D114999 added code to kill an immediate def if it was folded into its
only use by convertToThreeAddress. This patch updates LiveVariables when
that happens in order to fix verification failures exposed by D129213.
Differential Revision: https://reviews.llvm.org/D129661
This patch adds a test which shows that we may incorrectly register
allocate for RVV instructions which have no-overlap constraints on
source/dest registers of different LMUL groups.
The particular case shows that a vrgatherei16 instruction writes to a
LMUL=1 register group v11 and reads from an EMUL=2 register group
v10/v11. This breaks the overlap constraints of the vrgatherei16
instruction.
The test also shows that disabling subregister liveness fixes the test.
We use `early-clobber` on the `VR` dest and the `VRM2` source to enforce
the constraint but with subregister liveness this constraint is not met.
It's unclear to me at this point whether this is per-design of
early-clobber in conjunction with subregisters (meaning we should find
another way of expressing this constraint) or whether it's a bug in the
register allocator somewhere.
Reviewed By: rogfer01
Differential Revision: https://reviews.llvm.org/D129639
In the Cortex A57 Optimization Guide [1] VCVTAU (AArch32) is incorrectly
listed in the AArch64 instructions for instruction groups:
- ASIMD FP convert, other, D-form
- ASIMD FP convert, other, Q-form
It's meant to be FCVTAU, this will be fixed in future releases of the guide.
[1] https://developer.arm.com/documentation/uan0015/b
In preparation for follow up patch folding above to CNOT.
Reviewed By: paulwalker-arm, peterwaller-arm
Differential Revision: https://reviews.llvm.org/D129625
The constructor of PatternApplicator doesn't have a constructor that
accepts only a `RewritePatternSet` as currently used in the example
code in PatternRewriter.md. Instead, one has to turn it into a
`FrozenRewritePatternSet`.
Reviewed By: nicolasvasilache
Differential Revision: https://reviews.llvm.org/D125236
Make the implementation more similar to other functions, by
explicitly skipping an unknown/undef first, and always falling
back to overdefined at the end. I don't think it makes a difference
now, but could make one once the constant evaluation can fail. In
that case we would directly mark the result as overdefined now,
rather than keeping it unknown (and later making it overdefined
because we think it's undef-based).
D25618 added a method to verify the instruction predicates for an
emitted instruction, through verifyInstructionPredicates added into
<Target>MCCodeEmitter::encodeInstruction. This is a very useful idea,
but the implementation inside MCCodeEmitter made it only fire for object
files, not assembly which most of the llvm test suite uses.
This patch moves the code into the <Target>_MC::verifyInstructionPredicates
method, inside the InstrInfo. The allows it to be called from other
places, such as in this patch where it is called from the
<Target>AsmPrinter::emitInstruction methods which should trigger for
both assembly and object files. It can also be called from other places
such as verifyInstruction, but that is not done here (it tends to catch
errors earlier, but in reality just shows all the mir tests that have
incorrect feature predicates). The interface was also simplified
slightly, moving computeAvailableFeatures into the function so that it
does not need to be called externally.
The ARM, AMDGPU (but not R600), AVR, Mips and X86 backends all currently
show errors in the test-suite, so have been disabled with FIXME
comments.
Recommitted with some fixes for the leftover MCII variables in release
builds.
Differential Revision: https://reviews.llvm.org/D129506
The value lattice explicitly represents undef, and markConstant()
internally checks for UndefValue and will create an undef rather
than constant lattice element in that case.
This is mostly a code simplification, it has little practical impact
because we usually get undef results from undef operands, and those
don't get processed.
Only leave the check behind for the CmpInst case, because it
currently goes through this incorrect code in the getCompare()
implementation: f98697642c/llvm/include/llvm/Analysis/ValueLattice.h (L456-L457)
Differential Revision: https://reviews.llvm.org/D128330
jGetLoadedDynamicLibrariesInfos normally checks with dyld to find
the list of binaries loaded in the inferior, and getting the filepath,
before trying to parse the Mach-O binary in inferior memory.
This allows for debugserver to parse a Mach-O binary present in memory,
but not yet registered with dyld. This patch also adds some simple
sanity checks that we're reading a Mach-O header before we begin
stepping through load commands, because we won't have the sanity check
of consulting dyld for the list of loaded binaries before parsing.
Also adds a testcase.
[This patch was reverted after causing a testsuite failure on a CI bot;
I haven't been able to repro the failure outside the CI, but I have a
theory that my sanity check on cputype which only matched arm64 and
x86_64 - and the CI machine may have a watch simulator that is still
using i386.]
Differential Revision: https://reviews.llvm.org/D128956
rdar://95737734
This change removes the partial bufferization passes from the sparse compilation pipeline and replaces them with One-Shot Bufferize. One-Shot Analysis (and TensorCopyInsertion) is used to resolve all out-of-place bufferizations, dense and sparse. Dense ops are then bufferized with BufferizableOpInterface. Sparse ops are still bufferized in the Sparsification pass.
Details:
* Dense allocations are automatically deallocated, unless they are yielded from a block. (In that case the alloc would leak.) All test cases are modified accordingly. E.g., some funcs now have an "out" tensor argument that is returned from the function. (That way, the allocation happens at the call site.)
* Sparse allocations are *not* automatically deallocated. They must be "released" manually. (No change, this will be addressed in a future change.)
* Sparse tensor copies are not supported yet. (Future change)
* Sparsification no longer has to consider inplacability. If necessary, allocations and/or copies are inserted during TensorCopyInsertion. All tensors are inplaceable by the time Sparsification is running. Instead of marking a tensor as "not inplaceable", it can be marked as "not writable", which will trigger an allocation and/or copy during TensorCopyInsertion.
Differential Revision: https://reviews.llvm.org/D129356
The SI machine scheduler inherits from ScheduleDAGMI.
This patch adds support for a few features that are implemented
in ScheduleDAGMI (or its base classes) that were missing so far
because their support is implemented in overridden functions.
* Support cl::opt -view-misched-dags
This option allows to open a graphical window of the scheduling DAG.
* Support cl::opt -misched-print-dags
This option allows to print the scheduling DAG in text form.
* After constructing the scheduling DAG, call postprocessDAG()
to apply any registered DAG mutations.
Note that currently there are no mutations defined in AMDGPUTargetMachine.cpp
in case SIScheduler is used.
Still add this to avoid surprises in the future in case mutations are added.
Differential Revision: https://reviews.llvm.org/D128808
-o is very common among tools. yaml2obj supports -o and it surprised me that
obj2yaml doesn't support -o. Just add it which doesn't take much code.
Differential Revision: https://reviews.llvm.org/D129713
With my version of the MSVC tools (14.11.25503), this was failing to
build because of missing declarations of `std::isalnum` and
`std::isdigit`. Include `<cctype>` to get these.
Detect template specializations that should be handled specially.
In some cases it is allowed to extend the `std` namespace with
template specializations.
Reviewed By: aaron.ballman
Differential Revision: https://reviews.llvm.org/D129353
There are two assumptions regarding jump table:
(a) It is accessed by only one fragment, say, Parent
(b) All entries target instructions in Parent
For (a), BOLT stores jump table entries as relative offset to Parent.
For (b), BOLT treats jump table entries target somewhere out of Parent
as INVALID_OFFSET, including fragment of same split function.
In this update, we extend (a) and (b) to include fragment of same split
functinon. For (a), we store jump table entries in absolute offset
instead. In addition, jump table will store all fragments that access
it. A fragment uses this information to only create label for jump table
entries that target to that fragment.
For (b), using absolute offset allows jump table entries to target
fragments of same split function, i.e., extend support for split jump
table. This can be done using relocation (fragment start/size) and
fragment detection heuristics (e.g., using symbol name pattern for
non-stripped binaries).
For jump table targets that can only be reached by one fragment, we
mark them as local label; otherwise, they would be the secondary
function entry to the target fragment.
Test Plan
```
ninja check-bolt
```
Reviewed By: Amir
Differential Revision: https://reviews.llvm.org/D128474
WG21 approved delimited escape sequences and named escape
sequences.
Adjust the extension warnings accordingly, and update
the release notes.
Reviewed By: aaron.ballman
Differential Revision: https://reviews.llvm.org/D129664
- add `FindZSTD.cmake`
- add zstd to `llvm::compression` namespace
- add a CMake option `LLVM_ENABLE_ZSTD` with behavior mirroring that of `LLVM_ENABLE_ZLIB`
- add tests for zstd to `llvm/unittests/Support/CompressionTest.cpp`
Reviewed By: leonardchan, MaskRay
Differential Revision: https://reviews.llvm.org/D128465
New version of Intel LAM patches
(https://lore.kernel.org/linux-mm/20220712231328.5294-1-kirill.shutemov@linux.intel.com/)
uses a different interface based on arch_prctl():
- arch_prctl(ARCH_GET_UNTAG_MASK, &mask) returns the current mask for
untagging the pointers. We use it to detect kernel LAM support.
- arch_prctl(ARCH_ENABLE_TAGGED_ADDR, nr_bits) enables pointer tagging
for the current process.
Because __NR_arch_prctl is defined in different headers, and no other
platforms need it at the moment, we only declare internal_arch_prctl()
on x86_64.
Reviewed By: vitalybuka
Differential Revision: https://reviews.llvm.org/D129645
- add `FindZSTD.cmake`
- add zstd to `llvm::compression` namespace
- add a CMake option `LLVM_ENABLE_ZSTD` with behavior mirroring that of `LLVM_ENABLE_ZLIB`
- add tests for zstd to `llvm/unittests/Support/CompressionTest.cpp`
Reviewed By: leonardchan, MaskRay
Differential Revision: https://reviews.llvm.org/D128465
The new driver primarily allows us to support RDC-mode compilations with
proper linking. This is not needed for non-RDC mode compilation, but we
still would like the new driver to be able to handle this mode so we can
transition away from the old driver in the future. This patch adds the
necessary code to support creating a fatbinary for CUDA code generation
as well as removing old assumptions and errors about RDC-mode with the
new driver.
Reviewed By: tra
Differential Revision: https://reviews.llvm.org/D129655
This just removes the code that gates the logic. The main issue here is
perf impact: without {D122258}, LLD takes a significant perf hit because
it now has to do a lot more work in the input parsing phase. But with
that change to eliminate unnecessary EH frames from input object files,
the perf overhead here is minimal. Concretely, here are the numbers for
some builds as measured on my 16-core Mac Pro:
**chromium_framework**
This is without the use of `-femit-dwarf-unwind=no-compact-unwind`:
base diff difference (95% CI)
sys_time 1.826 ± 0.019 1.962 ± 0.034 [ +6.5% .. +8.4%]
user_time 9.306 ± 0.054 9.926 ± 0.082 [ +6.2% .. +7.1%]
wall_time 8.225 ± 0.068 8.947 ± 0.128 [ +8.0% .. +9.6%]
samples 15 22
With that flag enabled, the regression mostly disappears, as hoped:
base diff difference (95% CI)
sys_time 1.839 ± 0.062 1.866 ± 0.068 [ -0.9% .. +3.8%]
user_time 9.452 ± 0.068 9.490 ± 0.067 [ -0.1% .. +0.9%]
wall_time 8.383 ± 0.127 8.452 ± 0.114 [ -0.1% .. +1.8%]
samples 17 21
**Unnamed internal app**
Without `-femit-dwarf-unwind`, this is the perf hit:
base diff difference (95% CI)
sys_time 1.372 ± 0.029 1.317 ± 0.024 [ -4.6% .. -3.5%]
user_time 2.835 ± 0.028 2.980 ± 0.027 [ +4.8% .. +5.4%]
wall_time 3.205 ± 0.079 3.383 ± 0.066 [ +4.9% .. +6.2%]
samples 102 83
With `-femit-dwarf-unwind`, the perf hit almost disappears:
base diff difference (95% CI)
sys_time 1.274 ± 0.026 1.270 ± 0.025 [ -0.9% .. +0.3%]
user_time 2.812 ± 0.023 2.822 ± 0.035 [ +0.1% .. +0.7%]
wall_time 3.166 ± 0.047 3.174 ± 0.059 [ -0.2% .. +0.7%]
samples 95 97
Just for fun, I measured the impact of `-femit-dwarf-unwind` on ld64
(`base` has the extra DWARF unwind info in the input object files,
`diff` doesn't):
base diff difference (95% CI)
sys_time 1.128 ± 0.010 1.124 ± 0.023 [ -1.3% .. +0.6%]
user_time 7.176 ± 0.030 7.106 ± 0.094 [ -1.5% .. -0.4%]
wall_time 7.874 ± 0.041 7.795 ± 0.121 [ -1.7% .. -0.3%]
samples 16 25
And for LLD:
base diff difference (95% CI)
sys_time 1.315 ± 0.019 1.280 ± 0.019 [ -3.2% .. -2.0%]
user_time 2.980 ± 0.022 2.822 ± 0.016 [ -5.5% .. -5.0%]
wall_time 3.369 ± 0.038 3.175 ± 0.033 [ -6.2% .. -5.3%]
samples 47 47
So parsing the extra EH frames is a lot more expensive for us than for
ld64. But given that we are quite a lot faster than ld64 to begin with,
I guess this isn't entirely unexpected...
Reviewed By: #lld-macho, oontvoo
Differential Revision: https://reviews.llvm.org/D129540
clang-format's documentation documented the more general clang-format-diff.py
script. Add documentation for the less general but arguably easier-to-use
git integration as well.
Differential Revision: https://reviews.llvm.org/D129563
Currently the PPC64R2SaveStub thunk will produce Power 10 code by default.
This produced an issue when linking older code that made use of the st_other=1
bit but was never meant to be linked or run on Power 10.
This patch makes it so that only the R_PPC64_REL24_NOTOC relocation can produce
Power 10 code.
Reviewed By: MaskRay
Differential Revision: https://reviews.llvm.org/D129580
CStringChecker is using getByteLength to get the length of a string
literal. For targets where a "char" is 8-bits, getByteLength() and
getLength() will be equal for a C string, but for targets where a "char"
is 16-bits getByteLength() returns the size in octets.
This is verified in our downstream target, but we have no way to add a
test case for this case since there is no target supporting 16-bit
"char" upstream. Since this cannot have a test case, I'm asserted this
change is "correct by construction", and visually inspected to be
correct by way of the following example where this was found.
The case that shows this fails using a target with 16-bit chars is here.
getByteLength() for the string literal returns 4, which fails when
checked against "char x[4]". With the change, the string literal is
evaluated to a size of 2 which is a correct number of "char"'s for a
16-bit target.
```
void strcpy_no_overflow_2(char *y) {
char x[4];
strcpy(x, "12"); // with getByteLength(), returns 4 using 16-bit chars
}
```
This change exposed that embedded nulls within the string are not
handled. This is documented as a FIXME for a future fix.
```
void strcpy_no_overflow_3(char *y) {
char x[3];
strcpy(x, "12\0");
}
```
Reviewed By: martong
Differential Revision: https://reviews.llvm.org/D129269
For the moment, we're pretty conservative here. My motivating case is the vscale one (as that is idiomatic for scalable vectorized loops on RISCV). There are two obvious approaches to fixing this, and I tried to add reasonable coverage for both even though I'll likely only fix one.
Fortran permits forward references to derived types in contexts that don't
require knowledge of the derived type definition for semantic analysis,
such as in the declaration of a pointer or allocatable variable or component.
But when the forward-referenced derived type is used later for a component
reference, it is possible for the DerivedTypeSpec in he base variable or component
declaration to still have a null scope pointer even if the type has been defined,
since DerivedTypeSpec and TypeSpec objects are created in scopes of use
rather than in scopes of definition. The fix is to call
DerivedTypeSpec::Instantiate() in the name resolution of each component
name so that the scope gets filled in if it is still null.
Differential Revision: https://reviews.llvm.org/D129681
Add `pcm-info` to the `target module dump` subcommands.
This dump command shows information about clang .pcm files. This command
effectively runs `clang -module-file-info` and produces identical output.
The .pcm file format is tightly coupled to the clang version. The clang
embedded in lldb is not guaranteed to match the version of the clang executable
available on the local system.
There have been times when I've needed to view the details about a .pcm file
produced by lldb's embedded clang, but because the clang executable was a
slightly different version, the `-module-file-info` invocation failed. With
this command, users can inspect .pcm files generated by lldb too.
Differential Revision: https://reviews.llvm.org/D129456
When MOLD= is an array and there is no SIZE= in a call to TRANSFER(),
the size of an element of the MOLD= is used as the denominator in a
division to establish the extent of the vector result. When the
total storage size of the SOURCE= is known to be zero, the result is
empty and no division is needed.
To avoid a division by zero at runtime, we need to check for a zero-sized
MOLD= element type when the storage size of SOURCE= is nonzero and there
is no SIZE=. Further, in the compilation-time rewriting of calls to
SHAPE(TRANSFER(...)) and SIZE(TRANSFER(...)) for constant folding and
simplification purposes, we can't replace the call with an arithmetic
element count expression when the storage size of SOURCE= is not known
to be zero and the element size of MOLD= is not known to be nonzero at
compilation time.
These changes mostly affect tests using a MOLD= argument that is an
assumed-length character.
Differential Revision: https://reviews.llvm.org/D129680
Most of the infrastructure for DECIMAL='COMMA' mode was in place
in the I/O runtime support library, but I dropped the ball for
list-directed character input, which has its own detection of
input separators. Finish the job.
Differential Revision: https://reviews.llvm.org/D129679
Most modern Fortran programs declare procedure pointers with a
procedure-declaration-stmt, but it's also possible to declare one
with a type-declaration-stmt with a POINTER attribute. In this
case, e.g. "real, external, pointer :: p => null()" the initializer
is required to be a null-init. The parse tree traversal in name
resolution would visit the null-init if the symbol were an object
pointer only, leading to a crash in the case of a procedure pointer.
That explanation of the bug is longer than the fix. In short,
ensure that a null-init in an entity-decl is visited for both
species of pointers.
Differential Revision: https://reviews.llvm.org/D129676
It's more natural to use uint8_t * (std::byte needs C++17 and llvm has
too much uint8_t *) and most callers use uint8_t * instead of char *.
The functions are recently moved into `llvm::compression::zlib::`, so
downstream projects need to make adaption anyway.
When the I/O runtime is truncating an external file due to an
implied ENDFILE or explicit ENDFILE, ensure that the unit's frame
buffer for the file discards any data that have become obsolete.
This bug caused trouble with ACCESS='STREAM' I/O using POS= on
a WRITE, but it may have not been limited to that scenario.
Differential Revision: https://reviews.llvm.org/D129673
Jay Foad