Summary:
The changes introduced in D87946 changed the API for libomptarget
functions. `__kmpc_push_target_tripcount` was a function in Clang 11.x
but was not given a backward-compatible interface. This change will
require people using Clang 13.x or 12.x to recompile their offloading
programs.
Reviewed By: jdoerfert cchen
Differential Revision: https://reviews.llvm.org/D98358
Starting with Power 10 the instruction paddi is available to use.
The instruction allows for immediates that are 34 bits.
This patch adds exploitation of the paddi instruction to allow us
to materialize constants.
Reviewed By: lei, amyk
Differential Revision: https://reviews.llvm.org/D93300
This is the minimal port from ELF. Any extension should easy from here
Test plan: ninja check-all-macho
Reviewed By: #lld-macho, thakis
Differential Revision: https://reviews.llvm.org/D98419
Add documentation that explains how to extend clang with the new
extensions/features. The guidelines also detail clang's position
about the extension pragmas for the new functionality.
Differential Revision: https://reviews.llvm.org/D97072
4c973ae implemented reduction of vector swap for lane-insensitive
operations. This commit fixes it for checking number of uses of the
vector operation.
This patch improves salvageDebugInfoImpl by allowing it to salvage arithmetic
operations with two or more non-const operands; this includes the GetElementPtr
instruction, and most Binary Operator instructions. These salvages produce
DIArgList locations and are only valid for dbg.values, as currently variadic
DIExpressions must use DW_OP_stack_value. This functionality is also only added
for salvageDebugInfoForDbgValues; other functions that directly call
salvageDebugInfoImpl (such as in ISel or Coroutine frame building) can be
updated in a later patch.
Differential Revision: https://reviews.llvm.org/D91722
If we encounter a degenerate select node where both operands are
the same, then we can continue negating the condition while swapping
operands, resulting in an infinite loop. Avoid this by bailing out
if both operands are the same.
Fixes https://bugs.llvm.org/show_bug.cgi?id=49509.
Differential Revision: https://reviews.llvm.org/D98340
When building builtins, the toolchain might not yet be at a stage
when linking a test application works yet, as builtins aren't
available. Therefore set CMAKE_TRY_COMPILE_TARGET_TYPE to STATIC_LIBRARY,
to avoid failing the compiler sanity check.
Setting CMAKE_TRY_COMPILE_TARGET_TYPE to STATIC_LIBRARY has the risk
of making checks for library availability succeed falsely (e.g.
indicating that libs would be available that really aren't, as the
tests don't do any linking), but the builtins library doesn't try to
link against any external libraries (and only produces static libraries
anyway), so it should be safe here.
This avoids having to set CMAKE_C_COMPILER_WORKS when bootstrapping a
cross toolchain, when building the builtins.
Differential Revision: https://reviews.llvm.org/D91334
This enables unifying command line flags with config options in clangd
internals. This patch changes behaviour in 2 places:
- BackgroundIndex was previously disabled when -remote-index was
provided. After this patch, it will be enabled but all files will have
bkgindex policy set to Skip.
- -index-file was loaded at startup (at least load was initiated), now
the load will happen through ProjectAwareIndex with first index query.
Unfortunately this doesn't simplify any options initially, as
- CompileCommandsDir is also used by clangd --check workflow, which
doesn't use configs.
- EnableBackgroundIndex option controls whether the component will be
created at all, which implies creation of extra threads registering a
listener for compilation database discoveries.
Differential Revision: https://reviews.llvm.org/D98029
Explicit specifier can only be mentioned on the in-line declaration of a
constructor, so don't carry it over to the definition.
Differential Revision: https://reviews.llvm.org/D98164
verifyCompatibleShapes is not transitive. Create an n-ary version and
update SameOperandShapes and SameOperandAndResultShapes traits to use
it.
Differential Revision: https://reviews.llvm.org/D98331
Also give CanonicalIncludes a less powerful interface (canonicalizes
symbols vs headers separately) so we can cache its results better.
Prior to this:
- path->uri conversions were not consistently cached, this is
particularly cheap when we start from a FileEntry* (which we often can)
- only a small fraction of header-to-include calculation was cached
This is a significant speedup at least for dynamic indexing of preambles.
On my machine, opening XRefs.cpp:
```
PreambleCallback 1.208 -> 1.019 (-15.7%)
BuildPreamble 5.538 -> 5.214 (-5.8%)
```
Differential Revision: https://reviews.llvm.org/D98371
Recently gc.result has been marked with readnone instead of readonly and
this opens a door for different optimization to duplicate gc.result.
Statepoint lowering is not ready to see several gc.results.
The problem appears when there are gc.results with one located in the same
basic block and another located in other basic block.
In this case we need both export VR and fill local setValue.
Note that this case is not sufficient optimization done before CodeGen.
It is evident that local gc.result dominates all other gc.results and it is handled
by GVN and EarlyCSE.
But anyway, even if IR is not optimal Backend should not crash on a valid IR.
Reviewers: reames, dantrushin
Reviewed By: dantrushin
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D98393
As pointed out by Joel E. Denny in D97845, the OverflowErrorStr variable
is misnamed because the error is raised for any parsing error. Note that
in FileCheck proper this only happens in case of (under|over)flow
because the regex will ensure a number in the correct format is matched.
Reviewed By: jdenny
Differential Revision: https://reviews.llvm.org/D98342
This moves code that sets the architecture name
and Float ABI into two new functions in
ToolChains/Arch/ARM.cpp. Greatly simplifying ComputeLLVMTriple.
Some light refactoring in setArchNameInTriple to
move local variables closer to their first use.
Reviewed By: ostannard
Differential Revision: https://reviews.llvm.org/D98253
Relative to the previous implementation, this always uses
aliasesUnknownInst() instead of aliasesPointer() to correctly
handle atomics. The added test case was previously miscompiled.
-----
Even when MemorySSA-based LICM is used, an AST is still populated
for scalar promotion. As the AST has quadratic complexity, a lot
of time is spent in this step despite the existing access count
limit. This patch optimizes the identification of promotable stores.
The idea here is pretty simple: We're only interested in must-alias
mod sets of loop invariant pointers. As such, only populate the AST
with loop-invariant loads and stores (anything else is definitely
not promotable) and then discard any sets which alias with any of
the remaining, definitely non-promotable accesses.
If we promoted something, check whether this has made some other
accesses loop invariant and thus possible promotion candidates.
This is much faster in practice, because we need to perform AA
queries for O(NumPromotable^2 + NumPromotable*NumNonPromotable)
instead of O(NumTotal^2), and NumPromotable tends to be small.
Additionally, promotable accesses have loop invariant pointers,
for which AA is cheaper.
This has a signicant positive compile-time impact. We save ~1.8%
geomean on CTMark at O3, with 6% on lencod in particular and 25%
on individual files.
Conceptually, this change is NFC, but may not be so in practice,
because the AST is only an approximation, and can produce
different results depending on the order in which accesses are
added. However, there is at least no impact on the number of promotions
(licm.NumPromoted) in test-suite O3 configuration with this change.
Differential Revision: https://reviews.llvm.org/D89264
The call to "set_curterm" inside the "terminalHasColors" function breaks
the EditLine configuration on some Linux distributions, causing certain
characters that have functions bound to them to not show up and
backspace to stop deleting characters (only visually). This patch
ensures that term struct is restored after the routine for cheking if
terminal supports colors is done, which fixes the aforementioned issue.
Reviewed By: labath
Differential Revision: https://reviews.llvm.org/D95230
Clean-up after D98279, remove one call to createConvertGPUKernelToBlobPass().
Depends On D98203
Reviewed By: mehdi_amini
Differential Revision: https://reviews.llvm.org/D98360
These are incompatible with opaque pointers. This is in preparation
of dropping this API on the IRBuilder side as well.
Instead explicitly pass the loaded type.
The paciasp and autiasp instructions are only accepted by recent
compilers, but have the same encoding as hint instructions, so we can
use the hint menmonic to support older compilers.
If MLIR_CUDA_RUNNER_ENABLED, register a 'gpu-to-cubin' conversion pass to mlir-opt.
The next step is to switch CUDA integration tests from mlir-cuda-runner to mlir-opt + mlir-cpu-runner and remove mlir-cuda-runner.
Depends On D98279
Reviewed By: herhut, rriddle, mehdi_amini
Differential Revision: https://reviews.llvm.org/D98203
For locally scoped lambdas like this there's no particular benefit to
explicitly listing captures - or avoiding capturing this. Switch to [&]
and make it all easier to maintain.
(& driveby change std::function to llvm::function_ref)
Commit 1b04bdc2f3 added support for
capturing the 'this' pointer in a SEH context (__finally or __except),
But the case in which the 'this' pointer is part of a lambda capture
was not handled properly
Differential Revision: https://reviews.llvm.org/D97687
This avoids the `__NR_gettimeofday` syscall number, which does not exist on 32-bit musl (it has `__NR_gettimeofday_time32`).
This switched Android to `clock_gettime` as well, which should work according to the old code before D96925.
Tested on Alpine Linux x86-64 (musl) and FreeBSD x86-64.
Reviewed By: vitalybuka
Differential Revision: https://reviews.llvm.org/D98121
This is no longer needed, we can add __llvm_profile_runtime directly
to llvm.compiler.used or llvm.used to achieve the same effect.
Differential Revision: https://reviews.llvm.org/D98325
All check-tsan tests fail on aarch64-*-linux because HeapMemEnd() > ShadowBeg()
for the following code path:
```
#if defined(__aarch64__) && !HAS_48_BIT_ADDRESS_SPACE
ProtectRange(HeapMemEnd(), ShadowBeg());
```
Restore the behavior before D86377 for aarch64-*-linux.
Joseph Huber