In preperation of the deferred instantation progress, this patch
propagates the multi-level template argument lists further through the
API to reduce the size of that patch.
According to [basic.def.odr]p14, the same redeclarations in different TU
but not attached to a named module are allowed. But we didn't take care
of concept decl for this condition. This patch tries to fix this
problem.
Reviewed By: ilya-biryukov
Differention Revision: https://reviews.llvm.org/D130614
Add the ability to put __attribute__((maybe_undef)) on function arguments.
Clang codegen introduces a freeze instruction on the argument.
Differential Revision: https://reviews.llvm.org/D130224
HLSL Resource objects will have restrictions on use and codegen
requirements. This patch is fairly minimal just adding the attribute
with no spellings since it will only be attached by the
HLSLExternalSemaSource.
Depends on D1300017.
Differential Revision: https://reviews.llvm.org/D130018
DR2338 clarified that it was undefined behavior to set the value outside the
range of the enumerations values for an enum without a fixed underlying type.
We should diagnose this with a constant expression context.
Differential Revision: https://reviews.llvm.org/D130058
The "strict context hash" is insufficient to identify module
dependencies during scanning, leading to different module build commands
being produced for a single module, and non-deterministically choosing
between them. This commit switches to hashing the canonicalized
`CompilerInvocation` of the module. By hashing the invocation we are
converting these from correctness issues to performance issues, and we
can then incrementally improve our ability to canonicalize
command-lines.
This change can cause a regression in the number of modules needed. Of
the 4 projects I tested, 3 had no regression, but 1, which was
clang+llvm itself, had a 66% regression in number of modules (4%
regression in total invocations). This is almost entirely due to
differences between -W options across targets. Of this, 25% of the
additional modules are system modules, which we could avoid if we
canonicalized -W options when -Wsystem-headers is not present --
unfortunately this is non-trivial due to some warnings being enabled in
system headers by default. The rest of the additional modules are mostly
real differences in potential warnings, reflecting incorrect behaviour
in the current scanner.
There were also a couple of differences due to `-DFOO`
`-fmodule-ignore-macro=FOO`, which I fixed here.
Since the output paths for the module depend on its context hash, we
hash the invocation before filling in outputs, and rely on the build
system to always return the same output paths for a given module.
Note: since the scanner itself uses an implicit modules build, there can
still be non-determinism, but it will now present as different
module+hashes rather than different command-lines for the same
module+hash.
Differential Revision: https://reviews.llvm.org/D129884
This fills out the default constructor for RWBuffer to assign the
handle with the result of __builtin_hlsl_create_handle which we can
then treat as a pointer to the resource data through the mid-level of
the compiler.
Depends on D130016
Differential Revision: https://reviews.llvm.org/D130017
This builtin allows the creation of custom scheduling pipelines on a per-region
basis. Like the sched_barrier builtin this is intended to be used either for
testing, in situations where the default scheduler heuristics cannot be
improved, or in critical kernels where users are trying to get performance that
is close to handwritten assembly. Obviously using these builtins will require
extra work from the kernel writer to maintain the desired behavior.
The builtin can be used to create groups of instructions called "scheduling
groups" where ordering between the groups is enforced by the scheduler.
__builtin_amdgcn_sched_group_barrier takes three parameters. The first parameter
is a mask that determines the types of instructions that you would like to
synchronize around and add to a scheduling group. These instructions will be
selected from the bottom up starting from the sched_group_barrier's location
during instruction scheduling. The second parameter is the number of matching
instructions that will be associated with this sched_group_barrier. The third
parameter is an identifier which is used to describe what other
sched_group_barriers should be synchronized with. Note that multiple
sched_group_barriers must be added in order for them to be useful since they
only synchronize with other sched_group_barriers. Only "scheduling groups" with
a matching third parameter will have any enforced ordering between them.
As an example, the code below tries to create a pipeline of 1 VMEM_READ
instruction followed by 1 VALU instruction followed by 5 MFMA instructions...
// 1 VMEM_READ
__builtin_amdgcn_sched_group_barrier(32, 1, 0)
// 1 VALU
__builtin_amdgcn_sched_group_barrier(2, 1, 0)
// 5 MFMA
__builtin_amdgcn_sched_group_barrier(8, 5, 0)
// 1 VMEM_READ
__builtin_amdgcn_sched_group_barrier(32, 1, 0)
// 3 VALU
__builtin_amdgcn_sched_group_barrier(2, 3, 0)
// 2 VMEM_WRITE
__builtin_amdgcn_sched_group_barrier(64, 2, 0)
Reviewed By: jrbyrnes
Differential Revision: https://reviews.llvm.org/D128158
This is pretty straightforward, it just adds a builtin to return a
pointer to a resource handle. This maps to a dx intrinsic.
The shape of this builtin and the underlying intrinsic will likely
shift a bit as this implementation becomes more feature complete, but
this is a good basis to get started.
Depends on D128569.
Differential Revision: https://reviews.llvm.org/D130016
Most of the change here is fleshing out the HLSLExternalSemaSource with
builder implementations to build the builtin types. Eventually, I may
move some of this code into tablegen or a more managable declarative
file but I want to get the AST generation logic ready first.
This code adds two new types into the HLSL AST, `hlsl::Resource` and
`hlsl::RWBuffer`. The `Resource` type is just a wrapper around a handle
identifier, and is largely unused in source. It will morph a bit over
time as I work on getting the source compatability correct, but for now
it is a reasonable stand-in. The `RWBuffer` type is not ready for use.
I'm posting this change for review because it adds a lot of
infrastructure code and is testable.
There is one change to clang code outside the HLSL-specific logic here,
which addresses a behavior change introduced a long time ago in
967d438439. That change resulted in unintentionally breaking
situations where an incomplete template declaration was provided from
an AST source, and needed to be completed later by the external AST.
That situation doesn't happen in the normal AST importer flow, but can
happen when an AST source provides incomplete declarations of
templates. The solution is to annotate template specializations of
incomplete types with the HasExternalLexicalSource bit from the base
template.
Depends on D128012.
Differential Revision: https://reviews.llvm.org/D128569
These vdup and vmov float16 intrinsics are being defined in both the
general section and then again in fp16 under a !aarch64 flag. The
vdup_lane intrinsics were being defined in both aarch64 and !aarch64
sections, so have been commoned. They are defined as macros, so do not
give duplicate warnings, but removing the duplicates shouldn't alter the
available intrinsics.
DR2338 clarified that it was undefined behavior to set the value outside the
range of the enumerations values for an enum without a fixed underlying type.
We should diagnose this with a constant expression context.
Differential Revision: https://reviews.llvm.org/D130058
Lifting the core functionalities of the clang-offload-bundler into a
user-facing library/API. This will allow online and JIT compilers to
bundle and unbundle files without spawning a new process.
This patch lifts the classes and functions used to implement
the clang-offload-bundler into a separate OffloadBundler.cpp,
and defines three top-level API functions in OfflaodBundler.h.
BundleFiles()
UnbundleFiles()
UnbundleArchives()
This patch also introduces a Config class that locally stores the
previously global cl::opt options and arrays to allow users to call
the APIs in a multi-threaded context, and introduces an
OffloadBundler class to encapsulate the top-level API functions.
We also lift the BundlerExecutable variable, which is specific
to the clang-offload-bundler tool, from the API, and replace
its use with an ObjcopyPath variable. This variable must be set
in order to internally call llvm-objcopy.
Finally, we move the API files from
clang/tools/clang-offload-bundler into clang/lib/Driver and
clang/include/clang/Driver.
Differential Revision: https://reviews.llvm.org/D129873
Use a delegating constructor to remove the last use of the deprecated
ctor of `TypeErasedDataflowAnalysis`, and then delete it.
Differential Revision: https://reviews.llvm.org/D130653
Otherwise we get invalid results for ODR checks. See changed test for an
example: despite the fact that we merge the first concept, its **uses**
were considered different by `Profile`, leading to redefinition errors.
After this change, canonical decl for a concept can come from a
different module and may not be visible. This behavior looks suspicious,
but does not break any tests. We might want to add a mechanism to make
the canonical concept declaration visible if we find code that relies on
this invariant.
Additionally make sure we always merge with the canonical declaration to
avoid chains of merged concepts being reported as redefinitions. An
example was added to the test.
Also change the order of includes in the test. Importing a moduralized
header before its textual part causes the include guard macro to be
exported and the corresponding `#include` becomes a no-op.
Reviewed By: ChuanqiXu
Differential Revision: https://reviews.llvm.org/D130585
Without this patch, clang will not wrap in an ElaboratedType node types written
without a keyword and nested name qualifier, which goes against the intent that
we should produce an AST which retains enough details to recover how things are
written.
The lack of this sugar is incompatible with the intent of the type printer
default policy, which is to print types as written, but to fall back and print
them fully qualified when they are desugared.
An ElaboratedTypeLoc without keyword / NNS uses no storage by itself, but still
requires pointer alignment due to pre-existing bug in the TypeLoc buffer
handling.
---
Troubleshooting list to deal with any breakage seen with this patch:
1) The most likely effect one would see by this patch is a change in how
a type is printed. The type printer will, by design and default,
print types as written. There are customization options there, but
not that many, and they mainly apply to how to print a type that we
somehow failed to track how it was written. This patch fixes a
problem where we failed to distinguish between a type
that was written without any elaborated-type qualifiers,
such as a 'struct'/'class' tags and name spacifiers such as 'std::',
and one that has been stripped of any 'metadata' that identifies such,
the so called canonical types.
Example:
```
namespace foo {
struct A {};
A a;
};
```
If one were to print the type of `foo::a`, prior to this patch, this
would result in `foo::A`. This is how the type printer would have,
by default, printed the canonical type of A as well.
As soon as you add any name qualifiers to A, the type printer would
suddenly start accurately printing the type as written. This patch
will make it print it accurately even when written without
qualifiers, so we will just print `A` for the initial example, as
the user did not really write that `foo::` namespace qualifier.
2) This patch could expose a bug in some AST matcher. Matching types
is harder to get right when there is sugar involved. For example,
if you want to match a type against being a pointer to some type A,
then you have to account for getting a type that is sugar for a
pointer to A, or being a pointer to sugar to A, or both! Usually
you would get the second part wrong, and this would work for a
very simple test where you don't use any name qualifiers, but
you would discover is broken when you do. The usual fix is to
either use the matcher which strips sugar, which is annoying
to use as for example if you match an N level pointer, you have
to put N+1 such matchers in there, beginning to end and between
all those levels. But in a lot of cases, if the property you want
to match is present in the canonical type, it's easier and faster
to just match on that... This goes with what is said in 1), if
you want to match against the name of a type, and you want
the name string to be something stable, perhaps matching on
the name of the canonical type is the better choice.
3) This patch could expose a bug in how you get the source range of some
TypeLoc. For some reason, a lot of code is using getLocalSourceRange(),
which only looks at the given TypeLoc node. This patch introduces a new,
and more common TypeLoc node which contains no source locations on itself.
This is not an inovation here, and some other, more rare TypeLoc nodes could
also have this property, but if you use getLocalSourceRange on them, it's not
going to return any valid locations, because it doesn't have any. The right fix
here is to always use getSourceRange() or getBeginLoc/getEndLoc which will dive
into the inner TypeLoc to get the source range if it doesn't find it on the
top level one. You can use getLocalSourceRange if you are really into
micro-optimizations and you have some outside knowledge that the TypeLocs you are
dealing with will always include some source location.
4) Exposed a bug somewhere in the use of the normal clang type class API, where you
have some type, you want to see if that type is some particular kind, you try a
`dyn_cast` such as `dyn_cast<TypedefType>` and that fails because now you have an
ElaboratedType which has a TypeDefType inside of it, which is what you wanted to match.
Again, like 2), this would usually have been tested poorly with some simple tests with
no qualifications, and would have been broken had there been any other kind of type sugar,
be it an ElaboratedType or a TemplateSpecializationType or a SubstTemplateParmType.
The usual fix here is to use `getAs` instead of `dyn_cast`, which will look deeper
into the type. Or use `getAsAdjusted` when dealing with TypeLocs.
For some reason the API is inconsistent there and on TypeLocs getAs behaves like a dyn_cast.
5) It could be a bug in this patch perhaps.
Let me know if you need any help!
Signed-off-by: Matheus Izvekov <mizvekov@gmail.com>
Differential Revision: https://reviews.llvm.org/D112374
We are supporting quadword lock free atomics on AIX. For the situation that users on AIX are using a libatomic that is lock-based for quadword types, we can't enable quadword lock free atomics by default on AIX in case user's new code and existing code accessing the same shared atomic quadword variable, we can't guarentee atomicity. So we need an option to enable quadword lock free atomics on AIX, thus we can build a quadword lock-free libatomic(also for advanced users considering atomic performance critical) for users to make the transition smooth.
Reviewed By: shchenz
Differential Revision: https://reviews.llvm.org/D127189
This patch adds initial support for context-sensitive analysis of simple functions whose definition is available in the translation unit, guarded by the `ContextSensitive` flag in the new `TransferOptions` struct. When this option is true, the `VisitCallExpr` case in the builtin transfer function has a fallthrough case which checks for a direct callee with a body. In that case, it constructs a CFG from that callee body, uses the new `pushCall` method on the `Environment` to make an environment to analyze the callee, and then calls `runDataflowAnalysis` with a `NoopAnalysis` (disabling context-sensitive analysis on that sub-analysis, to avoid problems with recursion). After the sub-analysis completes, the `Environment` from its exit block is simply assigned back to the environment at the callsite.
The `pushCall` method (which currently only supports non-method functions with some restrictions) maps the `SourceLocation`s for all the parameters to the existing source locations for the corresponding arguments from the callsite.
This patch adds a few tests to check that this context-sensitive analysis works on simple functions. More sophisticated functionality will be added later; the most important next step is to explicitly model context in some fields of the `DataflowAnalysisContext` class, as mentioned in a `FIXME` comment in the `pushCall` implementation.
Reviewed By: ymandel, xazax.hun
Differential Revision: https://reviews.llvm.org/D130306
Depends On D130305
This patch adds initial support for context-sensitive analysis of simple functions whose definition is available in the translation unit, guarded by the `ContextSensitive` flag in the new `TransferOptions` struct. When this option is true, the `VisitCallExpr` case in the builtin transfer function has a fallthrough case which checks for a direct callee with a body. In that case, it constructs a CFG from that callee body, uses the new `pushCall` method on the `Environment` to make an environment to analyze the callee, and then calls `runDataflowAnalysis` with a `NoopAnalysis` (disabling context-sensitive analysis on that sub-analysis, to avoid problems with recursion). After the sub-analysis completes, the `Environment` from its exit block is simply assigned back to the environment at the callsite.
The `pushCall` method (which currently only supports non-method functions with some restrictions) first calls `initGlobalVars`, then maps the `SourceLocation`s for all the parameters to the existing source locations for the corresponding arguments from the callsite.
This patch adds a few tests to check that this context-sensitive analysis works on simple functions. More sophisticated functionality will be added later; the most important next step is to explicitly model context in some fields of the `DataflowAnalysisContext` class, as mentioned in a `TODO` comment in the `pushCall` implementation.
Reviewed By: ymandel, xazax.hun
Differential Revision: https://reviews.llvm.org/D130306
Lifting the core functionalities of the clang-offload-bundler into a
user-facing library/API. This will allow online and JIT compilers to
bundle and unbundle files without spawning a new process.
This patch lifts the classes and functions used to implement
the clang-offload-bundler into a separate OffloadBundler.cpp,
and defines three top-level API functions in OfflaodBundler.h.
BundleFiles()
UnbundleFiles()
UnbundleArchives()
This patch also introduces a Config class that locally stores the
previously global cl::opt options and arrays to allow users to call
the APIs in a multi-threaded context, and introduces an
OffloadBundler class to encapsulate the top-level API functions.
We also lift the BundlerExecutable variable, which is specific
to the clang-offload-bundler tool, from the API, and replace
its use with an ObjcopyPath variable. This variable must be set
in order to internally call llvm-objcopy.
Finally, we move the API files from
clang/tools/clang-offload-bundler into clang/lib/Driver and
clang/include/clang/Driver.
Differential Revision: https://reviews.llvm.org/D129873
Currently, the use of preferred_name would block implementing std
modules in libcxx. See https://github.com/llvm/llvm-project/issues/56490
for example.
The problem is pretty hard and it looks like we couldn't solve it in a
short time. So we sent this patch as a workaround to avoid blocking us
to modularize STL. This is intended to be fixed properly in the future.
Reviewed By: erichkeane, aaron.ballman, tahonermann
Differential Revision: https://reviews.llvm.org/D130331
Previously we used to desugar implications and biconditionals into
equivalent CNF/DNF as soon as possible. However, this desugaring makes
debug output (Environment::dump()) less readable than it could be.
Therefore, it makes sense to keep the sugared representation of a
boolean formula, and desugar it in the solver.
Reviewed By: sgatev, xazax.hun, wyt
Differential Revision: https://reviews.llvm.org/D130519
Without this patch when using CMAKE_CXX_STANDARD=20 Microsoft compiler produces following warnings
clang\include\clang/Basic/DiagnosticIDs.h(48): warning C5054: operator '+': deprecated between enumerations of different types
clang\include\clang/Basic/DiagnosticIDs.h(49): warning C5054: operator '+': deprecated between enumerations of different types
clang\include\clang/Basic/DiagnosticIDs.h(50): warning C5054: operator '+': deprecated between enumerations of different types
clang\include\clang/Basic/DiagnosticIDs.h(51): warning C5054: operator '+': deprecated between enumerations of different types
clang\include\clang/Basic/DiagnosticIDs.h(52): warning C5054: operator '+': deprecated between enumerations of different types
clang\include\clang/Basic/DiagnosticIDs.h(53): warning C5054: operator '+': deprecated between enumerations of different types
clang\include\clang/Basic/DiagnosticIDs.h(54): warning C5054: operator '+': deprecated between enumerations of different types
clang\include\clang/Basic/DiagnosticIDs.h(55): warning C5054: operator '+': deprecated between enumerations of different types
clang\include\clang/Basic/DiagnosticIDs.h(56): warning C5054: operator '+': deprecated between enumerations of different types
clang\include\clang/Basic/DiagnosticIDs.h(57): warning C5054: operator '+': deprecated between enumerations of different types
clang\include\clang/Basic/DiagnosticIDs.h(58): warning C5054: operator '+': deprecated between enumerations of different types
clang\include\clang/Basic/DiagnosticIDs.h(59): warning C5054: operator '+': deprecated between enumerations of different types
Patch By: Godin
Reviewed By: aaron.ballman
Differential Revision: https://reviews.llvm.org/D130476
Before the patch we calculated the NRVO candidate looking at the
variable's whole enclosing scope. The research in [P2025] shows that
looking at the variable's potential scope is better and covers more
cases where NRVO would be safe and desirable.
Many thanks to @Izaron for the original implementation.
Reviewed By: ChuanqiXu
Differential Revision: https://reviews.llvm.org/D119792
Remove MaskedPrototype and add several fields in RVVIntrinsicRecord,
compute Prototype in runtime.
Reviewed By: rogfer01
Differential Revision: https://reviews.llvm.org/D126741
Leverage the method OpenCL uses that adds C intrinsics when the lookup
failed. There is no need to define C intrinsics in the header file any
more. It could help to avoid the large header file to speed up the
compilation of RVV source code. Besides that, only the C intrinsics used
by the users will be added into the declaration table.
This patch is based on https://reviews.llvm.org/D103228 and inspired by
OpenCL implementation.
### Experimental Results
#### TL;DR:
- Binary size of clang increase ~200k, which is +0.07% for debug build and +0.13% for release build.
- Single file compilation speed up ~33x for debug build and ~8.5x for release build
- Regression time reduce ~10% (`ninja check-all`, enable all targets)
#### Header size change
```
| size | LoC |
------------------------------
Before | 4,434,725 | 69,749 |
After | 6,140 | 162 |
```
#### Single File Compilation Time
Testcase:
```
#include <riscv_vector.h>
vint32m1_t test_vadd_vv_vfloat32m1_t(vint32m1_t op1, vint32m1_t op2, size_t vl) {
return vadd(op1, op2, vl);
}
```
##### Debug build:
Before:
```
real 0m19.352s
user 0m19.252s
sys 0m0.092s
```
After:
```
real 0m0.576s
user 0m0.552s
sys 0m0.024s
```
~33x speed up for debug build
##### Release build:
Before:
```
real 0m0.773s
user 0m0.741s
sys 0m0.032s
```
After:
```
real 0m0.092s
user 0m0.080s
sys 0m0.012s
```
~8.5x speed up for release build
#### Regression time
Note: the failed case is `tools/llvm-debuginfod-find/debuginfod.test` which is unrelated to this patch.
##### Debug build
Before:
```
Testing Time: 1358.38s
Skipped : 11
Unsupported : 446
Passed : 75767
Expectedly Failed: 190
Failed : 1
```
After
```
Testing Time: 1220.29s
Skipped : 11
Unsupported : 446
Passed : 75767
Expectedly Failed: 190
Failed : 1
```
##### Release build
Before:
```
Testing Time: 381.98s
Skipped : 12
Unsupported : 1407
Passed : 74765
Expectedly Failed: 176
Failed : 1
```
After:
```
Testing Time: 346.25s
Skipped : 12
Unsupported : 1407
Passed : 74765
Expectedly Failed: 176
Failed : 1
```
#### Binary size of clang
##### Debug build
Before
```
text data bss dec hex filename
335261851 12726004 552812 348540667 14c64efb bin/clang
```
After
```
text data bss dec hex filename
335442803 12798708 552940 348794451 14ca2e53 bin/clang
```
+253K, +0.07% code size
##### Release build
Before
```
text data bss dec hex filename
144123975 8374648 483140 152981763 91e5103 bin/clang
```
After
```
text data bss dec hex filename
144255762 8447296 483268 153186326 9217016 bin/clang
```
+204K, +0.13%
Authored-by: Kito Cheng <kito.cheng@sifive.com>
Co-Authored-by: Hsiangkai Wang <kai.wang@sifive.com>
Reviewed By: khchen, aaron.ballman
Differential Revision: https://reviews.llvm.org/D111617
This patch introduces a new `ConstructionContext` for
lambda capture. This `ConstructionContext` allows the
analyzer to construct the captured object directly into
it's final region, and makes it possible to capture
non-POD arrays.
Differential Revision: https://reviews.llvm.org/D129967
This patch introduces the evaluation of ArrayInitLoopExpr
in case of structured bindings and implicit copy/move
constructor. The idea is to call the copy constructor for
every element in the array. The parameter of the copy
constructor is also manually selected, as it is not a part
of the CFG.
Differential Revision: https://reviews.llvm.org/D129496
Fix `MapLattice` API to return `std::pair<iterator, bool>`,
allowing users to detect when an element has been inserted without
performing a redundant map lookup.
Differential Revision: https://reviews.llvm.org/D130497
This includes the revised provisions of [basic.lookup.argdep] p4
1. ADL is amended to handle p 4.3 where functions in trasitively imported modules may
become visible when they are exported in the same namespace as a visible type.
2. If a function is in a different modular TU, and has internal-linkage, we invalidate
its entry in an overload set.
[basic.lookup.argdep] p5 ex 2 now passes.
Differential Revision: https://reviews.llvm.org/D129174
Currently the C++20 concepts are only merged in `ASTReader`, i.e. when
coming from different TU. This can causes ambiguious reference errors when
trying to access the same concept that should otherwise be merged.
Please see the added test for an example.
Note that we currently use `ASTContext::isSameEntity` to check for ODR
violations. However, it will not check that concept requirements match.
The same issue holds for mering concepts from different TUs, I added a
FIXME and filed a GH issue to track this:
https://github.com/llvm/llvm-project/issues/56310
Reviewed By: ChuanqiXu
Differential Revision: https://reviews.llvm.org/D128921
This patch rewords the static assert diagnostic output. Failing a
_Static_assert in C should not report that static_assert failed. This
changes the wording to be more like GCC and uses "static assertion"
when possible instead of hard coding the name. This also changes some
instances of 'static_assert' to instead be based on the token in the
source code.
Differential Revision: https://reviews.llvm.org/D129048
report an error when encountering 'while' token parsing declarator
```
clang/test/Parser/while-loop-outside-function.c:3:1: error: while loop outside of a function
while // expected-error {{while loop outside of a function}}
^
clang/test/Parser/while-loop-outside-function.c:7:1: error: while loop outside of a function
while // expected-error {{while loop outside of a function}}
^
```
Fixes: https://github.com/llvm/llvm-project/issues/34462
Differential Revision: https://reviews.llvm.org/D129573
As per P2327R1,
|=, &= and ^= are no longer deprecated in all languages mode.
Reviewed By: aaron.ballman
Differential Revision: https://reviews.llvm.org/D130421
The #warning directive is standard in C++2b and C2x,
this adjusts the pedantic and extensions warning accordingly.
Reviewed By: aaron.ballman
Differential Revision: https://reviews.llvm.org/D130415
Erich Keane