If there is nothing to map, skip the mapping and avoid attempting to
copy 0 bytes from addrs, sizes and kinds.
Additionally, it could happen that a non-allocated address was deallocated,
such as a pointer set, leading to a free for the actual data.
libgomp/
* target.c (gomp_target_rev): Handle mapnum == 0 and avoid
freeing not allocated memory.
* testsuite/libgomp.fortran/reverse-offload-6.f90: New test.
expand_omp_for_init_counts was using for the case where collapse(2)
inner loop has init expression dependent on non-constant multiple of
the outer iterator and the condition upper bound expression doesn't
depend on the outer iterator fold_unary (NEGATE_EXPR, ...). This
will just return NULL if it can't be folded, we need fold_build1
instead.
2023-01-19 Jakub Jelinek <jakub@redhat.com>
PR middle-end/108459
* omp-expand.cc (expand_omp_for_init_counts): Use fold_build1 rather
than fold_unary for NEGATE_EXPR.
* testsuite/libgomp.c/pr108459.c: New test.
The comment in the loop says that we shouldn't add a map clause if such
a clause exists already, but the loop was actually using OMP_CLAUSE_DECL
on any clause. Target construct can have various clauses which don't
have OMP_CLAUSE_DECL at all (e.g. nowait, device or if) or clause
where it means something different (e.g. privatization clauses, allocate,
depend).
So, only check OMP_CLAUSE_DECL on OMP_CLAUSE_MAP clauses.
2023-01-05 Jakub Jelinek <jakub@redhat.com>
PR c++/108286
* semantics.cc (finish_omp_target_clauses): Ignore clauses other than
OMP_CLAUSE_MAP.
* testsuite/libgomp.c++/pr108286.C: New test.
DECL_OMP_PRIVATIZED_MEMBER vars are artificial vars with DECL_VALUE_EXPR
of this->field used just during gimplification and omp lowering/expansion
to privatize individual fields in methods when needed.
As the following testcase shows, when not in templates, they were handled
right, but in templates we actually called cp_finish_decl on them and
that can result in their destruction, which is obviously undesirable,
we should only destruct the privatized copies of them created in omp
lowering.
Fixed thusly.
2022-12-21 Jakub Jelinek <jakub@redhat.com>
PR c++/108180
* pt.cc (tsubst_expr): Don't call cp_finish_decl on
DECL_OMP_PRIVATIZED_MEMBER vars.
* testsuite/libgomp.c++/pr108180.C: New test.
Commit r13-4716-ge205ec03f0794aeac3e8a89e947c12624d5a274e accidentally
included a testcase of another patch that is pending review:
https://gcc.gnu.org/pipermail/gcc-patches/2022-December/608401.html
libgomp/
PR libfortran/108056
* testsuite/libgomp.fortran/allocate-4.f90: Remove
accidentally added file.
Since GCC 12, the conversion between the array descriptors formats - the
internal (GFC) and the C binding one (CFI) - moved to the compiler itself
such that the cfi_desc_to_gfc_desc/gfc_desc_to_cfi_desc functions are only
used with older code (GCC 9 to 11). The newly added checks caused asserts
as older code did not pass the proper values (e.g. real(4) as effective
argument arrived as BT_ASSUME type as the effective type got lost inbetween).
As proposed in the PR, revert to the GCC 11 version - known bugs is better
than some fixes and new issues. Still, GCC 12 is much better in terms of
TS29113 support and should really be used.
This patch uses the current libgomp version of the GCC 11 branch, except
it fixes the GFC version number (which is 0), uses calloc instead of malloc,
and sets the lower bound to 1 instead of keeping it as is for
CFI_attribute_other.
libgfortran/ChangeLog:
PR libfortran/108056
* runtime/ISO_Fortran_binding.c (cfi_desc_to_gfc_desc,
gfc_desc_to_cfi_desc): Mostly revert to GCC 11 version for
those backward-compatiblity-only functions.
This patch fixes a case where a combined directive (e.g. "!$omp target
parallel ...") contains both a map and a firstprivate clause for the
same variable. When the combined directive is split into two nested
directives, the outer "target" gets the "map" clause, and the inner
"parallel" gets the "firstprivate" clause, like so:
!$omp target parallel map(x) firstprivate(x)
-->
!$omp target map(x)
!$omp parallel firstprivate(x)
...
When there is no map of the same variable, the firstprivate is distributed
to both directives, e.g. for 'y' in:
!$omp target parallel map(x) firstprivate(y)
-->
!$omp target map(x) firstprivate(y)
!$omp parallel firstprivate(y)
...
This is not a recent regression, but appear to fix a long-standing ICE.
(The included testcase is based on one by Tobias.)
2022-12-06 Julian Brown <julian@codesourcery.com>
gcc/fortran/
* trans-openmp.cc (gfc_add_firstprivate_if_unmapped): New function.
(gfc_split_omp_clauses): Call above.
libgomp/
* testsuite/libgomp.fortran/combined-directive-splitting-1.f90: New
test.
This commit enabled reverse offload for nvptx such that gomp_target_rev
actually gets called. And it fills the latter function to do all of
the following: finding the host function to the device func ptr and
copying the arguments to the host, processing the mapping/firstprivate,
calling the host function, copying back the data and freeing as needed.
The data handling is made easier by assuming that all host variables
either existed before (and are in the mapping) or that those are
devices variables not yet available on the host. Thus, the reverse
mapping can do without refcounts etc. Note that the spec disallows
inside a target region device-affecting constructs other than target
plus ancestor device-modifier and it also limits the clauses permitted
on this construct.
For the function addresses, an additional splay tree is used; for
the lookup of mapped variables, the existing splay-tree is used.
Unfortunately, its data structure requires a full walk of the tree;
Additionally, the just mapped variables are recorded in a separate
data structure an extra lookup. While the lookup is slow, assuming
that only few variables get mapped in each reverse offload construct
and that reverse offload is the exception and not performance critical,
this seems to be acceptable.
libgomp/ChangeLog:
* libgomp.h (struct target_mem_desc): Predeclare; move
below after 'reverse_splay_tree_node' and add rev_array
member.
(struct reverse_splay_tree_key_s, reverse_splay_compare): New.
(reverse_splay_tree_node, reverse_splay_tree,
reverse_splay_tree_key): New typedef.
(struct gomp_device_descr): Add mem_map_rev member.
* oacc-host.c (host_dispatch): NULL init .mem_map_rev.
* plugin/plugin-nvptx.c (GOMP_OFFLOAD_get_num_devices): Claim
support for GOMP_REQUIRES_REVERSE_OFFLOAD.
* splay-tree.h (splay_tree_callback_stop): New typedef; like
splay_tree_callback but returning int not void.
(splay_tree_foreach_lazy): Define; like splay_tree_foreach but
taking splay_tree_callback_stop as argument.
* splay-tree.c (splay_tree_foreach_internal_lazy,
splay_tree_foreach_lazy): New; but early exit if callback returns
nonzero.
* target.c: Instatiate splay_tree_c with splay_tree_prefix 'reverse'.
(gomp_map_lookup_rev): New.
(gomp_load_image_to_device): Handle reverse-offload function
lookup table.
(gomp_unload_image_from_device): Free devicep->mem_map_rev.
(struct gomp_splay_tree_rev_lookup_data, gomp_splay_tree_rev_lookup,
gomp_map_rev_lookup, struct cpy_data, gomp_map_cdata_lookup_int,
gomp_map_cdata_lookup): New auxiliary structs and functions for
gomp_target_rev.
(gomp_target_rev): Implement reverse offloading and its mapping.
(gomp_target_init): Init current_device.mem_map_rev.root.
* testsuite/libgomp.fortran/reverse-offload-2.f90: New test.
* testsuite/libgomp.fortran/reverse-offload-3.f90: New test.
* testsuite/libgomp.fortran/reverse-offload-4.f90: New test.
* testsuite/libgomp.fortran/reverse-offload-5.f90: New test.
* testsuite/libgomp.fortran/reverse-offload-5a.f90: New test without
mapping of on-device allocated variables.
This patch adds support for omp_get_max_teams, omp_set_num_teams, and
omp_{gs}et_teams_thread_limit on offload devices. That includes the usage of
device-specific ICV values (specified as environment variables or changed on a
device). In order to reuse device-specific ICV values, a copy back mechanism is
implemented that copies ICV values back from device to the host.
Additionally, a limitation of the number of teams on gcn offload devices is
implemented. The number of teams is limited by twice the number of compute
units (one team is executed on one compute unit). This avoids queueing
unnessecary many teams and a corresponding allocation of large amounts of
memory. Without that limitation the memory allocation for a large number of
user-specified teams can result in an "memory access fault".
A limitation of the number of teams is already also implemented for nvptx
devices (see nvptx_adjust_launch_bounds in libgomp/plugin/plugin-nvptx.c).
gcc/ChangeLog:
* gimplify.cc (optimize_target_teams): Set initial num_teams_upper
to "-2" instead of "1" for non-existing num_teams clause in order to
disambiguate from the case of an existing num_teams clause with value 1.
libgomp/ChangeLog:
* config/gcn/icv-device.c (omp_get_teams_thread_limit): Added to
allow processing of device-specific values.
(omp_set_teams_thread_limit): Likewise.
(ialias): Likewise.
* config/nvptx/icv-device.c (omp_get_teams_thread_limit): Likewise.
(omp_set_teams_thread_limit): Likewise.
(ialias): Likewise.
* icv-device.c (omp_get_teams_thread_limit): Likewise.
(ialias): Likewise.
(omp_set_teams_thread_limit): Likewise.
* icv.c (omp_set_teams_thread_limit): Removed.
(omp_get_teams_thread_limit): Likewise.
(ialias): Likewise.
* libgomp.texi: Updated documentation for nvptx and gcn corresponding
to the limitation of the number of teams.
* plugin/plugin-gcn.c (limit_teams): New helper function that limits
the number of teams by twice the number of compute units.
(parse_target_attributes): Limit the number of teams on gcn offload
devices.
* target.c (get_gomp_offload_icvs): Added teams_thread_limit_var
handling.
(gomp_load_image_to_device): Added a size check for the ICVs struct
variable.
(gomp_copy_back_icvs): New function that is used in GOMP_target_ext to
copy back the ICV values from device to host.
(GOMP_target_ext): Update the number of teams and threads in the kernel
args also considering device-specific values.
* testsuite/libgomp.c-c++-common/icv-4.c: Fixed an error in the reading
of OMP_TEAMS_THREAD_LIMIT from the environment.
* testsuite/libgomp.c-c++-common/icv-5.c: Extended.
* testsuite/libgomp.c-c++-common/icv-6.c: Extended.
* testsuite/libgomp.c-c++-common/icv-7.c: Extended.
* testsuite/libgomp.c-c++-common/icv-9.c: New test.
* testsuite/libgomp.fortran/icv-5.f90: New test.
* testsuite/libgomp.fortran/icv-6.f90: New test.
gcc/testsuite/ChangeLog:
* c-c++-common/gomp/target-teams-1.c: Adapt expected values for
num_teams from "1" to "-2" in cases without num_teams clause.
* g++.dg/gomp/target-teams-1.C: Likewise.
* gfortran.dg/gomp/defaultmap-4.f90: Likewise.
* gfortran.dg/gomp/defaultmap-5.f90: Likewise.
* gfortran.dg/gomp/defaultmap-6.f90: Likewise.
Add support for gfx803 as an alias for fiji.
Add test cases for all supported 'isa' values.
gcc/ChangeLog:
* config/gcn/gcn.cc (gcn_omp_device_kind_arch_isa): Add gfx803.
* config/gcn/t-omp-device: Add gfx803.
libgomp/ChangeLog:
* testsuite/libgomp.c/declare-variant-4-fiji.c: New test.
* testsuite/libgomp.c/declare-variant-4-gfx803.c: New test.
* testsuite/libgomp.c/declare-variant-4-gfx900.c: New test.
* testsuite/libgomp.c/declare-variant-4-gfx906.c: New test.
* testsuite/libgomp.c/declare-variant-4-gfx908.c: New test.
* testsuite/libgomp.c/declare-variant-4-gfx90a.c: New test.
* testsuite/libgomp.c/declare-variant-4.h: New header file.
This patch causes the IPA simdclone pass to generate clones for
functions with the "omp declare target" attribute as if they had
"omp declare simd", provided the function appears to be suitable for
SIMD execution. The filter is conservative, rejecting functions
that write memory or that call other functions not known to be safe.
A new option -fopenmp-target-simd-clone is added to control this
transformation; it's enabled for offload processing at -O2 and higher.
gcc/ChangeLog:
* common.opt (fopenmp-target-simd-clone): New option.
(target_simd_clone_device): New enum to go with it.
* doc/invoke.texi (-fopenmp-target-simd-clone): Document.
* flag-types.h (enum omp_target_simd_clone_device_kind): New.
* omp-simd-clone.cc (auto_simd_fail): New function.
(auto_simd_check_stmt): New function.
(plausible_type_for_simd_clone): New function.
(ok_for_auto_simd_clone): New function.
(simd_clone_create): Add force_local argument, make the symbol
have internal linkage if it is true.
(expand_simd_clones): Also check for cloneable functions with
"omp declare target". Pass explicit_p argument to
simd_clone.compute_vecsize_and_simdlen target hook.
* opts.cc (default_options_table): Add -fopenmp-target-simd-clone.
* target.def (TARGET_SIMD_CLONE_COMPUTE_VECSIZE_AND_SIMDLEN):
Add bool explicit_p argument.
* doc/tm.texi: Regenerated.
* config/aarch64/aarch64.cc
(aarch64_simd_clone_compute_vecsize_and_simdlen): Update.
* config/gcn/gcn.cc
(gcn_simd_clone_compute_vecsize_and_simdlen): Update.
* config/i386/i386.cc
(ix86_simd_clone_compute_vecsize_and_simdlen): Update.
gcc/testsuite/ChangeLog:
* g++.dg/gomp/target-simd-clone-1.C: New.
* g++.dg/gomp/target-simd-clone-2.C: New.
* gcc.dg/gomp/target-simd-clone-1.c: New.
* gcc.dg/gomp/target-simd-clone-2.c: New.
* gcc.dg/gomp/target-simd-clone-3.c: New.
* gcc.dg/gomp/target-simd-clone-4.c: New.
* gcc.dg/gomp/target-simd-clone-5.c: New.
* gcc.dg/gomp/target-simd-clone-6.c: New.
* gcc.dg/gomp/target-simd-clone-7.c: New.
* gcc.dg/gomp/target-simd-clone-8.c: New.
* lib/scanoffloadipa.exp: New.
libgomp/ChangeLog:
* testsuite/lib/libgomp.exp: Load scanoffloadipa.exp library.
* testsuite/libgomp.c/target-simd-clone-1.c: New.
* testsuite/libgomp.c/target-simd-clone-2.c: New.
* testsuite/libgomp.c/target-simd-clone-3.c: New.
OpenMP permits that a 'target device(ancestor:1)' is called without being
enclosed in a target region - using the current device (i.e. the host) in
that case. This commit adds a testcase for this.
In case of nvptx, the missing on-device 'GOMP_target_ext' call causes that
it and also the associated on-device GOMP_REV_OFFLOAD_VAR variable are not
linked in from nvptx's libgomp.a. Thus, handle the failing cuModuleGetGlobal
gracefully by disabling reverse offload and assuming that the failure is fine.
libgomp/ChangeLog:
* plugin/plugin-nvptx.c (GOMP_OFFLOAD_load_image): Use unsigned int
for 'i' to match 'fn_entries'; regard absent GOMP_REV_OFFLOAD_VAR
as valid and the code having no reverse-offload code.
* testsuite/libgomp.c-c++-common/reverse-offload-2.c: New test.
OpenMP 5.0 permits to use arrays with derived type components for the list
items to the 'from'/'to' clauses of the 'target update' directive.
gcc/fortran/ChangeLog:
* openmp.cc (gfc_match_omp_clauses): Permit derived types for
the 'to' and 'from' clauses of 'target update'.
* trans-openmp.cc (gfc_trans_omp_clauses): Fixes for
derived-type changes; fix size for scalars.
libgomp/ChangeLog:
* testsuite/libgomp.fortran/target-11.f90: New test.
* testsuite/libgomp.fortran/target-13.f90: New test.
... which is 'libgomp.oacc-fortran/declare-allocatable-1.f90' adjusted
for missing support for OpenACC "Changes from Version 2.0 to 2.5":
"The 'declare create' directive with a Fortran 'allocatable' has new behavior".
Thus, after 'allocate'/before 'deallocate', call 'acc_create'/'acc_delete'
manually.
libgomp/
* testsuite/libgomp.oacc-fortran/declare-allocatable-1-runtime.f90:
New.
This patch prevents compiler-generated artificial variables from being
treated as privatization candidates for OpenACC.
The rationale is that e.g. "gang-private" variables actually must be
shared by each worker and vector spawned within a particular gang, but
that sharing is not necessary for any compiler-generated variable (at
least at present, but no such need is anticipated either). Variables on
the stack (and machine registers) are already private per-"thread"
(gang, worker and/or vector), and that's fine for artificial variables.
We're restricting this to blocks, as we still need to understand what it
means for a 'DECL_ARTIFICIAL' to appear in a 'private' clause.
Several tests need their scan output patterns adjusted to compensate.
2022-10-14 Julian Brown <julian@codesourcery.com>
PR middle-end/90115
gcc/
* omp-low.cc (oacc_privatization_candidate_p): Artificial vars are not
privatization candidates.
libgomp/
* testsuite/libgomp.oacc-fortran/declare-1.f90: Adjust scan output.
* testsuite/libgomp.oacc-fortran/host_data-5.F90: Likewise.
* testsuite/libgomp.oacc-fortran/if-1.f90: Likewise.
* testsuite/libgomp.oacc-fortran/print-1.f90: Likewise.
* testsuite/libgomp.oacc-fortran/privatized-ref-2.f90: Likewise.
Co-authored-by: Thomas Schwinge <thomas@codesourcery.com>
That is, adjust for optimization introduced with recent
commit r13-3217-gc4d15dddf6b9eacb36f535807ad2ee364af46e04
"[PR107195] Set range to zero when nonzero mask is 0", where GCC now
understands that after 'r *= 2;', 'r & 1' will never hold here, and thus
transforms/optimizes/"disturbs" the original code such that GCC/nvptx's later
"Neuter whole SESE regions" optimization no longer is applicable to it:
UNSUPPORTED: libgomp.oacc-c/../libgomp.oacc-c-c++-common/nvptx-sese-1.c -DACC_DEVICE_TYPE_nvidia=1 -DACC_MEM_SHARED=0 -foffload=nvptx-none -O0
PASS: libgomp.oacc-c/../libgomp.oacc-c-c++-common/nvptx-sese-1.c -DACC_DEVICE_TYPE_nvidia=1 -DACC_MEM_SHARED=0 -foffload=nvptx-none -O2 (test for excess errors)
PASS: libgomp.oacc-c/../libgomp.oacc-c-c++-common/nvptx-sese-1.c -DACC_DEVICE_TYPE_nvidia=1 -DACC_MEM_SHARED=0 -foffload=nvptx-none -O2 execution test
[-PASS:-]{+FAIL:+} libgomp.oacc-c/../libgomp.oacc-c-c++-common/nvptx-sese-1.c -DACC_DEVICE_TYPE_nvidia=1 -DACC_MEM_SHARED=0 -foffload=nvptx-none -O2 scan-nvptx-none-offload-rtl-dump mach "SESE regions:.* [0-9]+{[0-9]+->[0-9]+(\\.[0-9]+)+}"
Same for C++.
It's unclear to me if this is an actual "problem", which optimization is "more
important", so I've filed PR107344 "GCC/nvptx SESE region optimization" to
capture this question, and here restore what we intend to be testing (to my
understanding) in 'libgomp.oacc-c-c++-common/nvptx-sese-1.c'.
PR tree-optimization/107195
PR target/107344
libgomp/
* testsuite/libgomp.oacc-c-c++-common/nvptx-sese-1.c: Restore SESE
regions checking.
Duplicate libgomp.c-c++-common/requires-4.c (as ...-4a.c) but
with using a heap-allocated instead of static memory for a variable.
This change and the added offload_device_gcn check prepare for
pseudo-USM, where the device hardware cannot access all host
memory but only managed and pinned memory; for those, requires-4.c
will fail and the new check permits to add
target { ! { offload_device_nvptx || offload_device_gcn } }
to requires-4.c; however, it has not been added yet as pseuo-USM
support is not yet on mainline. (Review is pending for the USM
patches.)
include/ChangeLog:
* gomp-constants.h (GOMP_DEVICE_HSA): Comment out unused define.
libgomp/ChangeLog:
* testsuite/lib/libgomp.exp (check_effective_target_offload_device_gcn):
New.
* testsuite/libgomp.c-c++-common/on_device_arch.h (device_arch_gcn,
on_device_arch_gcn): New.
* testsuite/libgomp.c-c++-common/requires-4a.c: New test; copied from
requires-4.c but using heap-allocated memory.
After commit r13-3404-g7c55755d4c760de326809636531478fd7419e1e5
"amdgcn: Use FLAT addressing for all functions with pointer arguments [PR105421]",
"big" private data now works for GCN offloading, too.
PR target/105421
libgomp/
* testsuite/libgomp.oacc-c-c++-common/private-big-1.c: New.
That is, '-mptx=_' is only valid in '-foffload-options=nvptx-none', too.
Fix test case added in recent
commit r13-2625-g6b43f556f392a7165582aca36a19fe7389d995b2 "nvptx/mkoffload.cc:
Warn instead of error when reverse offload is not possible".
libgomp/
* testsuite/libgomp.c/reverse-offload-sm30.c: Fix nvptx-specific
'-foffload-options' syntax.
Fortranized testcases of commits r13-3257-ga58a965eb73
and r13-3258-g0ec4e93fb9f.
libgomp/ChangeLog:
* testsuite/libgomp.fortran/task-7.f90: New test.
* testsuite/libgomp.fortran/task-8.f90: New test.
* testsuite/libgomp.fortran/task-in-explicit-1.f90: New test.
* testsuite/libgomp.fortran/task-in-explicit-2.f90: New test.
* testsuite/libgomp.fortran/task-in-explicit-3.f90: New test.
* testsuite/libgomp.fortran/task-reduction-17.f90: New test.
* testsuite/libgomp.fortran/task-reduction-18.f90: New test.
This is pretty straightforward, if gomp_thread ()->task is NULL,
it can't be explicit task, otherwise if
gomp_thread ()->task->kind == GOMP_TASK_IMPLICIT, it is an implicit
task, otherwise explicit task.
2022-10-12 Jakub Jelinek <jakub@redhat.com>
* omp.h.in (omp_in_explicit_task): Declare.
* omp_lib.h.in (omp_in_explicit_task): Likewise.
* omp_lib.f90.in (omp_in_explicit_task): New interface.
* libgomp.map (OMP_5.2): New symbol version, export
omp_in_explicit_task and omp_in_explicit_task_.
* task.c (omp_in_explicit_task): New function.
* fortran.c (omp_in_explicit_task): Add ialias_redirect.
(omp_in_explicit_task_): New function.
* libgomp.texi (OpenMP 5.2): Mark omp_in_explicit_task as implemented.
* testsuite/libgomp.c-c++-common/task-in-explicit-1.c: New test.
* testsuite/libgomp.c-c++-common/task-in-explicit-2.c: New test.
* testsuite/libgomp.c-c++-common/task-in-explicit-3.c: New test.
When not in explicit parallel/target/teams construct, we in some cases create
an artificial parallel with a single thread (either to handle target nowait
or for task reduction purposes). In those cases, it handled again artificially
created implicit task (created by gomp_new_icv for cases where we needed to write
to some ICVs), but as the testcases show, didn't take into account possibility
of this being done from explicit task(s). The code would destroy/free the previous
task and replace it with the new implicit task. If task is an explicit task
(when teams is NULL, all explicit tasks behave like if (0)), it is a pointer to
a local stack variable, so freeing it doesn't work, and additionally we shouldn't
lose the explicit tasks - the new implicit task should instead replace the
ancestor task which is the first implicit one.
2022-10-12 Jakub Jelinek <jakub@redhat.com>
* task.c (gomp_create_artificial_team): Fix up handling of invocations
from within explicit task.
* target.c (GOMP_target_ext): Likewise.
* testsuite/libgomp.c/task-7.c: New test.
* testsuite/libgomp.c/task-8.c: New test.
* testsuite/libgomp.c-c++-common/task-reduction-17.c: New test.
* testsuite/libgomp.c-c++-common/task-reduction-18.c: New test.
OpenMP 5.1 added has_device_addr and relaxed the restrictions for
use_device_ptr, including processing non-type(c_ptr) arguments as
if has_device_addr was used. (There is a semantic difference.)
For completeness, the likewise change was done for 'use_device_ptr',
where non-type(c_ptr) arguments now use use_device_addr.
Finally, a warning for 'device(omp_{initial,invalid}_device)' was
silenced on the way as affecting the new testcase.
PR fortran/105318
gcc/fortran/ChangeLog:
* openmp.cc (resolve_omp_clauses): Update is_device_ptr restrictions
for OpenMP 5.1 and map to has_device_addr where applicable; map
use_device_ptr to use_device_addr where applicable.
Silence integer-range warning for device(omp_{initial,invalid}_device).
libgomp/ChangeLog:
* testsuite/libgomp.fortran/is_device_ptr-2.f90: New test.
gcc/testsuite/ChangeLog:
* gfortran.dg/gomp/is_device_ptr-1.f90: Remove dg-error.
* gfortran.dg/gomp/is_device_ptr-2.f90: Likewise.
* gfortran.dg/gomp/is_device_ptr-3.f90: Update tree-scan-dump.
This patch changes c_tree_equal to work more like cp_tree_equal, be
more strict in what it accepts. The ICE on the first testcase was
due to INTEGER_CST wi::wide (t1) == wi::wide (t2) comparison which
ICEs if the two constants have different precision, but as the second
testcase shows, being too lenient in it can also lead to miscompilation
of valid OpenMP programs where we think certain expression is the same
even when it isn't and can be guaranteed at runtime to represent different
memory location. So, the patch looks through only NON_LVALUE_EXPRs
and for constants as well as casts requires that the types match before
actually comparing the constant values or recursing on the cast operands.
2022-09-24 Jakub Jelinek <jakub@redhat.com>
PR c/106981
gcc/c/
* c-typeck.cc (c_tree_equal): Only strip NON_LVALUE_EXPRs at the
start. For CONSTANT_CLASS_P or CASE_CONVERT: return false if t1 and
t2 have different types.
gcc/testsuite/
* c-c++-common/gomp/pr106981.c: New test.
libgomp/
* testsuite/libgomp.c-c++-common/pr106981.c: New test.
This patch refactors struct sibling-list processing in gimplify.cc, and
adjusts some related mapping-clause processing in the Fortran FE and
omp-low.cc accordingly.
2022-09-13 Julian Brown <julian@codesourcery.com>
gcc/fortran/
* trans-openmp.cc (gfc_trans_omp_clauses): Don't create
GOMP_MAP_TO_PSET mappings for class metadata, nor GOMP_MAP_POINTER
mappings for POINTER_TYPE_P decls.
gcc/
* gimplify.cc (gimplify_omp_var_data): Remove GOVD_MAP_HAS_ATTACHMENTS.
(GOMP_FIRSTPRIVATE_IMPLICIT): Renumber.
(insert_struct_comp_map): Refactor function into...
(build_omp_struct_comp_nodes): This new function. Remove list handling
and improve self-documentation.
(extract_base_bit_offset): Remove BASE_REF, OFFSETP parameters. Move
code to strip outer parts of address out of function, but strip no-op
conversions.
(omp_mapping_group): Add DELETED field for use during reindexing.
(omp_strip_components_and_deref, omp_strip_indirections): New functions.
(omp_group_last, omp_group_base): Add GOMP_MAP_STRUCT handling.
(omp_gather_mapping_groups): Initialise DELETED field for new groups.
(omp_index_mapping_groups): Notice DELETED groups when (re)indexing.
(omp_siblist_insert_node_after, omp_siblist_move_node_after,
omp_siblist_move_nodes_after, omp_siblist_move_concat_nodes_after): New
helper functions.
(omp_accumulate_sibling_list): New function to build up GOMP_MAP_STRUCT
node groups for sibling lists. Outlined from gimplify_scan_omp_clauses.
(omp_build_struct_sibling_lists): New function.
(gimplify_scan_omp_clauses): Remove struct_map_to_clause,
struct_seen_clause, struct_deref_set. Call
omp_build_struct_sibling_lists as pre-pass instead of handling sibling
lists in the function's main processing loop.
(gimplify_adjust_omp_clauses_1): Remove GOVD_MAP_HAS_ATTACHMENTS
handling, unused now.
* omp-low.cc (scan_sharing_clauses): Handle pointer-type indirect
struct references, and references to pointers to structs also.
gcc/testsuite/
* g++.dg/goacc/member-array-acc.C: New test.
* g++.dg/gomp/member-array-omp.C: New test.
* g++.dg/gomp/target-3.C: Update expected output.
* g++.dg/gomp/target-lambda-1.C: Likewise.
* g++.dg/gomp/target-this-2.C: Likewise.
* c-c++-common/goacc/deep-copy-arrayofstruct.c: Move test from here.
* c-c++-common/gomp/target-50.c: New test.
libgomp/
* testsuite/libgomp.oacc-c-c++-common/deep-copy-15.c: New test.
* testsuite/libgomp.oacc-c-c++-common/deep-copy-16.c: New test.
* testsuite/libgomp.oacc-c++/deep-copy-17.C: New test.
* testsuite/libgomp.oacc-c-c++-common/deep-copy-arrayofstruct.c: Move
test to here, make "run" test.
Reverse offload requests at least -misa=sm_35; with this patch, a warning
instead of an error is shown, still permitting reverse offload for all
other configured device types. This is achieved by not calling
GOMP_offload_register_ver (and stopping generating pointless 'static const char'
variables, once known.)
The tool_name as progname changes adds "nvptx " and "gcn " to the
"mkoffload: warning/error:" diagnostic.
gcc/ChangeLog:
* config/nvptx/mkoffload.cc (process): Replace a fatal_error by
a warning + not enabling offloading if -misa=sm_30 prevents
reverse offload.
(main): Use tool_name as progname for diagnostic.
* config/gcn/mkoffload.cc (main): Likewise.
libgomp/ChangeLog:
* libgomp.texi (Offload-Target Specifics: nvptx): Document
that reverse offload requires >= -march=sm_35.
* testsuite/libgomp.c-c++-common/requires-4.c: Build for nvptx
with -misa=sm_35.
* testsuite/libgomp.c-c++-common/requires-5.c: Likewise.
* testsuite/libgomp.c-c++-common/requires-6.c: Likewise.
* testsuite/libgomp.c-c++-common/reverse-offload-1.c: Likewise.
* testsuite/libgomp.fortran/reverse-offload-1.f90: Likewise.
* testsuite/libgomp.c/reverse-offload-sm30.c: New test.
The thing is,
make check
or
make check RUNTESTFLAGS="c.exp='icv-6.c' c++.exp='icv-6.c'"
in libgomp obj dir work fine, but
make -j32 -k check RUNTESTFLAGS="c.exp='icv-6.c' c++.exp='icv-6.c'"
fails.
The thing is that the testcase as written relies on OMP_NUM_THREADS not being
set in environment (as it takes priority over OMP_NUM_THREADS_ALL for the
host).
So, if either a user has OMP_NUM_THREADS=42 in the environment by himself, or
when doing make check with -jN, we trigger:
if test $$num_cpus -gt 8 && test -z "$$OMP_NUM_THREADS"; then \
OMP_NUM_THREADS=8; export OMP_NUM_THREADS; \
echo @@@ libgomp OMP_NUM_THREADS adjusted to 8 because of parallel
make check and too many CPUs; \
fi; \
in libgomp/testsuite/Makefile.am and so the test fails.
2022-09-12 Jakub Jelinek <jakub@redhat.com>
PR libgomp/106894
* testsuite/libgomp.c-c++-common/icv-6.c: Include string.h.
(main): Avoid tests for which corresponding non-_ALL suffixed variable
is in the environment, or for OMP_NUM_TEAMS on the device
OMP_NUM_TEAMS_DEV_?.
This patch considers the environment variable syntax extension for
device-specific variants of environment variables from OpenMP 5.1 (see
OpenMP 5.1 specification, p. 75 and p. 639). An environment variable (e.g.
OMP_NUM_TEAMS) can have different suffixes:
_DEV (e.g. OMP_NUM_TEAMS_DEV): affects all devices but not the host.
_DEV_<device> (e.g. OMP_NUM_TEAMS_DEV_42): affects only device with
number <device>.
no suffix (e.g. OMP_NUM_TEAMS): affects only the host.
In future OpenMP versions also suffix _ALL will be introduced (see discussion
https://github.com/OpenMP/spec/issues/3179). This is also considered in this
patch:
_ALL (e.g. OMP_NUM_TEAMS_ALL): affects all devices and the host.
The precedence is as follows (descending). For the host:
1. no suffix
2. _ALL
For devices:
1. _DEV_<device>
2. _DEV
3. _ALL
That means, _DEV_<device> is used whenever available. Otherwise _DEV is used if
available, and at last _ALL. If there is no value for any of the variable
variants, default values are used as already implemented before.
This patch concerns parsing (a), storing (b), output (c) and transmission to the
device (d):
(a) The actual number of devices and the numbering are not known when parsing
the environment variables. Thus all environment variables are iterated and
searched for device-specific ones.
(b) Only configured device-specific variables are stored. Thus, a linked list
is used.
(c) The output is done in omp_display_env (see specification p. 468f). Global
ICVs are tagged with [all], see https://github.com/OpenMP/spec/issues/3179.
ICVs which are not global but aren't handled device-specific yet are tagged
with [host]. omp_display_env outputs the initial values of the ICVs. That is
why a dedicated data structure is introduced for the inital values only
(gomp_initial_icv_list).
(d) Device-specific ICVs are transmitted to the device via GOMP_ADDITIONAL_ICVS.
libgomp/ChangeLog:
* config/gcn/icv-device.c (omp_get_default_device): Return device-
specific ICV.
(omp_get_max_teams): Added for GCN devices.
(omp_set_num_teams): Likewise.
(ialias): Likewise.
* config/nvptx/icv-device.c (omp_get_default_device): Return device-
specific ICV.
(omp_get_max_teams): Added for NVPTX devices.
(omp_set_num_teams): Likewise.
(ialias): Likewise.
* env.c (struct gomp_icv_list): New struct to store entries of initial
ICV values.
(struct gomp_offload_icv_list): New struct to store entries of device-
specific ICV values that are copied to the device and back.
(struct gomp_default_icv_values): New struct to store default values of
ICVs according to the OpenMP standard.
(parse_schedule): Generalized for different variants of OMP_SCHEDULE.
(print_env_var_error): Function that prints an error for invalid values
for ICVs.
(parse_unsigned_long_1): Removed getenv. Generalized.
(parse_unsigned_long): Likewise.
(parse_int_1): Likewise.
(parse_int): Likewise.
(parse_int_secure): Likewise.
(parse_unsigned_long_list): Likewise.
(parse_target_offload): Likewise.
(parse_bind_var): Likewise.
(parse_stacksize): Likewise.
(parse_boolean): Likewise.
(parse_wait_policy): Likewise.
(parse_allocator): Likewise.
(omp_display_env): Extended to output different variants of environment
variables.
(print_schedule): New helper function for omp_display_env which prints
the values of run_sched_var.
(print_proc_bind): New helper function for omp_display_env which prints
the values of proc_bind_var.
(enum gomp_parse_type): Collection of types used for parsing environment
variables.
(ENTRY): Preprocess string lengths of environment variables.
(OMP_VAR_CNT): Preprocess table size.
(OMP_HOST_VAR_CNT): Likewise.
(INT_MAX_STR_LEN): Constant for the maximal number of digits of a device
number.
(gomp_get_icv_flag): Returns if a flag for a particular ICV is set.
(gomp_set_icv_flag): Sets a flag for a particular ICV.
(print_device_specific_icvs): New helper function for omp_display_env to
print device specific ICV values.
(get_device_num): New helper function for parse_device_specific.
Extracts the device number from an environment variable name.
(get_icv_member_addr): Gets the memory address for a particular member
of an ICV struct.
(gomp_get_initial_icv_item): Get a list item of gomp_initial_icv_list.
(initialize_icvs): New function to initialize a gomp_initial_icvs
struct.
(add_initial_icv_to_list): Adds an ICV struct to gomp_initial_icv_list.
(startswith): Checks if a string starts with a given prefix.
(initialize_env): Extended to parse the new syntax of environment
variables.
* icv-device.c (omp_get_max_teams): Added.
(ialias): Likewise.
(omp_set_num_teams): Likewise.
* icv.c (omp_set_num_teams): Moved to icv-device.c.
(omp_get_max_teams): Likewise.
(ialias): Likewise.
* libgomp-plugin.h (GOMP_DEVICE_NUM_VAR): Removed.
(GOMP_ADDITIONAL_ICVS): New target-side struct that
holds the designated ICVs of the target device.
* libgomp.h (enum gomp_icvs): Collection of ICVs.
(enum gomp_device_num): Definition of device numbers for _ALL, _DEV, and
no suffix.
(enum gomp_env_suffix): Collection of possible suffixes of environment
variables.
(struct gomp_initial_icvs): Contains all ICVs for which we need to store
initial values.
(struct gomp_default_icv):New struct to hold ICVs for which we need
to store initial values.
(struct gomp_icv_list): Definition of a linked list that is used for
storing ICVs for the devices and also for _DEV, _ALL, and without
suffix.
(struct gomp_offload_icvs): New struct to hold ICVs that are copied to
a device.
(struct gomp_offload_icv_list): Definition of a linked list that holds
device-specific ICVs that are copied to devices.
(gomp_get_initial_icv_item): Get a list item of gomp_initial_icv_list.
(gomp_get_icv_flag): Returns if a flag for a particular ICV is set.
* libgomp.texi: Updated.
* plugin/plugin-gcn.c (GOMP_OFFLOAD_load_image): Extended to read
further ICVs from the offload image.
* plugin/plugin-nvptx.c (GOMP_OFFLOAD_load_image): Likewise.
* target.c (gomp_get_offload_icv_item): Get a list item of
gomp_offload_icv_list.
(get_gomp_offload_icvs): New. Returns the ICV values
depending on the device num and the variable hierarchy.
(gomp_load_image_to_device): Extended to copy further ICVs to a device.
* testsuite/libgomp.c-c++-common/icv-5.c: New test.
* testsuite/libgomp.c-c++-common/icv-6.c: New test.
* testsuite/libgomp.c-c++-common/icv-7.c: New test.
* testsuite/libgomp.c-c++-common/icv-8.c: New test.
* testsuite/libgomp.c-c++-common/omp-display-env-1.c: New test.
* testsuite/libgomp.c-c++-common/omp-display-env-2.c: New test.
This patch implements doacross(sink: omp_cur_iteration - 1) that the
previous patchset emitted a sorry on during omp expansion.
It can be implemented with existing library functions.
To recap, depend(source)/doacross(source:)/doacross(source:omp_cur_iteration)
is implemented calling GOMP_doacross_post or GOMP_doacross_ull_post,
called with an array of long or unsigned long long elements, one for
all collapsed loops together and one for each further ordered loop if any.
We initialize that array in each thread when grabbing further set of iterations
and update it at the end of loops, so that it represents the current iteration
(as 0 based counters). When the worksharing loop is created, we tell the
library through another similar array the counts (the loop needs to be
rectangular) in each dimension, first element is count of all logical iterations
in the collapsed loops.
depend(sink:v1 op N1, v2 op N2, ...) is then implemented by conditionally calling
GOMP_doacross_wait/GOMP_doacross_ull_wait. For N? of 0 there is no check,
otherwise if it wants to wait in a particular dimension for a previous iteration,
we check that the corresponding iterator isn't the first one (or first few),
where the previous iterator in that dimension would be out of range, and similarly
for checking of next iteration in a dimension that it isn't the last one (or last few)
where it would be similarly out of bounds. Then the collapsed loop counters are
folded into a single 0 based counter (first argument) and then other 0 based
iterations counters on what iteration it should wait for.
Now, doacross(sink: omp_cur_iteration - 1) is supposed to wait for the previous
logical iteration in the combined iteration space of all ordered loops.
For the very first iteration in that combined iteration space it does nothing,
there is no previous iteration. And similarly it does nothing if there
are more ordered loops than collapsed loop and it isn't the first logical
iteration of the combined loops inside of the collapsed loops, because as implemented
we know the previous iteration in that case is always executed by the same thread
as the current one.
In the implementation, we use the same value as is stored in the first element
of the array for GOMP_doacross_post/GOMP_doacross_ull_post, if that value is 0,
we do nothing. The rest is different based on if ordered argument is equal to
collapse or not. If it is, then we otherwise call
GOMP_doacross_wait/GOMP_doacross_ull_wait with a single argument, one less than
that counter we compare against 0.
If ordered argument is bigger than collapse, we add a per-thread boolean variable
.first.N, which we set to true at the start of the outermost ordered loop inside
of the collapsed set of loops and set to false at the end of the innermost
ordered loop. If .first.N is false, we don't do anything (we know the previous
iteration was handled by the current thread and by my reading of the spec we don't
need to emit even a memory barrier in that case, because it is just synchronization
with the same thread), otherwise we call GOMP_doacross_wait/GOMP_doacross_ull_wait
with the first argument one less than the counter we compare against 0, and then
one less than 2nd and following counts if iterations we pass to the workshare
initialization. If say .counts.N passed to the workshare initialization is
{ 256, 13, 5, 2 } for collapse(3) ordered(6) loop, then
GOMP_doacross_post/GOMP_doacross_ull_post is called with arguments equal to
.ordereda.N[0] - 1, 12, 4, 1.
2022-09-08 Jakub Jelinek <jakub@redhat.com>
gcc/
* omp-expand.cc (expand_omp_ordered_sink): Add CONT_BB argument.
Add doacross(sink:omp_cur_iteration-1) support.
(expand_omp_ordered_source_sink): Clear counts[fd->ordered + 1].
Adjust expand_omp_ordered_sink caller.
(expand_omp_for_ordered_loops): If counts[fd->ordered + 1] is
non-NULL, set that variable to true at the start of outermost
non-collapsed loop and set it to false at the end of innermost
ordered loop.
(expand_omp_for_generic): If fd->ordered, allocate
1 + (fd->ordered - fd->collapse) further elements in counts array.
Copy to counts + 2 + fd->ordered the counts of fd->collapse ..
fd->ordered - 1 loop if any.
gcc/testsuite/
* c-c++-common/gomp/doacross-7.c: New test.
libgomp/
* libgomp.texi (OpenMP 5.2): Mention that omp_cur_iteration is now
fully supported.
* testsuite/libgomp.c/doacross-4.c: New test.
* testsuite/libgomp.c/doacross-5.c: New test.
* testsuite/libgomp.c/doacross-6.c: New test.
* testsuite/libgomp.c/doacross-7.c: New test.
gcc/ChangeLog:
PR middle-end/106548
* omp-low.cc (lower_rec_input_clauses): Use build_outer_var_ref
for 'simd' linear-step values that are variable.
libgomp/ChangeLog:
PR middle-end/106548
* testsuite/libgomp.c/linear-2.c: New test.
This run-time test test pointer-based iteration with collapse,
similar to the '(parallel) simd' test for PR106449 but for 'for'.
libgomp/ChangeLog:
* testsuite/libgomp.c-c++-common/pr106449-2.c: New test.
There were 2 issues visible on this new testcase, one that we didn't have
special POINTER_TYPE_P handling in a few spots of expand_omp_simd - for
pointers we need to use POINTER_PLUS_EXPR and need to have the non-pointer
part in sizetype, for non-rectangular loop on the other side we can rely on
multiplication factor 1, pointers can't be multiplied, without those changes
we'd ICE. The other issue was that we put n2 expression directly into a
comparison in a condition and regimplified that, for the &a[512] case that
and with gimplification being destructed that unfortunately meant modification
of original fd->loops[?].n2. Fixed by unsharing the expression. This was
causing a runtime failure on the testcase.
2022-07-29 Jakub Jelinek <jakub@redhat.com>
PR middle-end/106449
* omp-expand.cc (expand_omp_simd): Fix up handling of pointer
iterators in non-rectangular simd loops. Unshare fd->loops[i].n2
or n2 before regimplifying it inside of a condition.
* testsuite/libgomp.c-c++-common/pr106449.c: New test.
Fix-up for recent commit 06b2a2abe2
"Enhance '_Pragma' diagnostics verification in OMP C/C++ test cases".
Supposedly it's the same issue as in
<https://gcc.gnu.org/bugzilla/show_bug.cgi?id=101551#c2>, where I'd
noted that:
| [...] with an offloading-enabled build of GCC we're losing
| "note: in expansion of macro '[...]'" diagnostics.
| (Effectively '-ftrack-macro-expansion=0'?)
PR middle-end/101551
libgomp/
* testsuite/libgomp.oacc-c-c++-common/reduction-5.c: XFAIL
'offloading_enabled' diagnostics issue.
The handling of #pragma GCC diagnostic uses input_location, which is not always
as precise as needed; in particular the relative location of some tokens and a
_Pragma directive will crucially determine whether a given diagnostic is enabled
or suppressed in the desired way. PR97498 shows how the C frontend ends up with
input_location pointing to the beginning of the line containing a _Pragma()
directive, resulting in the wrong behavior if the diagnostic to be modified
pertains to some tokens found earlier on the same line. This patch fixes that by
addressing two issues:
a) libcpp was not assigning a valid location to the CPP_PRAGMA token
generated by the _Pragma directive.
b) C frontend was not setting input_location to something reasonable.
With this change, the C frontend is able to change input_location to point to
the _Pragma token as needed.
This is just a two-line fix (one for each of a) and b)), the testsuite changes
were needed only because the location on the tested warnings has been somewhat
improved, so the tests need to look for the new locations.
gcc/c/ChangeLog:
PR preprocessor/97498
* c-parser.cc (c_parser_pragma): Set input_location to the
location of the pragma, rather than the start of the line.
libcpp/ChangeLog:
PR preprocessor/97498
* directives.cc (destringize_and_run): Override the location of
the CPP_PRAGMA token from a _Pragma directive to the location of
the expansion point, as is done for the tokens lexed from it.
gcc/testsuite/ChangeLog:
PR preprocessor/97498
* c-c++-common/pr97498.c: New test.
* c-c++-common/gomp/pragma-3.c: Adapt for improved warning locations.
* c-c++-common/gomp/pragma-5.c: Likewise.
* gcc.dg/pragma-message.c: Likewise.
libgomp/ChangeLog:
* testsuite/libgomp.oacc-c-c++-common/reduction-5.c: Adapt for
improved warning locations.
* testsuite/libgomp.oacc-c-c++-common/vred2d-128.c: Likewise.
Similar to how the other 'mkoffload's got changed in
recent commit 683f118439
"OpenMP: Move omp requires checks to libgomp".
This also means finally switching Intel MIC 'mkoffload' to
'GOMP_offload_register_ver', 'GOMP_offload_unregister_ver',
making 'GOMP_offload_register', 'GOMP_offload_unregister'
legacy entry points.
gcc/
* config/i386/intelmic-mkoffload.cc (generate_host_descr_file)
(prepare_target_image, main): Handle OpenMP 'requires'.
(generate_host_descr_file): Switch to 'GOMP_offload_register_ver',
'GOMP_offload_unregister_ver'.
libgomp/
* target.c (GOMP_offload_register, GOMP_offload_unregister):
Denote as legacy entry points.
* testsuite/lib/libgomp.exp
(check_effective_target_offload_target_any): New proc.
* testsuite/libgomp.c-c++-common/requires-1.c: Enable for
'offload_target_any'.
* testsuite/libgomp.c-c++-common/requires-3.c: Likewise.
* testsuite/libgomp.c-c++-common/requires-7.c: Likewise.
* testsuite/libgomp.fortran/requires-1.f90: Likewise.
These should compile and link and execute in all configurations; host-fallback
execution, which we may actually verify.
Follow-up to recent commit 683f118439
"OpenMP: Move omp requires checks to libgomp".
libgomp/
* testsuite/libgomp.c-c++-common/requires-4.c: Enhance testing.
* testsuite/libgomp.c-c++-common/requires-5.c: Likewise.
As documented, this one does "Check diagnostic by device-compiler's lto1".
Indeed there are none when compiling with '-foffload=disable' with an
offloading-enabled compiler, so we should use 'offload_target_[...]', as
used in other similar test cases.
Follow-up to recent commit 683f118439
"OpenMP: Move omp requires checks to libgomp".
libgomp/
* testsuite/libgomp.c-c++-common/requires-3.c: Adjust.
