Fixed the processing of the unsupported VLAs in the reduction clauses.
Used targetDiag if the diagnostics can be delayed and emit it
immediately if the target does not support VLAs and we're parsing target
directive with the reduction clauses.
llvm-svn: 365821
Summary:
Some OpenMP clauses rely on the values of the variables. If the variable
is not initialized and used in OpenMP clauses that depend on the
variables values, it should be reported that the uninitialized variable
is used in the OpenMP clause expression.
This patch adds initial processing for uninitialized variables in OpenMP
constructs. Currently, it checks for use of the uninitialized variables
in the structured blocks.
Reviewers: NoQ, Szelethus, dcoughlin, xazax.hun, a.sidorin, george.karpenkov, szepet
Subscribers: rnkovacs, guansong, jfb, jdoerfert, cfe-commits
Tags: #clang
Differential Revision: https://reviews.llvm.org/D64356
llvm-svn: 365786
by David Truby.
Summary:
This adds a zero length array section mapping for each pointer captured by a lambda that is used in a target region, as per section 2.19.7.1 of the OpenMP 5 specification.
Reviewers: ABataev
Reviewed By: ABataev
Subscribers: guansong, jdoerfert, cfe-commits
Tags: #clang
Differential Revision: https://reviews.llvm.org/D64558
llvm-svn: 365777
The device should use the same float point representation as the host.
Previous patch fixed the handling of the sizes of the float point types,
but did not fixed the fp semantics. This patch makes target device to
use the host fp semantics. this is required for the correct data
transfer between host and device and correct codegen.
llvm-svn: 365485
Previously, lambda captures were processed in the function called during
capturing the variables. It leads to the recursive functions calls and
may result in the compiler crash.
llvm-svn: 364820
If the variable is used in the OpenMP region implicitly, we need to
check the data-sharing attributes for such variables and generate
implicit clauses for them. Patch improves analysis of such variables for
better handling of data-sharing rules.
llvm-svn: 364683
Fixed handling of the data-sharing attributes for static members when
requesting top most attribute. Previously, it might return the incorrect
attributes for static members if they were overriden in the outer
constructs.
llvm-svn: 364655
According to the OpenMP 5.0 standard, the loop iteration variable in the associated
for-loop of a simd construct with just one associated for-loop may be
listed in a private, lastprivate, or linear clause with a linear-step
that is the increment of the associated for-loop. Also, the loop
teration variables in the associated for-loops of a simd construct with
multiple associated for-loops may be listed in a private or lastprivate
clause.
llvm-svn: 364650
The errors for incorrectly specified data-sharing attributes for simd
constructs must be emitted only for the explicitly provided clauses, not
the predetermined ones.
llvm-svn: 364647
Target-based runtime functions use int64_t type for sizes, while the
compiler uses size_t type. It leads to miscompilation in 32 bit mode.
llvm-svn: 364327
If the variably modified type is declared outside of the captured region
and then used in the cast expression along with array subscript
expression, the type is not captured and it leads to the compiler crash.
llvm-svn: 364080
Summary:
This patch adds support for the handling of the variables under the declare target to clause.
The variables in this case are handled like link variables are. A pointer is created on the host and then mapped to the device. The runtime will then copy the address of the host variable in the device pointer.
Reviewers: ABataev, AlexEichenberger, caomhin
Reviewed By: ABataev
Subscribers: guansong, jdoerfert, cfe-commits
Tags: #clang
Differential Revision: https://reviews.llvm.org/D63108
llvm-svn: 363959
Summary:
This patch strengthens the tests introduced in D63009 by:
- adding new test for default device ID.
- modifying existing tests to pass device ID local variable to the task allocation function.
Reviewers: ABataev, Hahnfeld, caomhin, jdoerfert
Reviewed By: ABataev
Subscribers: guansong, jdoerfert, cfe-commits
Tags: #clang
Differential Revision: https://reviews.llvm.org/D63454
llvm-svn: 363809
Device have to use the same mangling as the host for 128bit float types. Otherwise, the codegen for the device is unable to find the parent function when it tries to generate the outlined function for the target region and it leads to incorrect compilation and crash at the runtime.
llvm-svn: 363734
If the host uses 128 bit long doubles, the compiler should generate correct code for NVPTX devices. If the return type has 128 bit long doubles, in LLVM IR this type must be coerced to int array instead.
llvm-svn: 363720
The device code must use the same long double type as the host.
Otherwise the code cannot be linked and executed properly. Patch adds
only basic support and checks for supporting of the host long double
double on the device.
llvm-svn: 363717
Summary: This patch avoids the emission of maps for target link variables when unified memory is present.
Reviewers: ABataev, caomhin
Reviewed By: ABataev
Subscribers: guansong, jdoerfert, cfe-commits
Tags: #clang
Differential Revision: https://reviews.llvm.org/D60883
llvm-svn: 363435
LLVM IR recently added a Type parameter to the byval Attribute, so that
when pointers become opaque and no longer have an element type the
information will still be present in IR.
For now the Type parameter is optional (which is why Clang didn't need
this change at the time), but it will become mandatory soon.
llvm-svn: 362652
the captured region scope.
This removes a case where we would build expressions (and mark
declarations odr-used) in the wrong scope.
Remove the now-unused 'capture initializer' field on sema::Capture
(except for 'this' captures, which still need to be cleaned up).
No functionality change intended (except that we now very slightly more
precisely determine whether we need to use a capture or not when another
captured region encloses an OpenMP captured region).
llvm-svn: 362179
This patch adjusts `PragmaOpenMPHandler` to set the location of
`tok::annot_pragma_openmp` to the `#pragma` location instead of the
`omp` location so that the former becomes the start location of the
OpenMP AST node. This can be useful when, for example, rewriting a
directive using Clang's Rewrite facility. Most of this patch updates
tests for changes to locations in diagnostics and `-ast-dump` output.
Reviewed By: ABataev, lebedev.ri, Meinersbur, aaron.ballman
Differential Revision: https://reviews.llvm.org/D61509
llvm-svn: 361867
Summary:
This patch adds a test for requires with unified share memory clause when a declare target link is present.
This test needs to go in prior to changes to declare target link for comparison purposes.
Reviewers: ABataev, caomhin
Reviewed By: ABataev
Subscribers: guansong, jdoerfert, cfe-commits
Tags: #clang
Differential Revision: https://reviews.llvm.org/D62407
llvm-svn: 361658
If the variable is a firstprivate variable and it was not emitted beause
this a constant variable with the constant initializer, we can use the
initial value instead of the variable itself. It also fixes the problem
with the compiler crash in this case.
llvm-svn: 361564
Summary:
This patch adds support for the registration of the requires directives with the runtime.
Each requires directive clause will enable a particular flag to be set.
The set of flags is passed to the runtime to be checked for compatibility with other such flags coming from other object files.
The registration function is called whenever OpenMP is present even if a requires directive is not present. This helps detect cases in which requires directives are used inconsistently.
Reviewers: ABataev, AlexEichenberger, caomhin
Reviewed By: ABataev, AlexEichenberger
Subscribers: jholewinski, guansong, jfb, jdoerfert, cfe-commits
Tags: #clang
Differential Revision: https://reviews.llvm.org/D60568
llvm-svn: 361298
Currently, we ignore all dso locality attributes/info when building for
the device and thus all symblos are externally visible and can be
preemted at the runtime. It may lead to incorrect results. We need to
follow the same logic, compiler uses for static/pie builds.
llvm-svn: 361283
performance.
Internally generated functions must be marked as always_inlines in most
cases. Patch marks some extra reduction function + outlined parallel
functions as always_inline for better performance, but only if the
optimization is requested.
llvm-svn: 361269
If the combined directive has default(none) clause and has clauses for
inner directive that reference some variables, for which data-sharing
attributes are not specified, the error messages should be emitted for
such variables.
llvm-svn: 360365
If the default(none) was specified for the construct, we might miss
diagnostic for the globals without explicitly specified data-sharing
attributes. Patch fixes this problem.
llvm-svn: 360362
This has introduced (exposed?) a crash in clang sema,
that does not happen without this patch.
I'll followup in the original bugreport and commit with reproducer.
This reverts commit r360061.
llvm-svn: 360327
This implementation isn't sound as per the standard.
It erroneously diagnoses e.g. the following case:
```
$ cat test.cpp
void f(int n) {
#pragma omp parallel default(none) if(n)
;
}
```
```
$ ./bin/clang -fopenmp test.cpp
test.cpp:2:40: error: variable 'n' must have explicitly specified data sharing attributes
#pragma omp parallel default(none) if(n)
^
test.cpp:2:31: note: explicit data sharing attribute requested here
#pragma omp parallel default(none) if(n)
^
1 error generated.
```
As per OpenMP Application Programming Interface Version 5.0 November 2018:
* 2.19.4.1default Clause
The default clause explicitly determines the data-sharing attributes of
variables that are referenced *in a parallel, teams, or task generating
construct and would otherwise be implicitly determined
(see Section 2.19.1.1 on page 270).
* 2.6.1 Determining the Number of Threads for a parallel Region
Using a variable in an if or num_threads clause expression of a parallel
construct causes an implicit reference to the variable in all enclosing
constructs. The if clause expression and the num_threads clause expression
are evaluated in the context outside of the parallel construct,
This reverts commit r360073.
llvm-svn: 360326
default(none).
If the combined directive has default(none) clause and has clauses for
inner directive that reference some variables, for which data-sharing
attributes are not specified, the error messages should be emitted for
such variables.
llvm-svn: 360073
If the `default(none)` was specified for the construct, we might miss
diagnostic for the globals without explicitly specified data-sharing
attributes. Patch fixes this problem.
llvm-svn: 360061
counters.
According to the OpenMP 5.0, For any associated loop where the b or lb
expression is not loop invariant with respect to the outermost loop, the
var-outer that appears in the expression may not have a random access
iterator type.
llvm-svn: 359340
loop nests.
Added a checks that the initializer/condition expressions depend only
only of the single previous loop iteration variable.
llvm-svn: 359200
Without this patch, APSInt inherits APInt::isNegative, which merely
checks the sign bit without regard to whether the type is actually
signed. isNonNegative and isStrictlyPositive call isNegative and so
are also affected.
This patch adjusts APSInt to override isNegative, isNonNegative, and
isStrictlyPositive with implementations that consider whether the type
is signed.
A large set of Clang OpenMP tests are affected. Without this patch,
these tests assume that `true` is not a valid argument for clauses
like `collapse`. Indeed, `true` fails APInt::isStrictlyPositive but
not APSInt::isStrictlyPositive. This patch adjusts those tests to
assume `true` should be accepted.
This patch also adds tests revealing various other similar fixes due
to APSInt::isNegative calls in Clang's ExprConstant.cpp and
SemaExpr.cpp: `++` and `--` overflow in `constexpr`, evaluated object
size based on `alloc_size`, `<<` and `>>` shift count validation, and
OpenMP array section validation.
Reviewed By: lebedev.ri, ABataev, hfinkel
Differential Revision: https://reviews.llvm.org/D59712
llvm-svn: 359012
runtime.
target [teams distribute] simd costructs do not require full runtime for
the correct execution, we can run them without full runtime.
llvm-svn: 358766
Summary: The requires directive containing target related clauses must appear before any target region in the compilation unit.
Reviewers: ABataev, AlexEichenberger, caomhin
Reviewed By: ABataev
Subscribers: guansong, jfb, jdoerfert, cfe-commits
Tags: #clang
Differential Revision: https://reviews.llvm.org/D60875
llvm-svn: 358709
All target-parallel-based constructs can be run in SPMD mode from now
on. Even if num_threads clauses or if clauses are used, such constructs
can be executed in SPMD mode.
llvm-svn: 358595
Combined constructs with parallel and if clauses without modifiers may
be executed in SPMD mode since if the condition is true for the target
region, it is also true for parallel region and the threads must be run
in parallel.
llvm-svn: 358503
mode.
After the previous patch with the more correct handling of the number of
threads in parallel regions, the parallel regions with num_threads
clauses can be executed in SPMD mode.
llvm-svn: 358445
regions.
Added more complex analysis for number of teams and number of threads in
the target regions, also merged related common code between CGOpenMPRuntime
and CGOpenMPRuntimeNVPTX classes.
llvm-svn: 358126
Added special processing of the memory management directives/clauses for
NVPTX target. For private locals, omp_default_mem_alloc and
omp_thread_mem_alloc result in allocation in local memory.
omp_const_mem_alloc allocates const memory, omp_teams_mem_alloc
allocates shared memory, and omp_cgroup_mem_alloc and
omp_large_cap_mem_alloc allocate global memory.
llvm-svn: 357923
If the pointer is captured by reference, it must be mapped as
_PTR_AND_OBJ kind of mapping to correctly translate the pointer address
on the device.
llvm-svn: 357488
Before this patch, CGLoop would dump all transformations for a loop into
a single LoopID without encoding any order in which to apply them.
rL348944 added the possibility to encode a transformation order using
followup-attributes.
When a loop has more than one transformation, use the follow-up
attribute define the order in which they are applied. The emitted order
is the defacto order as defined by the current LLVM pass pipeline,
which is:
LoopFullUnrollPass
LoopDistributePass
LoopVectorizePass
LoopUnrollAndJamPass
LoopUnrollPass
MachinePipeliner
This patch should therefore not change the assembly output, assuming
that all explicit transformations can be applied, and no implicit
transformations in-between. In the former case,
WarnMissedTransformationsPass should emit a warning (except for
MachinePipeliner which is not implemented yet). The latter could be
avoided by adding 'llvm.loop.disable_nonforced' attributes.
Because LoopUnrollAndJamPass processes a loop nest, generation of the
MDNode is delayed to after the inner loop metadata have been processed.
A temporary LoopID is therefore used to annotate instructions and
RAUW'ed by the actual LoopID later.
Differential Revision: https://reviews.llvm.org/D57978
llvm-svn: 357415
According to OpenMP 5.0, 2.11.4 allocate Clause, Restrictions, allocate
clauses that appear on a target construct or on constructs in a target
region must specify an allocator expression unless a requires directive
with the dynamic_allocators clause is present in the same compilation
unit. Patch adds a check for this restriction.
llvm-svn: 357412
According to OpenMP 5.0 standard, 2.11.4 allocate Clause, Restrictions,
For any list item that is specified in the allocate clause on a
directive, a data-sharing attribute clause that may create a private
copy of that list item must be specified on the same directive. Patch
adds the checks for this restriction.
llvm-svn: 357390
target and task-based directives.
According to OpenMP 5.0, 2.11.4 allocate Clause, Restrictions, For task,
taskloop or target directives, allocation requests to memory allocators
with the trait access set to thread result in unspecified behavior.
Patch introduces a check for omp_thread_mem_alloc predefined allocator
on target- and trask-based directives.
llvm-svn: 357205
dynamic_allocators.
According to the OpenMP 5.0, 2.11.3 allocate Directive, Restrictions,
allocate directives that appear in a target region must specify an
allocator clause unless a requires directive with the dynamic_allocators
clause is present in the same compilation unit. Patch adds a check for a
presence of the requires directive with the dynamic_allocators clause.
llvm-svn: 356758
clause in target region.
According to the OpenMP 5.0, 2.11.3 allocate Directive, Restrictions,
allocate directives that appear in a target region must specify an
allocator clause unless a requires directive with the dynamic_allocators
clause is present in the same compilation unit.
llvm-svn: 356752
For the global variables the allocate directive must specify only the
predefined allocator. This allocator must be translated into the correct
form of the address space for the targets that support different address
spaces.
llvm-svn: 356702
allocators.
It is better to deduce omp_allocator_handle_t type from the predefined
allocators, because omp.h header might not define it explicitly. Plus,
it allows to identify the predefined allocators correctly when trying to
build the allcoator for the global variables.
llvm-svn: 356607
If the allocator was specified for the variable and next one is found
with the different allocator, the warning is emitted, and the allocator
is ignored.
llvm-svn: 356513
According to OpenMP, 2.11.3 allocate Directive, Restrictions, C / C++,
if a list item has a static storage type, the allocator expression in
the allocator clause must be a constant expression that evaluates to
one of the predefined memory allocator values. Added check for this
restriction.
llvm-svn: 356496
Added initial codegen for the local variables with the #pragma omp
allocate directive. Instead of allocating the variables on the stack,
__kmpc_alloc|__kmpc_free functions are used for memory (de-)allocation.
llvm-svn: 356472
If the doacross lop construct is used and the loop counter is declare
outside of the loop, the compiler might crash trying to get the address
of the loop counter. Patch fixes this problem.
llvm-svn: 356198
array.
If the firstprivate variable is a reference, we may incorrectly classify
the kind of the private copy. Use the type of the private copy instead
of the original shared variable.
llvm-svn: 356098
'_openmp_teams_reductions_buffer_$_.
nvlink does not handle weak linkage correctly, same symbols with the
different sizes are reported as erroneous though the largest size must
be chosen instead. Patch fixes this problem by using Internal linkage
instead of the Common.
llvm-svn: 356072
If the variable was declared and marked as declare target, a new offload
entry with size 0 is created. But if later a definition is created and
marked as declare target, this definition is not added to the entry set
and the definition remains not mapped to the target. Patch fixes this
problem allowing to redefine the size and linkage for
previously registered declaration.
llvm-svn: 355960
If the declare target link global is used in the target region
indirectly (used in the inner parallel, teams, etc. regions), we may
miss this variable and it leads to incorrect codegen.
llvm-svn: 355858
memory.
If the variable with the constant non-scalar type is firstprivatized in
the target region, the local copy is created with the data copying.
Instead, we allocate the copy in the constant memory and avoid extra
copying in the outlined target regions. This global copy is used in the
target regions without loss of the performance.
llvm-svn: 355418
statements.
If the assembler instruction is not generated and the delayed diagnostic
is emitted, we may end up with extra warning message for variables used
in the asm statement. Since the asm statement is not built, the
variables may be left non-referenced and it may produce a warning about
a use of the non-initialized variables.
llvm-svn: 354928
This patch implements the parsing and sema support for the OpenMP
'from'-clause with potential user-defined mappers attached.
User-defined mappers are a new feature in OpenMP 5.0. A 'from'-clause
can have an explicit or implicit associated mapper, which instructs the
compiler to generate and use customized mapping functions. An example is
shown below:
struct S { int len; int *d; };
#pragma omp declare mapper(id: struct S s) map(s, s.d[0:s.len])
struct S ss;
#pragma omp target update from(mapper(id): ss) // use the mapper with name 'id' to map ss from device
Contributed-by: Lingda Li <lildmh@gmail.com>
Differential Revision: https://reviews.llvm.org/D58638
llvm-svn: 354817
This patch implements the parsing and sema support for OpenMP to clause
with potential user-defined mappers attached. User defined mapper is a
new feature in OpenMP 5.0. A to/from clause can have an explicit or
implicit associated mapper, which instructs the compiler to generate and
use customized mapping functions. An example is shown below:
struct S { int len; int *d; };
#pragma omp declare mapper(id: struct S s) map(s, s.d[0:s.len])
struct S ss;
#pragma omp target update to(mapper(id): ss) // use the mapper with name 'id' to map ss to device
Contributed-by: <lildmh@gmail.com>
Differential Revision: https://reviews.llvm.org/D58523
llvm-svn: 354698
Summary:
Added the ability to emit target-specific builtin assembler error
messages only in case if the function is really is going to be emitted
for the device.
Reviewers: rjmccall
Subscribers: guansong, jdoerfert, cfe-commits
Tags: #clang
Differential Revision: https://reviews.llvm.org/D58243
llvm-svn: 354486
A faster way to reduce the values in teams reductions was found, the
codegen is updated to use this faster algorithm and new runtime functions.
llvm-svn: 354479
This patch implements the parsing and sema support for OpenMP map
clauses with potential user-defined mapper attached. User defined mapper
is a new feature in OpenMP 5.0. A map clause can have an explicit or
implicit associated mapper, which instructs the compiler to generate
extra data mapping. An example is shown below:
struct S { int len; int *d; };
#pragma omp declare mapper(id: struct S s) map(s, s.d[0:s.len])
struct S ss;
#pragma omp target map(mapper(id) tofrom: ss) // use the mapper with name 'id' to map ss
Contributed-by: Lingda Li <lildmh@gmail.com>
Differential Revision: https://reviews.llvm.org/D58074
llvm-svn: 354347
expression is a discarded-value expression.
Summary:
We used to get this wrong in three ways:
1) During parsing, an expression-statement followed by the }) ending a
statement expression was always treated as producing the value of the
statement expression. That's wrong for ({ if (1) expr; })
2) During template instantiation, various kinds of statement (most
statements not appearing directly in a compound-statement) were not
treated as discarded-value expressions, resulting in missing volatile
loads (etc).
3) In all contexts, an expression-statement with attributes was not
treated as producing the value of the statement expression, eg
({ [[attr]] expr; }).
Also fix incorrect enforcement of OpenMP rule that directives can "only
be placed in the program at a position where ignoring or deleting the
directive would result in a program with correct syntax". In particular,
a label (be it goto, case, or default) should not affect whether
directives are permitted.
Reviewers: aaron.ballman, rjmccall
Subscribers: cfe-commits
Tags: #clang
Differential Revision: https://reviews.llvm.org/D57984
llvm-svn: 354090
Fixed diagnostic emission for the exceptions support in case of the
compilation of OpenMP code for the devices. From now on, it uses delayed
diagnostics mechanism, previously used for CUDA only. It allow to
diagnose not allowed used of exceptions only in functions that are going
to be codegen'ed.
llvm-svn: 353542
It is important to delay the emission of the diagnostic messages for the
functions unless it is proved that the function is going to be used on
the device side. It is required to support compilation with some of the
target-specific system headers.
llvm-svn: 353540
Some of these functions take some extraneous arguments, e.g. EltSize,
Offset, which are computable from the Type and DataLayout.
Add some asserts to ensure that the computed values are consistent
with the passed-in values, in preparation for eliminating the
extraneous arguments. This also asserts that the Type is an Array for
the calls named "Array" and a Struct for the calls named "Struct".
Then, correct a couple of errors:
1. Using CreateStructGEP on an array type. (this causes the majority
of the test differences, as struct GEPs are created with i32
indices, while array GEPs are created with i64 indices)
2. Passing the wrong Offset to CreateStructGEP in TargetInfo.cpp on
x86-64 NACL (which uses 32-bit pointers).
Differential Revision: https://reviews.llvm.org/D57766
llvm-svn: 353529
The fix is to issue error messages if there are more than one
teams construct inside a target constructs.
#pragma omp target
{
#pragma omp teams
{ ... }
#pragma omp teams
{ ... }
}
llvm-svn: 353186
Summary: this commit adds support to a new dependence type introduced in OpenMP
5.0. The LLVM OpenMP RTL already supports this feature, so we only need to
modify CLANG to take advantage of them.
Differential Revision: https://reviews.llvm.org/D57576
llvm-svn: 353018
This patch implements parsing and sema for "omp declare mapper"
directive. User defined mapper, i.e., declare mapper directive, is a new
feature in OpenMP 5.0. It is introduced to extend existing map clauses
for the purpose of simplifying the copy of complex data structures
between host and device (i.e., deep copy). An example is shown below:
struct S { int len; int *d; };
#pragma omp declare mapper(struct S s) map(s, s.d[0:s.len]) // Memory region that d points to is also mapped using this mapper.
Contributed-by: Lingda Li <lildmh@gmail.com>
Differential Revision: https://reviews.llvm.org/D56326
llvm-svn: 352906
required.
Function __kmpc_push_target_tripcount should be emitted only if the
offloading entry is going to be emitted (for use in tgt_target...
functions). Otherwise, it should not be emitted.
llvm-svn: 352669
We don't need to use the predetermined data-sharing attributes for the
loop counters if the user explicitly specified correct data-sharing
attributes for such variables.
llvm-svn: 352543
According to the report, better to keep the original strict compare
operation as the loop condition with unsigned loop counters to make the
loop countable. This allows further loop transformations.
llvm-svn: 352526
With commit r351627, LLVM gained the ability to apply (existing) IPO
optimizations on indirections through callbacks, or transitive calls.
The general idea is that we use an abstraction to hide the middle man
and represent the callback call in the context of the initial caller.
It is described in more detail in the commit message of the LLVM patch
r351627, the llvm::AbstractCallSite class description, and the
language reference section on callback-metadata.
This commit enables clang to emit !callback metadata that is
understood by LLVM. It does so in three different cases:
1) For known broker functions declarations that are directly
generated, e.g., __kmpc_fork_call for the OpenMP pragma parallel.
2) For known broker functions that are identified by their name and
source location through the builtin detection, e.g.,
pthread_create from the POSIX thread API.
3) For user annotated functions that carry the "callback(callee, ...)"
attribute. The attribute has to include the name, or index, of
the callback callee and how the passed arguments can be
identified (as many as the callback callee has). See the callback
attribute documentation for detailed information.
Differential Revision: https://reviews.llvm.org/D55483
llvm-svn: 351629
Summary: Change the strategy for computing loop index variables after collapsing a loop nest via the collapse clause by replacing the expensive remainder operation with multiplications and additions.
Reviewers: ABataev, caomhin
Reviewed By: ABataev
Subscribers: guansong, arphaman, cfe-commits
Differential Revision: https://reviews.llvm.org/D56413
llvm-svn: 350759
Summary: Introduce a compiler flag for cases when the user knows that the collapsed loop counter can be safely represented using at most 32 bits. This will prevent the emission of expensive mathematical operations (such as the div operation) on the iteration variable using 64 bits where 32 bit operations are sufficient.
Reviewers: ABataev, caomhin
Reviewed By: ABataev
Subscribers: hfinkel, kkwli0, guansong, cfe-commits
Differential Revision: https://reviews.llvm.org/D55928
llvm-svn: 350758
Summary:
There is a minor issue in how the implicit data-sharings for nested tasks are computed.
For the following example:
```
int x;
#pragma omp task shared(x)
#pragma omp task
x++;
```
We compute an implicit data-sharing of shared for `x` in the second task although I think that it should be firstprivate. Below you can find the part of the OpenMP spec that covers this example:
- // In a task generating construct, if no default clause is present, a variable for which the data-sharing attribute is not determined by the rules above and that in the enclosing context is determined to be shared by all implicit tasks bound to the current team is shared.//
- //In a task generating construct, if no default clause is present, a variable for which the data-sharing attribute is not determined by the rules above is firstprivate.//
Since each implicit-task has its own copy of `x`, we shouldn't apply the first rule.
Reviewers: ABataev
Reviewed By: ABataev
Subscribers: cfe-commits, rogfer01
Tags: #openmp
Differential Revision: https://reviews.llvm.org/D56430
llvm-svn: 350734
Each we create the target regions with the teams distribute inner
region, we can better estimate number of the teams required to execute
the target region. Function __kmpc_push_target_tripcount() is used for
purpose, which accepts device_id and the number of the iterations,
performed by the associated loop.
llvm-svn: 350571
As discussed in D56113, this patch refactors the implementation of the
const restriction for linear to reuse a function introduced by D56113.
A side effect is that, if a variable has mutable members, this
diagnostic is now skipped, and the diagnostic for the variable not
being an integer or pointer is reported instead.
Reviewed By: ABataev
Differential Revision: https://reviews.llvm.org/D56299
llvm-svn: 350441
As discussed in D56113, this patch refactors the implementation of the
const restriction for reductions to reuse a function introduced by
D56113. A side effect is that diagnostics sometimes now say
"variable" instead of "list item" when a list item is a variable.
Reviewed By: ABataev
Differential Revision: https://reviews.llvm.org/D56298
llvm-svn: 350440
The following appears in OpenMP 3.1 sec. 2.9.1.1 as a predetermined
data-sharing attribute:
> Variables with const-qualified type having no mutable member are
> shared.
It does not appear in OpenmP 4.0, 4.5, or 5.0. This patch removes the
implementation of that attribute when the requested OpenMP version is
greater than 3.1.
One effect of that removal is that `default(none)` affects const
variables without mutable members.
Also, without this patch, if a const variable without mutable members
was explicitly lastprivate or private, it was an error because it was
predetermined shared. Now, clang instead complains that it's const
without mutable fields, which is a more intelligible diagnostic. That
should be fine for all of the above versions because they all have
something like the following, which is quoted from OpenMP 5.0
sec. 2.19.3:
> A variable that is privatized must not have a const-qualified type
> unless it is of class type with a mutable member. This restriction does
> not apply to the firstprivate clause.
reduction and linear clauses already have separate checks for const
variables. Future patches will merge the implementations.
Reviewed By: ABataev
Differential Revision: https://reviews.llvm.org/D56113
llvm-svn: 350439
nvvm_barrier0.
Use runtime functions instead of the direct call to the nvvm intrinsics.
It allows to prevent some dangerous LLVM optimizations, that breaks the
code for the NVPTX target.
llvm-svn: 350328
For constants with the predefined data-sharing clauses we may had
troubles with the target combined directives. It may cause compiler
crash in some corner cases.
llvm-svn: 350127
Not all %itanium_abi_triple values support TLS. Makes
OpenMP/declare_reduction_codegen.cpp, OpenMP/parallel_copyin_codegen.cpp for
%itanium_abi_triples without TLS support.
Alternatively we could pass -fnoopenmp-use-tls and tweak some of the CHECK
lines, but this seems a bit simpler.
Fixes PR40156.
Differential Revision: https://reviews.llvm.org/D56086
llvm-svn: 350067
Instead of generating llvm.mem.parallel_loop_access metadata, generate
llvm.access.group on instructions and llvm.loop.parallel_accesses on
loops. There is one access group per generated loop.
This is clang part of D52116/r349725.
Differential Revision: https://reviews.llvm.org/D52117
llvm-svn: 349823
Need to mark the loop as started when the initialization statement is
found. It is required to prevent possible incorrect loop iteraton
variable detection during template instantiation and fix the compiler
crash during the codegen.
llvm-svn: 349657
Without this patch, clang doesn't complain that X needs explicit data
sharing attributes in the following:
```
#pragma omp target teams default(none)
{
#pragma omp parallel num_threads(X)
;
}
```
However, clang does produce that complaint after the braces are
removed. With this patch, clang complains in both cases.
Reviewed By: ABataev
Differential Revision: https://reviews.llvm.org/D55861
llvm-svn: 349635
A map clause with the close map-type-modifier is a hint to
prefer that the variables are mapped using a copy into faster
memory.
Patch by Ahsan Saghir (saghir)
Differential Revision: https://reviews.llvm.org/D55719
llvm-svn: 349551
buffer.
Seems to me, nvlink has a bug with the proper support of the weakly
linked symbols. It does not allow to define several shared memory buffer
with the different sizes even with the weak linkage. Instead we always
use 128 bytes buffer to prevent nvlink from the error message emission.
llvm-svn: 349540
Inlined runtime with the current implementation of the interwarp copy
function leads to the undefined behavior because of the not quite
correct implementation of the barriers. Start using generic
__kmpc_barier function instead of the custom made barriers.
llvm-svn: 349192
If the array section is based on pointer and this sections is mapped in
target region + then it is used in the inner parallel region, it also
must be globalized as the pointer itself is passed by value, not by
reference.
llvm-svn: 348492
Critical regions in NVPTX are the constructs, which, generally speaking,
are not supported by the NVPTX target. Instead we're using special
technique to handle the critical regions. Currently they are supported
only within the loop and all the threads in the loop must execute the
same critical region.
Inside of this special regions the regions still must be emitted as
critical, to avoid possible data races between the teams +
synchronization must use __kmpc_barrier functions.
llvm-svn: 348272
__kmpc_barrier runtime functions must be marked as convergent to prevent
some dangerous optimizations. Also, for NVPTX target all barriers must
be emitted as simple barriers.
llvm-svn: 348271
This moves everything primarily testing the functionality of -ast-dump and -ast-print into their own directory, rather than leaving the tests spread around the testing directory.
llvm-svn: 348017
Summary: This patch adds a new runtime for the SPMD deinit kernel function which replaces the previous function. The new function takes as argument the flag which signals whether the runtime is required or not. This enables the compiler to optimize out the part of the deinit function which are not needed.
Reviewers: ABataev, caomhin
Reviewed By: ABataev
Subscribers: jholewinski, guansong, cfe-commits
Differential Revision: https://reviews.llvm.org/D54970
llvm-svn: 347915
Fixed emission of the target regions found in the virtual functions.
Previously we may end up with the situation when those regions could be
skipped.
llvm-svn: 347793
modes.
If the region is inside target|teams|distribute region, we can emit the
locations with the correct info for execution mode and runtime mode.
Patch adds this ability to the NVPTX codegen to help the optimizer to
produce better code.
llvm-svn: 347583
For the NVPTX target default locations should be emitted as constants +
additional info must be emitted in the reserved_2 field of the ident_t
structure. The 1st bit controls the execution mode and the 2nd bit
controls use of the lightweight runtime. The combination of the bits for
Non-SPMD mode + lightweight runtime represents special undefined mode,
used outside of the target regions for orphaned directives or functions.
Should allow and additional optimization inside of the target regions.
llvm-svn: 347425
Loop-control variables with the default data-sharing attributes should
not be captured in the OpenMP region as they are private by default.
Also, default attributes should be emitted for such variables in the
inner OpenMP regions for the correct data sharing during codegen.
llvm-svn: 347409
forms of random access iterator
In OpenMP 4.5, only 4 relational operators are supported: <, <=, >,
and >=. This work is to enable support for relational operator
!= (not-equal) as one of the canonical forms.
Patch by Anh Tuyen Tran
Differential Revision: https://reviews.llvm.org/D54441
llvm-svn: 347405
This patch adjusts a test not to depend on deprecated FileCheck
behavior that permits overlapping matches within a block of CHECK-DAG
directives. Thus, this patch also removes uses of FileCheck's
-allow-deprecated-dag-overlap command-line option.
There were two issues in this test:
1. There were sets of patterns for store instructions in which a
pattern X could match a superset of a pattern Y. While X appeared
before Y, Y's intended match appeared before X's intended match. The
result was that X matched Y's intended match. Under the old
overlapping behavior, Y also matched Y's intended match. Under the
new non-overlapping behavior, Y had nothing left to match. This patch
fixes this by gathering these sets in one place and putting the most
specific patterns (Y) before the more general patterns (X).
2. The CHECK-DAG patterns involving the variables CBPADDR3 and
CBPADDR4 were the same, but there was only one match in the text, so
CBPADDR4 patterns had nothing to match under the new non-overlapping
behavior. Moreover, a preceding related series of directives had
variables (SADDR0, BPADDR0, etc.) numbered only 0 through 4, but this
series had variables numbered 0 through 5. Assuming CBPADDR4's
directives were not intended, this patch removes them.
Reviewed By: ABataev
Differential Revision: https://reviews.llvm.org/D54765
llvm-svn: 347351
This patch adjusts a test not to depend on deprecated FileCheck
behavior that permits overlapping matches within a block of CHECK-DAG
directives. Thus, this patch also removes uses of FileCheck's
-allow-deprecated-dag-overlap command-line option.
Specifically, the FileCheck variables DBG_LOC_START, DBG_LOC_END, and
DBG_LOC_CANCEL were all set to the same value. As a result, three
TERM_DEBUG-DAG patterns, one for each variable, all matched the same
text under the old overlapping behavior. Under the new
non-overlapping behavior, that's not permitted. This patch's solution
is to replace these variables with one variable and replace these
patterns with one pattern.
Reviewed By: ABataev
Differential Revision: https://reviews.llvm.org/D54764
llvm-svn: 347350
Alexey Bataev