For unknown reasons, this had gotten added for the libgomp HSA plugin in commit
b8d89b03db (r242749) "Remove build dependence on
HSA run-time", and later propagated into the GCN plugin.
libgomp/
* plugin/plugin-gcn.c (init_environment_variables): Don't prepend
the 'HSA_RUNTIME_LIB' path to 'libhsa-runtime64.so'.
* plugin/configfrag.ac (HSA_RUNTIME_LIB): Clean up.
* config.h.in: Regenerate.
* configure: Likewise.
Some gcc configurations default to -m32 but support -m64 too. This patch
just makes the ILP32 tests more reliable by following what e.g. libsanitizer
configury does.
2021-03-04 Jakub Jelinek <jakub@redhat.com>
* configure.ac: Add AC_CHECK_SIZEOF([void *]).
* plugin/configfrag.ac: Check $ac_cv_sizeof_void_p value instead of
checking of -m32 or -mx32 options on the command line.
* config.h.in: Regenerated.
* configure: Regenerated.
As recently again discussed in <https://gcc.gnu.org/PR97436> "[nvptx] -m32
support", nvptx offloading other than for 64-bit host has never been
implemented, tested, supported. So we simply should buildn't the nvptx libgomp
plugin in this case.
This avoids build problems if, for example, in a (standard) bi-arch
x86_64-pc-linux-gnu '-m64'/'-m32' build, libcuda is available only in a 64-bit
variant but not in a 32-bit one, which, for example, is the case if you build
GCC against the CUDA toolkit's 'stubs/libcuda.so' (see
<https://stackoverflow.com/a/52784819>).
This amends PR65099 commit a92defdab7 (r225560)
"[nvptx offloading] Only 64-bit configurations are currently supported" to
match the way we're doing this for the HSA/GCN plugins.
libgomp/
PR libgomp/65099
* plugin/configfrag.ac (PLUGIN_NVPTX): Restrict to supported
configurations.
* configure: Regenerate.
* plugin/plugin-nvptx.c (nvptx_get_num_devices): Remove 64-bit
check.
These are the same header files that exist in the Radeon Open Compute Runtime
project (as of October 2020), but they have been specially relicensed by AMD
for use in GCC.
The header files retain AMD copyright.
include/ChangeLog:
* hsa.h: Replace whole file.
* hsa_ext_amd.h: New file.
* hsa_ext_image.h: New file.
libgomp/ChangeLog:
* plugin/plugin-gcn.c: Include hsa_ext_amd.h.
(HSA_AMD_AGENT_INFO_COMPUTE_UNIT_COUNT): Delete redundant definition.
Ensure the code will continue to compile when elf.h gets these definitions.
libgomp/ChangeLog:
* plugin/plugin-gcn.c: Don't redefine relocations if elf.h has them.
(reserved): Delete unused define.
Using this patch, when using GOMP_DEBUG=1 and launching a kernel in
GOMP_OFFLOAD_run (used by the omp implementation), we see the kernel launch
dimensions:
...
GOMP_OFFLOAD_run: kernel main$_omp_fn$0: \
launch [(teams: 1), 1, 1] [(lanes: 32), (threads: 1), 1]
...
Build on x86_64-linux with nvptx accelerator, tested libgomp.
libgomp/ChangeLog:
2020-10-08 Tom de Vries <tdevries@suse.de>
PR libgomp/81802
* plugin/plugin-nvptx.c (GOMP_OFFLOAD_run): Report launch
dimensions.
By running libgomp test-case libgomp.c/target-28.c with GOMP_NVPTX_PTXRW=w
(using a maintenance patch that adds support for this env var), we dump the
ptx in target-28.exe to file. By editing one ptx file to rename
gomp_nvptx_main to gomp_nvptx_main2 in both declaration and call, and
running with GOMP_NVPTX_PTXRW=r, we trigger a link error:
...
$ GOMP_NVPTX_PTXRW=r ./target-28.exe
libgomp: cuLinkComplete error: unknown error
...
The error is somewhat uninformative.
Fix this by dumping the error log returned by the failing cuda call, such
that we have instead:
...
$ GOMP_NVPTX_PTXRW=r ./target-28.exe
libgomp: Link error log error : \
Undefined reference to 'gomp_nvptx_main2' in ''
libgomp: cuLinkComplete error: unknown error
...
Build on x86_64 with nvptx accelerator, tested libgomp.
libgomp/ChangeLog:
* plugin/plugin-nvptx.c (link_ptx): Print elog if cuLinkComplete call
fails.
Change test for CUDA callback context in nvptx_free() from using
GOMP_PLUGIN_acc_thread () into checking for CUDA_ERROR_NOT_PERMITTED,
for the former only works for OpenACC, but not OpenMP offloading.
2020-08-20 Chung-Lin Tang <cltang@codesourcery.com>
libgomp/
* plugin/plugin-nvptx.c (nvptx_free):
Change "GOMP_PLUGIN_acc_thread () == NULL" test into check of
CUDA_ERROR_NOT_PERMITTED status for cuMemGetAddressRange. Adjust
comments.
This patch removes the generation of HSAIL from the compiler, the HSA
offloading plugin from libgomp and the associated testsuite tests and
infrastructure bits from the respective testsuites.
Apart from removal of the obvious files, I removed bits that I found
by searching for HSA related terms and by re-tracing my steps and
looking at the patches that introduced HSA in the first place. I did
not remove everything these patches brought in, for example:
- the mechanism to pass offload-target specific info from the application to
the offloading plugin - but the same mechanism is also used to
communicate number of teams and the thread limit to all offload targets.
- run_func hook in gomp_device_descr stays too, although now it is
not used. If some future offload target would like the ability to
refuse to offload some functions, it can use it. It is easy to
remove as a follow-up if it is considered clutter, though.
- configure options --with-hsa-runtime=PATH, -with-hsa-runtime-include=PATH
and --with-hsa-runtime-lib=PATH rmeain because GCN uses them too.
- Surprisingly, GOMP_TARGET_ARG_HSA_KERNEL_ATTRIBUTES (a constant
from gomp-constants.h) appears in the source of the amdgcn libgomp
plugin, although I tend to think that code path is not ever used
and this patch certainly removes it from the compiler.
Nevertheless, it seems it has potential value beyond HSAIL and so
I've kept it, it can of course always be easily removed in the
future of GCN folk abandon it too.
- I assume constants OFFLOAD_TARGET_TYPE_HSA and GOMP_DEVICE_HSA
need to stay indefinitely too just so that no future offload
target picks that number.
- I have kept dg-require-effective-target
offload_device_nonshared_as requirement of thests which have it.
It is quite probable I missed some small HSA artifacts but those
should be easy to remove later as we find them.
include/ChangeLog:
2020-07-24 Martin Jambor <mjambor@suse.cz>
* gomp-constants.h (GOMP_VERSION_HSA): Remove.
gcc/ChangeLog:
2020-07-24 Martin Jambor <mjambor@suse.cz>
* hsa-brig-format.h: Moved to brig/brigfrontend.
* hsa-brig.c: Removed.
* hsa-builtins.def: Likewise.
* hsa-common.c: Likewise.
* hsa-common.h: Likewise.
* hsa-dump.c: Likewise.
* hsa-gen.c: Likewise.
* hsa-regalloc.c: Likewise.
* ipa-hsa.c: Likewise.
* omp-grid.c: Likewise.
* omp-grid.h: Likewise.
* Makefile.in (BUILTINS_DEF): Remove hsa-builtins.def.
(OBJS): Remove hsa-common.o, hsa-gen.o, hsa-regalloc.o, hsa-brig.o,
hsa-dump.o, ipa-hsa.c and omp-grid.o.
(GTFILES): Removed hsa-common.c and omp-expand.c.
* builtins.def: Remove processing of hsa-builtins.def.
(DEF_HSA_BUILTIN): Remove.
* common.opt (flag_disable_hsa): Remove.
(-Whsa): Ignore.
* config.in (ENABLE_HSA): Removed.
* configure.ac: Removed handling configuration for hsa offloading.
(ENABLE_HSA): Removed.
* configure: Regenerated.
* doc/install.texi (--enable-offload-targets): Remove hsa from the
example.
(--with-hsa-runtime): Reword to reference any HSA run-time, not
specifically HSA offloading.
* doc/invoke.texi (Option Summary): Remove -Whsa.
(Warning Options): Likewise.
(Optimize Options): Remove hsa-gen-debug-stores.
* doc/passes.texi (Regular IPA passes): Remove section on IPA HSA
pass.
* gimple-low.c (lower_stmt): Remove GIMPLE_OMP_GRID_BODY case.
* gimple-pretty-print.c (dump_gimple_omp_for): Likewise.
(dump_gimple_omp_block): Likewise.
(pp_gimple_stmt_1): Likewise.
* gimple-walk.c (walk_gimple_stmt): Likewise.
* gimple.c (gimple_build_omp_grid_body): Removed function.
(gimple_copy): Remove GIMPLE_OMP_GRID_BODY case.
* gimple.def (GIMPLE_OMP_GRID_BODY): Removed.
* gimple.h (gf_mask): Removed GF_OMP_PARALLEL_GRID_PHONY,
OMP_FOR_KIND_GRID_LOOP, GF_OMP_FOR_GRID_PHONY,
GF_OMP_FOR_GRID_INTRA_GROUP, GF_OMP_FOR_GRID_GROUP_ITER and
GF_OMP_TEAMS_GRID_PHONY. Renumbered GF_OMP_FOR_KIND_SIMD and
GF_OMP_TEAMS_HOST.
(gimple_build_omp_grid_body): Removed declaration.
(gimple_has_substatements): Remove GIMPLE_OMP_GRID_BODY case.
(gimple_omp_for_grid_phony): Removed.
(gimple_omp_for_set_grid_phony): Likewise.
(gimple_omp_for_grid_intra_group): Likewise.
(gimple_omp_for_grid_intra_group): Likewise.
(gimple_omp_for_grid_group_iter): Likewise.
(gimple_omp_for_set_grid_group_iter): Likewise.
(gimple_omp_parallel_grid_phony): Likewise.
(gimple_omp_parallel_set_grid_phony): Likewise.
(gimple_omp_teams_grid_phony): Likewise.
(gimple_omp_teams_set_grid_phony): Likewise.
(CASE_GIMPLE_OMP): Remove GIMPLE_OMP_GRID_BODY case.
* lto-section-in.c (lto_section_name): Removed hsa.
* lto-streamer.h (lto_section_type): Removed LTO_section_ipa_hsa.
* lto-wrapper.c (compile_images_for_offload_targets): Remove special
handling of hsa.
* omp-expand.c: Do not include hsa-common.h and gt-omp-expand.h.
(parallel_needs_hsa_kernel_p): Removed.
(grid_launch_attributes_trees): Likewise.
(grid_launch_attributes_trees): Likewise.
(grid_create_kernel_launch_attr_types): Likewise.
(grid_insert_store_range_dim): Likewise.
(grid_get_kernel_launch_attributes): Likewise.
(get_target_arguments): Remove code passing HSA grid sizes.
(grid_expand_omp_for_loop): Remove.
(grid_arg_decl_map): Likewise.
(grid_remap_kernel_arg_accesses): Likewise.
(grid_expand_target_grid_body): Likewise.
(expand_omp): Remove call to grid_expand_target_grid_body.
(omp_make_gimple_edges): Remove GIMPLE_OMP_GRID_BODY case.
* omp-general.c: Do not include hsa-common.h.
(omp_maybe_offloaded): Do not check for HSA offloading.
(omp_context_selector_matches): Likewise.
* omp-low.c: Do not include hsa-common.h and omp-grid.h.
(build_outer_var_ref): Remove handling of GIMPLE_OMP_GRID_BODY.
(scan_sharing_clauses): Remove handling of OMP_CLAUSE__GRIDDIM_.
(scan_omp_parallel): Remove handling of the phoney variant.
(check_omp_nesting_restrictions): Remove handling of
GIMPLE_OMP_GRID_BODY and GF_OMP_FOR_KIND_GRID_LOOP.
(scan_omp_1_stmt): Remove handling of GIMPLE_OMP_GRID_BODY.
(lower_omp_for_lastprivate): Remove handling of gridified loops.
(lower_omp_for): Remove phony loop handling.
(lower_omp_taskreg): Remove phony construct handling.
(lower_omp_teams): Likewise.
(lower_omp_grid_body): Removed.
(lower_omp_1): Remove GIMPLE_OMP_GRID_BODY case.
(execute_lower_omp): Do not call omp_grid_gridify_all_targets.
* opts.c (common_handle_option): Do not handle hsa when processing
OPT_foffload_.
* params.opt (hsa-gen-debug-stores): Remove.
* passes.def: Remove pass_ipa_hsa and pass_gen_hsail.
* timevar.def: Remove TV_IPA_HSA.
* toplev.c: Do not include hsa-common.h.
(compile_file): Do not call hsa_output_brig.
* tree-core.h (enum omp_clause_code): Remove OMP_CLAUSE__GRIDDIM_.
(tree_omp_clause): Remove union field dimension.
* tree-nested.c (convert_nonlocal_omp_clauses): Remove the
OMP_CLAUSE__GRIDDIM_ case.
(convert_local_omp_clauses): Likewise.
* tree-pass.h (make_pass_gen_hsail): Remove declaration.
(make_pass_ipa_hsa): Likewise.
* tree-pretty-print.c (dump_omp_clause): Remove GIMPLE_OMP_GRID_BODY
case.
* tree.c (omp_clause_num_ops): Remove the element corresponding to
OMP_CLAUSE__GRIDDIM_.
(omp_clause_code_name): Likewise.
(walk_tree_1): Remove GIMPLE_OMP_GRID_BODY case.
* tree.h (OMP_CLAUSE__GRIDDIM__DIMENSION): Remove.
(OMP_CLAUSE__GRIDDIM__SIZE): Likewise.
(OMP_CLAUSE__GRIDDIM__GROUP): Likewise.
gcc/fortran/ChangeLog:
2020-07-24 Martin Jambor <mjambor@suse.cz>
* f95-lang.c (gfc_init_builtin_functions): Remove processing of
hsa-builtins.def.
gcc/brig/ChangeLog:
2020-07-24 Martin Jambor <mjambor@suse.cz>
* brigfrontend/brig-util.h (hsa_type_packed_p): Declared.
* brigfrontend/brig-util.cc (hsa_type_packed_p): Moved here from
removed gcc/hsa-common.c.
libgomp/ChangeLog:
2020-07-24 Martin Jambor <mjambor@suse.cz>
* plugin/Makefrag.am: Remove configuration of HSA plugin.
* aclocal.m4: Regenerated.
* Makefile.in: Regenerated.
* config.h.in: Regenerated.
* configure: Regenerated.
* plugin/configfrag.ac: Likewise.
* plugin/hsa_ext_finalize.h: Removed.
* plugin/plugin-hsa.c: Likewise.
* testsuite/Makefile.in: Regenerated.
* testsuite/lib/libgomp.exp
(offload_target_to_openacc_device_type): Remove hsa case.
(check_effective_target_hsa_offloading_selected_nocache): Removed
(check_effective_target_hsa_offloading_selected): Likewise.
(libgomp_init): Do not add -Wno-hsa to additional_flags.
* testsuite/libgomp.hsa.c/alloca-1.c: Removed test.
* testsuite/libgomp.hsa.c/bitfield-1.c: Likewise.
* testsuite/libgomp.hsa.c/bits-insns.c: Likewise.
* testsuite/libgomp.hsa.c/builtins-1.c: Likewise.
* testsuite/libgomp.hsa.c/c.exp: Likewise.
* testsuite/libgomp.hsa.c/complex-1.c: Likewise.
* testsuite/libgomp.hsa.c/complex-align-2.c: Likewise.
* testsuite/libgomp.hsa.c/formal-actual-args-1.c: Likewise.
* testsuite/libgomp.hsa.c/function-call-1.c: Likewise.
* testsuite/libgomp.hsa.c/get-level-1.c: Likewise.
* testsuite/libgomp.hsa.c/gridify-1.c: Likewise.
* testsuite/libgomp.hsa.c/gridify-2.c: Likewise.
* testsuite/libgomp.hsa.c/gridify-3.c: Likewise.
* testsuite/libgomp.hsa.c/gridify-4.c: Likewise.
* testsuite/libgomp.hsa.c/memory-operations-1.c: Likewise.
* testsuite/libgomp.hsa.c/pr69568.c: Likewise.
* testsuite/libgomp.hsa.c/pr82416.c: Likewise.
* testsuite/libgomp.hsa.c/rotate-1.c: Likewise.
* testsuite/libgomp.hsa.c/staticvar.c: Likewise.
* testsuite/libgomp.hsa.c/switch-1.c: Likewise.
* testsuite/libgomp.hsa.c/switch-branch-1.c: Likewise.
* testsuite/libgomp.hsa.c/switch-sbr-2.c: Likewise.
* testsuite/libgomp.hsa.c/tiling-1.c: Likewise.
* testsuite/libgomp.hsa.c/tiling-2.c: Likewise.
gcc/testsuite/ChangeLog:
2020-07-24 Martin Jambor <mjambor@suse.cz>
* lib/target-supports.exp (check_effective_target_offload_hsa):
Removed.
* c-c++-common/gomp/gridify-1.c: Removed test.
* c-c++-common/gomp/gridify-2.c: Likewise.
* c-c++-common/gomp/gridify-3.c: Likewise.
* c-c++-common/gomp/hsa-indirect-call-1.c: Likewise.
* gfortran.dg/gomp/gridify-1.f90: Likewise.
* gcc.dg/gomp/gomp.exp: Do not pass -Wno-hsa to tests.
* g++.dg/gomp/gomp.exp: Likewise.
* gfortran.dg/gomp/gomp.exp: Likewise.
This upgrades the compiler to emit HSA Code Object v3 binaries. This means
changing the assembler directives, and linker command line options.
The gcn-run and libgomp loaders need corresponding alterations. The
relocations no longer need to be fixed up manually, and the kernel symbol
names have changed slightly.
This move makes the binaries compatible with the new rocgdb from ROCm 3.5.
2020-06-17 Andrew Stubbs <ams@codesourcery.com>
gcc/
* config/gcn/gcn-hsa.h (TEXT_SECTION_ASM_OP): Use ".text".
(BSS_SECTION_ASM_OP): Use ".bss".
(ASM_SPEC): Remove "-mattr=-code-object-v3".
(LINK_SPEC): Add "--export-dynamic".
* config/gcn/gcn-opts.h (processor_type): Replace PROCESSOR_VEGA with
PROCESSOR_VEGA10 and PROCESSOR_VEGA20.
* config/gcn/gcn-run.c (HSA_RUNTIME_LIB): Use ".so.1" variant.
(load_image): Remove obsolete relocation handling.
Add ".kd" suffix to the symbol names.
* config/gcn/gcn.c (MAX_NORMAL_SGPR_COUNT): Set to 62.
(gcn_option_override): Update gcn_isa test.
(gcn_kernel_arg_types): Update all the assembler directives.
Remove the obsolete options.
(gcn_conditional_register_usage): Update MAX_NORMAL_SGPR_COUNT usage.
(gcn_omp_device_kind_arch_isa): Handle PROCESSOR_VEGA10 and
PROCESSOR_VEGA20.
(output_file_start): Rework assembler file header.
(gcn_hsa_declare_function_name): Rework kernel metadata.
* config/gcn/gcn.h (GCN_KERNEL_ARG_TYPES): Set to 16.
* config/gcn/gcn.opt (PROCESSOR_VEGA): Remove enum.
(PROCESSOR_VEGA10): New enum value.
(PROCESSOR_VEGA20): New enum value.
libgomp/
* plugin/plugin-gcn.c (init_environment_variables): Use ".so.1"
variant for HSA_RUNTIME_LIB name.
(find_executable_symbol_1): Delete.
(find_executable_symbol): Delete.
(init_kernel_properties): Add ".kd" suffix to symbol names.
(find_load_offset): Delete.
(create_and_finalize_hsa_program): Remove relocation handling.
Ensure that the returned status values are not ignored. The old code was
not broken, but this is both safer and satisfies static analysis.
2020-04-23 Andrew Stubbs <ams@codesourcery.com>
PR other/94629
libgomp/
* plugin/plugin-gcn.c (init_hsa_context): Check return value from
hsa_iterate_agents.
(GOMP_OFFLOAD_init_device): Check return values from both calls to
hsa_agent_iterate_regions.
An OpenMP "nowait" clause on a target construct currently leads to
a call to GOMP_OFFLOAD_async_run in the plugin that is used for
offloading at execution time. The nvptx plugin contains only a stub
of this function that always produces a fatal error if called.
This commit changes the "nowait" implementation to ignore the clause
if the executing device's plugin does not implement GOMP_OFFLOAD_async_run.
The stub in the nvptx plugin is removed which effectively means that
programs containing "nowait" can now be executed with nvptx offloading
as if the clause had not been used.
This behavior is consistent with the OpenMP specification which says that
"[...] execution of the target task *may* be deferred" (emphasis added),
cf. OpenMP 5.0, page 172.
libgomp/
* plugin/plugin-nvptx.c: Remove GOMP_OFFLOAD_async_run stub.
* target.c (gomp_load_plugin_for_device): Make "async_run" loading
optional.
(gomp_target_task_fn): Assert "devicep->async_run_func".
(clear_unsupported_flags): New function to remove unsupported flags
(right now only GOMP_TARGET_FLAG_NOWAIT) that can be be ignored.
(GOMP_target_ext): Apply clear_unsupported_flags to flags.
* testsuite/libgomp.c/target-33.c:
Remove xfail for offload_target_nvptx.
* testsuite/libgomp.c/target-34.c: Likewise.
2020-01-31 Kwok Cheung Yeung <kcy@codesourcery.com>
gcc/
* config/gcn/mkoffload.c (process_asm): Add sgpr_count and vgpr_count
to definition of hsa_kernel_description. Parse assembly to find SGPR
and VGPR count of kernel and store in hsa_kernel_description.
libgomp/
* plugin/plugin-gcn.c (struct hsa_kernel_description): Add sgpr_count
and vgpr_count fields.
(struct kernel_info): Add a field for a hsa_kernel_description.
(run_kernel): Reduce the number of threads/workers if the requested
number would require too many VGPRs.
(init_basic_kernel_info): Initialize description field with
the hsa_kernel_description entry for the kernel.
Add full support for the OpenACC 2.6 acc_get_property and
acc_get_property_string functions to the libgomp GCN plugin.
libgomp/
* plugin-gcn.c (struct agent_info): Add fields "name" and
"vendor_name" ...
(GOMP_OFFLOAD_init_device): ... and init from here.
(struct hsa_context_info): Add field "driver_version_s" ...
(init_hsa_contest): ... and init from here.
(GOMP_OFFLOAD_openacc_get_property): Replace stub with a proper
implementation.
* testsuite/libgomp.oacc-c-c++-common/acc_get_property.c:
Enable test execution for amdgcn and host offloading targets.
* testsuite/libgomp.oacc-fortran/acc_get_property.f90: Likewise.
* testsuite/libgomp.oacc-c-c++-common/acc_get_property-aux.c
(expect_device_properties): Split function into ...
(expect_device_string_properties): ... this new function ...
(expect_device_memory): ... and this new function.
* testsuite/libgomp.oacc-c-c++-common/acc_get_property-gcn.c:
Add test.
The HSA/ROCm runtime rejects binaries not built for the exact GPU device
present. So far, the libgomp amdgcn plugin does not verify that the GPU ISA
and the ISA specified at compile time match before handing over the binary to
the runtime. In case of a mismatch, the user is confronted with an unhelpful
runtime error.
This commit implements a runtime ISA check. In case of an ISA mismatch, the
execution is aborted with a clear error message and a hint at the correct
compilation parameters for the GPU on which the execution has been attempted.
libgomp/
* plugin/plugin-gcn.c (EF_AMDGPU_MACH): New enum.
* (EF_AMDGPU_MACH_MASK): New constant.
* (gcn_isa): New typedef.
* (gcn_gfx801_s): New constant.
* (gcn_gfx803_s): New constant.
* (gcn_gfx900_s): New constant.
* (gcn_gfx906_s): New constant.
* (gcn_isa_name_len): New constant.
* (elf_gcn_isa_field): New function.
* (isa_hsa_name): New function.
* (isa_gcc_name): New function.
* (isa_code): New function.
* (struct agent_info): Add field "device_isa" and remove field
"gfx900_p".
* (GOMP_OFFLOAD_init_device): Adapt agent init to "agent_info"
field changes, fail if device has unknown ISA.
* (parse_target_attributes): Replace "gfx900_p" by "device_isa".
* (isa_matches_agent): New function ...
* (create_and_finalize_hsa_program): ... used from here to check
that the GPU ISA and the code-object ISA match.
Add generic support for the OpenACC 2.6 `acc_get_property' and
`acc_get_property_string' routines, as well as full handlers for the
host and the NVPTX offload targets and minimal handlers for the HSA,
Intel MIC, and AMD GCN offload targets.
Included are C/C++ and Fortran tests that, in particular, print
the property values for acc_property_vendor, acc_property_memory,
acc_property_free_memory, acc_property_name, and acc_property_driver.
The output looks as follows:
Vendor: GNU
Name: GOMP
Total memory: 0
Free memory: 0
Driver: 1.0
with the host driver (where the memory related properties are not
supported for the host device and yield 0, conforming to the standard)
and output like:
Vendor: Nvidia
Total memory: 12651462656
Free memory: 12202737664
Name: TITAN V
Driver: CUDA Driver 9.1
with the NVPTX driver.
2019-12-22 Maciej W. Rozycki <macro@codesourcery.com>
Frederik Harwath <frederik@codesourcery.com>
Thomas Schwinge <tschwinge@codesourcery.com>
include/
* gomp-constants.h (gomp_device_property): New enum.
libgomp/
* libgomp.h (gomp_device_descr): Add `get_property_func' member.
* libgomp-plugin.h (gomp_device_property_value): New union.
(gomp_device_property_value): New prototype.
* openacc.h (acc_device_t): Add `acc_device_current' enumeration
constant.
(acc_device_property_t): New enum.
(acc_get_property, acc_get_property_string): New prototypes.
* oacc-init.c (acc_get_device_type): Also assert that result
is not `acc_device_current'.
(get_property_any, acc_get_property, acc_get_property_string):
New functions.
* openacc.f90 (openacc_kinds): Add `acc_device_current' and
`acc_property_memory', `acc_property_free_memory',
`acc_property_name', `acc_property_vendor' and
`acc_property_driver' constants. Add `acc_device_property' data
type.
(openacc_internal): Add `acc_get_property' and
`acc_get_property_string' interfaces. Add `acc_get_property_h',
`acc_get_property_string_h', `acc_get_property_l' and
`acc_get_property_string_l'.
* oacc-host.c (host_get_property): New function.
(host_dispatch): Wire it.
* target.c (gomp_load_plugin_for_device): Handle `get_property'.
* libgomp.map (OACC_2.6): Add `acc_get_property', `acc_get_property_h_',
`acc_get_property_string' and `acc_get_property_string_h_' symbols.
* libgomp.texi (OpenACC Runtime Library Routines): Add
`acc_get_property'.
(acc_get_property): New node.
* plugin/plugin-gcn.c (GOMP_OFFLOAD_get_property): New
function (stub).
* plugin/plugin-hsa.c (GOMP_OFFLOAD_get_property): New function.
* plugin/plugin-nvptx.c (CUDA_CALLS): Add `cuDeviceGetName',
`cuDeviceTotalMem', `cuDriverGetVersion' and `cuMemGetInfo'
calls.
(GOMP_OFFLOAD_get_property): New function.
(struct ptx_device): Add new field "name".
(cuda_driver_version_s): Add new static variable ...
(nvptx_init): ... and init from here.
* testsuite/libgomp.oacc-c-c++-common/acc_get_property.c: New test.
* testsuite/libgomp.oacc-c-c++-common/acc_get_property-2.c: New test.
* testsuite/libgomp.oacc-c-c++-common/acc_get_property-3.c: New test.
* testsuite/libgomp.oacc-c-c++-common/acc_get_property-aux.c: New file
with test helper functions.
* testsuite/libgomp.oacc-fortran/acc_get_property.f90: New test.
liboffloadmic/
* plugin/libgomp-plugin-intelmic.cpp (GOMP_OFFLOAD_get_property):
New function.
Reviewed-by: Thomas Schwinge <thomas@codesourcery.com>
Co-Authored-By: Frederik Harwath <frederik@codesourcery.com>
Co-Authored-By: Thomas Schwinge <tschwinge@codesourcery.com>
From-SVN: r279710
PR libgomp/90585
* plugin/plugin-hsa.c: Include gstdint.h. Include inttypes.h only if
HAVE_INTTYPES_H is defined.
(print_uint64_t): New typedef.
(PRIu64): Define if HAVE_INTTYPES_H is not defined.
(print_kernel_dispatch, run_kernel): Use PRIu64 macro instead of
"lu", cast uint64_t HSA_DEBUG and fprintf arguments to print_uint64_t.
(release_kernel_dispatch): Likewise. Cast shadow->debug to uintptr_t
before casting to void *.
* plugin/plugin-nvptx.c: Include gstdint.h instead of stdint.h.
* oacc-mem.c: Don't include config.h nor stdint.h.
* target.c: Don't include config.h.
* oacc-cuda.c: Likewise.
* oacc-host.c: Don't include stdint.h.
From-SVN: r271597
I wrote a test-case:
...
int
main (void)
{
for (unsigned i = 0; i < 128; ++i)
{
acc_init (acc_device_nvidia);
acc_shutdown (acc_device_nvidia);
}
return 0;
}
...
and ran it under valgrind. The only leak location reported with a frequency
of 128, was the allocation of ptx_devices in nvptx_init.
Fix this by freeing ptx_devices in GOMP_OFFLOAD_fini_device, once
instantiated_devices drops to 0.
2019-01-24 Tom de Vries <tdevries@suse.de>
* plugin/plugin-nvptx.c (GOMP_OFFLOAD_fini_device): Free ptx_devices
once instantiated_devices drops to 0.
From-SVN: r268237
Consider test-case:
...
int
main (void)
{
#pragma acc parallel async
;
#pragma acc parallel async
;
#pragma acc wait
return 0;
}
...
This fails with:
...
libgomp: cuMemAlloc error: invalid argument
Segmentation fault (core dumped)
...
The cuMemAlloc error is due to the fact that we're try to allocate 0 bytes.
Fix this by preventing calling map_push with size zero argument in nvptx_exec.
This also has the consequence that for the abort-1.c test-case, we end up
calling cuMemFree during map_fini for the struct cuda_map allocated in
map_init, which fails because an abort happened. Fix this by calling
cuMemFree with CUDA_CALL_NOCHECK in cuda_map_destroy.
2019-01-23 Tom de Vries <tdevries@suse.de>
PR target/PR88946
* plugin/plugin-nvptx.c (cuda_map_destroy): Use CUDA_CALL_NOCHECK for
cuMemFree.
(nvptx_exec): Don't call map_push if mapnum == 0.
* testsuite/libgomp.oacc-c-c++-common/pr88946.c: New test.
From-SVN: r268178
There are currently two situations where this assert triggers:
...
libgomp/plugin/plugin-nvptx.c: map_fini: Assertion `!s->map->active' failed.
...
First, in abort-1.c, a parallel region triggering an abort:
...
int
main (void)
{
#pragma acc parallel
abort ();
return 0;
}
...
The abort is detected in nvptx_exec as the CUDA_ERROR_ILLEGAL_INSTRUCTION
return status of the cuStreamSynchronize call after kernel launch, which is
then handled by calling non-returning function GOMP_PLUGIN_fatal.
Consequently, the map_pop in nvptx_exec that in case of cuStreamSynchronize
success would remove or inactive the element added by the map_push earlier in
nvptx_exec, does not trigger. With the element no longer active, but still
marked active and a member of s->map, we run into the assert during
GOMP_OFFLOAD_fini_device, which is triggered from atexit handler
gomp_target_fini (which is triggered by the GOMP_PLUGIN_fatal mentioned above
calling exit).
Second, in pr88941.c, an async parallel region without wait:
...
int
main (void)
{
#pragma acc parallel async
;
/* no #pragma acc wait */
return 0;
}
...
Because nvptx_exec is handling an async region, it does not call map_pop for
the element added by map_push, but schedules an kernel execution completion
event to call map_pop. Again, we run into the assert during
GOMP_OFFLOAD_fini_device, which is triggered from atexit handler
gomp_target_fini, but the exit in this case is triggered by returning from main.
So either the kernel is still running, or the kernel has completed but the
corresponding event that is supposed to call map_pop is stuck in the event
queue, waiting for an event_gc.
Fix this by removing the assert, and skipping the freeing of device memory if
the map is still marked active (though in the async case, this is more a
workaround than an fix).
2019-01-23 Tom de Vries <tdevries@suse.de>
PR target/88941
PR target/88939
* plugin/plugin-nvptx.c (cuda_map_destroy): Handle map->active case.
(map_fini): Remove "assert (!s->map->active)".
* testsuite/libgomp.oacc-c-c++-common/pr88941.c: New test.
From-SVN: r268177
The map field of a struct ptx_stream is a FIFO. The FIFO is implemented as a
single linked list, with pop-from-the-front semantics.
The function map_pop pops an element, either by:
- deallocating the element, if there is more than one element
- or marking the element inactive, if there's only one element
The responsibility of map_push is to push an element to the back, as well as
selecting the element to push, by:
- allocating an element, or
- reusing the element at the front if inactive and big enough, or
- dropping the element at the front if inactive and not big enough, and
allocating one that's big enough
The current implemention gets at least the first and most basic scenario wrong:
> map = cuda_map_create (size);
We create an element, and assign it to map.
> for (t = s->map; t->next != NULL; t = t->next)
> ;
We determine the last element in the fifo.
> t->next = map;
We append the new element.
> s->map = map;
But here, we throw away the rest of the FIFO, and declare the FIFO to be just
the new element.
This problem causes the test-case asyncwait-1.c to fail intermittently on some
systems. The pr87835.c test-case added here is a a minimized and modified
version of asyncwait-1.c (avoiding the kernel construct) that is more likely to
fail.
Fix this by rewriting map_pop more robustly, by:
- seperating the function in two phases: select element, push element
- when reusing or dropping an element, making sure that the element is cleanly
popped from the queue
- rewriting the push element part in such a way that it can handle all cases
without needing if statements, such that each line is exercised for each of
the three cases.
2019-01-23 Tom de Vries <tdevries@suse.de>
PR target/87835
* plugin/plugin-nvptx.c (map_push): Fix adding of allocated element.
* testsuite/libgomp.oacc-c-c++-common/pr87835.c: New test.
From-SVN: r268176
Update message in nvptx libgomp plugin about insufficient resources to launch
kernel, to accommodate for the fact the vector_length can now be variable.
2019-01-12 Tom de Vries <tdevries@suse.de>
* plugin/plugin-nvptx.c (nvptx_exec): Update insufficient hardware
resources diagnostic.
From-SVN: r267890
When using a compiler build with:
...
+#define PTX_DEFAULT_VECTOR_LENGTH PTX_CTA_SIZE
...
consider a test-case:
...
int
main (void)
{
#pragma acc parallel vector_length (64)
#pragma acc loop worker
for (unsigned int i = 0; i < 32; i++)
#pragma acc loop vector
for (unsigned int j = 0; j < 64; j++)
;
return 0;
}
...
If num_workers is 16, either because:
- we add a "num_workers (16)" clause on the parallel directive, or
- we set "GOMP_OPENACC_DIM=:16:", or
- the libgomp plugin chooses 16 num_workers
we run into an illegal instruction at runtime, because a bar.sync instruction
tries to use a barrier 16. The instruction is illegal, because ptx supports
only 16 barriers per CTA, and the valid range is 0..15.
The problem is that with a warp-multiple vector length, we use a code generation
scheme with a per-worker barrier. And because barrier zero is reserved for
per-cta barrier, only the remaining 15 barriers can be used as per-worker
barrier, and consequently we can't use num_workers larger than 15.
This problem occurs only for vector_length 64. For vector_length 32, we use a
different code generation scheme, and for vector_length >= 96, the maximum
num_workers is not big enough not to trigger this problem.
Also, this problem only occurs for num_workers 16. As explained above,
num_workers 15 is safe to use, and 16 is already the maximum num_workers for
vector_length 64.
This patch fixes the problem in both the compiler (handling "num_workers (16)")
and in the libgomp nvptx plugin (with and without "GOMP_OPENACC_DIM=:16:").
2019-01-11 Tom de Vries <tdevries@suse.de>
* config/nvptx/nvptx.c (PTX_CTA_NUM_BARRIERS, PTX_PER_CTA_BARRIER)
(PTX_NUM_PER_CTA_BARRIER, PTX_FIRST_PER_WORKER_BARRIER)
(PTX_NUM_PER_WORKER_BARRIERS): Define.
(nvptx_apply_dim_limits): Prevent vector_length 64 and
num_workers 16.
* plugin/plugin-nvptx.c (nvptx_exec): Prevent vector_length 64 and
num_workers 16.
From-SVN: r267838
When using a compiler build with:
...
+#define PTX_DEFAULT_VECTOR_LENGTH PTX_CTA_SIZE
+#define PTX_MAX_VECTOR_LENGTH PTX_CTA_SIZE
...
and running the libgomp testsuite, we run into an execution failure in
parallel-loop-1.c, due to a cuda launch failure:
...
nvptx_exec: kernel f6_none_none$_omp_fn$0: launch gangs=480, workers=0, \
vectors=1024
libgomp: cuLaunchKernel error: invalid argument
...
because workers == 0.
The workers variable is set to 0 here in nvptx_exec:
...
workers = blocks / actual_vectors;
...
because actual_vectors is 1024, and blocks is 768:
...
cuOccupancyMaxPotentialBlockSize: grid = 10, block = 768
...
Fix this by ensuring that workers is at least one.
2019-01-09 Tom de Vries <tdevries@suse.de>
* plugin/plugin-nvptx.c (nvptx_exec): Make sure to launch with at least
one worker.
From-SVN: r267746
An OpenACC async queue is always synchronized with itself, so invocations like
"#pragma acc wait(0) async(0)", or "acc_wait_async (0, 0)" don't make a lot of
sense, but are still valid.
libgomp/
PR libgomp/88495
* plugin/plugin-nvptx.c (nvptx_wait_async): Don't refuse
"identical parameters".
* testsuite/libgomp.oacc-c-c++-common/asyncwait-nop-1.c: Update.
* testsuite/libgomp.oacc-c-c++-common/lib-80.c: Remove.
From-SVN: r267152
Per my reading of the OpenACC specification (and as supported by secondary
documentation, such as code examples, or presentations), it's valid to call
"acc_get_cuda_stream"/"acc_set_cuda_stream" also with "acc_async_sync",
"acc_async_noval" arguments, not just with the nonnegative values as currently
implemented.
libgomp/
PR libgomp/88370
* libgomp.texi (acc_get_current_cuda_context, acc_get_cuda_stream)
(acc_set_cuda_stream): Clarify.
* oacc-cuda.c (acc_get_cuda_stream, acc_set_cuda_stream): Use
"async_valid_p".
* plugin/plugin-nvptx.c (nvptx_set_cuda_stream): Refuse "async ==
acc_async_sync".
* testsuite/libgomp.oacc-c-c++-common/acc_set_cuda_stream-1.c: New file.
* testsuite/libgomp.oacc-c-c++-common/async_queue-1.c: Likewise.
* testsuite/libgomp.oacc-c-c++-common/lib-84.c: Update.
* testsuite/libgomp.oacc-c-c++-common/lib-85.c: Likewise.
From-SVN: r267147
