In the Vector to LLVM conversion, the `replaceTransferOp` function calls
into a type converter that may fail and suppresses the status. Change
the function to return the failure status instead, Since it is called
from a pattern, the failure can be readily propagated to the rest of
infrastructure.
Summary:
This patch handles illegal scalable types when lowering IR operations,
addressing several places where the value of isScalableVector() is
ignored.
For types such as <vscale x 8 x i32>, this means splitting the
operations. In this example, we would split it into two
operations of type <vscale x 4 x i32> for the low and high halves.
In cases such as <vscale x 2 x i32>, the elements in the vector
will be promoted. In this case they will be promoted to
i64 (with a vector of type <vscale x 2 x i64>)
Reviewers: sdesmalen, efriedma, huntergr
Reviewed By: efriedma
Subscribers: david-arm, tschuett, hiraditya, rkruppe, psnobl, cfe-commits, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D78812
Splitting critical edges for indirect branches
the SplitIndirectBrCriticalEdges() function may break branch
probabilities if target basic block happens to have unset
a probability for any of its successors. That is because in
such cases the getEdgeProbability(Target) function returns
probability 1/NumOfSuccessors and it is called after Target
was split (thus Target has a single successor). As the result
the correspondent successor of the split block gets
probability 100% but 1/NumOfSuccessors is expected (or better
be left unset).
Reviewers: yamauchi
Differential Revision: https://reviews.llvm.org/D78806
If tail-folding of the scalar remainder loop is applied, the primary induction
variable is splat to a vector and used by the masked load/store vector
instructions, thus the IV does not remain scalar. Because we now mark
that the IV does not remain scalar for these cases, we don't emit the vector IV
if it is not used. Thus, the vectoriser produces less dead code.
Thanks to Ayal Zaks for the direction how to fix this.
Differential Revision: https://reviews.llvm.org/D78911
When calculating the natural alignment for scalable vectors it
is acceptable to calculate an allocation size based on the minimum
number of elements in the vector.
This code path is exercised by an existing test:
CodeGen/AArch64/sve-intrinsics-int-arith.ll
Differential Revision: https://reviews.llvm.org/D79475
The list of destination load ops while evaluating producer-consumer
fusion wasn't being maintained as a set, and as such, duplicate load ops
were being added to it. Although this is harmless correctness-wise, it's
a killer efficiency-wise and it prevents interesting/useful fusions
(including for eg. reshapes into a matmul). The reason the latter
fusions would be missed is that a slice union would be unnecessarily
needed due to the duplicate load ops on a memref added to the 'dst
loads' list. Since slice union is unimplemented for the local var case,
a single destination load op that leads to local vars (like a floordiv /
mod producing fusion), a common case, would not get fused due to an
unnecessary union being tried with itself. (The union would actually be
the same thing but we would bail out.)
Besides the above, this would also significantly speed up fusion as all
the unnecessary slice computations / unions, checks, etc. due to the
duplicates go away.
Differential Revision: https://reviews.llvm.org/D79547
The source file used to test Objective-C data-formatters didn't respected
any formatting (mixed tab and spaces, lines exceed column 80, etc...).
This patch reformats the file using clang-format to make it easier to
work with.
Signed-off-by: Med Ismail Bennani <medismail.bennani@gmail.com>
This patch adds statistics about the contribution of each object file to
the linked debug info. When --statistics is passed to dsymutil, it
prints a table after linking as illustrated below.
It lists the object file name, the size of the debug info in the object
file in bytes, and the absolute size contribution to the linked dSYM and
the percentage difference. The table is sorted by the output size, so
the object files contributing the most to the link are listed first.
.debug_info section size (in bytes)
-------------------------------------------------------------------------------
Filename Object dSYM Change
-------------------------------------------------------------------------------
basic2.macho.x86_64.o 210b 165b -24.00%
basic3.macho.x86_64.o 177b 150b -16.51%
basic1.macho.x86_64.o 125b 129b 3.15%
-------------------------------------------------------------------------------
Total 512b 444b -14.23%
-------------------------------------------------------------------------------
Differential revision: https://reviews.llvm.org/D79513
f8990feb12 enabled installing PIC version of both libclang.a and
libclang.so when LIBCLANG_BUILD_STATIC is ON. But it broke the no-PIC
build when LLVM_ENABLE_PIC=OFF with the following error:
```
CMake Error at
/b/s/w/ir/cache/builder/src/third_party/llvm/clang/tools/libclang/CMakeLists.txt:123
(target_compile_definitions):
target_compile_definitions called with non-compilable target type
```
This is because as the code loops through ${name} and ${name}_static, it
introduced a side effect, which is adding an empty libclang_static to
targets. Later target_compile_definitions is called on libclang_static.
That function requires that target must have been created by a command
such as add_executable() or add_library(), so it crashed.
The solution is to not naively loop through both libclang and
libclang_static, but only the ones that are actually added by
llvm_add_library(). Here's the library build type to library target name
mapping:
| SHARED only | libclang |
| STATIC only | libclang |
| SHARED and STATIC | libclang and libclang_static |
So only when SHARED and STATIC are both set should we loop through two
targets. Explicitly parse the STATIC argument and set the list
accordingly.
Reviewed By: smeenai
Differential Revision: https://reviews.llvm.org/D79059
When reporting diagnostics from ASan's (and other sanitizer's) signal
handlers we should strip the "invalid signature" bit before printing
addresses. This makes the report less confusing and let's the user
focus on the real issue.
rdar://62615826
Reviewed By: kubamracek, delcypher
Differential Revision: https://reviews.llvm.org/D79132
Now using patterns, since there's a single-instruction lowering. (We
could convert to VSELECT and pattern-match that, but there doesn't seem
to be much point.)
I think this might be the first instruction to use nested multiclasses
this way? It seems like a good way to reduce duplication between
different integer widths. Let me know if it seems like an improvement.
Also, while I'm here, fix the return type of SETCC so we don't try to
merge a sign-extend with a SETCC.
Differential Revision: https://reviews.llvm.org/D79193
For Swift LLDB (but potentially also for module support in Clang-land)
we need a way to accumulate the path remappings produced by
Module::RegisterXcodeSDK(). In order to make this work for
SymbolFileDebugMaps, registering the search path remapping with both
modules is necessary.
Differential Revision: https://reviews.llvm.org/D79384
<rdar://problem/62750529>
Summary:
Add --relative to the suggested git-diff one liner. If the user does not
pass this argument, then git will produce a diff with the path relative
to the repository root. If the user's working directory is not the
repository root, then clang-format will complain that the file is not
found. The --relative argument makes git produce a diff with the files
relative to the working directory.
Add note to doc string to warn users about the fact that filenames
embedded in the diff are used as-is with no attempts to "do what they
mean, not what they say"
Reviewers: djasper, alexfh, efriedma, klimek, thakis
Reviewed By: efriedma
Subscribers: cfe-commits
Tags: #clang
Differential Revision: https://reviews.llvm.org/D79054
Summary: I updated the code that produces the message and the associated test.
Reviewers: sscalpone, DavidTruby, kiranchandramohan
Subscribers: llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D79514
Neither gcc or icc support this. Split out from D79472. I want
to remove more, but it looks like icc does support some things
gcc doesn't and I need to double check our internal test suites.
The Python used to run lit can be different from the Python LLDB was
build with. One scenario where this happens is when LLVM can find the
Python 3 interpreter, but not the Python 3 libraries, in which case LLDB
build and links against Python 3. Without this change, you end up with
an ModuleNotFoundError because of the mismatch.
Instead of looking at the Python interpreter that's used to run lit,
lldbtest should use the interpreter that matches the Python version LLDB
was build against.
Differential revision: https://reviews.llvm.org/D79519
loop nest.
Summary: As discussed in https://reviews.llvm.org/D73129.
Example
Before unroll and jam:
for
A
for
B
for
C
D
E
After unroll and jam (currently):
for
A
A'
for
B
for
C
D
B'
for
C'
D'
E
E'
After unroll and jam (Ideal):
for
A
A'
for
B
B'
for
C
C'
D
D'
E
E'
This is the first patch to change unroll and jam to work in the ideal
way.
This patch change the safety checks needed to make sure is safe to
unroll and jam in the ideal way.
Reviewer: dmgreen, jdoerfert, Meinersbur, kbarton, bmahjour, etiotto
Reviewed By: Meinersbur
Subscribers: fhahn, hiraditya, zzheng, llvm-commits, anhtuyen, prithayan
Tag: LLVM
Differential Revision: https://reviews.llvm.org/D76132
Y is the start of several 2 letter constraints, but we also had
partial support to recognize it by itself. But it doesn't look
like it can get through clang as a single letter so the backend
support for this was effectively dead.
A few documentation clarifications and moving one part of the
docs around to be closer to the first mention of display so that
it's easier to spot based on some user feedback.
Differential Revision: https://reviews.llvm.org/D79443
When debugging a remote platform, the platform you get from
GetPlatformForArchitecture doesn't inherit from PlatformDarwin.
HostInfoMacOSX seems like the right place to have a global store of
local paths.
Differential Revision: https://reviews.llvm.org/D79364
When debugging from a SymbolMap the creation of CompileUnits for the
individual object files is so lazy that RegisterXcodeSDK() is not
invoked at all before the Swift TypeSystem wants to read it. This
patch fixes this by introducing an explicit
SymbolFile::ParseXcodeSDK() call that can be invoked deterministically
before the result is required.
<rdar://problem/62532151+62326862>
https://reviews.llvm.org/D79273
passed to __builtin_os_log_format to extend its lifetime to the end of
its enclosing block
Extend only lifetimes of pointers returned by function calls or message
sends instead. In the long term, we should lifetime-extend pointers in
more complex expressions and non-ARC objects (e.g., C++ temporaries)
too.
rdar://problem/61846261
Summary: I updated the code that produces the message and the associated test.
Reviewers: sscalpone, DavidTruby
Subscribers: llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D79514
Before this patch, the debug record S_GTHREAD32 which represents global thread_local symbols, was emitted by LLD into the respective module stream. This makes Visual Studio unable to display thread_local symbols in the debugger.
After this patch, S_GTHREAD32 is moved into the globals stream. This matches MSVC behavior.
Differential Revision: https://reviews.llvm.org/D79005
Add a way to get mutable equivalence sets to Scope so that they can
have sizes and offsets assigned to them.
Change CommonBlockDetails to have mutable symbols so that they can have
sizes and offets assigned to them. This also allows the removal of some
`const_cast`s.
Add MutableSymbolRef and MutableSymbolVector as mutable analogs to
SymbolRef and SymbolVector. Replace uses of equivalent types with those
names.
Differential Revision: https://reviews.llvm.org/D79346
When using vec_load/store_len_r with an immediate length operand
of 16 or larger, LLVM will currently emit an VLRL/VSTRL instruction
with that immediate. This creates a valid encoding (which should be
supported by the assembler), but always traps at runtime. This patch
fixes this by not creating VLRL/VSTRL in those cases.
This would result in loading the length into a register and
calling VLRLR/VSTRLR instead. However, these operations with
a length of 15 or larger are in fact simply equivalent to a
full vector load or store. And in fact the same holds true for
vec_load/store_len as well.
Therefore, add a DAGCombine rule to replace those operations with
plain vector loads or stores if the length is known at compile
time and equal or larger to 15.
Summary: systemd recently added a clang-format file. One issue I
encountered in using clang-format on systemd is that systemd does
not add a space before the parens of their foreach macros but
clang-format always adds a space. This does not seem to be
configurable in clang-format. This revision adds the
ControlStatementsExceptForEachMacros option to SpaceBeforeParens
which puts a space before all control statement parens except
ForEach macros. This drastically reduces the amount of changes
when running clang-format on systemd's source code.
Reviewers: MyDeveloperDay, krasimir, mitchell-stellar
Reviewed By: MyDeveloperDay
Subscribers: cfe-commits
Tags: #clang-format, #clang
Differential Revision: https://reviews.llvm.org/D78869
Calling getShiftAmountTy with LegalTypes set may return a type that's too narrow to hold the shift amount for integer type it's applied to.
Fixes the regression introduced by D79096
Differential Revision: https://reviews.llvm.org/D79405
Objects in common blocks have offsets relative to the start of the
common block, independent of the enclosing scope, so they are processed
first. Add alignment to CommonBlockDetails to record the required
alignment of the common block.
For equivalence sets, each object depends on the one that is forced to
occur first in memory. The rest are recorded in the dependents_ map and
have offsets assigned after the other symbols are done.
Differential Revision: https://reviews.llvm.org/D79347
When the folding is performed in place, the `::fold` function does not populate
its `results` argument to indicate that. (In the folding hook for single-result
operations, the result of the original operation is expected to be returned,
but it is then ignored by the wrapper.) `OperationFolder::create` would
erronously rely on the _operation_ having zero results instead of on the
_folding_ producing zero new results to populate the list of results with those
of the original operation. This would lead to a crash for single-result ops
with in-place folds where the first result is accessed uncondtionally because
the list of results was not properly populated. Use the list of values produced
by the folding instead.
Differential Revision: https://reviews.llvm.org/D79497
Wen-Heng (Jack) Chung