The script used to fix up LLDB's header for use in the macOS framework
contained 2 bugs that this commit addreses:
1. The output contents were appended to the output file multiple times
instead of only being written once.
2. The script was not considering LLDB includes that were *not* from the
SB API.
This commit addresses and fixes both of these bugs and updates the
corresponding test to match.
Currently, when VPSlotTracker is initialized with a VPlan, its
assignName method calls printAsOperand on each underlying instruction.
Each such call recomputes slot numbers for the entire function, leading
to O(N × M) complexity, where M is the number of instructions in the
loop and N is the number of instructions in the function.
This results in slow debug output for large loops. For example, printing
costs of all instructions becomes O(M² × N), which is especially painful
when enabling verbose dumps.
This patch improves debugging performance by caching slot numbers using
ModuleSlotTracker. It avoids redundant recomputation and makes debug
output significantly faster.
In GFX10, a number of VOP3 instructions should allow opsel, including
V_MAX_I16, V_MAX_U16, V_MIN_I16, V_MIN_U16, V_MUL_LO_U16, V_LSHLREV_B16,
V_LSHRREV_B16, and V_ASHRREV_I16.
This patch adds a release note that explains the current status of
OpenACC in Clang. Currently we cannot actually make an executable
because the OpenACC dialect of MLIR doesn't support any amount of
lowering to LLVM-IR, so the usefulness of OpenACC is entirely for
front-end related uses, such as tooling or semantic checking.
Summary:
There's a new one called the AIE (AI Engine). We could handle this, but
since we don't use it currently I'm just making it future-proof. Adding
the AIE check would require checking the HSA version which isn't
worthwhile just yet.
ArrayRef has a constructor that accepts std::nullopt. This
constructor dates back to the days when we still had llvm::Optional.
Since the use of std::nullopt outside the context of std::optional is
kind of abuse and not intuitive to new comers, I would like to move
away from the constructor and eventually remove it.
This patch replaces std::nullopt with {}. There are a couple of
places where std::nullopt is replaced with TypeRange() to accommodate
perfect forwarding.
Convert all binary calls of min/max to extremum operations, so that
extremums generated by the compiler compare equal, and user min/max
calls also compare equal.
Fixes#133646
Originally opened as #144162 but I accidentally pushed a merge in such a
way that a bunch of code owners got added to the review. This is just
rebasing the original work on main and fixing the failing tests.
More closely match improveShuffleKindFromMask's shuffle ordering by
trying to match a SK_InsertSubvector shuffles patterns before SK_Select
- both can match many of the same patterns, but its much easier to
recognise when a SK_InsertSubvector can be converted to SK_Select than
vice-versa.
Another step towards #145335 - which I'm hoping will allow us to
generalise improveShuffleKindFromMask and remove getInstructionCost's
shuffle matching entirely.
Allow users to set the minimum absolute count for indirect call promotion. This is primarily meant to be control indirect call promotion for synthetic vp metadata introduced in #141164 for use by MemProf.
Refactor the code for NewTextSegmentAddress to correctly point at the
true start of the segment when PHDR table is placed at the beginning. We
used to offset NewTextSegmentAddress by PHDR table plus cache line
alignment.
NFC for proper binaries. Some YAML binaries from our tests will diverge
due to bad segment address/offset alignment.
In preparation to unify the codegen of [l]lrint and [l]lround, making
the latter go through custom-lowering for vector-codegen, add some tests
showing the current fixed-vector-unrolled codegen of [l]lround.
This is the culmination of a series of changes described in [1].
Although somewhat large by line count, it is almost entirely mechanical,
creating a new library in DebugInfo/DWARF/LowLevel. This new library has
very minimal dependencies, allowing it to be used from more places than
the normal DebugInfo/DWARF library--in particular from MC.
I am happy to put it in another location, or to structure it differently
if that makes sense. Some have suggested in BinaryFormat, but it is not
a great fit there. But if that makes more sense to the reviewers, I can
do that.
Another possibility would be to use pass-through headers to allow
clients who don't care to depend only on DebugInfo/DWARF. This would be
a much less invasive change, and perhaps easier for clients. But also a
system that hides details.
Either way, I'm open.
1.
https://discourse.llvm.org/t/rfc-debuginfo-dwarf-refactor-into-to-lower-and-higher-level-libraries/86665/2
This PR also sets `alwayslink=True` for `//libc/test:LibcUnitTest`. This
ensures that the `main` function provided always gets linked into a test
target. While not strictly necessary, it makes it so tests like
45d8759cbe/libc/test/include/signbit_test.c
will give a duplicate symbol error if they incorrectly depend on
`//libc/test:LibcUnitTest`.
This PR is missing tests for generated header includes since the current
Bazel setup lacks generated headers or a mechanism to run hermetic
tests. CMake version of the header include tests:
a2ce564720/libc/test/include/CMakeLists.txt (L515)
See issue https://github.com/llvm/llvm-project/issues/134780
Review #145600 and #145770 crossed, which caused compute-copy and
combined-copy tests to fail because of an insufficiently written 'check'
line for a cir.func, which didn't account for the linkage spec being
added. This patch adds that to fix the build.
The test `test/ClangScanDeps/P1689.cppm` was previously not supported on
Windows due to the differences in path separators between Windows and
Linux. This normalizes the paths, allowing this test to run on Windows
as well.
This is motivated by the review feedback for #145221, where the same
change was suggested for a test derived from this one.
Sections which are not allowed to carry data are marked as virtual. Only
complication when writing out the text is that it must be written in
chunks of 32k-1 bytes, which is done by having a wrapper stream writing
those records.
Data of BSS sections is not written, since the contents is known to be
zero. Instead, the fill byte value is used.
…ess" (#144810)
This relands commit e0933ab5ae. The
original commit was causing the test TestCreateAfterAttach.py to fail on
ARM Ubuntu bots. It's possible that this could've been happening because
the test for wait-attach progress reporting is waiting on a process
named "a.out" which could be too generic as multiple other tests (when
run in parallel on the bots) could also be using processes named
"a.out". This commit changes the wait-attach progress report test to
wait on a unique process name.
Original PR description:
This commit adds a progress report when wait-attaching to a process as
well as a test for this.
Original PR link: https://github.com/llvm/llvm-project/pull/144768
This patch is very similar to #134819 in nature. Before this patch, we
were dereferencing pointers to objects which were never constructed. Now
we always assume that nodes store `pair<const KeyT, ValueT>` for
unordered_maps instead, as they actually do.
Add SDPatternMatch matcher and unit test coverage for `ISD::LOAD`
opcode.
This only matches the loaded value i.e. ResNo 0 and not the output
chain.
e.g.
```
m_Load(m_Value(), m_Value(), m_Value())
```
The first value is the input chain, the second is the base pointer, and
the last value is the offset.
In DXC, there is an option to enable all KHR extension. This is added by
passing the KHR option to the SPIR-V backend, which will enable all of
the appropriate extensions.
Part of https://github.com/llvm/llvm-project/issues/137650.
While doing lowering, I discovered that the restriction onthe allowed
modifiers for 'copyout' didn't make sense! After discussion on the
OpenACC standards mailing list I discovered that this was a copy/paste
error during standardization that they intend to fix, and really meant
for copyout to allow alwaysout instead of alwaysin.
When implementing, I blindly followed the standard :)
This patch corrects the implementation to do what was meant.
The
[protocol](https://microsoft.github.io/debug-adapter-protocol//specification.html#Types_Source)
expects that `sourceReference` be less than `(2^31)-1`, but we currently
represent memory address as source reference, this can be truncated
either when sending through json or by the client. Instead, generate new
source references based on the memory address.
Make the `ResolveSource` function return an optional source.
### Summary
Currently, if the setting `interpreter.save-transcript` is enabled,
whenever we call "statistics dump", it'll default to reporting a huge
list of transcripts which can be a bit noisy. This is because the
current check `GetIncludeTranscript` returns `!GetSummaryOnly()` by
default if no specific transcript-setting option is given in the
statistics dump command (ie. `statistics dump --transcripts=false` or
`statistics dump --transcripts=true`). Then when
`interpreter.save-transcript` is enabled, this saves a list of
transcripts, and the transcript list ends up getting logged by default.
These changes default the option to log transcripts in the `statistics
dump` command to "false". This can still be enabled via the
`--transcripts` option if users want to see a transcript. Since
`interpreter.save-transcript` is false by default, the main delta is
that if `interpreter.save-transcript` is true and summary mode is false,
we now disable saving the transcript.
This also adds a warning to 'statistics dump --transcript=true' when
interpreter.save-transcript is disabled, which should help users
understand
why transcript data is empty.
### Testing
#### Manual testing
Tested with `settings set interpreter.save-transcript true` enabled at
startup on a toy hello-world program:
```
(lldb) settings set interpreter.save-transcript true
(lldb) target create "/home/qxy11/hello-world/a.out"
Current executable set to '/home/qxy11/hello-world/a.out' (x86_64).
(lldb) statistics dump
{
/* no transcript */
}
(lldb) statistics dump --transcript=true
{
"transcript": [
{
"command": "statistics dump",
"commandArguments": "",
"commandName": "statistics dump",
"durationInSeconds": 0.0019650000000000002,
"error": "",
"output": "{...
},
{
"command": "statistics dump --transcript=true",
"commandArguments": "--transcript=true",
"commandName": "statistics dump",
"timestampInEpochSeconds": 1750720021
}
]
}
```
Without `settings set interpreter.save-transcript true`:
```
(lldb) target create "/home/qxy11/hello-world/a.out"
Current executable set to '/home/qxy11/hello-world/a.out' (x86_64).
(lldb) statistics dump
{
/* no transcript */
}
(lldb) statistics dump --transcript=true
{
/* no transcript */
}
warning: transcript requested but none was saved. Enable with 'settings set interpreter.save-transcript true'
```
#### Unit tests
Changed unit tests to account for new expected default behavior to
`false`, and added a couple new tests around expected behavior with
`--transcript=true`.
```
lldb-dotest -p TestStats ~/llvm-sand/external/llvm-project/lldb/test/API/commands/statistics/basic/
```
