroy,
+ unsigned int nowait, void **depend)
+{
The GNU coding standards say:
"Please put a comment on each function saying what the function does,
what sorts of arguments it gets, and what the possible values of
arguments mean and are used for. It is not necessary to duplicate in
words the meaning of the C argument declarations, if a C type is being
used in its customary fashion. If there is anything nonstandard about
its use (such as an argument of type char * which is really the address
of the second character of a string, not the first), or any possible
values that would not work the way one would expect (such as, that
strings containing newlines are not guaranteed to work), be sure to say so."
Here we certainly have nonstandard use of C types (those void *
arguments, which might either be a pointer to a thing or a pointer to an
array of pointers to things, at least in the case of prefer_type the
thing turns out to be some sort of byte encoding represented in either a
char * or char **). On top of that, this routine is an ABI between the
compiler and library and is especially important to document for that
reason as well.
I'm also thinking this is a pretty messy interface for an ABI... I'm
not sure where the bottlenecks are but it seems like having code gen for
the OpenMP constructs that call this routine create a stack variable to
hold the "pointer to thing" and uniformly making the interface "thing**"
would not add as much overhead as needing to handle both "thing*" and
"thing **" at runtime. Plus there is a benefit to using the actual type
of the argument in the function declaration in terms of making the code
on both the GCC and libgomp sides simpler, more understandable, and and
more maintainable.
-Sandra
