Codegen support for reduction over private variable with reduction
clause. Section 7.6.10 in in OpenMP 6.0 spec.
- An internal shared copy is initialized with an initializer value.
- The shared copy is updated by combining its value with the values from
the private copies created by the clause.
- Once an encountering thread verifies that all updates are complete,
its original list item is updated by merging its value with that of the
shared copy and then broadcast to all threads.
Sample Test Case from OpenMP 6.0 Example
```
#include <assert.h>
#include <omp.h>
#define N 10
void do_red(int n, int *v, int &sum_v)
{
sum_v = 0; // sum_v is private
#pragma omp for reduction(original(private),+: sum_v)
for (int i = 0; i < n; i++)
{
sum_v += v[i];
}
}
int main(void)
{
int v[N];
for (int i = 0; i < N; i++)
v[i] = i;
#pragma omp parallel num_threads(4)
{
int s_v; // s_v is private
do_red(N, v, s_v);
assert(s_v == 45);
}
return 0;
}
```
Expected Codegen:
```
// A shared global/static variable is introduced for the reduction result.
// This variable is initialized (e.g., using memset or a UDR initializer)
// e.g., .omp.reduction.internal_private_var
// Barrier before any thread performs combination
call void @__kmpc_barrier(...)
// Initialization block (executed by thread 0)
// e.g., call void @llvm.memset.p0.i64(...) or call @udr_initializer(...)
call void @__kmpc_critical(...)
// Inside critical section:
// Load the current value from the shared variable
// Load the thread-local private variable's value
// Perform the reduction operation
// Store the result back to the shared variable
call void @__kmpc_end_critical(...)
// Barrier after all threads complete their combinations
call void @__kmpc_barrier(...)
// Broadcast phase:
// Load the final result from the shared variable)
// Store the final result to the original private variable in each thread
// Final barrier after broadcast
call void @__kmpc_barrier(...)
```
---------
Co-authored-by: Chandra Ghale <ghale@pe31.hpc.amslabs.hpecorp.net>
This patch introduces the `OpenMPIRBuilder::TargetKernelDefaultAttrs`
structure used to simplify passing default and constant values for
number of teams and threads, and possibly other target kernel-related
information in the future.
This is used to forward values passed to `createTarget` to
`createTargetInit`, which previously used a default unrelated set of
values.
This patch migrates the OpenMP UserDefinedMapper codegen from Clang to
the OpenMPIRBuilder. I will be adding further patches in the near future
so that OpenMP dialect in MLIR can make use of these.
Initial parsing/sema for 'strict' modifier with 'num_tasks' and
‘grainsize’ clause is present in these commits
[grainsize_parsing](ab9eac762c)
and
[num_tasks_parsing](56c1660170 (diff-4184486638e85284c3a2c961a81e7752231022daf97e411007c13a6732b50db9R6545))
. However, this implementation appears incomplete as it lacks code
generation support. A runtime patch was introduced in this runtime
commit
[runtime_patch](540007b427 (diff-5e95f9319910d6965d09c301359dbe6b23f3eef5ce4d262ef2c2d2137875b5c4R374))
, which adds a new API, _kmpc_taskloop_5, to accommodate the strict
modifier.
In this patch I have added codegen support. When the strict modifier is
present alongside the grainsize or num_tasks clauses of taskloop
construct, the code now emits a call to _kmpc_taskloop_5, which includes
an additional parameter of type i32 with the value 1 to indicate the
strict modifier. If the strict modifier is not present, it falls back to
the existing _kmpc_taskloop API call.
---------
Co-authored-by: Chandra Ghale <ghale@pe31.hpc.amslabs.hpecorp.net>
To authenticate pointers, CodeGen needs access to the key and
discriminators that were used to sign the pointer. That information is
sometimes known from the context, but not always, which is why `Address`
needs to hold that information.
This patch adds methods and data members to `Address`, which will be
needed in subsequent patches to authenticate signed pointers, and uses
the newly added methods throughout CodeGen. Although this patch isn't
strictly NFC as it causes CodeGen to use different code paths in some
cases (e.g., `mergeAddressesInConditionalExpr`), it doesn't cause any
changes in functionality as it doesn't add any information needed for
authentication.
In addition to the changes mentioned above, this patch introduces class
`RawAddress`, which contains a pointer that we know is unsigned, and
adds several new functions for creating `Address` and `LValue` objects.
This reapplies d9a685a9dd589486e882b722e513ee7b8c84870c, which was
reverted because it broke ubsan bots. There seems to be a bug in
coroutine code-gen, which is causing EmitTypeCheck to use the wrong
alignment. For now, pass alignment zero to EmitTypeCheck so that it can
compute the correct alignment based on the passed type (see function
EmitCXXMemberOrOperatorMemberCallExpr).
To authenticate pointers, CodeGen needs access to the key and
discriminators that were used to sign the pointer. That information is
sometimes known from the context, but not always, which is why `Address`
needs to hold that information.
This patch adds methods and data members to `Address`, which will be
needed in subsequent patches to authenticate signed pointers, and uses
the newly added methods throughout CodeGen. Although this patch isn't
strictly NFC as it causes CodeGen to use different code paths in some
cases (e.g., `mergeAddressesInConditionalExpr`), it doesn't cause any
changes in functionality as it doesn't add any information needed for
authentication.
In addition to the changes mentioned above, this patch introduces class
`RawAddress`, which contains a pointer that we know is unsigned, and
adds several new functions for creating `Address` and `LValue` objects.
This reapplies 8bd1f9116aab879183f34707e6d21c7051d083b6. The commit
broke msan bots because LValue::IsKnownNonNull was uninitialized.
To authenticate pointers, CodeGen needs access to the key and
discriminators that were used to sign the pointer. That information is
sometimes known from the context, but not always, which is why `Address`
needs to hold that information.
This patch adds methods and data members to `Address`, which will be
needed in subsequent patches to authenticate signed pointers, and uses
the newly added methods throughout CodeGen. Although this patch isn't
strictly NFC as it causes CodeGen to use different code paths in some
cases (e.g., `mergeAddressesInConditionalExpr`), it doesn't cause any
changes in functionality as it doesn't add any information needed for
authentication.
In addition to the changes mentioned above, this patch introduces class
`RawAddress`, which contains a pointer that we know is unsigned, and
adds several new functions for creating `Address` and `LValue` objects.
Summary:
Currently, OpenMP handles the `omp requires` clause by emitting a global
constructor into the runtime for every translation unit that requires
it. However, this is not a great solution because it prevents us from
having a defined order in which the runtime is accessed and used.
This patch changes the approach to no longer use global constructors,
but to instead group the flag with the other offloading entires that we
already handle. This has the effect of still registering each flag per
requires TU, but now we have a single constructor that handles
everything.
This function removes support for the old `__tgt_register_requires` and
replaces it with a warning message. We just had a recent release, and
the OpenMP policy for the past four releases since we switched to LLVM
is that we do not provide strict backwards compatibility between major
LLVM releases now that the library is versioned. This means that a user
will need to recompile if they have an old binary that relied on
`register_requires` having the old behavior. It is important that we
actively deprecate this, as otherwise it would not solve the problem of
having no defined init and shutdown order for `libomptarget`. The
problem of `libomptarget` not having a define init and shutdown order
cascades into a lot of other issues so I have a strong incentive to be
rid of it.
It is worth noting that the current `__tgt_offload_entry` only has space
for a 32-bit integer here. I am planning to overhaul these at some point
as well.
Summary:
This patch reworks how we handle global constructors in OpenMP.
Previously, we emitted individual kernels that were all registered and
called individually. In order to provide more generic support, this
patch moves all handling of this to the target backend and the runtime
plugin. This has the benefit of supporting the GNU extensions for
constructors an destructors, removing a class of failures related to
shared library destruction order, and allows targets other than OpenMP
to use the same support without needing to change the frontend.
This is primarily done by calling kernels that the backend emits to
iterate a list of ctor / dtor functions. For x64, this is automatic and
we get it for free with the standard `dlopen` handling. For AMDGPU, we
emit `amdgcn.device.init` and `amdgcn.device.fini` functions which
handle everything atuomatically and simply need to be called. For NVPTX,
a patch https://github.com/llvm/llvm-project/pull/71549 provides the
kernels to call, but the runtime needs to set up the array manually by
pulling out all the known constructor / destructor functions.
One concession that this patch requires is the change that for GPU
targets in OpenMP offloading we will use `llvm.global_dtors` instead of
using `atexit`. This is because `atexit` is a separate runtime function
that does not mesh well with the handling we're trying to do here. This
should be equivalent in all cases except for cases where we would need
to destruct manually such as:
```
struct S { ~S() { foo(); } };
void foo() {
static S s;
}
```
However this is broken in many other ways on the GPU, so it is not
regressing any support, simply increasing the scope of what we can
handle.
This changes the handling of ctors / dtors. This patch now outputs a
information message regarding the deprecation if the old format is used.
This will be completely removed in a later release.
Depends on: https://github.com/llvm/llvm-project/pull/71549
We used to pass the min/max threads/teams values through different paths
from the frontend to the middle end. This simplifies the situation by
passing the values once, only when we will create the KernelEnvironment,
which contains the values. At that point we also manifest the metadata,
as appropriate. Some footguns have also been removed, e.g., our target
check is now triple-based, not calling convention-based, as the latter
is dependent on the ordering of operations. The types of the values have
been unified to int32_t.
We now provide the information about the min/max thread and team count
from to the OMPIRBuilder, no matter what the source was. That means we
unify `thread_limit`, `num_teams`, `num_threads` handling with the
target specific attriutes (`__launch_bounds__` and
`amdgpu_flat_work_group_size`). This is in preparation to pass the
values to the runtime, and to allow the middle-end (OpenMP-opt) to
tighten the values if it seems appropriate. There is no "real" change
after this commit.
offloading
- This patch adds support for thread_limit clause on target directive according to OpenMP 51 [2.14.5]
- The idea is to create an outer task for target region, when there is a thread_limit clause, and manipulate the thread_limit of task instead. This way, thread_limit will be applied to all the relevant constructs enclosed by the target region.
Differential Revision: https://reviews.llvm.org/D152054
OpenMP 5.1 allows emission of the `indirect` clause on declare target
functions, see https://www.openmp.org/spec-html/5.1/openmpsu70.html#x98-1080002.14.7.
The intended use of this is to permit calling device functions via their
associated host pointer. In order to do this the first step will be
building a map associating these variables. Doing this will require the
same offloading entry handling we use for other kernels and globals.
We intentionally emit a new global on the device side. Although it's
possible to look up the device function's address directly, this would
require changing the visibility and would prevent us from making static
functions indirect. Also, the CUDA toolchain will optimize out unused
functions and using a global prevents that. The downside is that the
runtime will need to read the global and copy its value, but there
shouldn't be any other costs.
Note that this patch just performs the codegen, currently this new
offloading entry type is unused and will be ignored by the runtime.
Reviewed By: jdoerfert
Differential Revision: https://reviews.llvm.org/D157738
We used to have two separate implementations to derive the number of
threads used in a target region. This lead us to sometimes miss out on
user provided thread bounds (num_threads, or thread_limit) when we
looked for "constant default values". If we might miss out on the
presence of those bounds, we cannot set the thread_limit statically
since the runtime will try to honor user input rather than cap it at the
"preferred default". This patch replaces the secondary implementation
with the primary in a mode that will not emit code but just look for the
presence, and potentially upper bounds, of thread limiting clauses.
The runtime test would not pass without this rewrite as we missed some
clauses, set the static limit on the device to the preferred value, but
then violated that value at runtime.
Fixes: https://github.com/llvm/llvm-project/issues/64845
Differential Revision: https://reviews.llvm.org/D158381
Migrate createForStaticInitFunction, createDispatchInitFunction, createDispatchNextFunction and createDispatchFiniFunction from Clang CodeGen to OMPIRBuilder.
Differential Revision: https://reviews.llvm.org/D157994
This patch migrates the UseDevicePtr and UseDeviceAddr clause related code for handling privatisation from Clang codegen to the OMPIRBuilder
Depends on D150860
Reviewed By: jdoerfert
Differential Revision: https://reviews.llvm.org/D152554
This patch renames the `OpenMPIRBuilderConfig` flags to reduce confusion over
their meaning. `IsTargetCodegen` becomes `IsGPU`, whereas `IsEmbedded` becomes
`IsTargetDevice`. The `-fopenmp-is-device` compiler option is also renamed to
`-fopenmp-is-target-device` and the `omp.is_device` MLIR attribute is renamed
to `omp.is_target_device`. Getters and setters of all these renamed properties
are also updated accordingly. Many unit tests have been updated to use the new
names, but an alias for the `-fopenmp-is-device` option is created so that
external programs do not stop working after the name change.
`IsGPU` is set when the target triple is AMDGCN or NVIDIA PTX, and it is only
valid if `IsTargetDevice` is specified as well. `IsTargetDevice` is set by the
`-fopenmp-is-target-device` compiler frontend option, which is only added to
the OpenMP device invocation for offloading-enabled programs.
Differential Revision: https://reviews.llvm.org/D154591
clause.
This is just syntax to make it easier for the user. It doesn't add any
new functionality.
for
doacross(sink: omp_cur_iteration - 1)
Equivalent to
doacross(sink: ConterVar - 1, ...)
doacross(source: omp_cur_iteration)
Equivalent to
doacross(source)
And restriction is:
OMP5.2 p.327
If vector is specified with the omp_cur_iteration keyword and with
sink as the dependence-type then it must be omp_cur_iteration - 1.
If vector is specified with source as the dependence-type then it must be
omp_cur_iteration.
Differential Revision: https://reviews.llvm.org/D154556
This patch refactors the code generation that emits the offloading kernel
launch and moves the core portion to the OpenMPIRBuilder so that it can be used
from flang in the future.
Reviewed By: jdoerfert
Differential Revision: https://reviews.llvm.org/D151035
This allows the generation of OpenMP offload metadata for the OpenMP
dialect when lowering to LLVM-IR and moves some of the shared logic
between the OpenMP Dialect and Clang into the IRBuilder.
Reviewers: jsjodin, jdoerfert, kiranchandramohan
Differential Revision: https://reviews.llvm.org/D148370
This patch prefixes omp outlined helpers and reduction funcs
with the original function's name.
Reviewed By: jdoerfert
Differential Revision: https://reviews.llvm.org/D140722
This patch attempts to prefix omp outlined helpers and reduction funcs
with the original function's name.
Reviewed By: jdoerfert
Differential Revision: https://reviews.llvm.org/D140722
This patch adds the OffloadEntriesInfoManager to the OpenMPIRBuilder, and
allows the OffloadEntriesInfoManager to access the Configuration in the
OpenMPIRBuilder. With the shared Config there is no risk for inconsistencies,
and there is no longer the need for clang to have a separate
OffloadEntriesInfoManager.
Reviewed By: jdoerfert
Differential Revision: https://reviews.llvm.org/D146549
Support for taskwait nowait clause with placeholder for runtime changes.
Reviewed By: cchen, ABataev
Differential Revision: https://reviews.llvm.org/D131830
Added codegen for `omp error` directive.
This is to generate IR to call:
void __kmpc_error(ident_t *loc, int severity, const char *message);
Differential Revision: https://reviews.llvm.org/D139166
This patch mechanically replaces None with std::nullopt where the
compiler would warn if None were deprecated. The intent is to reduce
the amount of manual work required in migrating from Optional to
std::optional.
This is part of an effort to migrate from llvm::Optional to
std::optional:
https://discourse.llvm.org/t/deprecating-llvm-optional-x-hasvalue-getvalue-getvalueor/63716
This change moves the getName function from clang and moves the separator class
members from CGOpenMPRuntime into OMPIRBuilder. Also enusre all the getters
in the config class are const.
Reviewed By: jdoerfert
Differential Revision: https://reviews.llvm.org/D137725
This patch introudces the OpenMPIRBuilderConfig class which contains various
flags that are needed to lower OMP constructs to LLVM-IR. The purpose is to
keep the flags in one place so they do not have to be passed in every time.
The flags can be set optionally since some uses cases don't rely on functions
that depend on these flags.
Reviewed By: jdoerfert, tschuett
Differential Revision: https://reviews.llvm.org/D138220
This patch removes getOrCreateInternalVariable from Clang OMP CodeGen and replaces it's uses with OMPBuilder::getOrCreateInternalVariable. Also refactors OMPBuilder::getOrCreateInternalVariable to change type of name from Twine to StringRef
Differential Revision: https://reviews.llvm.org/D137720
This revision fixes typos where there are 2 consecutive words which are
duplicated. There should be no code changes in this revision (only
changes to comments and docs). Do let me know if there are any
undesirable changes in this revision. Thanks.
This patch moves the createOffloadEntriesAndInfoMetadata to OpenMPIRBuilder,
the createOffloadEntry helper function. The clang specific error handling is
invoked using a callback. This code will also be used by flang in the future.
This patch moves the implementation of the OffloadEntriesInfoManager
to the OMPIRbuilder. This class will later be used by flang as well.
Reviewed By: jdoerfert
Differential Revision: https://reviews.llvm.org/D135786
This patch moves the emitOffloadingArraysArgument function and
supporting data structures to OpenMPIRBuilder. This will later be used
in flang as well. The TargetDataInfo class was split up into generic
information and clang-specific data, which remain in clang. Further
migration will be done in in the future.
Reviewed By: jdoerfert
Differential Revision: https://reviews.llvm.org/D134662
Parallel regions are outlined as functions with capture variables explicitly generated as distinct parameters in the function's argument list. That complicates the fork_call interface in the OpenMP runtime: (1) the fork_call is variadic since there is a variable number of arguments to forward to the outlined function, (2) wrapping/unwrapping arguments happens in the OpenMP runtime, which is sub-optimal, has been a source of ABI bugs, and has a hardcoded limit (16) in the number of arguments, (3) forwarded arguments must cast to pointer types, which complicates debugging. This patch avoids those issues by aggregating captured arguments in a struct to pass to the fork_call.
Reviewed By: jdoerfert, jhuber6, ABataev
Differential Revision: https://reviews.llvm.org/D102107
Previously we added the `push_target_tripcount` function to send the
loop tripcount to the device runtime so we knew how to configure the
teams / threads for execute the loop for a teams distribute construct.
This was implemented as a separate function mostly to avoid changing the
interface for backwards compatbility. Now that we've changed it anyway
and the new interface can take an arbitrary number of arguments via the
struct without changing the ABI, we can move this to the new interface.
This will simplify the runtime by removing unnecessary state between
calls.
Depends on D128550
Reviewed By: jdoerfert
Differential Revision: https://reviews.llvm.org/D128816
This patch moves the logic for generating the offloading entries to the
OpenMPIRBuilder. This makes it easier to re-use in other places, such as
for OpenMP support in Flang or using the same method for generating
offloading entires for other languages like Cuda.
Reviewed By: tianshilei1992
Differential Revision: https://reviews.llvm.org/D123460
