merge main into amd-staging #1018
Merged
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
SPIR-V does not have a production-grade linker so it is often necessary to stay in LLVM-IR as long as possible and only convert to SPIR-V at the very end. As such, it is common that we want to create a BC library that is linked into a user program (motivating example here is the OpenMP device RTL). We only convert to SPIR-V at the very end when we have a fully linked program. Other targets can achieve a similar goal by using LTO to get a linked-BC file, but the SPIR-V linker does not support LTO, so we have no way to do it with `clang`. We can do it with `llvm-link` directly, but my understanding is that it is not intended to be directly called in production workflows and does not fit well into code that supports multiple targets. Extend the existing behavior for `HIP` that allows `-emit-llvm` without `-c`. --------- Signed-off-by: Nick Sarnie <nick.sarnie@intel.com>
Add a lit test that exercises RocmInstallationDetector when clang is installed under /opt/rocm/lib/llvm/bin, using a synthetic sysroot so it does not depend on any real /opt/rocm installation. The test copies the built clang binary into the test tree so the driver observes the expected InstalledDir and real path, following the precedent of other driver tests that copy %clang into a temporary layout. This is for testing llvm#138928
- AA is not actually used.
Static analysis flagged a case where we had defined the destructor but not copy ctor or copy assignment, which violates the Rule of Three. I applied this to all the cases I found. With the exception to destructors with empty bodies. These should be defaulted but that should be done seperately.
…e `hash` specializations (llvm#174120) In this release: - [x] `file_clock` - [x] `steady_clock` - [x] `system_clock` - [x] `time_point` - [x] some `hash` specializations Any other existing clocks are already annotated. Annotated some `hash` specializations for already annotated classes. `[[nodiscard]]` should be applied to functions where discarding the return value is most likely a correctness issue. - https://libcxx.llvm.org/CodingGuidelines.html - https://wg21.link/time.point - https://wg21.link/time.clock.file - https://wg21.link/time.clock.steady - https://wg21.link/time.clock.system Towards llvm#172124
…llvm#173345) Clang provides lambdas as an extension in C++03 now, so we can use them to simplify our code a bit.
This macro provides relatively little value. Using the standard FTM directly removes one level of indicretion, making it easier to understand what is actually happening.
…lvm#171625) Clang provides these macros since LLVM 19, so we can assume they are available.
The scale is a signed quantity. This avoids an assertion failure with github.com/llvm/pull/171456.
…lvm#174243) add test case to test lib call are used for the ___strcpy milicode.
…llvm#173509) Add lowering support for the OpenMP `device` clause on the `target` directive in Flang. The device expression is propagated through MLIR OpenMP and passed to the host-side `__tgt_target_kernel` call.
…lvm#174522) Darwin supports running tests on targets other than the host machine e.g. simulators. In such configurations, tests are run via a wrapper script (e.g. iossim_run.py). This refactors the lit test config to use a dedicated option to distinguish test suites that run on the host from those that run in any other configuration. This will allow the test suites to distinguish a local osx configuration from a "remote" one. rdar://167591463
llvm#174518) Casting a pointer to a different address space is a nice trick to prevent an access from being instrumented by sanitizers such as ASan. However, this trick is currently broken, as the test demonstrates. A minimal C PoC of the issue is: ``` void access(void *p) { auto tmp = (unsigned long __attribute__((address_space(1))) volatile *)p; *tmp = 0; } ``` Under fsanitize=address, the access does not get instrumented, but under fsanitize=thread, it does. (Note: Before this patch, this test will actually hit a CallInst assertion if you have them turned on, because the TSan runtime functions are defined to take pointer arguments, and non-default address space pointers are not compatible). rdar://166743781
Based on the name of the function, I think it's safe to assume `sys::fs::createTemporaryFile()` is not to be used for formal compiler outputs and should only be used for compiler-internal outputs (caches, debugging output, etc.). Thus, it should be fine to disable the IO sandbox and allow bypassing `vfs::OutputBackend` here.
…s link (llvm#174616) The old link is still dead/protected. Use the new link.
PR llvm#142970 Added for Windows targets to emit minimal codeview metadata even when debug info is disabled. This crashes the backend for UEFI x86_64-uefi triple as llvm.dbg.cu is expected unconditionally there. Handling it correctly in AsmPrinter and adding a regression test.
…:' string (llvm#174572) When the crashing thread was submitted to a libdispatch queue, the `Triggered by Thread:` string looks like: ``` Triggered by Thread: 0, Dispatch Queue: com.apple.main-thread ``` Trying to run the `crashlog` command on such crashlog file fails with: ``` (lldb) crashlog -i '/tmp/lldb.crash' Traceback (most recent call last): File "Git/llvm-worktrees/main/builds/release/lib/python3.13/site-packages/lldb/macosx/crashlog.py", line 1444, in __call__ SymbolicateCrashLogs(debugger, shlex.split(command), result, True) ~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "Git/llvm-worktrees/main/builds/release/lib/python3.13/site-packages/lldb/macosx/crashlog.py", line 1888, in SymbolicateCrashLogs load_crashlog_in_scripted_process( ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~^ debugger, crashlog_path, options, result ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ ) ^ File "Git/llvm-worktrees/main/builds/release/lib/python3.13/site-packages/lldb/macosx/crashlog.py", line 1495, in load_crashlog_in_scripted_process crashlog = CrashLogParser.create(debugger, crashlog_path, options).parse() File "Git/llvm-worktrees/main/builds/release/lib/python3.13/site-packages/lldb/macosx/crashlog.py", line 1043, in parse self.parsers[self.parse_mode](line) ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~^^^^^^ File "Git/llvm-worktrees/main/builds/release/lib/python3.13/site-packages/lldb/macosx/crashlog.py", line 1121, in parse_normal self.crashlog.crashed_thread_idx = int(line[20:].strip().split()[0]) ~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ ValueError: invalid literal for int() with base 10: '0,' ``` This patch strips the possibly trailing comma in when parsing this string. Tested locally that this fixes the issue. LLDB test suggestions welcome
In W64, size_t and ul are different sizes. So, std::max cannot correctly deduce the operator type, resulting an error at compile time. The fix was originally added in llvm#173802, but as suggested it's best to keep unrelated changes separate.
…ve pre 20 support (llvm#173157) It makes sense to decouple traits from the things they describe, since that's the purpose of traits. Furthermore, this will make it possible to reuse these traits from other places in the library (e.g. <semaphore>) outside of the synchronization library. Pre-C++20 support is removed as a drive-by since the synchronization library does not support pre-C++20 anymore. This makes it possible to simplify a few constructs by using concepts. Requested in llvm#172214 (comment)
Clang's rewriter currently violates the IO sandbox due to a call to `llvm::writeToOutput()`. Since the "blessed" `llvm::vfs::OutputBackend` isn't easily available in that particular spot, this PR instead disables the sandbox and leaves a FIXME behind.
…llvm#173238) This patch adds the bare-minimum tests for the LLDB `llvm::ArrayRef` formatters. Since this would be the first LLVM data fromatter test for LLDB, I wanted to keep the test itself minimal and mainly set up/agree on the infrastructure (i.e., CMake machinery, etc.). The setup mimicks that of GDB. The main differences are: 1. the GDB formatter tests put all the test cases in one monolithic test file, whereas I'm planning on having one test-file per LLVM container. 2. the GDB formatter tests currently only get run if LLVM was built with debug-info. Not sure we have any build-bots out there running this configuration. In this patch we rely on just the debug-info from the LLVM headers (see latest commit). That way the tests get more consistently run (at least our public macOS buildbot runs the cross-project-tests in Release+Asserts).
Move all lldb-dap vscode extension files to the extension directory. Follow up from llvm#162635 (comment)
There was a typo introduced in llvm#168637, this PR fixes it. Thanks to Nikolas for pointing it out.
…vm#153338) This tests that we don't ICE if a struct hasn't been forward declared. Originally fixed in 160fb11 Link: ClangBuiltLinux/linux#2114
This PR adds more value types for future use.
This patch adds libstdcpp string view summary provider.
Follow-up to llvm#171204 and 1f331e4 to only rely on isAddressSCEVForCost in legacy isAddressSCEVForCost, completely aligning the decisions of VPlan and legacy cost model.
…eturn type is valid (llvm#174603) Fixes: llvm#173943
…174373) This avoids the GOT slot being generated which is undesirable when using static linking. A better solution would be to reorganize the code to avoid the use of weak symbols altogether. Fixes llvm#173409
…clarators with identifier-list parameters (llvm#174017) Fixes llvm#173826 --- This patch resolves an issue where Clang could crash while parsing the `enable_if` attribute on C function declarators with identifier-list parameters. In C, identifier-list function declarators use parameter identifiers rather than full parameter declarations, as specified by the C spec. Since `enable_if` is parsed early to participate in redeclaration checks https://github.com/llvm/llvm-project/blob/0f3a9f658a7f459b9aa701ec42a999b645653862/clang/lib/Parse/ParseDecl.cpp#L674-L675 the parser could encounter such identifiers before their parameter declarations exist and incorrectly assume they were already formed, leading to an assertion failure. This change makes the early `enable_if` parsing path handle parameters that have not yet been declared, preventing the crash while preserving existing behavior for well-formed function prototypes.
Trimming the content of the progress event is the responsibility of the consumer, not the producer. For example, when using the statusline, there's plenty of space to show longer expressions. rdar://166879951
)" (llvm#174614) This change adds a new intrinsic for AMDGPU that implements a wave shuffle, allowing arbitrary swizzling between lanes using an index. In the initial version of this commit, there was an issue in one of the tests added that returned a signal, causing testing to fail when combined with another recent change to 'not'. For context on the initial commit see llvm#167372 --------- Signed-off-by: Domenic Nutile <domenic.nutile@gmail.com> Co-authored-by: Jay Foad <jay.foad@gmail.com>
The files in the lldb-dap local .gitignore are only relevant to the extension. Moving it into the extension dir also makes it easier to mirror to the lldb-vscode repo.
) Support Type with PointerRepresentation in CatchParam Issue llvm#154992
Keeping them conservatively simple and coarse grained.
…oadcast` (llvm#174206) Use the same repacking logic as for shuffle/swizzle.
Collaborator
dpalermo
approved these changes
Jan 6, 2026
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
20 participants
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
No description provided.