This implements the clang side of D116531. The elementtype
attribute is added for all indirect constraints (*) and tests are
updated accordingly.
Differential Revision: https://reviews.llvm.org/D116666
In 32bit mode, attaching TBAA metadata to the store following the call
to inline assembler results in describing the wrong type by making a
fake lvalue(i.e., whatever the inline assembler happens to leave in
EAX:EDX.) Even if inline assembler somehow describes the correct type,
setting TBAA information on return type of call to inline assembler is
likely not correct, since TBAA rules need not apply to inline assembler.
Differential Revision: https://reviews.llvm.org/D115320
Extension of D112504. Lower amdgpu printf to `__llvm_omp_vprintf`
which takes the same const char*, void* arguments as cuda vprintf and also
passes the size of the void* alloca which will be needed by a non-stub
implementation of `__llvm_omp_vprintf` for amdgpu.
This removes the amdgpu link error on any printf in a target region in favour
of silently compiling code that doesn't print anything to stdout.
Reviewed By: jdoerfert
Differential Revision: https://reviews.llvm.org/D112680
at the start of the entry block, which in turn would aid better code transformation/optimization.
Reviewed By: rnk
Differential Revision: https://reviews.llvm.org/D110257
Extension of D112504. Lower amdgpu printf to `__llvm_omp_vprintf`
which takes the same const char*, void* arguments as cuda vprintf and also
passes the size of the void* alloca which will be needed by a non-stub
implementation of `__llvm_omp_vprintf` for amdgpu.
This removes the amdgpu link error on any printf in a target region in favour
of silently compiling code that doesn't print anything to stdout.
The exact set of changes to check-openmp probably needs revision before commit
Reviewed By: jdoerfert
Differential Revision: https://reviews.llvm.org/D112680
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
Differential Revision: https://reviews.llvm.org/D102107
This patch supports OpenMP 5.0 metadirective features.
It is implemented keeping the OpenMP 5.1 features like dynamic user condition in mind.
A new function, getBestWhenMatchForContext, is defined in llvm/Frontend/OpenMP/OMPContext.h
Currently this function return the index of the when clause with the highest score from the ones applicable in the Context.
But this function is declared with an array which can be used in OpenMP 5.1 implementation to select all the valid when clauses which can be resolved in runtime. Currently this array is set to null by default and its implementation is left for future.
Reviewed By: jdoerfert
Differential Revision: https://reviews.llvm.org/D91944
This patch supports OpenMP 5.0 metadirective features.
It is implemented keeping the OpenMP 5.1 features like dynamic user condition in mind.
A new function, getBestWhenMatchForContext, is defined in llvm/Frontend/OpenMP/OMPContext.h
Currently this function return the index of the when clause with the highest score from the ones applicable in the Context.
But this function is declared with an array which can be used in OpenMP 5.1 implementation to select all the valid when clauses which can be resolved in runtime. Currently this array is set to null by default and its implementation is left for future.
Reviewed By: jdoerfert
Differential Revision: https://reviews.llvm.org/D91944
This patch supports OpenMP 5.0 metadirective features.
It is implemented keeping the OpenMP 5.1 features like dynamic user condition in mind.
A new function, getBestWhenMatchForContext, is defined in llvm/Frontend/OpenMP/OMPContext.h
Currently this function return the index of the when clause with the highest score from the ones applicable in the Context.
But this function is declared with an array which can be used in OpenMP 5.1 implementation to select all the valid when clauses which can be resolved in runtime. Currently this array is set to null by default and its implementation is left for future.
Reviewed By: jdoerfert
Differential Revision: https://reviews.llvm.org/D91944
Recommit of 707ce34b06. Don't introduce a
dependency to the LLVMPasses component, instead register the required
passes individually.
Add methods for loop unrolling to the OpenMPIRBuilder class and use them in Clang if `-fopenmp-enable-irbuilder` is enabled. The unrolling methods are:
* `unrollLoopFull`
* `unrollLoopPartial`
* `unrollLoopHeuristic`
`unrollLoopPartial` and `unrollLoopHeuristic` can use compiler heuristics to automatically determine the unroll factor. If possible, that is if no CanonicalLoopInfo is required to pass to another method, metadata for LLVM's LoopUnrollPass is added. Otherwise the unroll factor is determined using the same heurstics as user by LoopUnrollPass. Not requiring a CanonicalLoopInfo, especially with `unrollLoopHeuristic` allows greater flexibility.
With full unrolling and partial unrolling with known unroll factor, instead of duplicating instructions by the OpenMPIRBuilder, the full unroll is still delegated to the LoopUnrollPass. In case of partial unrolling the loop is first tiled using the existing `tileLoops` methods, then the inner loop fully unrolled using the same mechanism.
Reviewed By: jdoerfert, kiranchandramohan
Differential Revision: https://reviews.llvm.org/D107764
Add support for ordered directive in the OpenMPIRBuilder.
This patch also modidies clang to use the ordered directive when the
option -fopenmp-enable-irbuilder is enabled.
Also fix one ICE when parsing one canonical for loop with the relational
operator LE or GE in openmp region by replacing unary increment
operation of the expression of the variable "Expr A" minus the variable
"Expr B" (++(Expr A - Expr B)) with binary addition operation of the
experssion of the variable "Expr A" minus the variable "Expr B" and the
expression with constant value "1" (Expr A - Expr B + "1").
Reviewed By: Meinersbur, kiranchandramohan
Differential Revision: https://reviews.llvm.org/D107430
Breaks build with -DBUILD_SHARED_LIBS=ON
```
CMake Error: The inter-target dependency graph contains the following strongly connected component (cycle):
"LLVMFrontendOpenMP" of type SHARED_LIBRARY
depends on "LLVMPasses" (weak)
"LLVMipo" of type SHARED_LIBRARY
depends on "LLVMFrontendOpenMP" (weak)
"LLVMCoroutines" of type SHARED_LIBRARY
depends on "LLVMipo" (weak)
"LLVMPasses" of type SHARED_LIBRARY
depends on "LLVMCoroutines" (weak)
depends on "LLVMipo" (weak)
At least one of these targets is not a STATIC_LIBRARY. Cyclic dependencies are allowed only among static libraries.
CMake Generate step failed. Build files cannot be regenerated correctly.
```
This reverts commit 707ce34b06.
Add methods for loop unrolling to the OpenMPIRBuilder class and use them in Clang if `-fopenmp-enable-irbuilder` is enabled. The unrolling methods are:
* `unrollLoopFull`
* `unrollLoopPartial`
* `unrollLoopHeuristic`
`unrollLoopPartial` and `unrollLoopHeuristic` can use compiler heuristics to automatically determine the unroll factor. If possible, that is if no CanonicalLoopInfo is required to pass to another method, metadata for LLVM's LoopUnrollPass is added. Otherwise the unroll factor is determined using the same heurstics as user by LoopUnrollPass. Not requiring a CanonicalLoopInfo, especially with `unrollLoopHeuristic` allows greater flexibility.
With full unrolling and partial unrolling with known unroll factor, instead of duplicating instructions by the OpenMPIRBuilder, the full unroll is still delegated to the LoopUnrollPass. In case of partial unrolling the loop is first tiled using the existing `tileLoops` methods, then the inner loop fully unrolled using the same mechanism.
Reviewed By: jdoerfert, kiranchandramohan
Differential Revision: https://reviews.llvm.org/D107764
The purpose of __attribute__((disable_sanitizer_instrumentation)) is to
prevent all kinds of sanitizer instrumentation applied to a certain
function, Objective-C method, or global variable.
The no_sanitize(...) attribute drops instrumentation checks, but may
still insert code preventing false positive reports. In some cases
though (e.g. when building Linux kernel with -fsanitize=kernel-memory
or -fsanitize=thread) the users may want to avoid any kind of
instrumentation.
Differential Revision: https://reviews.llvm.org/D108029
@kpn pointed out that the global variable initialization functions didn't
have the "strictfp" metadata set correctly, and @rjmccall said that there
was buggy code in SetFPModel and StartFunction, this patch is to solve
those problems. When Sema creates a FunctionDecl, it sets the
FunctionDeclBits.UsesFPIntrin to "true" if the lexical FP settings
(i.e. a combination of command line options and #pragma float_control
settings) correspond to ConstrainedFP mode. That bit is used when CodeGen
starts codegen for a llvm function, and it translates into the
"strictfp" function attribute. See bugs.llvm.org/show_bug.cgi?id=44571
Reviewed By: Aaron Ballman
Differential Revision: https://reviews.llvm.org/D102343
Summary:
The AIX linker will produce errors on unresolved weak symbols. Change the
generated code to not check for the initialization function but just call
it and ensure that it always exists. Also, the AIX atexit routine has a
different name (and signature) so call it correctly. Update the lit tests
to test on AIX appropriately.
Author: Jamie Schmeiser <schmeise@ca.ibm.com>
Reviewed By: hubert.reinterpretcast (Hubert Tong)
Differential Revision: https://reviews.llvm.org/D104420
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
Differential Revision: https://reviews.llvm.org/D102107
The codegen for simd constructs was affected by the presence (or
absence) of the 'monotonic' schedule modifier for worksharing
loops. The modifier is only intended to apply to the scheduling of
chunks for a thread, not iterations of a loop inside a chunk.
In addition, the monotonic modifier was applied to worksharing loops
by default if no schedule clause was present; the referenced part of
the OpenMP 4.5 spec in the code (section 2.7.1) only applies if the
user specified a schedule clause with a static kind but no modifier.
Without a user-specified schedule clause we should default to
nonmonotonic scheduling.
Reviewed By: ABataev
Differential Revision: https://reviews.llvm.org/D103793
Implementation of the unroll directive introduced in OpenMP 5.1. Follows the approach from D76342 for the tile directive (i.e. AST-based, not using the OpenMPIRBuilder). Tries to use `llvm.loop.unroll.*` metadata where possible, but has to fall back to an AST representation of the outer loop if the partially unrolled generated loop is associated with another directive (because it needs to compute the number of iterations).
Reviewed By: ABataev
Differential Revision: https://reviews.llvm.org/D99459
If the memory object is scalable type, we do not know the exact size of
it at compile time. Set the size of lifetime marker to unknown if the
object is scalable one.
Differential Revision: https://reviews.llvm.org/D102822
This patch is the Part-1 (FE Clang) implementation of HW Exception handling.
This new feature adds the support of Hardware Exception for Microsoft Windows
SEH (Structured Exception Handling).
This is the first step of this project; only X86_64 target is enabled in this patch.
Compiler options:
For clang-cl.exe, the option is -EHa, the same as MSVC.
For clang.exe, the extra option is -fasync-exceptions,
plus -triple x86_64-windows -fexceptions and -fcxx-exceptions as usual.
NOTE:: Without the -EHa or -fasync-exceptions, this patch is a NO-DIFF change.
The rules for C code:
For C-code, one way (MSVC approach) to achieve SEH -EHa semantic is to follow
three rules:
* First, no exception can move in or out of _try region., i.e., no "potential
faulty instruction can be moved across _try boundary.
* Second, the order of exceptions for instructions 'directly' under a _try
must be preserved (not applied to those in callees).
* Finally, global states (local/global/heap variables) that can be read
outside of _try region must be updated in memory (not just in register)
before the subsequent exception occurs.
The impact to C++ code:
Although SEH is a feature for C code, -EHa does have a profound effect on C++
side. When a C++ function (in the same compilation unit with option -EHa ) is
called by a SEH C function, a hardware exception occurs in C++ code can also
be handled properly by an upstream SEH _try-handler or a C++ catch(...).
As such, when that happens in the middle of an object's life scope, the dtor
must be invoked the same way as C++ Synchronous Exception during unwinding
process.
Design:
A natural way to achieve the rules above in LLVM today is to allow an EH edge
added on memory/computation instruction (previous iload/istore idea) so that
exception path is modeled in Flow graph preciously. However, tracking every
single memory instruction and potential faulty instruction can create many
Invokes, complicate flow graph and possibly result in negative performance
impact for downstream optimization and code generation. Making all
optimizations be aware of the new semantic is also substantial.
This design does not intend to model exception path at instruction level.
Instead, the proposed design tracks and reports EH state at BLOCK-level to
reduce the complexity of flow graph and minimize the performance-impact on CPP
code under -EHa option.
One key element of this design is the ability to compute State number at
block-level. Our algorithm is based on the following rationales:
A _try scope is always a SEME (Single Entry Multiple Exits) region as jumping
into a _try is not allowed. The single entry must start with a seh_try_begin()
invoke with a correct State number that is the initial state of the SEME.
Through control-flow, state number is propagated into all blocks. Side exits
marked by seh_try_end() will unwind to parent state based on existing
SEHUnwindMap[].
Note side exits can ONLY jump into parent scopes (lower state number).
Thus, when a block succeeds various states from its predecessors, the lowest
State triumphs others. If some exits flow to unreachable, propagation on those
paths terminate, not affecting remaining blocks.
For CPP code, object lifetime region is usually a SEME as SEH _try.
However there is one rare exception: jumping into a lifetime that has Dtor but
has no Ctor is warned, but allowed:
Warning: jump bypasses variable with a non-trivial destructor
In that case, the region is actually a MEME (multiple entry multiple exits).
Our solution is to inject a eha_scope_begin() invoke in the side entry block to
ensure a correct State.
Implementation:
Part-1: Clang implementation described below.
Two intrinsic are created to track CPP object scopes; eha_scope_begin() and eha_scope_end().
_scope_begin() is immediately added after ctor() is called and EHStack is pushed.
So it must be an invoke, not a call. With that it's also guaranteed an
EH-cleanup-pad is created regardless whether there exists a call in this scope.
_scope_end is added before dtor(). These two intrinsics make the computation of
Block-State possible in downstream code gen pass, even in the presence of
ctor/dtor inlining.
Two intrinsic, seh_try_begin() and seh_try_end(), are added for C-code to mark
_try boundary and to prevent from exceptions being moved across _try boundary.
All memory instructions inside a _try are considered as 'volatile' to assure
2nd and 3rd rules for C-code above. This is a little sub-optimized. But it's
acceptable as the amount of code directly under _try is very small.
Part-2 (will be in Part-2 patch): LLVM implementation described below.
For both C++ & C-code, the state of each block is computed at the same place in
BE (WinEHPreparing pass) where all other EH tables/maps are calculated.
In addition to _scope_begin & _scope_end, the computation of block state also
rely on the existing State tracking code (UnwindMap and InvokeStateMap).
For both C++ & C-code, the state of each block with potential trap instruction
is marked and reported in DAG Instruction Selection pass, the same place where
the state for -EHsc (synchronous exceptions) is done.
If the first instruction in a reported block scope can trap, a Nop is injected
before this instruction. This nop is needed to accommodate LLVM Windows EH
implementation, in which the address in IPToState table is offset by +1.
(note the purpose of that is to ensure the return address of a call is in the
same scope as the call address.
The handler for catch(...) for -EHa must handle HW exception. So it is
'adjective' flag is reset (it cannot be IsStdDotDot (0x40) that only catches
C++ exceptions).
Suppress push/popTerminate() scope (from noexcept/noTHrow) so that HW
exceptions can be passed through.
Original llvm-dev [RFC] discussions can be found in these two threads below:
https://lists.llvm.org/pipermail/llvm-dev/2020-March/140541.htmlhttps://lists.llvm.org/pipermail/llvm-dev/2020-April/141338.html
Differential Revision: https://reviews.llvm.org/D80344/new/
Currently Clang does not add mustprogress to inifinite loops with a
known constant condition, matching C11 behavior. The forward progress
guarantee in C++11 and later should allow us to add mustprogress to any
loop (http://eel.is/c++draft/intro.progress#1).
This allows us to simplify the code dealing with adding mustprogress a
bit.
Reviewed By: aaron.ballman, lebedev.ri
Differential Revision: https://reviews.llvm.org/D96418
This is a Clang-only change and depends on the existing "musttail"
support already implemented in LLVM.
The [[clang::musttail]] attribute goes on a return statement, not
a function definition. There are several constraints that the user
must follow when using [[clang::musttail]], and these constraints
are verified by Sema.
Tail calls are supported on regular function calls, calls through a
function pointer, member function calls, and even pointer to member.
Future work would be to throw a warning if a users tries to pass
a pointer or reference to a local variable through a musttail call.
Reviewed By: rsmith
Differential Revision: https://reviews.llvm.org/D99517
The first one is the real parameters of the coroutine function, the
other one just for copying parameters to the coroutine frame.
Considering the following c++ code:
```
struct coro {
...
};
coro foo(struct test & t) {
...
co_await suspend_always();
...
co_await suspend_always();
...
co_await suspend_always();
}
int main(int argc, char *argv[]) {
auto c = foo(...);
c.handle.resume();
...
}
```
Function foo is the standard coroutine function, and it has only
one parameter named t (ignoring this at first),
when we use the llvm code to compile this function, we can get the
following ir:
```
!2921 = distinct !DISubprogram(name: "foo", linkageName:
"_ZN6Object3fooE4test", scope: !2211, file: !45, li\
ne: 48, type: !2329, scopeLine: 48, flags: DIFlagPrototyped |
DIFlagAllCallsDescribed, spFlags: DISPFlagDefi\
nition | DISPFlagOptimized, unit: !44, declaration: !2328,
retainedNodes: !2922)
!2924 = !DILocalVariable(name: "t", arg: 2, scope: !2921, file: !45,
line: 48, type: !838)
...
!2926 = !DILocalVariable(name: "t", scope: !2921, type: !838, flags:
DIFlagArtificial)
```
We can find there are two `the same` DIVariable named t in the same
dwarf scope for foo.resume.
And when we try to use llvm-dwarfdump to dump the dwarf info of this
elf, we get the following output:
```
0x00006684: DW_TAG_subprogram
DW_AT_low_pc (0x00000000004013a0)
DW_AT_high_pc (0x00000000004013a8)
DW_AT_frame_base (DW_OP_reg7 RSP)
DW_AT_object_pointer (0x0000669c)
DW_AT_GNU_all_call_sites (true)
DW_AT_specification (0x00005b5c "_ZN6Object3fooE4test")
0x000066a5: DW_TAG_formal_parameter
DW_AT_name ("t")
DW_AT_decl_file ("/disk1/yifeng.dongyifeng/my_code/llvm/build/bin/coro-debug-1.cpp")
DW_AT_decl_line (48)
DW_AT_type (0x00004146 "test")
0x000066ba: DW_TAG_variable
DW_AT_name ("t")
DW_AT_type (0x00004146 "test")
DW_AT_artificial (true)
```
The elf also has two 't' in the same scope.
But unluckily, it might let the debugger
confused. And failed to print parameters for O0 or above.
This patch will make coroutine parameters and move
parameters use the same DIVar and try to fix the problems
that I mentioned before.
Test Plan: check-clang
Reviewed By: aprantl, jmorse
Differential Revision: https://reviews.llvm.org/D97533
tl;dr Correct implementation of Corouintes requires having lifetime intrinsics available.
Coroutine functions are functions that can be suspended and resumed latter. To do so, data that need to stay alive after suspension must be put on the heap (i.e. the coroutine frame).
The optimizer is responsible for analyzing each AllocaInst and figure out whether it should be put on the stack or the frame.
In most cases, for data that we are unable to accurately analyze lifetime, we can just conservatively put them on the heap.
Unfortunately, there exists a few cases where certain data MUST be put on the stack, not on the heap. Without lifetime intrinsics, we are unable to correctly analyze those data's lifetime.
To dig into more details, there exists cases where at certain code points, the current coroutine frame may have already been destroyed. Hence no frame access would be allowed beyond that point.
The following is a common code pattern called "Symmetric Transfer" in coroutine:
```
auto tmp = await_suspend();
__builtin_coro_resume(tmp.address());
return;
```
In the above code example, `await_suspend()` returns a new coroutine handle, which we will obtain the address and then resume that coroutine. This essentially "transfered" from the current coroutine to a different coroutine.
During the call to `await_suspend()`, the current coroutine may be destroyed, which should be fine because we are not accessing any data afterwards.
However when LLVM is emitting IR for the above code, it needs to emit an AllocaInst for `tmp`. It will then call the `address` function on tmp. `address` function is a member function of coroutine, and there is no way for the LLVM optimizer to know that it does not capture the `tmp` pointer. So when the optimizer looks at it, it has to conservatively assume that `tmp` may escape and hence put it on the heap. Furthermore, in some cases `address` call would be inlined, which will generate a bunch of store/load instructions that move the `tmp` pointer around. Those stores will also make the compiler to think that `tmp` might escape.
To summarize, it's really difficult for the mid-end to figure out that the `tmp` data is short-lived.
I made some attempt in D98638, but it appears to be way too complex and is basically doing the same thing as inserting lifetime intrinsics in coroutines.
Also, for reference, we already force emitting lifetime intrinsics in O0 for AlwaysInliner: https://github.com/llvm/llvm-project/blob/main/llvm/lib/Passes/PassBuilder.cpp#L1893
Differential Revision: https://reviews.llvm.org/D99227
Arthur Eubanks