This patch adds each pass' pass argument in the header for IR dumps.
For example:
Before:
```
*** IR Dump Before InstructionSelect ***
```
After:
```
*** IR Dump Before InstructionSelect (instruction-select) ***
```
The goal is to make it easier to know what argument to pass to
command line options like `debug-only` or `run-pass` to further
investigate a given pass.
This change simplifies `clang/lib/Frontend/CompilerInvocation.cpp`
because we no longer need to manually parse the flag and set codegen
options in the frontend. However, we still need to manually parse the
flag in the driver because:
* The marshalling infrastructure doesn't operate there.
* We need to do some platform specific checks in the driver
that will likely never be supported by any kind of marshalling
infrastructure.
rdar://71609176
Differential Revision: https://reviews.llvm.org/D97327
The new `-fsanitize-address-destructor-kind=` option allows control over how module
destructors are emitted by ASan.
The new option is consumed by both the driver and the frontend and is propagated into
codegen options by the frontend.
Both the legacy and new pass manager code have been updated to consume the new option
from the codegen options.
It would be nice if the new utility functions (`AsanDtorKindToString` and
`AsanDtorKindFromString`) could live in LLVM instead of Clang so they could be
consumed by other language frontends. Unfortunately that doesn't work because
the clang driver doesn't link against the LLVM instrumentation library.
rdar://71609176
Differential Revision: https://reviews.llvm.org/D96572
This commit adds checks for the following:
* labels
* block expressions
* random integers cast to `void*`
* function pointers cast to `void*`
Differential Revision: https://reviews.llvm.org/D94640
This test case was causing a PowerPC buildbot to fail as it happened to
be named lld-multistage,
which matches with the original regex and therefore fails the check-not.
This should better represent the desired check.
Differential Revision: https://reviews.llvm.org/D97423
This happens in codebases a lot, which use xor where both sides are
macros. Using xor in that case is not the common error-prone 2^6 code
that the warning was introduced for.
Don't diagnose such a use of xor.
Differential Revision: https://reviews.llvm.org/D97445
Allow users to use a non-system version of perl, python and awk, which is useful
in certain package managers.
Reviewed By: JDevlieghere, MaskRay
Differential Revision: https://reviews.llvm.org/D95119
Finally, this patch moves from round-tripping one `CompilerInvocation` at a time to round-tripping the invocation as a whole.
This patch includes only the code required to make round-tripping the whole invocation work. More cleanups will be done in a follow-up patch.
Depends on D96847, D97041 & D97042.
Reviewed By: dexonsmith
Differential Revision: https://reviews.llvm.org/D96280
There are two preconditions to reproduce the issue,
1. Use -save-temps option
2. Provide the -o option with name equal to the input file name
without the file extension. For e.g. clang a.c -o a
With the -o specified, the AssembleJobAction after OffloadWrapperJobAction
will produce the object file with same name as host code object file.
Due to this clash, the OffloadWrapperAction overwrites the initial host
object file, which results in lld error. This also fixes the `multiple definition of __dummy.omp_offloading.entry'` issue in D96769 .
Reviewed By: jdoerfert
Differential Revision: https://reviews.llvm.org/D97273
Adding support for intrinsics of AMX-BF16.
This patch alse fix a bug that AMX-INT8 instructions will be selected with wrong
predicate.
Differential Revision: https://reviews.llvm.org/D97358
For -fgpu-rdc mode, static device vars in different TU's may have the same name.
To support accessing file-scope static device variables in host code, we need to give them
a distinct name and external linkage. This can be done by postfixing each static device variable with
a distinct CUID (Compilation Unit ID) hash.
Since the static device variables have different name across compilation units, now we let
them have external linkage so that they can be looked up by the runtime.
Reviewed by: Artem Belevich, and Jon Chesterfield
Differential Revision: https://reviews.llvm.org/D85223
When '__cl_clang_function_pointers' extension is enabled
the parser should allow obtaining the function address.
This fixes PR49264!
Differential Revision: https://reviews.llvm.org/D97203
Add the remaining missing builtin function declarations that have enum
or typedef argument or return types.
Differential Revision: https://reviews.llvm.org/D96860
-O1 and above do dont call real optimizer pipeline in ThinLTO PreLink.
Also clang can't add PostLink OptimizerLastEPCallbacks for in-process ThinLTO.
This results in missing sanitizer passes with ThinLTO.
Simple working solution is just call OptimizerLastEPCallbacks
at the end of buildThinLTOPreLinkDefaultPipeline.
Differential Revision: https://reviews.llvm.org/D96320
Patch takes advantage of the implicit default behavior to reduce the number of attributes, which in turns reduces compilation time.
Reviewed By: serge-sans-paille
Differential Revision: https://reviews.llvm.org/D97116
Currently managed variables are emitted as undefined symbols, which
causes difficulty for diagnosing undefined symbols for non-managed
variables.
This patch transforms managed variables in device compilation so that
they can be emitted as normal variables.
Reviewed by: Artem Belevich
Differential Revision: https://reviews.llvm.org/D96195
Demonstrate how to add RISC-V V builtins and lower them to IR intrinsics for V extension.
Authored-by: Roger Ferrer Ibanez <rofirrim@gmail.com>
Co-Authored-by: Hsiangkai Wang <kai.wang@sifive.com>
Differential Revision: https://reviews.llvm.org/D93446
https://bugs.llvm.org/show_bug.cgi?id=40858
CheckShadow is now called for each binding in the structured binding to make sure it does not shadow any other variable in scope. This does use a custom implementation of getShadowedDeclaration though because a BindingDecl is not a VarDecl
Added a few unit tests for this. In theory though all the other shadow unit tests should be duplicated for the structured binding variables too but whether it is probably not worth it as they use common code. The MyTuple and std interface code has been copied from live-bindings-test.cpp
Reviewed By: rsmith
Differential Revision: https://reviews.llvm.org/D96147
When targeting a MSVC triple, --dependant-libs with the name of the clang runtime library for profiling is added to the command line args. In it's current implementations clang_rt.profile-<ARCH> is chosen as the name. When building a distribution using LLVM_ENABLE_PER_TARGET_RUNTIME_DIR this fails, due to the runtime file names not having an architecture suffix in the filename.
This patch refactors getCompilerRT and getCompilerRTBasename to always consider per-target runtime directories. getCompilerRTBasename now simply returns the filename component of the path found by getCompilerRT
Differential Revision: https://reviews.llvm.org/D96638
This commit prevents warnings from -Wconversion when a clang vector type
is implicitly converted to a sizeless builtin type -- for example, when
implicitly converting a fixed-predicate to a scalable predicate.
The code below:
1 #include <arm_sve.h>
2
3 #define N __ARM_FEATURE_SVE_BITS
4 #define FIXED_ATTR __attribute__((arm_sve_vector_bits (N)))
5 typedef svbool_t fixed_svbool_t FIXED_ATTR;
6
7 inline fixed_svbool_t foo(fixed_svbool_t p) {
8 return svnot_z(svptrue_b64(), p);
9 }
would previously raise this warning:
warning: implicit conversion turns vector to scalar: \
'fixed_svbool_t' (vector of 8 'unsigned char' values) to 'svbool_t' \
(aka '__SVBool_t') [-Wconversion]
Note that many cases of these implicit conversions were already
permitted because many functions inside arm_sve.h are spawned via
preprocessor macros, and the call to isInSystemMacro would cover us in
this case. This commit fixes the remaining cases.
Differential Revision: https://reviews.llvm.org/D97053
Follow-up to fe2dcd89ac.
Update test per review comments, restoring the "D" type to its
original state, and adding new "L" type. (Sorry, this was intended to
be included in the prior commit)
Differential Revision: https://reviews.llvm.org/D96044
Previously, the definition was so-marked, but the declaration was
not. This resulted in LLVM's dwarf emission treating the function as
being external, and incorrectly emitting DW_AT_external.
Differential Revision: https://reviews.llvm.org/D96044
Currently TypePrinter lumps anonymous classes and unnamed classes in one group "anonymous" this is not correct and can be confusing in some contexts.
Differential Revision: https://reviews.llvm.org/D96807
If a static assert has a message as the right side of an and condition, suggest a fix it of replacing the '&&' to ','.
`static_assert(cond && "Failed Cond")` -> `static_assert(cond, "Failed cond")`
This use case comes up when lazily replacing asserts with static asserts.
Reviewed By: aaron.ballman
Differential Revision: https://reviews.llvm.org/D89065
In current implementation of `deviceRTLs`, we're using some functions
that are CUDA version dependent (if CUDA_VERSION < 9, it is one; otheriwse, it
is another one). As a result, we have to compile one bitcode library for each
CUDA version supported. A worse problem is forward compatibility. If a new CUDA
version is released, we have to update CMake file as well.
CUDA 9.2 has been released for three years. Instead of using various weird tricks
to make `deviceRTLs` work with different CUDA versions and still have forward
compatibility, we can simply drop support for CUDA 9.1 or lower version. It has at
least two benifits:
- We don't need to generate bitcode libraries for each CUDA version;
- Clang driver doesn't need to search for the bitcode lib based on CUDA version.
We can claim that starting from LLVM 12, OpenMP offloading on NVPTX target requires
CUDA 9.2+.
Reviewed By: jdoerfert, JonChesterfield
Differential Revision: https://reviews.llvm.org/D97003
This change enables the builtin function declarations
in clang driver by default using the Tablegen solution
along with the implicit include of 'opencl-c-base.h'
header.
A new flag '-cl-no-stdinc' disabling all default
declarations and header includes is added. If any other
mechanisms were used to include the declarations (e.g.
with -Xclang -finclude-default-header) and the new default
approach is not sufficient the, `-cl-no-stdinc` flag has
to be used with clang to activate the old behavior.
Tags: #clang
Differential Revision: https://reviews.llvm.org/D96515
This patch adds the following SHA3 Intrinsics:
vsha512hq_u64,
vsha512h2q_u64,
vsha512su0q_u64,
vsha512su1q_u64
veor3q_u8
veor3q_u16
veor3q_u32
veor3q_u64
veor3q_s8
veor3q_s16
veor3q_s32
veor3q_s64
vrax1q_u64
vxarq_u64
vbcaxq_u8
vbcaxq_u16
vbcaxq_u32
vbcaxq_u64
vbcaxq_s8
vbcaxq_s16
vbcaxq_s32
vbcaxq_s64
Note need to include +sha3 and +crypto when building from the front-end
Reviewed By: DavidSpickett
Differential Revision: https://reviews.llvm.org/D96381
Removes `CrossTranslationUnitContext::getImportedFromSourceLocation`
Removes the corresponding unit-test segment.
Introduces the `CrossTranslationUnitContext::getMacroExpansionContextForSourceLocation`
which will return the macro expansion context for an imported TU. Also adds a
few implementation FIXME notes where applicable, since this feature is
not implemented yet. This fact is also noted as Doxygen comments.
Uplifts a few CTU LIT test to match the current **incomplete** behavior.
It is a regression to some extent since now we don't expand any
macros in imported TUs. At least we don't crash anymore.
Note that the introduced function is already covered by LIT tests.
Eg.: Analysis/plist-macros-with-expansion-ctu.c
Reviewed By: balazske, Szelethus
Differential Revision: https://reviews.llvm.org/D94673
Removes the obsolete ad-hoc macro expansions during bugreport constructions.
It will skip the macro expansion if the expansion happened in an imported TU.
Also removes the expected plist file, while expanding matching context for
the tests.
Adds a previously crashing `plist-macros-with-expansion.c` testfile.
Temporarily marks `plist-macros-with-expansion-ctu.c ` to `XFAIL`.
Reviewed By: xazax.hun, Szelethus
Differential Revision: https://reviews.llvm.org/D93224
This patch moves the creation of the '-Wspir-compat' argument from cc1 to the driver.
Without this change, generating command line arguments from `CompilerInvocation` cannot be done reliably: there's no way to distinguish whether '-Wspir-compat' was passed to cc1 on the command line (should be generated), or if it was created within `CompilerInvocation::CreateFromArgs` (should not be generated).
This is also in line with how other '-W' flags are handled.
(This was introduced in D21567.)
Reviewed By: Anastasia
Differential Revision: https://reviews.llvm.org/D97041
`sm_35` is the minimum requirement for OpenMP offloading on NVPTX device.
Current driver test case is using `sm_20`. D97003 is going to switch the minimum
CUDA version to 9.2, which only supports `sm_30+`. This patch makes step for the
change.
Reviewed By: JonChesterfield
Differential Revision: https://reviews.llvm.org/D97120
When compiling UEFI applications, the main function is named
efi_main() instead of main(). Let's exclude efi_main() from
-Wmissing-prototypes as well to avoid warnings when working
on UEFI applications.
Differential Revision: https://reviews.llvm.org/D95746
When WPD is enabled, via WholeProgramVTables, emit type metadata for
available_externally vtables. Additionally, add the vtables to the
llvm.compiler.used global so that they are not prematurely eliminated
(before *LTO analysis).
This is needed to avoid devirtualizing calls to a function overriding a
class defined in a header file but with a strong definition in a shared
library. Without type metadata on the available_externally vtables from
the header, the WPD analysis never sees what a derived class is
overriding. Even if the available_externally base class functions are
pure virtual, because shared library definitions are already treated
conservatively (committed patches D91583, D96721, and D96722) we will
not devirtualize, which would be unsafe since the library might contain
overrides that aren't visible to the LTO unit.
An example is std::error_category, which is overridden in LLVM
and causing failures after a self build with WPD enabled, because
libstdc++ contains hidden overrides of the virtual base class methods.
Differential Revision: https://reviews.llvm.org/D96919
This enables use of MemorySSA instead of MemDep in MemCpyOpt. To
allow this without significant compile-time impact, the MemCpyOpt
pass is moved directly before DSE (in the cases where this was not
already the case), which allows us to reuse the existing MemorySSA
analysis.
Unlike the MemDep-based implementation, the MemorySSA-based MemCpyOpt
can also perform simple optimizations across basic blocks.
Differential Revision: https://reviews.llvm.org/D94376
This enables AES fusion and the post RA scheduler for the Neoverse cores.
And while we are it also for the A55 that we had missed earlier.
Differential Revision: https://reviews.llvm.org/D96866
Add option -fgpu-sanitize to enable sanitizer for AMDGPU target.
Since it is experimental, it is off by default.
Reviewed by: Artem Belevich
Differential Revision: https://reviews.llvm.org/D96835
mode.
We use that mode when evaluating ICEs in C, and those shortcuts could
result in ICE evaluation producing the wrong answer, specifically if we
evaluate a statement-expression as part of evaluating the ICE.
Currently TypePrinter lumps anonymous classes and unnamed classes in one group "anonymous" this is not correct and can be confusing in some contexts.
Differential Revision: https://reviews.llvm.org/D96807
We currently always store absolute filenames in coverage mapping. This
is problematic for several reasons. It poses a problem for distributed
compilation as source location might vary across machines. We are also
duplicating the path prefix potentially wasting space.
This change modifies how we store filenames in coverage mapping. Rather
than absolute paths, it stores the compilation directory and file paths
as given to the compiler, either relative or absolute. Later when
reading the coverage mapping information, we recombine relative paths
with the working directory. This approach is similar to handling
ofDW_AT_comp_dir in DWARF.
Finally, we also provide a new option, -fprofile-compilation-dir akin
to -fdebug-compilation-dir which can be used to manually override the
compilation directory which is useful in distributed compilation cases.
Differential Revision: https://reviews.llvm.org/D95753
We currently always store absolute filenames in coverage mapping. This
is problematic for several reasons. It poses a problem for distributed
compilation as source location might vary across machines. We are also
duplicating the path prefix potentially wasting space.
This change modifies how we store filenames in coverage mapping. Rather
than absolute paths, it stores the compilation directory and file paths
as given to the compiler, either relative or absolute. Later when
reading the coverage mapping information, we recombine relative paths
with the working directory. This approach is similar to handling
ofDW_AT_comp_dir in DWARF.
Finally, we also provide a new option, -fprofile-compilation-dir akin
to -fdebug-compilation-dir which can be used to manually override the
compilation directory which is useful in distributed compilation cases.
Differential Revision: https://reviews.llvm.org/D95753
The recent commit 00a6254 "Stop traping on sNaN in builtin_isnan" changed the
lowering in constrained FP mode of builtin_isnan from an FP comparison to
integer operations to avoid trapping.
SystemZ has a special instruction "Test Data Class" which is the preferred
way to do this check. This patch adds a new target hook "testFPKind()" that
lets SystemZ emit the s390_tdc intrinsic instead.
testFPKind() takes the BuiltinID as an argument and is expected to soon
handle more opcodes than just 'builtin_isnan'.
Review: Thomas Preud'homme, Ulrich Weigand
Differential Revision: https://reviews.llvm.org/D96568
would otherwise include template specialization types
This helps reduce the size of the encoded C++ type strings in the binary.
This is enabled by default only on Darwin, but can be enabled/disabled
via command line options.
rdar://63288571
Differential Revision: https://reviews.llvm.org/D96816
The following commits added commandline arguments to control following the Arm
Procedure Call Standard for certain volatile bitfield operations:
- https://reviews.llvm.org/D67399
- https://reviews.llvm.org/D72932
This commit fixes the oversight that these args weren't passed from the driver
to cc1 if appropriate.
Where *appropriate* means:
- `-faapcs-bitfield-width`: is the default, so won't be passed
- `-fno-aapcs-bitfield-width`: should be passed
- `-faapcs-bitfield-load`: should be passed
Differential Revision: https://reviews.llvm.org/D96784
Added -mrop-protection for Power PC to turn on codegen that provides some
protection from ROP attacks.
The option is off by default and can be turned on for Power 8, Power 9 and
Power 10.
This patch is for the option only. The feature will be implemented by a later
patch.
Reviewed By: amyk
Differential Revision: https://reviews.llvm.org/D96512
This fixes an issue when "-gdwarf-N" switch was ignored if it was given
before another debug option.
Differential Revision: https://reviews.llvm.org/D96865
Add the types for the RISC-V V extension builtins.
These types will be used by the RISC-V V intrinsics which require
types of the form <vscale x 1 x i64>(LMUL=1 element size=64) or
<vscale x 4 x i32>(LMUL=2 element size=32), etc. The vector_size
attribute does not work for us as it doesn't create a scalable
vector type. We want these types to be opaque and have no operators
defined for them. We want them to be sizeless. This makes them
similar to the ARM SVE builtin types. But we will have quite a bit
more types. This patch adds around 60. Later patches will add
another 230 or so types representing tuples of these types similar
to the x2/x3/x4 types in ARM SVE. But with extra complexity that
these types are combined with the LMUL concept that is unique to
RISCV.
For more background see this RFC
http://lists.llvm.org/pipermail/llvm-dev/2020-October/145850.html
Authored-by: Roger Ferrer Ibanez <roger.ferrer@bsc.es>
Co-Authored-by: Hsiangkai Wang <kai.wang@sifive.com>
Differential Revision: https://reviews.llvm.org/D92715
This matches the platform default for GCC. It primarily matters when the
integrated assembler is not used as there is no default CPU defined for
ARMv7-A and GNU as is upset with -mcpu=generic.
The new spec does not have `exnref` so EH does not have dependency of
the reference types proposal anymore.
Reviewed By: dschuff
Differential Revision: https://reviews.llvm.org/D96903
Drop the `Separate` form of `-fmodule-name X`, `-fprofile-remapping-file X`, and `-frewrite-map-file X`.
To the best of my knowledge they are not used. Their conventional Joined forms (`-fFOO=`) should be used instead.
`-fdebug-compilation-dir X` is used in several places, e.g. chromium/infra/goma.
It is also advertised in http://blog.llvm.org/2019/11/deterministic-builds-with-clang-and-lld.html
So we keep it but make the EQ form canonical and the Separate form an alias.
Differential Revision: https://reviews.llvm.org/D96886
Basic block sections enables function sections implicitly, this is not needed
and is inefficient with "=list" option.
We had basic block sections enable function sections implicitly in clang. This
is particularly inefficient with "=list" option as it places functions that do
not have any basic block sections in separate sections. This causes unnecessary
object file overhead for large applications.
This patch disables this implicit behavior. It only creates function sections
for those functions that require basic block sections.
This patch is the second of two patches and this patch removes the implicit
enabling of function sections with basic block sections in clang.
Differential Revision: https://reviews.llvm.org/D93876
Add enum and typedef argument support to `-fdeclare-opencl-builtins`,
which was the last major missing feature.
Adding the remaining missing builtins is left as future work.
Differential Revision: https://reviews.llvm.org/D96051
The option was added in D90507 for C/C++ source files. This patch adds
support for assembly files.
Differential Revision: https://reviews.llvm.org/D96783
This allows the option to affect the LTO output. Module::Max helps to
generate debug info for all modules in the same format.
Differential Revision: https://reviews.llvm.org/D96597
This change affects 'SemaOpenCLCXX/newdelete.cl' test,
thus the patch contains adjustments in types validation of
operators new and delete
Reviewed By: Anastasia
Differential Revision: https://reviews.llvm.org/D96178
Displaying the problem range could crash if the begin and end of a
range is in different files or macros. After the change such range
is displayed only as the beginning location.
There is a bug for this problem:
https://bugs.llvm.org/show_bug.cgi?id=46540
Reviewed By: steakhal
Differential Revision: https://reviews.llvm.org/D95860
OpenMP 5.0 removed a lot of restriction for overlapped mapped items
comparing to OpenMP 4.5. Patch restricts the checks for overlapped data
mappings only for OpenMP 4.5 and less and reorders mapping of the
arguments so, that present and alloc mappings are processed first and
then all others.
Differential Revision: https://reviews.llvm.org/D86119
Implement all of P1825R0:
- implicitly movable entity can be an rvalue reference to non-volatile
automatic object.
- operand of throw-expression can be a function or catch-clause parameter
(support for function parameter has already been implemented).
- in the first overload resolution, the selected function no need to be
a constructor.
- in the first overload resolution, the first parameter of the selected
function no need to be an rvalue reference to the object's type.
This patch also removes the diagnostic `-Wreturn-std-move-in-c++11`.
Differential Revision: https://reviews.llvm.org/D88220
The tile directive is in OpenMP's Technical Report 8 and foreseeably will be part of the upcoming OpenMP 5.1 standard.
This implementation is based on an AST transformation providing a de-sugared loop nest. This makes it simple to forward the de-sugared transformation to loop associated directives taking the tiled loops. In contrast to other loop associated directives, the OMPTileDirective does not use CapturedStmts. Letting loop associated directives consume loops from different capture context would be difficult.
A significant amount of code generation logic is taking place in the Sema class. Eventually, I would prefer if these would move into the CodeGen component such that we could make use of the OpenMPIRBuilder, together with flang. Only expressions converting between the language's iteration variable and the logical iteration space need to take place in the semantic analyzer: Getting the of iterations (e.g. the overload resolution of `std::distance`) and converting the logical iteration number to the iteration variable (e.g. overload resolution of `iteration + .omp.iv`). In clang, only CXXForRangeStmt is also represented by its de-sugared components. However, OpenMP loop are not defined as syntatic sugar. Starting with an AST-based approach allows us to gradually move generated AST statements into CodeGen, instead all at once.
I would also like to refactor `checkOpenMPLoop` into its functionalities in a follow-up. In this patch it is used twice. Once for checking proper nesting and emitting diagnostics, and additionally for deriving the logical iteration space per-loop (instead of for the loop nest).
Differential Revision: https://reviews.llvm.org/D76342
This takes advantage of the implicit default behavior to reduce the number of
attributes, which in turns reduces compilation time. I've observed -3% in
instruction count when compiling sqlite3 amalgamation with -O0
Differential Revision: https://reviews.llvm.org/D96400
This patch generates the `-f[no-]finite-loops` arguments from `CompilerInvocation` (added in D96419), fixing test failures of Clang built with `-DCLANG_ROUND_TRIP_CC1_ARGS=ON`.
Reviewed By: fhahn
Differential Revision: https://reviews.llvm.org/D96761
Even code in target and declare target regions might not be emitted.
With this patch we delay more diagnostics and use laziness and linkage
to determine if a function is emitted (for the device). Note that we
still eagerly emit diagnostics for target regions, unfortunately, see
the TODO for the reason.
This hopefully fixes PR48933.
Reviewed By: JonChesterfield
Differential Revision: https://reviews.llvm.org/D95928
Type errors in function declarations were not (always) diagnosed prior
to this patch. Furthermore, certain remarks did not get associated
properly which caused them to be emitted multiple times.
Reviewed By: JonChesterfield
Differential Revision: https://reviews.llvm.org/D95912
This commit fixes bug #48739. The bug was caused by the way static_casts
on pointer-to-member caused the CXXBaseSpecifier list of a
MemberToPointer to grow instead of shrink.
The list is now grown by implicit casts and corresponding entries are
removed by static_casts. No-op static_casts cause no effect.
Reviewed By: vsavchenko
Differential Revision: https://reviews.llvm.org/D95877
This is a follow up of D92940.
We have successfully converted fadd/fmul _mm_reduce_* intrinsics to
llvm.reduction + reassoc flag. We can do the same approach for fmin/fmax
too, i.e. llvm.reduction + nnan flag.
Reviewed By: spatel
Differential Revision: https://reviews.llvm.org/D93179
This patch ensures that vector predication and vectorization width
pragmas work together correctly/as expected. Specifically, this patch
fixes the issue that when vectorization_width > 1, the vector
predication behaviour (this would matter if it has NOT been disabled
explicitly by a pragma) was getting ignored, which was incorrect.
The fix here removes the dependence of vector predication on the
vectorization width. The loop metadata corresponding to clang loop
pragma vectorize_predicate is always emitted, if the pragma is
specified, even if vectorization is disabled by vectorize_width(1)
or vectorize(disable) since the option is also used for interleaving
by the LoopVectorize pass.
Reviewed By: dmgreen, Meinersbur
Differential Revision: https://reviews.llvm.org/D94779
vec_xl() and vec_xst() should not emit alignment hints since they take a
scalar pointer and also add a byte offset if passed.
This patch uses memcpy to achieve the desired result.
Review: Ulrich Weigand
Differential Revision: https://reviews.llvm.org/D96471
This patch adds 2 new options to control when Clang adds `mustprogress`:
1. -ffinite-loops: assume all loops are finite; mustprogress is added
to all loops, regardless of the selected language standard.
2. -fno-finite-loops: assume no loop is finite; mustprogress is not
added to any loop or function. We could add mustprogress to
functions without loops, but we would have to detect that in Clang,
which is probably not worth it.
Reviewed By: jdoerfert
Differential Revision: https://reviews.llvm.org/D96419
class types.
The goal is to provide a way to bypass constructor homing when emitting
class definitions and force class definitions in the debug info.
Not sure about the wording of the attribute, or whether it should be
specific to classes with constructors
explicitly emitting retainRV or claimRV calls in the IR
Background:
This fixes a longstanding problem where llvm breaks ARC's autorelease
optimization (see the link below) by separating calls from the marker
instructions or retainRV/claimRV calls. The backend changes are in
https://reviews.llvm.org/D92569.
https://clang.llvm.org/docs/AutomaticReferenceCounting.html#arc-runtime-objc-autoreleasereturnvalue
What this patch does to fix the problem:
- The front-end adds operand bundle "clang.arc.attachedcall" to calls,
which indicates the call is implicitly followed by a marker
instruction and an implicit retainRV/claimRV call that consumes the
call result. In addition, it emits a call to
@llvm.objc.clang.arc.noop.use, which consumes the call result, to
prevent the middle-end passes from changing the return type of the
called function. This is currently done only when the target is arm64
and the optimization level is higher than -O0.
- ARC optimizer temporarily emits retainRV/claimRV calls after the calls
with the operand bundle in the IR and removes the inserted calls after
processing the function.
- ARC contract pass emits retainRV/claimRV calls after the call with the
operand bundle. It doesn't remove the operand bundle on the call since
the backend needs it to emit the marker instruction. The retainRV and
claimRV calls are emitted late in the pipeline to prevent optimization
passes from transforming the IR in a way that makes it harder for the
ARC middle-end passes to figure out the def-use relationship between
the call and the retainRV/claimRV calls (which is the cause of
PR31925).
- The function inliner removes an autoreleaseRV call in the callee if
nothing in the callee prevents it from being paired up with the
retainRV/claimRV call in the caller. It then inserts a release call if
claimRV is attached to the call since autoreleaseRV+claimRV is
equivalent to a release. If it cannot find an autoreleaseRV call, it
tries to transfer the operand bundle to a function call in the callee.
This is important since the ARC optimizer can remove the autoreleaseRV
returning the callee result, which makes it impossible to pair it up
with the retainRV/claimRV call in the caller. If that fails, it simply
emits a retain call in the IR if retainRV is attached to the call and
does nothing if claimRV is attached to it.
- SCCP refrains from replacing the return value of a call with a
constant value if the call has the operand bundle. This ensures the
call always has at least one user (the call to
@llvm.objc.clang.arc.noop.use).
- This patch also fixes a bug in replaceUsesOfNonProtoConstant where
multiple operand bundles of the same kind were being added to a call.
Future work:
- Use the operand bundle on x86-64.
- Fix the auto upgrader to convert call+retainRV/claimRV pairs into
calls with the operand bundles.
rdar://71443534
Differential Revision: https://reviews.llvm.org/D92808
This patch uses the existing logic of CUDA for searching libomptarget
and extracts it to a common method.
Reviewed By: JonChesterfield, tianshilei1992
Differential Revision: https://reviews.llvm.org/D96248
EarlyCSEPass called after msan redices code size by about 10%.
Similar optimization exists for legacy pass manager in
addGeneralOptsForMemorySanitizer.
Reviewed By: eugenis
Differential Revision: https://reviews.llvm.org/D96406
The ability to specify alignment was recently added, and it's an
important property which we should ensure is set as expected by
Clang. (Especially before making further changes to Clang's code in
this area.) But, because it's on the end of the lines, the existing
tests all ignore it.
Therefore, update all the tests to also verify the expected alignment
for atomicrmw and cmpxchg. While I was in there, I also updated uses
of 'load atomic' and 'store atomic', and added the memory ordering,
where that was missing.
In D91442, @MaskRay commented about a failure. This commit does the following to
address his comments:
1. Replace %T with %t as former is deprecated.
2. Add an explicit --sysroot argument in a test.
Some tests were failing when gcc-10-riscv64-linux-gnu is installed on test machine.
This was happening because the test was checking a case when --gcc-toolchain is not
provided. But if --sysroot was also not provided then code could pick a toolchain
installed in /usr. So to make the test more robust, I have provided an explicit --sysroot
argument. Its value has been chosen to match the existing patterns.
Reviewed By: MaskRay
Differential Revision: https://reviews.llvm.org/D93023
This change removes the XFAIL from the original test and duplicates the test into sanitize-coverage-old-pm.c
which uses the old pass manager and has the corresponding XFAIL.
This should fix the XPASS from this and similar runs:
http://lab.llvm.org:8011/#/builders/60/builds/1875
Multi-configuration generators (such as Visual Studio and Xcode) allow the specification of a build flavor at build time instead of config time, so the lit configuration files need to support that - and they do for the most part. There are several places that had one of two issues (or both!):
1) Paths had %(build_mode)s set up, but then not configured, resulting in values that would not work correctly e.g. D:/llvm-build/%(build_mode)s/bin/dsymutil.exe
2) Paths did not have %(build_mode)s set up, but instead contained $(Configuration) (which is the value for Visual Studio at configuration time, for Xcode they would have had the equivalent) e.g. "D:/llvm-build/$(Configuration)/lib".
This seems to indicate that we still have a lot of fragility in the configurations, but also that a number of these paths are never used (at least on Windows) since the errors appear to have been there a while.
This patch fixes the configurations and it has been tested with Ninja and Visual Studio to generate the correct paths. We should consider removing some of these settings altogether.
Reviewed By: JDevlieghere, mehdi_amini
Differential Revision: https://reviews.llvm.org/D96427
With https://reviews.llvm.org/D63376, we began storing the APValue
directly into the ConstantExpr object so that we could reuse the
calculated value later. However, it missed a case when not in C++11
mode but the expression is known to be constant.
Before this commit, expression statements could not be annotated
with statement attributes. Whenever parser found attribute, it
unconditionally assumed that it was followed by a declaration.
This not only doesn't allow expression attributes to have attributes,
but also produces spurious error diagnostics.
In order to maintain all previously compiled code, we still assume
that GNU attributes are followed by declarations unless ALL of those
are statement attributes. And even in this case we are not forcing
the parser to think that it should parse a statement, but rather
let it proceed as if no attributes were found.
Differential Revision: https://reviews.llvm.org/D93630
The swift_bridge attribute warns when the attribute is applied multiple
times to the same declaration. However, it warns about the arguments
being different to the attribute without ever checking if the arguments
actually are different. If the arguments are different, diagnose,
otherwise silently accept the code. Either way, drop the duplicated
attribute.
Today, inside a template, you can get completion for:
Foo<T> t;
t.^
t has dependent type Foo<T>, and we use the primary template to find its members.
However we also want this to work:
t.foo.bar().^
The type of t.foo.bar() is DependentTy, so we attempt to resolve using similar
heuristics (e.g. primary template).
Differential Revision: https://reviews.llvm.org/D96376
Add the builtin functions brought by the
cl_khr_subgroup_extended_types extension to
`-fdeclare-opencl-builtins`.
Differential Revision: https://reviews.llvm.org/D96279
After D93264, using both -fdebug-info-for-profiling and
-fpseudo-probe-for-profiling will cause the compiler to crash.
Diagnose these conflicting options in the driver.
Also, the existing CodeGen test was using the driver when it should be
running cc1.
Differential Revision: https://reviews.llvm.org/D96354
Not all platforms accept -stdlib or -rtlib. Instead of complaining about
the wrong argument to these options, clang complains about the option
itself being present.
Pass an appropriate -target to the clang invocations.
Add the builtin functions brought by the
cl_khr_subgroup_non_uniform_arithmetic extension to
`-fdeclare-opencl-builtins`.
Differential Revision: https://reviews.llvm.org/D95951
This reverts commit 3500cc8d89.
This old commit was made over a completely false premise. OSSymbols
aren't different from other OSObjects and we shouldn't treat them
differently for the purposes of static analysis.
Since ToolChain::GetCXXStdlibType() is a simple getter that might emit
the "invalid library name in argument" warning, it can conceivably be
called several times while initializing the build pipeline.
Before this patch, a simple 'clang++ -stdlib=foo ./test.cpp' would print
the warning twice, -rt=lib=foo would print 6 times.
Change this and always only print the warning once. Keep the rest of the
semantics of the functions.
Differential Revision: https://reviews.llvm.org/D95915
Signed prefix is removed and the single word spelling is
printed for the scalar types.
Tags: #clang
Differential Revision: https://reviews.llvm.org/D96161
Intrinsics *reduce_add/mul_ps/pd have assumption that the elements in
the vector are reassociable. So we need to always assign the reassoc
flag when we call _mm_reduce_* intrinsics.
Reviewed By: spatel
Differential Revision: https://reviews.llvm.org/D96231
It is very common to check callbacks and completion handlers for null.
This patch supports such checks using built-in functions:
* __builtin_expect
* __builtin_expect_with_probablity
* __builtin_unpredictable
rdar://73455388
Differential Revision: https://reviews.llvm.org/D96268
This patch added a distinct CUID for each input file, which is represented by InputAction.
clang initially creates an InputAction for each input file for the host compilation. In CUDA/HIP action
builder, each InputAction is given a CUID and cloned for each GPU arch, and the CUID is also cloned. In this way,
we guarantee the corresponding device and host compilation for the same file shared the
same CUID. On the other hand, different compilation units have different CUID.
-fuse-cuid=random|hash|none is added to control the method to generate CUID. The default
is hash. -cuid=X is also added to specify CUID explicitly, which overrides -fuse-cuid.
Reviewed by: Artem Belevich
Differential Revision: https://reviews.llvm.org/D95007
during the same evaluation.
It looks like the only case for which this matters is determining
whether mutable subobjects of a heap allocation can be modified during
constant evaluation.
variable's destruction if it didn't do so during construction.
The standard doesn't give any guidance as to what to do here, but this
approach seems reasonable and conservative, and has been proposed to the
standard committee.
A module with errors would be marked as out-of-date, then the `compilerModule` action would produce it, but due to the error it would be treated as failure and the resulting PCM would not get used.
rdar://74087062
Differential Revision: https://reviews.llvm.org/D96246
Currently -fgpu-rdc is not passed to host clang -cc1.
This causes issue because -fgpu-rdc affects shadow
variable linkage in host compilation.
Reviewed by: Artem Belevich
Differential Revision: https://reviews.llvm.org/D96105
Link: https://lists.llvm.org/pipermail/llvm-dev/2020-October/146162.html "[RFC] FileCheck: (dis)allowing unused prefixes"
If a downstream project using lit needs time for transition,
add the following to `lit.local.cfg`:
```
from lit.llvm.subst import ToolSubst
fc = ToolSubst('FileCheck', unresolved='fatal')
config.substitutions.insert(0, (fc.regex, 'FileCheck --allow-unused-prefixes'))
```
Differential Revision: https://reviews.llvm.org/D95849
As Itanium ABI[http://itanium-cxx-abi.github.io/cxx-abi/abi.html#once-ctor]
points out:
"The size of the guard variable is 64 bits. The first byte (i.e. the byte at
the address of the full variable) shall contain the value 0 prior to
initialization of the associated variable, and 1 after initialization is complete."
Differential Revision: https://reviews.llvm.org/D95822
Pipe element type spelling for arg info metadata
should follow the same behavior as normal type spelling.
We should only use the canonical type spelling in the
base type field.
This patch also removed duplication in type handling.
Tags: #clang
Differential Revision: https://reviews.llvm.org/D96151
This reverts commit 6039f821 and reapplies bff6d9bb.
Clang's Index/implicit-attrs.m test invokes c-index-test with -fobjc-arc. This flag is not compatible with -fobjc-runtime=gcc, which gets implied on Linux.
The original commit uncovered this by correctly reporting issues when parsing -cc1 command line.
This commit fixes the test to explicitly provide ObjectiveC runtime compatible with ARC.
`CompilerInvocation::CreateFromArgs` doesn't always report command line parsing failures through the return value. Sometimes, errors are only reported via diagnostics.
Some clients like `c-index-test` only check the return value and don't check the state of `DiagnosticsEngine`.
If we were to start returning the correct return value from `CreateFromArgs`, this index test starts to fail, because it specifies `-std=c++11` for a C input, which is invalid.
This patch fixes that issue by adding forgotten `-x c++` argument.
Reviewed By: dexonsmith
Differential Revision: https://reviews.llvm.org/D95879
Currently the emscripten frontend driver injects this when building
with thread support. Moving this into the clang driver itself makes
the emscripten python driver less magical.
Differential Revision: https://reviews.llvm.org/D96171
When a function or a file is excluded using -fprofile-list= option,
don't emit coverage mapping as doing so confuses users since those
functions would always have zero count. This also reduces the binary
size considerably in cases where only a few functions or files are
being instrumented.
Differential Revision: https://reviews.llvm.org/D96000
For -fgpu-rdc, shadow variables should not be internalized, otherwise
they cannot be accessed by other TUs. This is necessary because
the shadow variable of external device variables are always
emitted as undefined symbols, which need to resolve to a global
symbols.
Managed variables need to be emitted as undefined symbols
in device compilations.
Reviewed by: Artem Belevich
Differential Revision: https://reviews.llvm.org/D95901
__builtin_isnan currently generates a floating-point compare operation
which triggers a trap when faced with a signaling NaN in StrictFP mode.
This commit uses integer operations instead to not generate any trap in
such a case.
Reviewed By: kpn
Differential Revision: https://reviews.llvm.org/D95948
- The failures are all cc1-based tests due to the missing `-aux-triple` options,
which is always prepared by the driver in CUDA/HIP compilation.
- Add extra check on the missing aux-targetinfo to prevent crashing.
[hip][cuda] Enable extended lambda support on Windows.
- On Windows, extended lambda has extra issues due to the numbering
schemes are different between the host compilation (Microsoft C++ ABI)
and the device compilation (Itanium C++ ABI. Additional device side
lambda number is required per lambda for the host compilation to
correctly mangle the device-side lambda name.
- A hybrid numbering context `MSHIPNumberingContext` is introduced to
number a lambda for both host- and device-compilations.
Reviewed By: rnk
Differential Revision: https://reviews.llvm.org/D69322
This reverts commit 4874ff0241.
This patch adds possibility to define OpenCL C 3.0 feature macros
via command line option or target setting.
Reviewed By: Anastasia
Differential Revision: https://reviews.llvm.org/D95776
emitting retainRV or claimRV calls in the IR
This reapplies 3fe3946d9a without the
changes made to lib/IR/AutoUpgrade.cpp, which was violating layering.
Original commit message:
Background:
This patch makes changes to the front-end and middle-end that are
needed to fix a longstanding problem where llvm breaks ARC's autorelease
optimization (see the link below) by separating calls from the marker
instructions or retainRV/claimRV calls. The backend changes are in
https://reviews.llvm.org/D92569.
https://clang.llvm.org/docs/AutomaticReferenceCounting.html#arc-runtime-objc-autoreleasereturnvalue
What this patch does to fix the problem:
- The front-end adds operand bundle "clang.arc.rv" to calls, which
indicates the call is implicitly followed by a marker instruction and
an implicit retainRV/claimRV call that consumes the call result. In
addition, it emits a call to @llvm.objc.clang.arc.noop.use, which
consumes the call result, to prevent the middle-end passes from changing
the return type of the called function. This is currently done only when
the target is arm64 and the optimization level is higher than -O0.
- ARC optimizer temporarily emits retainRV/claimRV calls after the calls
with the operand bundle in the IR and removes the inserted calls after
processing the function.
- ARC contract pass emits retainRV/claimRV calls after the call with the
operand bundle. It doesn't remove the operand bundle on the call since
the backend needs it to emit the marker instruction. The retainRV and
claimRV calls are emitted late in the pipeline to prevent optimization
passes from transforming the IR in a way that makes it harder for the
ARC middle-end passes to figure out the def-use relationship between
the call and the retainRV/claimRV calls (which is the cause of
PR31925).
- The function inliner removes an autoreleaseRV call in the callee if
nothing in the callee prevents it from being paired up with the
retainRV/claimRV call in the caller. It then inserts a release call if
the call is annotated with claimRV since autoreleaseRV+claimRV is
equivalent to a release. If it cannot find an autoreleaseRV call, it
tries to transfer the operand bundle to a function call in the callee.
This is important since ARC optimizer can remove the autoreleaseRV
returning the callee result, which makes it impossible to pair it up
with the retainRV/claimRV call in the caller. If that fails, it simply
emits a retain call in the IR if the implicit call is a call to
retainRV and does nothing if it's a call to claimRV.
Future work:
- Use the operand bundle on x86-64.
- Fix the auto upgrader to convert call+retainRV/claimRV pairs into
calls annotated with the operand bundles.
rdar://71443534
Differential Revision: https://reviews.llvm.org/D92808
emitting retainRV or claimRV calls in the IR
Background:
This patch makes changes to the front-end and middle-end that are
needed to fix a longstanding problem where llvm breaks ARC's autorelease
optimization (see the link below) by separating calls from the marker
instructions or retainRV/claimRV calls. The backend changes are in
https://reviews.llvm.org/D92569.
https://clang.llvm.org/docs/AutomaticReferenceCounting.html#arc-runtime-objc-autoreleasereturnvalue
What this patch does to fix the problem:
- The front-end adds operand bundle "clang.arc.rv" to calls, which
indicates the call is implicitly followed by a marker instruction and
an implicit retainRV/claimRV call that consumes the call result. In
addition, it emits a call to @llvm.objc.clang.arc.noop.use, which
consumes the call result, to prevent the middle-end passes from changing
the return type of the called function. This is currently done only when
the target is arm64 and the optimization level is higher than -O0.
- ARC optimizer temporarily emits retainRV/claimRV calls after the calls
with the operand bundle in the IR and removes the inserted calls after
processing the function.
- ARC contract pass emits retainRV/claimRV calls after the call with the
operand bundle. It doesn't remove the operand bundle on the call since
the backend needs it to emit the marker instruction. The retainRV and
claimRV calls are emitted late in the pipeline to prevent optimization
passes from transforming the IR in a way that makes it harder for the
ARC middle-end passes to figure out the def-use relationship between
the call and the retainRV/claimRV calls (which is the cause of
PR31925).
- The function inliner removes an autoreleaseRV call in the callee if
nothing in the callee prevents it from being paired up with the
retainRV/claimRV call in the caller. It then inserts a release call if
the call is annotated with claimRV since autoreleaseRV+claimRV is
equivalent to a release. If it cannot find an autoreleaseRV call, it
tries to transfer the operand bundle to a function call in the callee.
This is important since ARC optimizer can remove the autoreleaseRV
returning the callee result, which makes it impossible to pair it up
with the retainRV/claimRV call in the caller. If that fails, it simply
emits a retain call in the IR if the implicit call is a call to
retainRV and does nothing if it's a call to claimRV.
Future work:
- Use the operand bundle on x86-64.
- Fix the auto upgrader to convert call+retainRV/claimRV pairs into
calls annotated with the operand bundles.
rdar://71443534
Differential Revision: https://reviews.llvm.org/D92808
Commit 6bf29dbb enables float128 feature by default for Power9 targets.
But float128 may cause build failure in libcxx testing. Revert this
commit first to unblock LLVM 12 release.
The attribute definition claimed the attribute was inheritable (which
only applies to declaration attributes) and not a statement attribute.
Further, it treats subject appertainment errors as being parse errors
rather than semantic errors, which leads to us accepting invalid code.
For instance, we currently fail to reject:
void foo() {
int i = 1000;
__attribute__((nomerge, opencl_unroll_hint(8)))
if (i) { foo(); }
}
This addresses the issues by clarifying that opencl_unroll_hint is a
statement attribute and handles its appertainment checks in the
semantic layer instead of the parsing layer. This changes the output of
the diagnostic text to be more consistent with other appertainment
errors.
When the -matomics feature is not enabled, disable POSIXThreads
mode and set the thread model to Single, so that we don't predefine
macros like `__STDCPP_THREADS__`.
Differential Revision: https://reviews.llvm.org/D96091
Defer constant checking of dependent initializer to template instantiation
since it cannot be done for dependent values.
Reviewed by: Artem Belevich
Differential Revision: https://reviews.llvm.org/D95840
explicitly qualified as members of the current instantiation.
Despite the nested name specifier being fully-dependent in this case,
the elaborated type might only be instantiation-dependent, because the
type is a member of the current instantiation.
This fixes Bugzilla #48894 for Arm, where it
was reported that -Wa,-march was not being handled
by the integrated assembler.
This was previously fixed for -Wa,-mthumb by
parsing the argument in ToolChain::ComputeLLVMTriple
instead of CollectArgsForIntegratedAssembler.
It has to be done in the former because the Triple
is read only by the time we get to the latter.
Previously only mcpu would work via -Wa but only because
"-target-cpu" is it's own option to cc1, which we were
able to modify. Target architecture is part of "-target-triple".
This change applies the same workaround to -march and cleans up
handling of -Wa,-mcpu at the same time. There were some
places where we were not using the last instance of an argument.
The existing -Wa,-mthumb code was doing this correctly,
so I've just added tests to confirm that.
Now the same rules will apply to -Wa,-march/-mcpu as would
if you just passed them to the compiler:
* -Wa/-Xassembler options only apply to assembly files.
* Architecture derived from mcpu beats any march options.
* When there are multiple mcpu or multiple march, the last
one wins.
* If there is a compiler option and an assembler option of
the same type, we prefer the one that fits the input type.
* If there is an applicable mcpu option but it is overruled
by an march, the cpu value is still used for the "-target-cpu"
cc1 option.
Reviewed By: nickdesaulniers
Differential Revision: https://reviews.llvm.org/D95872
When deducing a reference type for forwarding references prevent
adding default address space of a template argument if it is given.
This got reported in PR48896 because in OpenCL all parameters are
in private address space and therefore when we initialize a
forwarding reference with a parameter we should just inherit the
address space from it i.e. keep __private instead of __generic.
Tags: #clang
Differential Revision: https://reviews.llvm.org/D95624
This patch implements generation of remaining header search arguments.
It's done manually in C++ as opposed to TableGen, because we need the flexibility and don't anticipate reuse.
This patch also tests the generation of header search options via a round-trip. This way, the code gets exercised whenever Clang is built and tested in asserts mode. All `check-clang` tests pass.
Reviewed By: dexonsmith
Differential Revision: https://reviews.llvm.org/D94472
doubly-nested implicit CXXConstructExprs.
Ensure that we transform the parameter initializer using
TransformInitializer rather than TransformExpr so that we properly strip
down and rebuild the initialization, including any necessary
CXXBindTemporaryExprs. Otherwise we can end up forgetting to destroy
temporary objects used to construct a constructor parameter.
- On Windows, extended lambda has extra issues due to the numbering
schemes are different between the host compilation (Microsoft C++ ABI)
and the device compilation (Itanium C++ ABI. Additional device side
lambda number is required per lambda for the host compilation to
correctly mangle the device-side lambda name.
- A hybrid numbering context `MSHIPNumberingContext` is introduced to
number a lambda for both host- and device-compilations.
Reviewed By: rnk
Differential Revision: https://reviews.llvm.org/D69322
A module in the cache with an error should just be a cache miss. If
allowing errors (with -fallow-pcm-with-compiler-errors), a rebuild is
needed so that the appropriate diagnostics are output and in case search
paths have changed. If not allowing errors, the module was built
*allowing* errors and thus should be rebuilt regardless.
Reviewed By: akyrtzi
Differential Revision: https://reviews.llvm.org/D95989
when rewriting 'a < b' as '(a <=> b) < 0'.
It's pretty common for comparison category types to use a pointer or
pointer-to-member type as their '0' parameter.
This is a corner of the differences between C99 designators and C++20
designators that we'd previously overlooked. As with other such cases,
this continues to be permitted as an extension and allowed by default,
behind the -Wc99-designators warning flag, except in cases where it
leads to a conformance difference (such as in overload resolution and in
a SFINAE context).
Clang usually propagates counter mapping region for conditions of `if`, `while`,
`for`, etc from parent counter. We should do the same for condition of conditional operator.
Differential Revision: https://reviews.llvm.org/D95918
Currently clang is not correctly retrieving from the AST the metadata for
constrained FP builtins. This patch fixes that for the X86 specific builtins.
Differential Revision: https://reviews.llvm.org/D94614
On z/OS, other error messages are not matched correctly in lit tests.
```
EDC5121I Invalid argument.
EDC5111I Permission denied.
```
This patch adds a lit substitution to fix it.
Reviewed By: jhenderson
Differential Revision: https://reviews.llvm.org/D95808
Prevent materializing temporaries in the address space of the references
they are bind to. The temporaries should always be in the same address
space - private for OpenCL.
Tags: #clang
Differential Revision: https://reviews.llvm.org/D95608
Add the builtin functions brought by the cl_khr_subgroup_ballot
extension to `-fdeclare-opencl-builtins`.
Also add placeholder comments for the other Extended Subgroup
Functions from the OpenCL Extension Specification.
Add a comment clarifying the scope of the test.
Differential Revision: https://reviews.llvm.org/D95523
This patch adds AMDGPUOpenMPToolChain for supporting OpenMP
offloading to AMD GPU's.
Originally authored by Greg Rodgers
Reviewed By: jdoerfert
Differential Revision: https://reviews.llvm.org/D94961
Sample re-annotation is required in LTO time to achieve a reasonable post-inline profile quality. However, we have seen that such LTO-time re-annotation degrades profile quality. This is mainly caused by preLTO code duplication that is done by passes such as loop unrolling, jump threading, indirect call promotion etc, where samples corresponding to a source location are aggregated multiple times due to the duplicates. In this change we are introducing a concept of distribution factor for pseudo probes so that samples can be distributed for duplicated probes scaled by a factor. We hope that optimizations duplicating code well-maintain the branch frequency information (BFI) based on which probe distribution factors are calculated. Distribution factors are updated at the end of preLTO pipeline to reflect an estimated portion of the real execution count.
This change also introduces a pseudo probe verifier that can be run after each IR passes to detect duplicated pseudo probes.
A saturated distribution factor stands for 1.0. A pesudo probe will carry a factor with the value ranged from 0.0 to 1.0. A 64-bit integral distribution factor field that represents [0.0, 1.0] is associated to each block probe. Unfortunately this cannot be done for callsite probes due to the size limitation of a 32-bit Dwarf discriminator. A 7-bit distribution factor is used instead.
Changes are also needed to the sample profile inliner to deal with prorated callsite counts. Call sites duplicated by PreLTO passes, when later on inlined in LTO time, should have the callees’s probe prorated based on the Prelink-computed distribution factors. The distribution factors should also be taken into account when computing hotness for inline candidates. Also, Indirect call promotion results in multiple callisites. The original samples should be distributed across them. This is fixed by adjusting the callisites' distribution factors.
Reviewed By: wmi
Differential Revision: https://reviews.llvm.org/D93264
The number of iterations calculation was failing in some cases with more
than two collpased loops. Now the LoopIterationSpace selected matches
InitDependOnLC and CondDependOnLC.
Differential Revision: https://reviews.llvm.org/D95834
Restrict use of references to functions as they can
result in non-conforming behavior.
Tags: #clang
Differential Revision: https://reviews.llvm.org/D95442
Nicolas Guillemot