As it contains a self-reference, the default copy/move ctors
would not be safe.
Move the DSEState::get() method into the ctor to make sure no move
occurs here even without NRVO.
This is a speculative fix for test failures on
llvm-clang-x86_64-expensive-checks-win.
Store queues in unique_ptr so they are destroyed when the global DeviceInfo is. Currently they leak which raises an assert in debug builds of hsa.
Reviewed By: pdhaliwal
Differential Revision: https://reviews.llvm.org/D109511
This is another regression noted with the proposal to canonicalize
to the min/max intrinsics in D98152.
Here are Alive2 attempts to show correctness without specifying
exact constants:
https://alive2.llvm.org/ce/z/bvfCwh (smax)
https://alive2.llvm.org/ce/z/of7eqy (smin)
https://alive2.llvm.org/ce/z/2Xtxoh (umax)
https://alive2.llvm.org/ce/z/Rm4Ad8 (umin)
(if you comment out the assume and/or no-wrap, you should see failures)
The different output for the umin test is due to a fold added with
c4fc2cb5b2 :
// umin(x, 1) == zext(x != 0)
We probably want to adjust that, so it applies more generally
(umax --> sext or patterns where we can fold to select-of-constants).
Some folds that were ok when starting with cmp+select may increase
instruction count for the equivalent intrinsic, so we have to decide
if it's worth altering a min/max.
Differential Revision: https://reviews.llvm.org/D110038
Inline assembly was not handled at all and treated like a llvm::Value.
In particular, it tried to create a pointer it which is not allowed.
Fix by handling like a llvm::Constant such that it is just reused when
required, instead of trying to marshall it in memory.
Fixes llvm.org/PR51960
VirtualUse ensures consistency over different source of values with
Polly. In particular, this enables its use of instructions moved between
Statement. Before the patch, the code wrongly assumed that the BB's
instructions are also the ScopStmt's instructions. Reference are
determined for OpenMP outlining and GPGPU kernel extraction.
GPGPU CodeGen had some problems. For one, it generated GPU kernel
parameters for constants. Second, it emitted GPU-side invariant loads
which have already been loaded by the host. This has been partially
fixed, it still generates a store for the invariant load result, but
using the value that the host has already written.
WARNING: I did not test the generated PollyACC code on an actual GPU.
The improved consistency will be made use of in the next patch.
The function was intended to catch OpenMP functions such as
get_thread_id(). If matched, the call would be considered synthesizable.
There were a few problems with this:
* get_thread_id() is not 'const' in the sense of have the gcc manual
defines it: "do not examine any values except their arguments".
get_thread_id() reads OpenCL runtime libreary global state.
What was inteded was probably 'speculable'.
* isConstCall was implemented using mayReadOrWriteMemory(). 'const' is
stricter than that, mayReadOrWriteMemory is e.g. true for malloc(),
since it may only read/write addresses that are considered
inaccessible fro the application. However, malloc is certainly not
speculable.
* Values that are isConstCall were not handled consistently throughout
Polly. In particular, it was not considered for referenced values
(OpenMP outlining and PollyACC).
Fix by removing special handling for isConstCall entirely.
Disconnect should block until handleDisconnect completes, otherwise we might
destroy the SimpleRemoteEPC instance while it's still in use.
Thanks to Dave Blaikie for helping me track this down.
Most architectures use .got instead of .got.plt, so switching the default can
minimize customization.
This fixes an issue for SPARC V9 which uses .got .
AVR, AMDGPU, and MSP430 don't seem to use _GLOBAL_OFFSET_TABLE_.
This is a followup to D109844 (and alternative to D109907), which
integrates the new "earliest escape" tracking into AliasAnalysis.
This is done by replacing the pre-existing context-free capture
cache in AAQueryInfo with a replaceable (virtual) object with two
implementations: The SimpleCaptureInfo implements the previous
behavior (check whether object is captured at all), while
EarliestEscapeInfo implements the new behavior from DSE.
This combines the "earliest escape" analysis with the full power of
BasicAA: It subsumes the call handling from D109907, considers a
wider range of escape sources, and works with AA recursion. The
compile-time cost is slightly higher than with D109907.
Differential Revision: https://reviews.llvm.org/D110368
It is sufficient that the object has not been captured before the
load that produces the pointer we're loading. A capture after that
can not affect the already loaded pointer.
This is small part of D110368 applied separately.
The case of an Argument and an identified function local is already
handled earlier, because we don't care about captures in that case.
As such, we don't need to additionally consider the combination of
an Argument with a non-escaping identified function local.
This ensures that isEscapeSource() only returns true for
instructions, which is necessary for D110368.
Renames StartAddress and EndAddress members to Start and End.
Adds contains and overlap methods.
Adds a constructor from an address and size.
These changes are counterparts to LLVM commits ef391df2b6, c0d889995e, and
37f1b7a3f3.
This reverts commit bef55a2b47 while I investigate
failures on some bots. Also reverts "[lli] Add ChildTarget dependence on
OrcTargetProcess library." (7a219d801b) which was
a fallow-up to bef55a2b47.
EPCGenericRTDyldMemoryMnaager is an EPC-based implementation of the
RuntimeDyld::MemoryManager interface. It enables remote-JITing via EPC (backed
by a SimpleExecutorMemoryManager instance on the executor side) for RuntimeDyld
clients.
The lli and lli-child-target tools are updated to use SimpleRemoteEPC and
SimpleRemoteEPCServer (rather than OrcRemoteTargetClient/Server), and
EPCGenericRTDyldMemoryManager for MCJIT tests.
By enabling remote-JITing for MCJIT and RuntimeDyld-based ORC clients,
EPCGenericRTDyldMemoryManager allows us to deprecate older remote-JITing
support, including OrcTargetClient/Server, OrcRPCExecutorProcessControl, and the
Orc RPC system itself. These will be removed in future patches.
One of the cases identified in PR45116 - we don't need to limit store narrowing to ABI alignment, we can use allowsMemoryAccess - which tests using getABITypeAlign, but also checks if a target permits (fast) misaligned memory access by checking allowsMisalignedMemoryAccesses as a fallback.
Earlier during the development of {D69764} I felt it was no longer necessary to
ensure we were not trying to change code which didn't need to change
and we felt this could be removed, however I'd like to bring this back for now
as I am seeing some false positives in terms of the "replacements"
What I see is the generation of a replacement which is a "No Op" on the original
code, I think this comes about because of the merging of replacements:
```
static const a;
->
const static a;
->
static const a;
```
The replacements don't really merge, in such a way as to identify when we have gone
back to the original
Also remove the Penalty as I'm not using it (and it became marked as set and no used,
I'd rather get rid of it if it means nothing)
I think we need to do this step for now, as many people use the --output-replacements-xml
to identify that the file "needs a clang-format"
The same can be seen with the -n or --dry-run option as this uses the replacements
to drive the error/warning output.
Reviewed By: HazardyKnusperkeks
Differential Revision: https://reviews.llvm.org/D110392
Update the costs to match the codegen from combineMulToPMADDWD - not only can we use PMADDWD is its zero-extended, but also if its a constant or sign-extended from a vXi16 (which can be replaced with a zero-extension).
If we are multiplying by a sign-extended vXi32 constant, then we can mask off the upper 16 bits to allow folding to PMADDWD and make use of its implicit sign-extension from i16
We already do this on SSE41 targets where we have sext/zext instructions, now that combineShiftToPMULH handles SSE2 targets, we can enable this here as well.
With improved shuffle combines (in particular canonicalizeShuffleWithBinOps), we can now usefully perform this on any SSE2+ target.
We should be able to remove this entirely and just use DAGCombiner's combineShiftToMULH if we can someday get it to support illegal (pre-widened) types.
Replace misc. StringConvert uses with llvm::to_integer()
and llvm::to_float(), except for cases where further refactoring is
planned. The purpose of this change is to eliminate the StringConvert
API that is duplicate to LLVM, and less correct in behavior at the same
time.
Differential Revision: https://reviews.llvm.org/D110447
This patch is part of the upstreaming effort from fir-dev branch.
Add a description for the fir.array_load opeartion and rename lenParams to typeparams.
Reviewed By: kiranchandramohan
Differential Revision: https://reviews.llvm.org/D110393
Co-authored-by: Eric Schweitz <eschweitz@nvidia.com>
Linking against the LibXml2::LibXml2 target has the advantage of not only importing the library, but also adding the include path as well as any definitions the library requires. In case of a static build of libxml2, eg. a define is set on Windows to remove any DLL imports and export.
LLVM already makes use of the target, but c-index-test and lldb were still linking against the library only.
The workaround for Mac OS-X that I removed seems to have also been made redundant since https://reviews.llvm.org/D84563 I believe
Differential Revision: https://reviews.llvm.org/D109975
This patch adds functionality to FlatAffineConstraints to remove local
variables using equalities. This helps in keeping output representation of
FlatAffineConstraints smaller.
This patch is part of a series of patches aimed at generalizing affine
dependence analysis.
Reviewed By: bondhugula
Differential Revision: https://reviews.llvm.org/D110056
Given:
- A jump table
- Which jumps to the next block
- The next block ends in a WLS
- Where the WLS conditionally jumps to block earlier in the program.
The Arm block placement pass would attempt to move the block containing
the WLS earlier, as the WLS instruction can only branch forward. In
doing so it would add a branch from the jumptable block to the WLS
block, thinking it previously fell-through.
This in itself would be fine, if a little inefficient, but the constant
island pass expects all instructions after a jump-table branch to have
been removed by analyzeBranch. So it gets confused and can assign the
same labels to multiple jump table blocks.
I've changed the condition to the same as used in analyzeBranch.
It's only used during race reporting.
There is no point in polluting the main header file with it.
Reviewed By: xgupta
Differential Revision: https://reviews.llvm.org/D110470
See PR51872 for the original repro.
This fixes a crash when converting a templated constructor into a deduction
guide, in case any of the template parameters were invalid.
Signed-off-by: Matheus Izvekov <mizvekov@gmail.com>
Reviewed By: rsmith
Differential Revision: https://reviews.llvm.org/D110460
We currently, incorrectly, assume that a range always has at least
one element when building a contiguous range. This commit adds
a proper empty check to avoid crashing.
Differential Revision: https://reviews.llvm.org/D110457
Nikita Popov