merge main into amd-staging #621
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
This patch fixes most of the ASan tests that were failing on Darwin when running under the internal shell. There are still a couple left that are more interesting cases that I'll do in a follow up patch. The tests that still need to be done: ``` TestCases/Darwin/duplicate_os_log_reports.cpp TestCases/Darwin/dyld_insert_libraries_reexec.cpp TestCases/Darwin/interface_symbols_darwin.cpp ``` Reviewers: thetruestblue, fhahn, vitalybuka, DanBlackwell, ndrewh Reviewed By: DanBlackwell Pull Request: llvm#168545
Only the fortran source files in flang/test/Lower/PowerPC and some in flang/test/Lower have been modified. The other files in the directory will be cleaned up in subsequent commits
…ific address spaces (llvm#167770) For some backends, e.g., BPF, it is desirable to only sanitize memory belonging to specific address spaces. More specifically, it is sometimes desirable to only apply address sanitization for arena memory belonging to address space 1. However, AddressSanitizer currently does not support selectively sanitizing address spaces. Add a new option to select which address spaces to apply AddressSanitizer to. No functional change for existing targets (namely AMD GPU) that hardcode which address spaces to sanitize
In this PR we are proposing to change LLDB codebase so that LLDB is able to print values of integer registers that have more than 64-bits (even if the number of bits is not equal to 128). --------- Co-authored-by: Matej Košík <matej.kosik@codasip.com> Co-authored-by: Jonas Devlieghere <jonas@devlieghere.com>
…to non-vectors (llvm#168081) Updates the demanded elements before recursing through copies in case the type of the source register changes from a non-vector register to a vector register. Fixes llvm#167842.
) * original change llvm#162730 * with windows fix llvm#164843 * remove timeout that was pointed out in the comment above * Remove test that starts and listens on a socket to avoid timeout issues
…vm#168165) and make (llvm#165264) Truely recover Executor::getDefaultExecutor. The previous change missed std::unique_ptr, which is needed in a normal program exit, since only with that ThreadPoolExecutor destructor will be called in a normal program exit, where it ensures the executor has been stopped and waits for worker threads to finish. The wait is important as it prevents intermittent crashes on Windows when the process is doing a full exit.
In line with a std proposal to introduce std::clmul, and in preparation to introduce a clmul intrinsic, implement carry-less multiply primitives for APIntOps, clmul[rh]. Ref: https://isocpp.org/files/papers/P3642R3.html
Identified with modernize-loop-convert.
https://alive2.llvm.org/ce/z/YGT5SN https://alive2.llvm.org/ce/z/PVDxCw https://alive2.llvm.org/ce/z/8buR2N This is tricky because with positive numbers, we only go up, so we can in fact always hit the signed_max boundary. This is important because the intrinsic we use has the behavior of going the OTHER way, aka clamp to INT_MIN if it goes in that direction. And the range checking we do only works for positive numbers. Because of this issue, we can only do this for constants as well.
When building just the runtimes (eg a patch only touches compiler-rt), we do not actually run any normal check targets. This ends up causing an empty ninja invocation, which builds more targets than necessary. Gate the ninja build for normal check-* targets under an if statement to fix this.
The AArch64 backend converts trees formed by conjunctions/disjunctions of comparisons into sequences of `CCMP` instructions. The implementation before this change checks whether a sub-tree must be processed first. If not, it processes the operations in the order they occur in the DAG. This may not be optimal if there is a corresponding `SUB` node for one of the comparisons. In this case, we should process this comparison first because we can then use the same instruction for the `SUB` node and the comparison. To achieve this, this commit comprises the following changes: - Extend `canEmitConjunction` with a new output parameter `PreferFirst`, which reports to the caller whether the sub-tree should preferably be processed first. - Set `PreferFirst` to `true` if we can find a corresponding `SUB` node in the DAG. - If we can process a sub-tree with `PreferFirst = true` first (i.e., we do not violate any `MustBeFirst` constraint by doing so), we swap the sub-trees. - The already existing code for performing the common subexpression elimination takes care to use only a single instruction for the comparison and the `SUB` node if possible. Closes llvm#149685.
…37170) In general, "Flat instructions look at the per-workitem address and determine for each work item if the target memory address is in global, private or scratch memory." (RDNA2 ISA) That means that FLAT instructions need to be considered for VMEM hazards even without "specific segment". Also, LDS DMA should be considered for LDS hazard detection. See also llvm#137148
…m#168549) Move `GetInnermostExecPart` and `IsStrictlyStructuredBlock` from Semantics/openmp-utils.* to Parser/openmp-utils.*. These two only depend on the AST contents and properties.
This reverts commit b3d6264. This broke the workflow because the sync-labels flag was set to a zero-length string to work around an issue. The underlying issue has been fixed and the value is now required to be a boolean. We can just drop the value because we want the default behavior anyways. This should be the last remaining breaking change from v5 that we need to migrate.
This reverts commit bd8c941. This still broke things and evidently needs more testing on a fork before relanding. https://github.com/llvm/llvm-project/actions/runs/19475911086
…t could be converted to vector loads plus shuffles (llvm#168571) This is turning up in some legalisation code when shuffling vectors bitcast from illegal loads. Ideally we'd handle more complex shuffles, but reverse is a start.
Reverts llvm#164356 The bots are broken.
They were still running because the conditional was not correct. This patch fixes that so they do not interefere with the results of the job.
These are more idiomatic in bash.
This allows SDNodes to be validated against their expected type profiles and reduces the number of changes required to add a new node. Some nodes fail validation, those are enumerated in `ARMSelectionDAGInfo::verifyTargetNode()`. Some of the bugs are easy to fix, but probably they should be fixed separately, this patch is already big. Part of llvm#119709. Pull Request: llvm#168212
…vm#168559) Fixes the assertion in llvm#168523 This patch lifts the small, odd-sized integer to 8 bits, ensuring that the following lowering code behaves correctly.
This patch adds `LLDBLog::InstrumentationRuntime` as a log channel to provide an appropriate channel for instrumentation runtime plugins as previously one did not exist. A small use of the channel is added to illustrate its use. The logging added is not intended to be comprehensive. This is primarily motivated by an `-fbounds-safety` instrumentation plugin (swiftlang#11835). rdar://164920875
…y-Name Lookups (llvm#168143) This PR adds some test coverage for `StableDirs` during by-name lookups.
…vm#167904) Always rely on local scopes to enforce the lifetime of these helper objects and by extension where the "closing" of various C++ code constructs happens.
…to interleave3-8. (llvm#168473)
…lvm#167745) If vector-unaligned-mem support is not enabled, we should not generate loads/stores that are not aligned to their element size. We already do this for non-VP vector loads/stores. This code has been in our downstream for about a year and a half after finding the vectorizer generating misaligned loads/stores. I don't think that is unique to our downstream. Doing this for masked vp.load/store requires widening the mask as well which is harder to do. NOTE: Because we have to scale the VL, this will introduce additional vsetvli and the VL optimizer will not be effective at optimizing any arithmetic that is consumed by the store.
This test was failing on chromium builds with error: ``` /Volumes/Work/s/w/ir/x/w/llvm_build/bin/llc -o - /Volumes/Work/s/w/ir/x/w/llvm-llvm-project/llvm/test/DebugInfo/AArch64/instr-ref-target-hooks-sp-clobber.mir -run-pass=livedebugvalues | /Volumes/Work/s/w/ir/x/w/llvm_build/bin/FileCheck /Volumes/Work/s/w/ir/x/w/llvm-llvm-project/llvm/test/DebugInfo/AArch64/instr-ref-target-hooks-sp-clobber.mir # RUN: at line 8 + /Volumes/Work/s/w/ir/x/w/llvm_build/bin/llc -o - /Volumes/Work/s/w/ir/x/w/llvm-llvm-project/llvm/test/DebugInfo/AArch64/instr-ref-target-hooks-sp-clobber.mir -run-pass=livedebugvalues + /Volumes/Work/s/w/ir/x/w/llvm_build/bin/FileCheck /Volumes/Work/s/w/ir/x/w/llvm-llvm-project/llvm/test/DebugInfo/AArch64/instr-ref-target-hooks-sp-clobber.mir error: YAML:121:3: unknown key 'stackSizePPR' stackSizePPR: 0 ^~~~~~~~~~~~ FileCheck error: '<stdin>' is empty. FileCheck command line: /Volumes/Work/s/w/ir/x/w/llvm_build/bin/FileCheck /Volumes/Work/s/w/ir/x/w/llvm-llvm-project/llvm/test/DebugInfo/AArch64/instr-ref-target-hooks-sp-clobber.mir ``` This is an attempt to reland the failing test
LLVM IR verifier checks for `extraData` in debug info metadata. This is a follow-up PR based on discussions in llvm#165023
Currently the test cfi-multiple-location.mir is marked as XFAIL. This causes failures on some build bots because the test unexpectedly passes. Mark this test as UNSUPPORTED for now. Later I plan to merge an MR which fixes an issue in CFIInstrInserter and this test will be enabled.
Fixes unsigned int underflows in `MFMASmallGemmSingleWaveOpt::applyIGLPStrategy`.
…lvm#165375) Identity masks can be treated as free when scalable vectorization is possible making the check agnostic of the vectorization policy fixed/scalable, This allows for aggressive vector combines for identity shuffle masks.
This should really check if the libcall is known supported. For now mips doesn't configure its RuntimeLibcallsInfo correctly, and does not have any of the mips16 calls in it. For now there isn't a way to add them without triggering conflicting cases in tablegen, so keep parsing the raw name as it was before.
We previously added support for marking GlobalOp operations as constant, but the handling to actually do so was left mostly unimplemented. This fills in the missing pieces.
…vm#168584) For vectors, CTLZ, CTTZ, CTPOP all operate on individual elements. The lowering should be based on the element width. I noticed this by inspection. No tests in tree are currently affected, but I thought it would be good to fix so someone doesn't have to debug it in the future.
…#168586) The compiler doesn't emit a diagnostics when the signature of a function defined in a namespace gets out-of-sync with its declaration. Let's use qualified names for function definitions instead of nesting them in a namespace so that mismatches are diagnosed by the compiler rather than by the (less understandable) linker.
…lvm#168187) A function prologue can begin with a pre-index STR instruction for a floating-point register. To construct an unwind plan from assembly correctly, the instruction emulator must support such instructions.
Follow up on a cse OpType-mismatch crash reported due to ef023ca (Reland [VPlan] Expand WidenInt inductions with nuw/nsw), setting the OpType correctly when returning from getFlagsFromIndDesc.
…lvm#167703) In this PR I'm changing the way we provide the missing functions like strnlen() on z/OS from the separate header file to a wrapper around the system headers that declare these functions. This will be less intrusive. --------- Co-authored-by: Zibi Sarbinowski <zibi@ca.ibm.com>
1. Handle transformed awaitables for `AllowedCallees`, which generate temporaries and weren't being handled by llvm#167778. 1. Fix name mismatches in `storeOptions`.
This change drops the use of the "Layout" type and instead uses explicit padding throughout the compiler to represent types in HLSL buffers. There are a few parts to this, though it's difficult to split them up as they're very interdependent: 1. Refactor HLSLBufferLayoutBuilder to allow us to calculate the padding of arbitrary types. 2. Teach Clang CodeGen to use HLSL specific paths for cbuffers when generating aggregate copies, array accesses, and structure accesses. 3. Simplify DXILCBufferAccesses such that it directly replaces accesses with dx.resource.getpointer rather than recalculating the layout. 4. Basic infrastructure for SPIR-V handling, but the implementation itself will need work in follow ups. Fixes several issues, including llvm#138996, llvm#144573, and llvm#156084. Resolves llvm#147352.
The pstl top-level directory was removed, but we forgot to remove pstl from the list of valid subdirectories.
FCmp instructions have both a predicate and fast-math flags. Introduce a new FCmp kind, that combines both to model this correctly in the current system. This should be NFC modulo VPlan printing which now includes the correct fast-math flags.
…from its lower 32-bit (llvm#168458) On some targets, a packed f32 instruction can only read 32 bits from a scalar operand (SGPR or literal) and replicates the bits to both channels. In this case, we should not fold an immediate value if it can't be replicated from its lower 32-bit. Fixes SWDEV-567139.
…lvm#162362) This patch provides definitions for `pkey_*` functions for linux x86_64. `pkey_alloc`, `pkey_free`, and `pkey_mprotect` are simple syscall wrappers. `pkey_set` and `pkey_get` modify architecture-specific registers. The logic for these live in architecture specific directories: * `libc/src/sys/mman/linux/x86_64/pkey_common.h` has a real implementation * `libc/src/sys/mman/linux/generic/pkey_common.h` contains stubs that just return `ENOSYS`.
This allows SDNodes to be validated against their expected type profiles and reduces the number of changes required to add a new node. The verification functionality detected a few issues, two of them were fixed (missing `SDNPMemOperand` property on `TCGEN05_MMA` nodes and extra glue operand/result on `CallPrototype`), the one remaining is with `ProxyReg` node, see `NVPTXSelectionDAGInfo::verifyTargetNode()`. Part of llvm#119709. Pull Request: llvm#168367
…lvm#168514) Introduces the Task and TaskDispatcher interfaces (TaskDispatcher.h), ThreadPoolTaskDispatcher implementation (ThreadPoolTaskDispatch.h), and updates Session to include a TaskDispatcher instance that can be used to run tasks. TaskDispatcher's introduction is motivated by the need to handle calls to JIT'd code initiated from the controller process: Incoming calls will be wrapped in Tasks and dispatched. Session shutdown will wait on TaskDispatcher shutdown, ensuring that all Tasks are run or destroyed prior to the Session being destroyed.
Add support for marking global variables with common linkage.
…m#168072) In this case, the value is a constant, not an addend to a relocation. So the "Relocation Not In Range" error must not be triggered. Regression from PR llvm#112877 Fixes llvm#132322
Collaborator
dpalermo
approved these changes
Nov 19, 2025
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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.