Summary:
Previously critical regions were emitted with the barrier making it a
worksharing construct though it is not. Also, it leads to incorrect
behavior in Cuda9+. Patch fixes this problem.
Reviewers: ABataev, jdoerfert
Subscribers: jholewinski, guansong, cfe-commits, grokos
Tags: #clang
Differential Revision: https://reviews.llvm.org/D66673
llvm-svn: 369946
non-ordered loops.
According to OpenMP 5.0, 2.9.2 Worksharing-Loop Construct, Desription, If the static schedule kind is specified or if the ordered clause is specified, and if the nonmonotonic modifier is not specified, the effect is as if the monotonic modifier is specified. Otherwise, unless the monotonic modifier is specified, the effect is as if the nonmonotonic modifier is specified.
The first part of this requirement is implemented in runtime. Patch adds
support for the second, nonmonotonic, part of this requirement.
llvm-svn: 369801
construct.
OpenMP 5.0 introduced new clause for declare target directive, device_type clause, which may accept values host, nohost, and any. Host means
that the function must be emitted only for the host, nohost - only for
the device, and any - for both, device and the host.
llvm-svn: 369775
For `map`, the following restriction changed in OpenMP 5.0:
* OpenMP 4.5 [2.15.5.1, Restrictions]: "A list item cannot appear in
both a map clause and a data-sharing attribute clause on the same
construct.
* OpenMP 5.0 [2.19.7.1, Restrictions]: "A list item cannot appear in
both a map clause and a data-sharing attribute clause on the same
construct unless the construct is a combined construct."
This patch removes this restriction in the case of combined constructs
and OpenMP 5.0, and it updates Sema not to capture a scalar by copy in
the target region when `firstprivate` and `map` appear for that scalar
on a combined target construct.
This patch also adds a fixme to a test that now reveals that a
diagnostic about loop iteration variables is dropped in the case of
OpenMP 5.0. That bug exists regardless of this patch's changes.
Reviewed By: ABataev, jdoerfert, hfinkel, kkwli0
Differential Revision: https://reviews.llvm.org/D65835
llvm-svn: 369619
If the function is marked as declare target in a standalone directive,
the delayed diagnostics is not emitted. Patch fixes this problem.
llvm-svn: 369432
target.
According to OpenMP 5.0, if a lambda declaration and definition appears between a declare target directive and the matching end declare target directive, all variables that are captured by the lambda expression must also appear in a to clause.
llvm-svn: 369146
Added basic support for non-rectangular loops. It requires an additional
analysis of min/max boundaries for non-rectangular loops. Since only
linear dependency is allowed, we can do this analysis.
llvm-svn: 368903
Summary:
This patch adds support for the close map modifier in Clang.
This ensures that the new map type is marked and passed to the OpenMP runtime appropriately.
Additional regression tests have been merged from patch D55892 (author @saghir).
Reviewers: ABataev, caomhin, jdoerfert, kkwli0
Reviewed By: ABataev
Subscribers: kkwli0, Hahnfeld, saghir, guansong, cfe-commits
Tags: #clang
Differential Revision: https://reviews.llvm.org/D65341
llvm-svn: 368491
Summary:
Added support for basic analysis of the linear variables and linear step
expression. Linear loop iteration variables must be excluded from this
analysis, only non-loop iteration variables must be analyzed.
Reviewers: NoQ
Subscribers: guansong, cfe-commits, caomhin, kkwli0
Tags: #clang
Differential Revision: https://reviews.llvm.org/D65461
llvm-svn: 368295
According to the OpenMP standard, compiler must define _OPENMP macro,
which has value in format yyyymm, where yyyy is the year of the standard
and mm is the month of the standard. For OpenMP 5.0 this value must be
set to 201811.
llvm-svn: 368170
This patch implements the code generation for OpenMP 5.0 declare mapper
(user-defined mapper) constructs. For each declare mapper, a mapper
function is generated. These mapper functions will be called by the
runtime and/or other mapper functions to achieve user defined mapping.
The design slides can be found at
https://github.com/lingda-li/public-sharing/blob/master/mapper_runtime_design.pptx
Re-commit after revert in r367773 because r367755 changed the LLVM-IR
output such that a CHECK line failed.
Patch by Lingda Li <lildmh@gmail.com>
Differential Revision: https://reviews.llvm.org/D59474
llvm-svn: 367905
This patch implements the code generation for OpenMP 5.0 declare mapper
(user-defined mapper) constructs. For each declare mapper, a mapper
function is generated. These mapper functions will be called by the
runtime and/or other mapper functions to achieve user defined mapping.
The design slides can be found at
https://github.com/lingda-li/public-sharing/blob/master/mapper_runtime_design.pptx
Patch by Lingda Li <lildmh@gmail.com>
Differential Revision: https://reviews.llvm.org/D59474
llvm-svn: 367773
For consistency with normal instructions and clarity when reading IR,
it's best to print the %0, %1, ... names of function arguments in
definitions.
Also modifies the parser to accept IR in that form for obvious reasons.
llvm-svn: 367755
Summary:
This patch fixes the case where variables in different compilation units or the same compilation unit are under the declare target link clause AND have the same name.
This also fixes the name clash error that occurs when unified memory is activated.
The changes in this patch include:
- Pointers to internal variables are given unique names.
- Externally visible variables are given the same name as before.
- All pointer variables (external or internal) are weakly linked.
Reviewers: ABataev, jdoerfert, caomhin
Reviewed By: ABataev
Subscribers: lebedev.ri, guansong, cfe-commits
Tags: #clang
Differential Revision: https://reviews.llvm.org/D64592
llvm-svn: 367613
Summary:
Reduction variables are the variables, for which the private copies
must be created in the OpenMP regions. Then they are initialized with
the predefined values depending on the reduction operation. After exit
from the OpenMP region the original variable is updated using the
reduction value and the value of the original reduction variable.
Reviewers: NoQ
Subscribers: guansong, jdoerfert, caomhin, kkwli0, cfe-commits
Tags: #clang
Differential Revision: https://reviews.llvm.org/D65106
llvm-svn: 367116
Summary:
Firstprivate variables are the variables, for which the private copies
must be created in the OpenMP regions and must be initialized with the
original values. Thus, we must report if the uninitialized variable is
used as firstprivate.
Reviewers: NoQ
Subscribers: guansong, jdoerfert, caomhin, kkwli0, cfe-commits
Tags: #clang
Differential Revision: https://reviews.llvm.org/D64765
llvm-svn: 366689
If the threadprivate variable is used in the copyin clause on inner
parallel directive with TLS support, we capture this variable in all
outer OpenMP scopes. It leads to the fact that in all scopes we're
working with the original variable, not the threadprivate copies.
llvm-svn: 366483
variables.
Loop control variables are private in loop-based constructs and we shall
take this into account when generate the code for inner constructs.
Currently, those variables are reported as shared in many cases. Moved
the analysis of the data-sharing attributes of the loop control variable
to an early semantic stage to correctly handle their attributes.
llvm-svn: 366474
checkDecl is only valid for VarDecls or FieldDecls, since getCanonicalDecl
expects only these. Prevent other Decl kinds (such as CXXMethodDecls and
EnumConstantDecls) from entering and asserting.
Differential Revision: https://reviews.llvm.org/D64842
llvm-svn: 366336
Summary:
Added support for analysis of if clauses in the OpenMP directives to be
able to check for the use of uninitialized variables.
Reviewers: NoQ
Subscribers: guansong, jfb, jdoerfert, caomhin, kkwli0, cfe-commits
Tags: clang
Differential Revision: https://reviews.llvm.org/D64646
llvm-svn: 366211
Summary:
Added support for analysis of if clauses in the OpenMP directives to be
able to check for the use of uninitialized variables.
Reviewers: NoQ
Subscribers: guansong, jfb, jdoerfert, caomhin, kkwli0, cfe-commits
Tags: #clang
Differential Revision: https://reviews.llvm.org/D64646
llvm-svn: 366068
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
Alexey Bataev