The analysis for const-ness of local variables required a view generally useful
matchers that are extracted into its own patch.
They are decompositionDecl and forEachArgumentWithParamType, that works
for calls through function pointers as well.
This is a reupload of https://reviews.llvm.org/D72505, that already landed,
but had to be reverted due to a GCC crash on powerpc
(https://reviews.llvm.org/rG4c48ea68e491cb42f1b5d43ffba89f6a7f0dadc4)
Because this took a long time to adress, i decided to redo this patch and
have a clean workflow.
I try to coordinate with someone that has a PPC to apply this patch and
test for the crash. If everything is fine, I intend to just commit.
If the crash is still happening, i hope to at least find the cause.
Differential Revision: https://reviews.llvm.org/D87588
Code was added that used llvm error checking to parse .debug_aranges, but the error check after parsing the DWARFDebugArangesSet was reversed and was causing no error to be returned with no valid address ranges being actually used. This meant we always would fall back onto creating out own address ranges by parsing the compile unit's ranges. This was causing problems for cases where the DW_TAG_compile_unit had a single address range by using a DW_AT_low_pc and DW_AT_high_pc attribute pair (not using a DW_AT_ranges attribute), but the .debug_aranges had correct split ranges. In this case we would end up using the single range for the compile unit that encompassed all of the ranges from the .debug_aranges section and would cause address resolving issues in LLDB where address lookups would fail for certain addresses.
Differential Revision: https://reviews.llvm.org/D87626
This patch is a first draft of a new pass that adds a more flexible way
to eliminate compares based on more complex constraints collected from
dominating conditions.
In particular, it aims at simplifying conditions of the forms below
using a forward propagation approach, rather than instcomine-style
ad-hoc backwards walking of def-use chains.
if (x < y)
if (y < z)
if (x < z) <- simplify
or
if (x + 2 < y)
if (x + 1 < y) <- simplify assuming no wraps
The general approach is to collect conditions and blocks, sort them by
dominance and then iterate over the sorted list. Conditions are turned
into a linear inequality and add it to a system containing the linear
inequalities that hold on entry to the block. For blocks, we check each
compare against the system and see if it is implied by the constraints
in the system.
We also keep a stack of processed conditions and remove conditions from
the stack and the constraint system once they go out-of-scope (= do not
dominate the current block any longer).
Currently there still are the least the following areas for improvements
* Currently large unsigned constants cannot be added to the system
(coefficients must be represented as integers)
* The way constraints are managed currently is not very optimized.
Reviewed By: spatel
Differential Revision: https://reviews.llvm.org/D84547
Bugpoint has lots of assumptions and hacks around the legacy PM, put off migrating it to NPM until later.
Fixes tests under BugPoint under NPM.
Reviewed By: arsenm
Differential Revision: https://reviews.llvm.org/D87655
Otherwise, we're declaring a non-static member function, and that
gives errors in C++11 because of the change of semantics between
C++11 and C++14 for non-const constexpr member functions.
This was always intended to be a friend declaration.
Was missing MODULE_ALIAS_ANALYSIS, previously only FUNCTION_ALIAS_ANALYSIS was taken into account.
Reviewed By: asbirlea
Differential Revision: https://reviews.llvm.org/D87664
Pin RUN lines with -analyze to legacy PM, add corresponding NPM RUN line if missing.
Reviewed By: asbirlea
Differential Revision: https://reviews.llvm.org/D87658
Local vars, marked with pragma allocate, mustbe allocate by the call of
the runtime function and cannot be allocated as other local variables.
Instead, we allocate a space for the pointer in private record and store
the address, returned by kmpc_alloc call in this pointer.
So, for untied tasks
```
#pragma omp task untied
{
S s;
#pragma omp allocate(s) allocator(allocator)
s = x;
}
```
compiler generates something like this:
```
struct task_with_privates {
S *ptr;
};
void entry(task_with_privates *p) {
S *s = p->s;
switch(partid) {
case 1:
p->s = (S*)kmpc_alloc();
kmpc_omp_task();
br exit;
case 2:
*s = x;
kmpc_omp_task();
br exit;
case 2:
~S(s);
kmpc_free((void*)s);
br exit;
}
exit:
}
```
Reviewed By: jdoerfert
Differential Revision: https://reviews.llvm.org/D86558
The test case has a check line for the option on a line that includes
the string lld surrounded by any characters. This causes failures
when said string is in the build path. What the test case presumably
means to test is the actual invocation of the LLD linker (i.e. a
linker that has that string as a suffix). This patch simply removes
the erroneous wildcard after the string.
Similar to the tsan suppression in
`Utils/VNCoercion.cpp:getLoadLoadClobberFullWidthSize` (rL175034; load widening used by GVN),
the D81766 optimization should be suppressed under tsan due to potential
spurious data race reports:
struct A {
int i;
const short s; // the load cannot be vectorized because
int modify; // it overlaps with bytes being concurrently modified
long pad1, pad2;
};
// __tsan_read16 does not know that some bytes are undef and accessing is safe
Similarly, under asan, users can mark memory regions with
`__asan_poison_memory_region`. A widened load can lead to a spurious
use-after-poison error. hwasan/memtag should be similarly suppressed.
`mustSuppressSpeculation` suppresses asan/hwasan/tsan but not memtag, so
we need to exclude memtag in `vectorizeLoadInsert`.
Note, memtag suppression can be relaxed if the load is aligned to the
its granule (usually 16), but that is out of scope of this patch.
Reviewed By: spatel, vitalybuka
Differential Revision: https://reviews.llvm.org/D87538
Make it possible to run the script command with a different language
than currently selected.
$ ./bin/lldb -l python
(lldb) script -l lua
>>> io.stdout:write("Hello, World!\n")
Hello, World!
When passing the language option and a raw command, you need to separate
the flag from the script code with --.
$ ./bin/lldb -l python
(lldb) script -l lua -- io.stdout:write("Hello, World!\n")
Hello, World!
Differential revision: https://reviews.llvm.org/D86996
PR47534 exposes a case where calling lowerShuffleWithSHUFPS directly from a derived repeated mask (found by is128BitLaneRepeatedShuffleMask) results in us using an non-canonicalized mask.
The missed canonicalization in this case is trivial - just commute the mask so we have more (swapped) LHS than RHS references so lowerShuffleWithSHUFPS can handle it.
The summary and very short discussion in D82122 summarizes whats happening here.
In short, liveness talks about variables, or expressions, anything that
has a value. Well, statements just simply don't have a one.
Differential Revision: https://reviews.llvm.org/D82598
Add a verifier for the loop op in the OpenACC dialect. Check basic restriction
from 2.9 Loop construct from the OpenACC 3.0 specs.
Reviewed By: ftynse
Differential Revision: https://reviews.llvm.org/D87546
Forward declare AAResults instead of the (old) AliasAnalysis type.
Remove includes from SLPVectorizer.cpp that are already included in SLPVectorizer.h.
https://llvm.org/PR45099 notes (correctly) that we're inconsistent in memory
allocation in `std::any`. We allocate memory with `std::allocator<T>::allocate`,
construct with placement new, destroy by calling the destructor directly, and
deallocate by calling `delete`. Most of those are customizable by the user,
but in different ways.
The standard is silent on how these things are to be accomplished.
This patch makes it so we use `allocator_traits<allocator<T>>` for all
of these operations (allocate, construct, destruct, deallocate).
This is, at least, consistent.
Fixes https://llvm.org/PR45099.
Differential Revision: https://reviews.llvm.org/D81133
If the constant operand is the opposite of the min/max value,
then the result must be the other value.
This is based on the similar codegen transform proposed in:
D87571
Currently we might derive the dynamic symbol table size from the DT_HASH hash table (using its `nchain` field).
It is possible to crash dumpers with a broken relocation that refers to a symbol with an index
that is too large. To trigger it, the inferred size of the dynamic symbol table should go past the end of the object.
This patch adds a size validation + warning.
Differential revision: https://reviews.llvm.org/D86923
Add the BufferSize argument constraint to fread and fwrite. This change
itself makes it possible to discover a security critical case, described
in SEI-CERT ARR38-C.
We also add the not-null constraint on the 3rd arguments.
In this patch, I also remove those lambdas that don't take any
parameters (Fwrite, Fread, Getc), thus making the code better
structured.
Differential Revision: https://reviews.llvm.org/D87081
Add Thread Local Storage Local Exec support to LLD. This is to support PC Relative addressing of Local Exec.
The patch teaches LLD to handle:
```
paddi r9, r13, x1@tprel
```
The relocation is:
```
R_PPC_TPREL34
```
Reviewed By: NeHuang, MaskRay
Differential Revision: https://reviews.llvm.org/D86608
2508ef01 doesn't totally fix the issue since we did not handle the case
when unused temporary negated result is the same with the result, which
is found by address sanitizer.
GetElementPtrInst::Create returns a GetElementPtrInst* so we don't need to cast. Similarly IntegerType inherits from the Type base class.
Also, I've used auto* in a few places to cleanup the code.
Helps fix some clang-tidy warnings which saw the dyn_casts and warned that these can return null.
The FailureReason input parameter maybe null, we check this in all other cases in the method but this one was missed somehow.
Fixes clang-tidy warning.
This patch adds the missing print for the vector_length in the parallel operation.
Reviewed By: ftynse
Differential Revision: https://reviews.llvm.org/D87630
This add canonicalizer for
- extracting an element from a dynamic_tensor_from_elements
- propagating constant operands to the type of dynamic_tensor_from_elements
Differential Revision: https://reviews.llvm.org/D87525
Muhammad Asif Manzoor