[all-commits] [llvm/llvm-project] ae374f: [lldb] Remove the unused Process::ReadMemoryInChun...
Alexey Bataev via All-commits
all-commits at lists.llvm.org
Fri Sep 18 16:02:43 PDT 2026
Branch: refs/heads/users/alexey-bataev/spr/slpemit-loop-carried-horizontal-reductions-as-loop-vector-accumulator
Home: https://github.com/llvm/llvm-project
Commit: ae374f89dafb3ad6f534064688cb023d5bfafca1
https://github.com/llvm/llvm-project/commit/ae374f89dafb3ad6f534064688cb023d5bfafca1
Author: Yao Qi <yao_qi at apple.com>
Date: 2026-09-18 (Fri, 18 Sep 2026)
Changed paths:
M lldb/include/lldb/Target/Process.h
M lldb/source/Target/Process.cpp
M lldb/test/API/functionalities/process_save_core_minidump/size_mismatch/TestProcessSaveCoreMinidumpSizeMismatch.py
Log Message:
-----------
[lldb] Remove the unused Process::ReadMemoryInChunks (#224656)
`Process::ReadMemoryInChunks` and its `ReadMemoryChunkCallback` typedef
have no callers left in the tree, no unit tests, and no SB API exposure.
This commit removes it.
The method was added in commit f869d6efeec8 (#129307) for the minidump
writer, which was its only caller. Then, commit 7f04a40a220e (#212641)
replaced that call with a direct loop over `ReadMemoryFromInferior` in
`MinidumpFileBuilder::ReadWriteMemoryInChunks`, leaving this method
dead.
This commit also drop the reference to this method from the description
of
`TestProcessSaveCoreMinidumpSizeMismatch.py`. The test is unchanged.
Commit: 216acfbf7b4599b41a56a9023b024678b6d202af
https://github.com/llvm/llvm-project/commit/216acfbf7b4599b41a56a9023b024678b6d202af
Author: alexey-gusarov <alexey.gusarov at gmail.com>
Date: 2026-09-18 (Fri, 18 Sep 2026)
Changed paths:
M lldb/include/lldb/Target/ThreadPlan.h
M lldb/source/Target/Thread.cpp
A lldb/test/API/functionalities/thread_plan/step_over_breakpoint_site/Makefile
A lldb/test/API/functionalities/thread_plan/step_over_breakpoint_site/TestStepOverBreakpointSite.py
A lldb/test/API/functionalities/thread_plan/step_over_breakpoint_site/main.c
A lldb/test/API/functionalities/thread_plan/step_over_breakpoint_site/run_to_address_plan.py
Log Message:
-----------
[lldb] Don't consult the plan stack for a stop an auto-continue will discard (#215522)
Fixes part (2) of #215189. The sibling PR for part (1) is #215521. The
two are
independent, but are best landed together.
When a completed thread plan auto-continues, `Thread::ShouldStop()`
still
consults subsequent plans and then discards their stop decisions. Those
calls
can change the state of the plans being consulted. In the failing case,
`ThreadPlanRunToAddress` loses the breakpoint it was running to, so the
process
runs past the requested address.
This does not require a scripted plan:
```python
# thread stopped on an enabled breakpoint site
thread.RunToAddress(next_instruction_address) # before: process runs away
```
`ThreadPlan.h` already says that an auto-continuing plan overrides the
`ShouldStop` of subsequently processed plans. Tighten that contract so
their
`ShouldStop` is not consulted at all, and make `Thread::ShouldStop()`
follow it
by ending the plan walk after popping the auto-continuing plan.
The subsequent plans therefore keep their state, and an unexecuted
`BreakpointSite` at the current PC is hit normally on the following
resume.
### Testing
Adds
`lldb/test/API/functionalities/thread_plan/step_over_breakpoint_site/`
with coverage for:
* `RunToAddress` from an enabled breakpoint site;
* the same plan queued by a scripted thread plan;
* the stepped-over breakpoint being hit again on a later pass;
* `RunToAddress` with the PC already off the breakpoint site;
* a target two instructions away;
* a user breakpoint at the target address.
The first three fail before the change on both Linux and Windows. The
remaining
three pass before and after the change and cover the surrounding
behavior.
Whole `check-lldb` passes on both Linux x86_64 and Windows x86_64 with 0
failed
and 0 unresolved tests.
macOS is not covered because no host was available.
---
Tool-use disclosure, per the LLVM AI tool policy: this change was
prepared with
the assistance of Claude Code (`Assisted-by:` trailer on the commit).
Everything
was read and reviewed before posting.
rdar://184560595
Commit: 5f23b5e1475236f51e614ea971e22bf800fbe7c8
https://github.com/llvm/llvm-project/commit/5f23b5e1475236f51e614ea971e22bf800fbe7c8
Author: Augusto Noronha <anoronha at apple.com>
Date: 2026-09-18 (Fri, 18 Sep 2026)
Changed paths:
M lldb/source/Plugins/Instruction/ARM64/EmulateInstructionARM64.cpp
M lldb/source/Plugins/Instruction/ARM64/EmulateInstructionARM64.h
M lldb/source/Plugins/UnwindAssembly/InstEmulation/UnwindAssemblyInstEmulation.cpp
M lldb/source/Plugins/UnwindAssembly/InstEmulation/UnwindAssemblyInstEmulation.h
A lldb/test/API/functionalities/unwind/outlined-prologue/Makefile
A lldb/test/API/functionalities/unwind/outlined-prologue/TestOutlinedPrologueUnwind.py
A lldb/test/API/functionalities/unwind/outlined-prologue/main.c
M lldb/unittests/UnwindAssembly/ARM64/CMakeLists.txt
M lldb/unittests/UnwindAssembly/ARM64/TestArm64InstEmulation.cpp
Log Message:
-----------
[lldb] Follow serial outlined functions when scanning for an unwind plan (#222800)
For code-size-constrained builds the compiler's instruction outliner
finds
identical instruction sequences across functions, moves them into a
single
OUTLINED_FUNCTION_<n>, and has every user call it. Instructions from a
function's prologue, body or epilogue may all be outlined.
lldb's instruction-emulation UnwindPlan creator scans the instructions
of a
function and, following the ABI's calling convention, assumes a call
leaves the
stack and registers in a known state. That assumption does not hold for
outlined functions: when prologue and epilogue instructions are outlined
out of
the function body, the UnwindPlan misses those register spills and
restores,
which leads to inaccurate variable tracking in the caller's frames.
This patch looks for call instructions in the function body, and when
the
callee is an OUTLINED_FUNCTION_<n> it evaluates the callee's
instructions too,
folding the UnwindPlan-relevant state changes -- register saves,
restores, CFA
changes -- into the plan of the function being scanned.
Only straight-line helpers are followed: a single basic block whose sole
control flow is the terminator returning to the caller. Anything else
has more
than one path through the callee is skipped entirely. In practice that
means
mid-function outlined code, which rarely changes the unwind state.
Outlined
prologues, the common and important case, are covered.
Assisted-by: Claude
rdar://187149822
Commit: fedbd3bdbe650fce8fe92a020698cc2db55a4ed7
https://github.com/llvm/llvm-project/commit/fedbd3bdbe650fce8fe92a020698cc2db55a4ed7
Author: Amr Hesham <amr96 at programmer.net>
Date: 2026-09-18 (Fri, 18 Sep 2026)
Changed paths:
M clang/lib/CIR/CodeGen/CIRGenModule.cpp
Log Message:
-----------
[CIR][NFC] Fix compilation error CIRGenModule (#224694)
Fix deduced conflicting types for parameter when calling std::max with
unsigned long and unsigned long long, can be fixed by
`std::max(f.getAlignment().value_or(1), 2ull)` but in this line, the
alignment is always less than 2 because of the parent if condition, so
max is always 2
Commit: b91795ba4d79b109cc8abe3d1d8255379b217754
https://github.com/llvm/llvm-project/commit/b91795ba4d79b109cc8abe3d1d8255379b217754
Author: Kewen Meng <Kewen.Meng at amd.com>
Date: 2026-09-18 (Fri, 18 Sep 2026)
Changed paths:
M libsycl/src/detail/queue_impl.cpp
M libsycl/src/usm_functions.cpp
M libsycl/unittests/handler/test_helpers.hpp
M libsycl/unittests/mock/helpers.cpp
M libsycl/unittests/mock/helpers.hpp
M libsycl/unittests/mock/mock.cpp
M libsycl/unittests/queue/memcpy.cpp
M libsycl/unittests/usm/alloc.cpp
M llvm/tools/llvm-gpu-loader/llvm-gpu-loader.cpp
M llvm/tools/llvm-gpu-loader/llvm-gpu-loader.h
M offload/languages/kernel/src/LanguageRuntime.cpp
M offload/liboffload/API/Memory.td
M offload/liboffload/src/OffloadImpl.cpp
M offload/plugins-nextgen/amdgpu/dynamic_hsa/hsa_ext_amd.h
M offload/plugins-nextgen/amdgpu/src/rtl.cpp
M offload/plugins-nextgen/common/include/PluginInterface.h
M offload/plugins-nextgen/common/src/PluginInterface.cpp
M offload/plugins-nextgen/cuda/dynamic_cuda/cuda.cpp
M offload/plugins-nextgen/cuda/dynamic_cuda/cuda.h
M offload/plugins-nextgen/cuda/src/rtl.cpp
M offload/plugins-nextgen/host/src/rtl.cpp
M offload/plugins-nextgen/level_zero/include/L0Memory.h
M offload/plugins-nextgen/level_zero/include/L0Plugin.h
M offload/plugins-nextgen/level_zero/src/L0Context.cpp
M offload/plugins-nextgen/level_zero/src/L0Device.cpp
M offload/plugins-nextgen/level_zero/src/L0Memory.cpp
M offload/plugins-nextgen/level_zero/src/L0Plugin.cpp
M offload/unittests/Conformance/include/mathtest/DeviceContext.hpp
M offload/unittests/Conformance/include/mathtest/DeviceResources.hpp
M offload/unittests/Conformance/lib/DeviceContext.cpp
M offload/unittests/Conformance/lib/DeviceResources.cpp
M offload/unittests/OffloadAPI/event/olGetEventElapsedTime.cpp
M offload/unittests/OffloadAPI/kernel/olLaunchKernel.cpp
M offload/unittests/OffloadAPI/memory/olGetMemInfo.cpp
M offload/unittests/OffloadAPI/memory/olGetMemInfoSize.cpp
M offload/unittests/OffloadAPI/memory/olMemAlloc.cpp
M offload/unittests/OffloadAPI/memory/olMemAllocAligned.cpp
M offload/unittests/OffloadAPI/memory/olMemFill.cpp
M offload/unittests/OffloadAPI/memory/olMemFree.cpp
M offload/unittests/OffloadAPI/memory/olMemPrefetch.cpp
M offload/unittests/OffloadAPI/memory/olMemcpy.cpp
M offload/unittests/OffloadAPI/queue/olLaunchHostFunction.cpp
M offload/unittests/OffloadAPI/queue/olWaitEvents.cpp
Log Message:
-----------
Revert "[offload] add context to olMemAlloc*" (#224703)
Revert: https://github.com/llvm/llvm-project/pull/222677 and
https://github.com/llvm/llvm-project/pull/224622
Bot has been down: https://lab.llvm.org/buildbot/#/builders/234
errors:
```
Failed Tests (5):
libomptarget :: amdgpu-amd-amdhsa :: tools/omp-kernel-replay/record-replay-advanced.cpp
libomptarget :: amdgpu-amd-amdhsa :: tools/omp-kernel-replay/record-replay-basic.cpp
libomptarget :: amdgpu-amd-amdhsa :: tools/omp-kernel-replay/record-replay-diff-teams-threads.cpp
libomptarget :: amdgpu-amd-amdhsa :: tools/omp-kernel-replay/record-replay-diff-threads.cpp
libomptarget :: amdgpu-amd-amdhsa :: tools/omp-kernel-replay/record-replay-ir-bitcode.cpp
```
After revert:
```
Total Discovered Tests: 3643
Skipped : 84 (2.31%)
Unsupported : 529 (14.52%)
Passed : 3026 (83.06%)
Expectedly Failed: 4 (0.11%)
```
Commit: 4cc7b28c66c88d69e7a778b0a33449c9b57640ed
https://github.com/llvm/llvm-project/commit/4cc7b28c66c88d69e7a778b0a33449c9b57640ed
Author: Mircea Trofin <mtrofin at google.com>
Date: 2026-09-18 (Fri, 18 Sep 2026)
Changed paths:
M llvm/utils/profcheck-xfail.txt
Log Message:
-----------
Exclude tailcallelim test (#224708)
keeping the bot green. This failure was introduced in PR #221694, but
tailcallelimination isn't yet clean wrt profcheck
Commit: b2b38d4b335293edca03764dd2bcfa314a880933
https://github.com/llvm/llvm-project/commit/b2b38d4b335293edca03764dd2bcfa314a880933
Author: geoffreygaren <ggaren at apple.com>
Date: 2026-09-18 (Fri, 18 Sep 2026)
Changed paths:
M clang/lib/StaticAnalyzer/Checkers/WebKit/PtrTypesSemantics.cpp
Log Message:
-----------
[WebKit Checkers][NFC] Extract hasPublicMethodInHierarchy from isSmartPtrCompatible (#224697)
So we can reuse it in borrow checking.
hasPublicMethodInHierarchy identifies RefCounted via 'ref' and 'deref'
public
member functions.
Borrow checking will identify CanBorrow via a 'crashIfBorrowed' public
member
function.
Assisted-by: Claude
Commit: f29371ad92f1b47279a3b46f7d336f3e48fa4538
https://github.com/llvm/llvm-project/commit/f29371ad92f1b47279a3b46f7d336f3e48fa4538
Author: Keshav Vinayak Jha <31160700+keshavvinayak01 at users.noreply.github.com>
Date: 2026-09-18 (Fri, 18 Sep 2026)
Changed paths:
M llvm/lib/Target/AMDGPU/AMDGPUArgumentUsageInfo.cpp
M llvm/lib/Target/AMDGPU/AMDGPURegBankLegalizeHelper.cpp
M llvm/lib/Target/AMDGPU/SIISelLowering.cpp
M llvm/test/CodeGen/AMDGPU/GlobalISel/irtranslator-call-abi-attribute-hints.ll
M llvm/test/CodeGen/AMDGPU/GlobalISel/irtranslator-call-implicit-args.ll
M llvm/test/CodeGen/AMDGPU/GlobalISel/irtranslator-call-return-values.ll
M llvm/test/CodeGen/AMDGPU/GlobalISel/irtranslator-call-sret.ll
M llvm/test/CodeGen/AMDGPU/GlobalISel/irtranslator-call.ll
M llvm/test/CodeGen/AMDGPU/GlobalISel/irtranslator-indirect-call.ll
M llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-amdgcn.workitem.id.mir
M llvm/test/CodeGen/AMDGPU/GlobalISel/regbankselect-add.mir
M llvm/test/CodeGen/AMDGPU/GlobalISel/regbankselect-amdgcn.ballot.i64.mir
M llvm/test/CodeGen/AMDGPU/GlobalISel/regbankselect-and-s1.mir
M llvm/test/CodeGen/AMDGPU/GlobalISel/regbankselect-anyext.mir
M llvm/test/CodeGen/AMDGPU/GlobalISel/regbankselect-brcond.mir
M llvm/test/CodeGen/AMDGPU/GlobalISel/regbankselect-freeze.mir
M llvm/test/CodeGen/AMDGPU/GlobalISel/regbankselect-mui-regbanklegalize.mir
M llvm/test/CodeGen/AMDGPU/GlobalISel/regbankselect-mui.mir
M llvm/test/CodeGen/AMDGPU/GlobalISel/regbankselect-sadde.mir
M llvm/test/CodeGen/AMDGPU/GlobalISel/regbankselect-select.mir
M llvm/test/CodeGen/AMDGPU/GlobalISel/regbankselect-sext.mir
M llvm/test/CodeGen/AMDGPU/GlobalISel/regbankselect-ssube.mir
M llvm/test/CodeGen/AMDGPU/GlobalISel/regbankselect-sub.mir
M llvm/test/CodeGen/AMDGPU/GlobalISel/regbankselect-trunc.mir
M llvm/test/CodeGen/AMDGPU/GlobalISel/regbankselect-uadde.mir
M llvm/test/CodeGen/AMDGPU/GlobalISel/regbankselect-usube.mir
M llvm/test/CodeGen/AMDGPU/GlobalISel/regbankselect-zext.mir
M llvm/test/CodeGen/AMDGPU/build_vector.gfx11plus.ll
M llvm/test/CodeGen/AMDGPU/carryout-selection.ll
M llvm/test/CodeGen/AMDGPU/fcopysign.gfx11plus.ll
M llvm/test/CodeGen/AMDGPU/fma-mix.gfx11plus.ll
M llvm/test/CodeGen/AMDGPU/fma.f16.gfx11plus.ll
M llvm/test/CodeGen/AMDGPU/indirect-addressing-term.ll
M llvm/test/CodeGen/AMDGPU/permlane16_opsel.ll
M llvm/test/CodeGen/AMDGPU/scalar_to_vector.gfx11plus.ll
M llvm/test/CodeGen/AMDGPU/vgpr-liverange-ir.ll
Log Message:
-----------
[AMDGPU] Use integer LLTs for work-item IDs (#224519)
Type work-item ID preloaded values and special entry-function live-ins
as i32, matching the intrinsic result and the existing GlobalISel and
SelectionDAG lowering. This prevents an s32 live-in type from reaching
the divergent G_AND and lets VgprToVccCopy preserve its source LLT
directly.
The original failure was a GlobalISel cannot-select abort on a 32-bit
VGPR G_AND produced from `llvm.amdgcn.workitem.id.x`. This follows up
#220987 by fixing the type at its producer instead of coercing it in
register-bank legalization.
A lot of tests need to be migrated because of this, owing to the
Extended LLTs change in https://github.com/llvm/llvm-project/pull/207419
Assisted-by: Codex
Signed-off-by: Keshav Vinayak Jha <keshavvinayakjha at gmail.com>
Co-authored-by: GPT-5 <noreply at openai.com>
Commit: 551af79c903169d0ee3ac35222a5af175424205d
https://github.com/llvm/llvm-project/commit/551af79c903169d0ee3ac35222a5af175424205d
Author: Michael G. Kazakov <mike.kazakov at gmail.com>
Date: 2026-09-18 (Fri, 18 Sep 2026)
Changed paths:
M libcxx/docs/ReleaseNotes/24.rst
M libcxx/docs/Status/Cxx17Papers.csv
M libcxx/include/CMakeLists.txt
M libcxx/include/__memory/pstl.h
M libcxx/include/__pstl/backend_fwd.h
M libcxx/include/__pstl/backends/default.h
M libcxx/include/__pstl/backends/libdispatch.h
M libcxx/include/__pstl/backends/serial.h
M libcxx/include/__pstl/backends/std_thread.h
M libcxx/include/__pstl/cpu_algos/for_each.h
A libcxx/include/__pstl/cpu_algos/uninitialized_algorithms.h
M libcxx/include/module.modulemap.in
M libcxx/test/libcxx/algorithms/pstl.iterator-requirements.verify.cpp
M libcxx/test/std/algorithms/pstl.exception_handling.pass.cpp
A libcxx/test/std/utilities/memory/specialized.algorithms/uninitialized.copy/pstl.uninitialized_copy.pass.cpp
A libcxx/test/std/utilities/memory/specialized.algorithms/uninitialized.copy/pstl.uninitialized_copy_n.pass.cpp
A libcxx/test/std/utilities/memory/specialized.algorithms/uninitialized.move/pstl.uninitialized_move.pass.cpp
A libcxx/test/std/utilities/memory/specialized.algorithms/uninitialized.move/pstl.uninitialized_move_n.pass.cpp
Log Message:
-----------
[libc++][pstl] Implementation of parallel std::uninitialized_(copy|move)(_n) based on parallel __for_each (#223229)
This PR adds an implementation of parallel versions of
`std::uninitialized_copy`, `std::uninitialized_copy_n`,
`std::uninitialized_move` and `std::uninitialized_move_n`.
Implementations are mostly one-liners under the hood, but require
dealing with iterator ranges, hence `std::uninitialized_copy` and
`std::uninitialized_move` are implemented in `cpu_algos`.
`std::uninitialized_copy_n` and `std::uninitialized_move_n` are
implemented in the `default` backend.
Fixes #103640
Fixes #134592
Fixes #134593
Part of #99938
Commit: 25e00ff449d6c2a2d21a0b109f576be88847ce1f
https://github.com/llvm/llvm-project/commit/25e00ff449d6c2a2d21a0b109f576be88847ce1f
Author: DeanSturtevant1 <dsturtevant at google.com>
Date: 2026-09-18 (Fri, 18 Sep 2026)
Changed paths:
M clang/include/clang/AST/DeclBase.h
M clang/lib/AST/DeclBase.cpp
Log Message:
-----------
[Clang] Move DeclContext::getEnclosingFunction/castEnclosingFunction out of line (#224711)
In DeclBase.h, clang::FunctionDecl is only forward-declared. Commit
c9074cfd0d95 defined getEnclosingFunction() and castEnclosingFunction()
inline in DeclBase.h, which instantiates dyn_cast<FunctionDecl> and
cast<FunctionDecl> on an incomplete type when DeclBase.h is compiled
standalone (e.g. with -fmodules / header units).
Move the non-const definitions to DeclBase.cpp where FunctionDecl is
complete, and delegate the const overloads via const_cast, matching
getOuterLexicalRecordContext() and
getEnclosingNonExpansionStatementContext().
Change prepared by Jetski and reviewed by me.
Commit: a6d5d3f8bacba58aaba3a9eb65b5b91dcce5b505
https://github.com/llvm/llvm-project/commit/a6d5d3f8bacba58aaba3a9eb65b5b91dcce5b505
Author: Florian Hahn <flo at fhahn.com>
Date: 2026-09-18 (Fri, 18 Sep 2026)
Changed paths:
M llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp
M llvm/test/Transforms/LoopVectorize/AArch64/force-target-instruction-cost.ll
M llvm/test/Transforms/LoopVectorize/AArch64/interleave-with-gaps.ll
M llvm/test/Transforms/LoopVectorize/AArch64/sve-gather-scatter.ll
M llvm/test/Transforms/LoopVectorize/AArch64/sve-interleaved-access-low-vf.ll
M llvm/test/Transforms/LoopVectorize/AArch64/sve-tail-folding-option.ll
M llvm/test/Transforms/LoopVectorize/AArch64/sve-widen-phi.ll
M llvm/test/Transforms/LoopVectorize/AArch64/transform-narrow-interleave-vscale-x-UF-step.ll
M llvm/test/Transforms/LoopVectorize/RISCV/strided-accesses.ll
Log Message:
-----------
[VPlan] Remove dead per-lane loop from VPScalarIVStepsRecipe::execute. (#224399)
replicateByVF already materializes every lane of a VPScalarIVStepsRecipe
as its own single-scalar clone, folding the lane offset into the clone's
start index (addLaneToStartIndex).
During execute, each VPScalarIVStepsRecipe has exactly its single lane
used.
Replace it with the lane-0 body and an assert. Note that this removes a
few redundant add start, 0.
PR: https://github.com/llvm/llvm-project/pull/224399
Commit: 660dbc9ef14ec76259b82a6d8118d4850d3b7d3d
https://github.com/llvm/llvm-project/commit/660dbc9ef14ec76259b82a6d8118d4850d3b7d3d
Author: Ben Shi <2283975856 at qq.com>
Date: 2026-09-18 (Fri, 18 Sep 2026)
Changed paths:
M llvm/lib/Target/AVR/AVRFrameLowering.cpp
M llvm/lib/Target/AVR/AVRISelLowering.cpp
M llvm/lib/Target/AVR/AVRISelLowering.h
A llvm/test/CodeGen/AVR/frameaddress.ll
A llvm/test/CodeGen/AVR/returnaddress.ll
Log Message:
-----------
[AVR] Implement selection of 'ISD::RETURNADDR' and 'ISD::FRAMEADDR' (#222909)
Commit: 1680496f872c941249f555d8a5e555f17dd4fb76
https://github.com/llvm/llvm-project/commit/1680496f872c941249f555d8a5e555f17dd4fb76
Author: Chinmay Deshpande <chdeshpa at amd.com>
Date: 2026-09-18 (Fri, 18 Sep 2026)
Changed paths:
M llvm/lib/Target/AMDGPU/AMDGPU.td
M llvm/lib/Target/AMDGPU/AMDGPUFeatures.td
A llvm/test/TableGen/AMDGPUTargetDefGenericClassification.td
M llvm/utils/TableGen/Basic/AMDGPUTargetDefEmitter.cpp
Log Message:
-----------
[AMDGPU] Classify generic-target features (#223179)
Classify features used by generic targets but absent from some covered
GPUs as AMDGPUGenericAnyFeature. This covers the gfx11 workarounds,
gfx12.5 register fixup and WMMA timing, and LDS size and bank count.
Derive the classified features directly from the SubtargetFeature
subclass and preserve their existing predicates. Keep RequiresCOV6
outside this policy because it is specific to generic targets and
absent from their covered GPUs.
Test the real target definitions, missing support for classified
features, and the existing all-member requirement for instruction
capabilities.
Commit: ab76cce30b9d16bed79fc4658562e53dd129aefe
https://github.com/llvm/llvm-project/commit/ab76cce30b9d16bed79fc4658562e53dd129aefe
Author: Dmitry Sidorov <Dmitry.Sidorov at amd.com>
Date: 2026-09-18 (Fri, 18 Sep 2026)
Changed paths:
A llvm/test/CodeGen/AMDGPU/mad-mix-fptrunc-rounding-bf16.ll
A llvm/test/CodeGen/AMDGPU/mad-mix-fptrunc-rounding.ll
Log Message:
-----------
[NFC][AMDGPU] Add tests for fptrunc folded into V_MAD/FMA_MIX (#219723)
V_MAD/FMA_MIX{LO,HI} compute the multiply or multiply-add in f32, round
that to f32, and then convert the f32 result to the 16-bit destination.
That is the same pair of roundings the IR performs, so the instructions
are a bit-exact match for an fptrunc of an f32 multiply or multiply-add
and need no contract flag.
The tests pin that across gfx803, gfx900, gfx906, gfx90a and gfx1100
with +real-true16 for f16, and gfx1250 in both true16 modes for bf16.
Where a shape appears both unflagged and with contract, the two must
select the same instructions.
Assisted-By: Claude Code Opus 5
Commit: 841dbaab188ecfa756d4c946f33b58758d34d339
https://github.com/llvm/llvm-project/commit/841dbaab188ecfa756d4c946f33b58758d34d339
Author: Ian Anderson <iana at apple.com>
Date: 2026-09-18 (Fri, 18 Sep 2026)
Changed paths:
M clang/lib/Driver/ToolChains/Darwin.cpp
A clang/test/Driver/darwin-static-lib-universal.c
M clang/test/Driver/darwin-static-lib.c
Log Message:
-----------
[clang][driver][darwin] Add more libtool arguments used by SwiftBuild (#224428)
SwiftBuild passes -l to libtool, but I didn't catch it the first time
around because it's not in the xcspec. It is in the unit test though. Go
through all of the linker arguments and inputs and add the missing ones
supported by libtool.
Assisted-by: Claude Code
rdar://187418335
Commit: eefe7721b114a71c94c966b1792ebd3b2fb8544e
https://github.com/llvm/llvm-project/commit/eefe7721b114a71c94c966b1792ebd3b2fb8544e
Author: Florian Hahn <flo at fhahn.com>
Date: 2026-09-18 (Fri, 18 Sep 2026)
Changed paths:
M llvm/lib/Transforms/Scalar/ConstraintElimination.cpp
M llvm/test/Transforms/ConstraintElimination/decompose-signed-preconditions.ll
M llvm/test/Transforms/ConstraintElimination/induction-exit-condition-or-and.ll
M llvm/test/Transforms/ConstraintElimination/materialize-flags.ll
M llvm/test/Transforms/ConstraintElimination/reproducer-remarks.ll
Log Message:
-----------
[ConstraintElim] Add signed no-wrap rule for sub to isKnownNoWrap. (#224621)
Port the existing logic from usub_sat simplification to isKnownNoWrap:
Op0 - Op1 does not wrap signed if Op1 is non-negative and Op0 s>= Op1.
https://alive2.llvm.org/ce/z/HsUyo_
This is in preparation to update simplifications to also use
isKnownNoWrap.
Improves results in a few cases
https://github.com/dtcxzyw/llvm-opt-benchmark-nightly/pull/1358.
Compile-time impact is in the noise:
https://llvm-compile-time-tracker.com/compare.php?from=0401d338f425ab4d85738d48e9f5432685bf70fb&to=5f9d1cae5fe4222b9ddaf59605e12420df09808e&stat=instructions:u
PR: https://github.com/llvm/llvm-project/pull/224621
Commit: 4a2306bb4be443d8ca89844d1625d0de01a924e8
https://github.com/llvm/llvm-project/commit/4a2306bb4be443d8ca89844d1625d0de01a924e8
Author: Min-Yih Hsu <min.hsu at sifive.com>
Date: 2026-09-18 (Fri, 18 Sep 2026)
Changed paths:
M llvm/lib/Target/RISCV/RISCVISelLowering.cpp
M llvm/test/CodeGen/RISCV/rvv/vector-interleave-fixed.ll
M llvm/test/CodeGen/RISCV/rvv/vector-interleave.ll
Log Message:
-----------
[RISCV] Lower VECTOR_INTERLEAVE 4 and 8 with Zvzip via DAGCombiner (#223861)
This is the second take of #222432, and to some extent #222750
The full context of why moving to this implementation can be found here:
https://github.com/llvm/llvm-project/pull/222432#discussion_r3992576835
Basically, #222432 has the downside of generating redundant
vslideup/down, as no DAG combine would be run in between different
operation lowering. Although #222750 tried to address that problem, the
fact that it has to call RAUW in the middle of lowering might result in
unexpected result.
This iteration of approach use a DAG combine to break
vector_interleave4/8 into vector_interleave2 early in the ISel, so that
any redundant insert/extract_subvector or concat_vectors should be
cleaned up automatically. Though some redundant pairs of vslideup/down
still show up but they are relatively easy to clean up, which will be
done in the following patch.
Commit: 34eeb2320ff2b991ddb3e3511bbe01ba14478d93
https://github.com/llvm/llvm-project/commit/34eeb2320ff2b991ddb3e3511bbe01ba14478d93
Author: Kazu Hirata <kazu at google.com>
Date: 2026-09-18 (Fri, 18 Sep 2026)
Changed paths:
M llvm/include/llvm/CodeGen/Passes.h
M llvm/lib/CodeGen/BranchFolding.cpp
M llvm/lib/CodeGen/LibcallLoweringInfo.cpp
M llvm/lib/CodeGen/ShadowStackGCLowering.cpp
Log Message:
-----------
[CodeGen] Remove unused pass creation functions (NFC) (#224646)
createBranchFolder:
Introduced on July 9, 2026 in commit
f1daa4cd5377d61c6b12d59188048259d646e395 without any callers.
createShadowStackGCLoweringPass:
The last caller was removed on July 7, 2021 in commit
0fdb25cd954c7d0d082379374092929e06184206.
createLibcallLoweringInfoWrapper:
Introduced on January 16, 2026 in commit
539412914a391d4dd2b74a959afcf1ac49f9e47a without any callers.
Assisted-by: Antigravity
Commit: e8cd5ce3183244e9f8d95e3211d106b056c89c75
https://github.com/llvm/llvm-project/commit/e8cd5ce3183244e9f8d95e3211d106b056c89c75
Author: Adrian Prantl <aprantl at apple.com>
Date: 2026-09-18 (Fri, 18 Sep 2026)
Changed paths:
M lldb/test/API/commands/frame/var-dil/expr/Comparison/TestFrameVarDILExprComparison.py
Log Message:
-----------
[lldb] Skip TestFrameVarDILExprComparison with DWARF < 4 (#224755)
The test relies on scoped enum types, which need DW_AT_enum_class and
thus DWARF 4 or newer.
Commit: fd823ddc6faa0f34c799d21449b54bc4492decc2
https://github.com/llvm/llvm-project/commit/fd823ddc6faa0f34c799d21449b54bc4492decc2
Author: YongKang Zhu <yongzhu at fb.com>
Date: 2026-09-18 (Fri, 18 Sep 2026)
Changed paths:
M bolt/include/bolt/Rewrite/RewriteInstance.h
M bolt/lib/Rewrite/RewriteInstance.cpp
M bolt/test/AArch64/use-old-text-zero-padding.c
Log Message:
-----------
[BOLT] Zero all stale bytes in reused file region (#222515)
`--use-old-text` reuses an input file region for new content. We have
only zeroed the padding that followed each newly written section in
#202375, which may leave stale bytes from input in front of the new
content. This could happen if the region is not on the new alignment
boundary, or when `--hot-functions-at-end` is used.
Replace the section name based forward scan with an interval coverage
over the region - collect the file extents of sections that contain
output bytes and then zero all the remaining holes.
Assisted-by: opus
Commit: 53ff33af0943139d83938c582c8c154665a40716
https://github.com/llvm/llvm-project/commit/53ff33af0943139d83938c582c8c154665a40716
Author: Adrian Prantl <aprantl at apple.com>
Date: 2026-09-18 (Fri, 18 Sep 2026)
Changed paths:
M lldb/test/API/android/platform/TestDefaultCacheLineSize.py
M lldb/test/API/commands/process/reverse-continue/TestReverseContinueCommandNotSupported.py
M lldb/test/API/commands/watchpoints/multiple_hits/TestMultipleHits.py
M lldb/test/API/functionalities/postmortem/minidump/TestMiniDump.py
M lldb/test/API/functionalities/process_save_core/TestProcessSaveCore.py
M lldb/test/API/functionalities/process_save_core_minidump/partial_read/TestProcessSaveCoreMinidumpPartialRead.py
M lldb/test/API/functionalities/process_save_core_minidump/size_mismatch/TestProcessSaveCoreMinidumpSizeMismatch.py
M lldb/test/API/functionalities/reverse-execution/TestReverseExecutionNotSupported.py
M lldb/test/API/functionalities/scripted_frame_provider/circular_dependency/TestFrameProviderCircularDependency.py
M lldb/test/API/functionalities/tail_call_frames/cross_object/TestCrossObjectTailCalls.py
M lldb/test/API/functionalities/thread/crash_during_step/TestCrashDuringStep.py
M lldb/test/API/lang/c/local_variables/TestLocalVariables.py
M lldb/test/API/lang/cpp/class_static/TestStaticVariables.py
M lldb/test/API/lang/objc/foundation/TestObjCMethods.py
M lldb/test/API/lang/objc/objc-ivar-offsets/TestObjCIvarOffsets.py
M lldb/test/API/lang/objc/objc-static-method/TestObjCStaticMethod.py
M lldb/test/API/lang/objc/objc-struct-argument/TestObjCStructArgument.py
M lldb/test/API/lang/objc/objc-struct-return/TestObjCStructReturn.py
M lldb/test/API/lang/objc/objc-super/TestObjCSuper.py
M lldb/test/API/python_api/findvalue_duplist/TestSBFrameFindValue.py
M lldb/test/API/python_api/lldbutil/frame/TestFrameUtils.py
M lldb/test/API/python_api/objc_type/TestObjCType.py
M lldb/test/API/python_api/process/TestProcessAPI.py
M lldb/test/API/python_api/thread/TestThreadAPI.py
M lldb/test/API/python_api/value/empty_class/TestValueAPIEmptyClass.py
M lldb/test/API/python_api/watchpoint/TestSetWatchpoint.py
M lldb/test/API/python_api/watchpoint/TestWatchpointIgnoreCount.py
M lldb/test/API/python_api/watchpoint/TestWatchpointIter.py
M lldb/test/API/python_api/watchpoint/TestWatchpointRead.py
Log Message:
-----------
[LLDB] Convert straightforward test launches to lldbutil (NFC) (#224404)
Replace the CreateTarget + breakpoint + LaunchSimple + stop-assert
boilerplate with the matching lldbutil.run_to_*_breakpoint helper at 35
call sites where the translation needs no extra arguments.
Besides being shorter, the helpers check that the breakpoint resolved to
at least one location and that the process stopped at that specific
breakpoint, which most of the original code did not do.
It also improves und unifies the error reporting when the launch fails.
Assisted-by: Claude
Commit: 666edf2186f29fae04fd38cd0f93608cd0ec06e0
https://github.com/llvm/llvm-project/commit/666edf2186f29fae04fd38cd0f93608cd0ec06e0
Author: Adrian Prantl <aprantl at apple.com>
Date: 2026-09-18 (Fri, 18 Sep 2026)
Changed paths:
M lldb/test/API/functionalities/conditional_break/TestConditionalBreak.py
M lldb/test/API/functionalities/return-value/TestReturnValue.py
M lldb/test/API/lang/cpp/dynamic-value/TestCppValueCast.py
M lldb/test/API/lang/objc/foundation/TestObjectDescriptionAPI.py
M lldb/test/API/linux/add-symbols/TestTargetSymbolsAddCommand.py
M lldb/test/API/linux/mix-dwo-and-regular-objects/TestMixedDwarfBinary.py
M lldb/test/API/macosx/safe-to-func-call/TestSafeFuncCalls.py
M lldb/test/API/python_api/class_members/TestSBTypeClassMembers.py
M lldb/test/API/python_api/frame/TestFrames.py
M lldb/test/API/python_api/frame/inlines/TestInlinedFrame.py
M lldb/test/API/python_api/symbol-context/TestSymbolContext.py
M lldb/test/API/python_api/target/TestTargetAPI.py
M lldb/test/API/python_api/type/TestTypeList.py
M lldb/test/API/python_api/value/TestValueAPI.py
M lldb/test/API/python_api/value/linked_list/TestValueAPILinkedList.py
M lldb/test/API/python_api/watchpoint/condition/TestWatchpointConditionAPI.py
Log Message:
-----------
[LLDB] Convert test launches with extra arguments to lldbutil (NFC) (#224412)
Replace the CreateTarget + breakpoint + LaunchSimple + stop-assert
boilerplate with the matching lldbutil.run_to_*_breakpoint helper at 19
call sites.
Besides being shorter, the helpers check that the breakpoint resolved to
at least one location and that the process stopped at that specific
breakpoint, which most of the original code did not do.
It also improves and unifies the error reporting when the launch fails.
Note for reviewers: In this patch, all invocations passed additional
arguments that needed to be carried over to the lldbutil replacement.
Assisted-by: Claude
Commit: 8282b18b2e998d215698c327184d87c63dc0c4e6
https://github.com/llvm/llvm-project/commit/8282b18b2e998d215698c327184d87c63dc0c4e6
Author: Adrian Prantl <aprantl at apple.com>
Date: 2026-09-18 (Fri, 18 Sep 2026)
Changed paths:
M lldb/test/API/python_api/lldbutil/iter/TestLLDBIterator.py
M lldb/test/API/python_api/lldbutil/iter/TestRegistersIterator.py
M lldb/test/API/python_api/watchpoint/watchlocation/TestSetWatchlocation.py
M lldb/test/API/python_api/watchpoint/watchlocation/TestTargetWatchAddress.py
Log Message:
-----------
[LLDB] Convert multi-threaded test launches to lldbutil (NFC) (#224415)
Replace the CreateTarget + breakpoint + LaunchSimple + stop-assert
boilerplate with the matching lldbutil.run_to_*_breakpoint helper at 7
call sites.
Besides being shorter, the helpers check that the breakpoint resolved to
at least one location and that the process stopped at that specific
breakpoint, which most of the original code did not do.
It also improves and unifies the error reporting when the launch fails.
Note for reviewers: These inferiors are multi-threaded and the original
code used get_stopped_thread(), which tolerates several threads stopping
at once. The helpers require exactly one by default, so every call here
passes only_one_thread=False to preserve the previous behavior. Without
it TestLLDBIterator fails with "Expected 1 thread to stop at breakpoint,
3 did" - but only under load, which makes this easy to miss when running
a single test.
Assisted-by: Claude
Commit: 5fbeec8d587bcb6c33cd0c194b60981e6357ded6
https://github.com/llvm/llvm-project/commit/5fbeec8d587bcb6c33cd0c194b60981e6357ded6
Author: Alexey Bataev <a.bataev at outlook.com>
Date: 2026-09-18 (Fri, 18 Sep 2026)
Changed paths:
M llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
A llvm/test/Transforms/SLPVectorizer/X86/or-disjoint-copyable-self-op.ll
Log Message:
-----------
[SLP]Drop disjoint flag for copyable lanes modeled as op(V, V)
or disjoint(V, V) is poison unless V is zero, so propagating the
flag from the scalar instructions miscompiles the synthesized
self-op lanes. Keep it for identity-modeled and zero lanes.
Fixes https://github.com/llvm/llvm-project/pull/220252#issuecomment-5736121992
Reviewers:
Pull Request: https://github.com/llvm/llvm-project/pull/224760
Commit: d93df21ccc49e4796126614cea7097d4ee433a41
https://github.com/llvm/llvm-project/commit/d93df21ccc49e4796126614cea7097d4ee433a41
Author: Adrian Prantl <aprantl at apple.com>
Date: 2026-09-18 (Fri, 18 Sep 2026)
Changed paths:
M lldb/packages/Python/lldbsuite/test/lldbutil.py
M lldb/test/API/commands/watchpoints/watchpoint_events/TestWatchpointEvents.py
M lldb/test/API/functionalities/inline-stepping/TestInlineStepping.py
M lldb/test/API/functionalities/thread/thread_specific_break_plus_condition/TestThreadSpecificBpPlusCondition.py
M lldb/test/API/functionalities/unwind/sigtramp/TestSigtrampUnwind.py
M lldb/test/API/linux/thread/create_during_instruction_step/TestCreateDuringInstructionStep.py
M lldb/test/API/macosx/add-dsym/TestAddDsymDownload.py
M lldb/test/API/python_api/thread/TestThreadAPI.py
Log Message:
-----------
[LLDB] Convert the remaining test launches to lldbutil (NFC) (#224416)
Replace the CreateTarget + breakpoint + LaunchSimple + stop-assert
boilerplate with the matching lldbutil.run_to_*_breakpoint helper.
Besides being shorter, the helpers check that the breakpoint resolved to
at least one location and that the process stopped at that specific
breakpoint, which most of the original code did not do.
It also improves and unifies the error reporting when the launch fails.
Note for reviewers: These did not follow the common pattern and had to
be converted by hand, so they are worth a closer look than the earlier
batches:
- Five of them called get_threads_stopped_at_breakpoint() and indexed
the resulting list. The helper already returns the single stopped
thread, so the list, its length assertions and the [0] subscript all
collapse into it.
- TestSigtrampUnwind checked the stop state with an if/fail block rather
than an assertion.
- TestAddDsymDownload built its target from self.exe, which is just
getBuildArtifact("a.out"), so it needs no exe_name argument.
- TestThreadAPI looks up the thread inside a while loop, so only the
target creation and launch could be folded into the helper.
Assisted-by: Claude
Commit: 7662a0e1c4034bd13ca46dd72c5e0f0d2614b6f5
https://github.com/llvm/llvm-project/commit/7662a0e1c4034bd13ca46dd72c5e0f0d2614b6f5
Author: Vitaly Buka <vitalybuka at google.com>
Date: 2026-09-18 (Fri, 18 Sep 2026)
Changed paths:
M llvm/test/Transforms/LowerTypeTests/cfi-jumptable-hotness-summary.ll
Log Message:
-----------
[NFC][LowerTypeTests] Update CFI jump table summary tests (#224761)
Update CHECK lines to match the new jump table layout in
* https://github.com/llvm/llvm-project/pull/221046
Implementation will not recognize `critical`.
Assisted-by: Gemini
Commit: 3fa94abb2f19065bd280a737558b6fd01f8a4168
https://github.com/llvm/llvm-project/commit/3fa94abb2f19065bd280a737558b6fd01f8a4168
Author: Paul Osmialowski <pawel.osmialowski at arm.com>
Date: 2026-09-18 (Fri, 18 Sep 2026)
Changed paths:
M compiler-rt/cmake/config-ix.cmake
Log Message:
-----------
[compiler-rt] Fix the use of COMPILER_RT_BUILTINS_LIBRARY (#224681)
When `COMPILER_RT_BUILTINS_LIBRARY` is set and a path to the builtins
library is being provided, it is still not being used, causing an
error in the COMPILER_RT_TARGET_HAS_ATOMICS test:
```
ld.lld: error: undefined symbol: __aarch64_cas8_acq_rel
>>> referenced by src.cxx
>>> CMakeFiles/cmTC_7f91d.dir/src.cxx.o:(main)
ld.lld: error: undefined symbol: __aarch64_ldadd8_acq_rel
>>> referenced by src.cxx
>>> CMakeFiles/cmTC_7f91d.dir/src.cxx.o:(main)
>>> referenced by src.cxx
>>> CMakeFiles/cmTC_7f91d.dir/src.cxx.o:(main)
clang++: error: linker command failed with exit code 1
```
Although it is visible in the
build/runtimes/runtimes-bins/CMakeFiles/CMakeConfigureLog.yaml log,
it is being ignored. This causes a cascade of subtle problems down the
line, including a faulty PGO instrumentation code being injected into
instrumented binaries, which causes race conditions whenever the
instrumented code is multi-threaded (a stub is being used instead of
the real atomics).
Commit: 87a926c0d736192ad8e149759347bfa8bd46d636
https://github.com/llvm/llvm-project/commit/87a926c0d736192ad8e149759347bfa8bd46d636
Author: Andy Kaylor <akaylor at nvidia.com>
Date: 2026-09-18 (Fri, 18 Sep 2026)
Changed paths:
M clang/include/clang/CIR/Dialect/IR/CIROps.td
M clang/lib/CIR/CodeGen/CIRGenException.cpp
M clang/lib/CIR/CodeGen/CIRGenFunction.cpp
M clang/lib/CIR/CodeGen/CIRGenFunction.h
M clang/lib/CIR/Dialect/IR/CIRDialect.cpp
A clang/test/CIR/CodeGen/ehspec-noexcept.cpp
A clang/test/CIR/CodeGen/ehspec.cpp
M clang/test/CIR/CodeGen/rtti-qualfn.cpp
M clang/test/CIR/IR/invalid-try-catch.cir
M clang/test/CIR/IR/try-catch.cir
M clang/test/CIR/Transforms/eh-abi-lowering-eh-spec.cir
M clang/test/CIR/Transforms/flatten-eh-spec.cir
Log Message:
-----------
[CIR][EH] Implement codegen for dynamic exception specification (#224375)
This implements CIR generation for C++ functions with dynamic exception
specifications. The exception filter is implemented as a cir.try
operation that wraps the entire function body. CFG flattening and EH ABI
lowerting were implemented in previous changes, so they are already in
place.
This also adds an NYI diagnostic for functions marked `noexcept`. That
will be implemented in a follow-up change, but the new diagnostic forced
updates to a few tests that were using that form.
Assisted-by: Cursor / various models
Commit: 4c31f641d008e0fe03d84cdc05e5000c1b530729
https://github.com/llvm/llvm-project/commit/4c31f641d008e0fe03d84cdc05e5000c1b530729
Author: Andy Kaylor <akaylor at nvidia.com>
Date: 2026-09-18 (Fri, 18 Sep 2026)
Changed paths:
M clang/include/clang/CIR/Dialect/Analysis/CIRAliasAnalysis.h
M clang/include/clang/CIR/Dialect/Analysis/CIRBasicAliasAnalysis.h
M clang/lib/CIR/Dialect/Analysis/CIRAliasAnalysis.cpp
M clang/lib/CIR/Dialect/Analysis/CIRBasicAliasAnalysis.cpp
M clang/test/CIR/Analysis/alias-analysis-underlying-object.cir
M clang/test/CIR/lib/Analysis/TestCIRAliasAnalysis.cpp
Log Message:
-----------
[CIR] Add object offset calculation for basic AA (#219047)
This adds code to compute the offset of a pointer into a base object
when the base object is computed by CIR's basic alias analysis. The
offset will be used in a future change to determine partial alias, but
at this point the analysis still reports MayAlias for objects with
different offsets.
Assisted-by: Cursor / various models
Commit: 0c2f772be92289447f477552e232cbc5306c4659
https://github.com/llvm/llvm-project/commit/0c2f772be92289447f477552e232cbc5306c4659
Author: Vitaly Buka <vitalybuka at google.com>
Date: 2026-09-18 (Fri, 18 Sep 2026)
Changed paths:
M llvm/lib/Transforms/IPO/LowerTypeTests.cpp
M llvm/test/Transforms/ThinLTOBitcodeWriter/cfi-functions-hotness.ll
Log Message:
-----------
[LowerTypeTests] Add -reorder-cfi-jump-tables-profiles kill switch (#224752)
Add `-reorder-cfi-jump-tables-profiles` (default: true) to allow
disabling CFI jump table hotness metadata generation and profiling
queries introduced in 50c563998bac.
When set to false, `createCfiFunctionsMetadata` avoids querying
`ProfileSummaryInfo` and `BlockFrequencyInfo`, leaving hotness bits in
`cfi.functions` metadata as 0 (unknown).
Assisted-by: Gemini
Commit: 400cdf6f379c0177e6259eeaf65ea8efa93326b3
https://github.com/llvm/llvm-project/commit/400cdf6f379c0177e6259eeaf65ea8efa93326b3
Author: Jianhui Li <jian.hui.li at intel.com>
Date: 2026-09-18 (Fri, 18 Sep 2026)
Changed paths:
M mlir/lib/Conversion/VectorToXeGPU/VectorToXeGPU.cpp
M mlir/test/Conversion/VectorToXeGPU/transfer-read-to-xegpu.mlir
M mlir/test/Conversion/VectorToXeGPU/transfer-write-to-xegpu.mlir
Log Message:
-----------
[mlir][VectorToXeGPU] Lower out-of-bounds 1-D transfers (#220490)
A vector.transfer_read / vector.transfer_write whose vector is 1-D and
whose access is out of bounds was not lowered.
This PR now builds a real in-bounds mask instead of handing
xegpu.load/xegpu.store an all-ones one. An element is in bounds when its
index is still inside the memref dimension it reads, i.e. i < dim(d) -
indices[d], and dimensions the transfer already declares in-bounds are
left alone. Since masked-off lanes of an xegpu.load hold unspecified
values rather than zeros, a read adds an arith.select to fill them with
the transfer's padding.
It also lowers a transfer_read of a single-element vector (any rank) to
a scalar xegpu.load, when that vector is only ever extracted to a
scalar. The load takes a scalar offset and a scalar mask, and the mask
is what keeps an out-of-bounds transfer from touching memory; because a
masked-off load is unspecified, the padding is applied with an
arith.select, as the scattered path already does. This avoids
special-casing single-element vectors in layout propagation: a scalar
consumer requests no layout, so the one-lane xegpu.load would otherwise
be left without one.
---------
Co-authored-by: Claude Opus 5 (1M context) <noreply at anthropic.com>
Commit: 3dc4cbeae3ebb2a6091fe86851f62615970ea17d
https://github.com/llvm/llvm-project/commit/3dc4cbeae3ebb2a6091fe86851f62615970ea17d
Author: Alexey Bataev <a.bataev at outlook.com>
Date: 2026-09-18 (Fri, 18 Sep 2026)
Changed paths:
M llvm/test/Transforms/SLPVectorizer/X86/loop-accumulator-reduction.ll
Log Message:
-----------
[SLP][NFC]Add a test with three-exit blocks, NFC
Reviewers:
Pull Request: https://github.com/llvm/llvm-project/pull/224768
Commit: d5574d64020b82246940b2c086e8ba88e454b0b2
https://github.com/llvm/llvm-project/commit/d5574d64020b82246940b2c086e8ba88e454b0b2
Author: Helena Kotas <hekotas at microsoft.com>
Date: 2026-09-18 (Fri, 18 Sep 2026)
Changed paths:
M llvm/lib/Target/SPIRV/CMakeLists.txt
M llvm/lib/Target/SPIRV/SPIRV.h
M llvm/lib/Target/SPIRV/SPIRVCodeGenPassBuilder.cpp
R llvm/lib/Target/SPIRV/SPIRVLegalizeImplicitBinding.cpp
A llvm/lib/Target/SPIRV/SPIRVLegalizeResourceBinding.cpp
M llvm/lib/Target/SPIRV/SPIRVPassRegistry.def
M llvm/lib/Target/SPIRV/SPIRVTargetMachine.cpp
M llvm/test/CodeGen/SPIRV/llc-pipeline.ll
R llvm/test/CodeGen/SPIRV/passes/SPIRVLegalizeImplicitBinding.ll
A llvm/test/CodeGen/SPIRV/passes/SPIRVLegalizeResourceBinding-Implicit.ll
Log Message:
-----------
[SPIRV] Rename pass to SPIRVLegalizeResourceBinding (#224747)
Rename `SPIRVLegalizeImplicitBinding` pass to
`SPIRVLegalizeResourceBinding` since it will soon be extended to handle
heap resource bindings as well. Renaming is as a separate PR will make
the upcoming change easier to review.
Related to #208054
Commit: fa98c5013c111efe858694dd7416e5d42421182d
https://github.com/llvm/llvm-project/commit/fa98c5013c111efe858694dd7416e5d42421182d
Author: Alexey Bataev <a.bataev at outlook.com>
Date: 2026-09-18 (Fri, 18 Sep 2026)
Changed paths:
M bolt/include/bolt/Rewrite/RewriteInstance.h
M bolt/lib/Rewrite/RewriteInstance.cpp
M bolt/test/AArch64/use-old-text-zero-padding.c
M clang/include/clang/AST/DeclBase.h
M clang/include/clang/CIR/Dialect/Analysis/CIRAliasAnalysis.h
M clang/include/clang/CIR/Dialect/Analysis/CIRBasicAliasAnalysis.h
M clang/include/clang/CIR/Dialect/IR/CIROps.td
M clang/lib/AST/DeclBase.cpp
M clang/lib/CIR/CodeGen/CIRGenException.cpp
M clang/lib/CIR/CodeGen/CIRGenFunction.cpp
M clang/lib/CIR/CodeGen/CIRGenFunction.h
M clang/lib/CIR/CodeGen/CIRGenModule.cpp
M clang/lib/CIR/Dialect/Analysis/CIRAliasAnalysis.cpp
M clang/lib/CIR/Dialect/Analysis/CIRBasicAliasAnalysis.cpp
M clang/lib/CIR/Dialect/IR/CIRDialect.cpp
M clang/lib/Driver/ToolChains/Darwin.cpp
M clang/lib/StaticAnalyzer/Checkers/WebKit/PtrTypesSemantics.cpp
M clang/test/CIR/Analysis/alias-analysis-underlying-object.cir
A clang/test/CIR/CodeGen/ehspec-noexcept.cpp
A clang/test/CIR/CodeGen/ehspec.cpp
M clang/test/CIR/CodeGen/rtti-qualfn.cpp
M clang/test/CIR/IR/invalid-try-catch.cir
M clang/test/CIR/IR/try-catch.cir
M clang/test/CIR/Transforms/eh-abi-lowering-eh-spec.cir
M clang/test/CIR/Transforms/flatten-eh-spec.cir
M clang/test/CIR/lib/Analysis/TestCIRAliasAnalysis.cpp
A clang/test/Driver/darwin-static-lib-universal.c
M clang/test/Driver/darwin-static-lib.c
M compiler-rt/cmake/config-ix.cmake
M libcxx/docs/ReleaseNotes/24.rst
M libcxx/docs/Status/Cxx17Papers.csv
M libcxx/include/CMakeLists.txt
M libcxx/include/__memory/pstl.h
M libcxx/include/__pstl/backend_fwd.h
M libcxx/include/__pstl/backends/default.h
M libcxx/include/__pstl/backends/libdispatch.h
M libcxx/include/__pstl/backends/serial.h
M libcxx/include/__pstl/backends/std_thread.h
M libcxx/include/__pstl/cpu_algos/for_each.h
A libcxx/include/__pstl/cpu_algos/uninitialized_algorithms.h
M libcxx/include/module.modulemap.in
M libcxx/test/libcxx/algorithms/pstl.iterator-requirements.verify.cpp
M libcxx/test/std/algorithms/pstl.exception_handling.pass.cpp
A libcxx/test/std/utilities/memory/specialized.algorithms/uninitialized.copy/pstl.uninitialized_copy.pass.cpp
A libcxx/test/std/utilities/memory/specialized.algorithms/uninitialized.copy/pstl.uninitialized_copy_n.pass.cpp
A libcxx/test/std/utilities/memory/specialized.algorithms/uninitialized.move/pstl.uninitialized_move.pass.cpp
A libcxx/test/std/utilities/memory/specialized.algorithms/uninitialized.move/pstl.uninitialized_move_n.pass.cpp
M libsycl/src/detail/queue_impl.cpp
M libsycl/src/usm_functions.cpp
M libsycl/unittests/handler/test_helpers.hpp
M libsycl/unittests/mock/helpers.cpp
M libsycl/unittests/mock/helpers.hpp
M libsycl/unittests/mock/mock.cpp
M libsycl/unittests/queue/memcpy.cpp
M libsycl/unittests/usm/alloc.cpp
M lldb/include/lldb/Target/Process.h
M lldb/include/lldb/Target/ThreadPlan.h
M lldb/packages/Python/lldbsuite/test/lldbutil.py
M lldb/source/Plugins/Instruction/ARM64/EmulateInstructionARM64.cpp
M lldb/source/Plugins/Instruction/ARM64/EmulateInstructionARM64.h
M lldb/source/Plugins/UnwindAssembly/InstEmulation/UnwindAssemblyInstEmulation.cpp
M lldb/source/Plugins/UnwindAssembly/InstEmulation/UnwindAssemblyInstEmulation.h
M lldb/source/Target/Process.cpp
M lldb/source/Target/Thread.cpp
M lldb/test/API/android/platform/TestDefaultCacheLineSize.py
M lldb/test/API/commands/frame/var-dil/expr/Comparison/TestFrameVarDILExprComparison.py
M lldb/test/API/commands/process/reverse-continue/TestReverseContinueCommandNotSupported.py
M lldb/test/API/commands/watchpoints/multiple_hits/TestMultipleHits.py
M lldb/test/API/commands/watchpoints/watchpoint_events/TestWatchpointEvents.py
M lldb/test/API/functionalities/conditional_break/TestConditionalBreak.py
M lldb/test/API/functionalities/inline-stepping/TestInlineStepping.py
M lldb/test/API/functionalities/postmortem/minidump/TestMiniDump.py
M lldb/test/API/functionalities/process_save_core/TestProcessSaveCore.py
M lldb/test/API/functionalities/process_save_core_minidump/partial_read/TestProcessSaveCoreMinidumpPartialRead.py
M lldb/test/API/functionalities/process_save_core_minidump/size_mismatch/TestProcessSaveCoreMinidumpSizeMismatch.py
M lldb/test/API/functionalities/return-value/TestReturnValue.py
M lldb/test/API/functionalities/reverse-execution/TestReverseExecutionNotSupported.py
M lldb/test/API/functionalities/scripted_frame_provider/circular_dependency/TestFrameProviderCircularDependency.py
M lldb/test/API/functionalities/tail_call_frames/cross_object/TestCrossObjectTailCalls.py
M lldb/test/API/functionalities/thread/crash_during_step/TestCrashDuringStep.py
M lldb/test/API/functionalities/thread/thread_specific_break_plus_condition/TestThreadSpecificBpPlusCondition.py
A lldb/test/API/functionalities/thread_plan/step_over_breakpoint_site/Makefile
A lldb/test/API/functionalities/thread_plan/step_over_breakpoint_site/TestStepOverBreakpointSite.py
A lldb/test/API/functionalities/thread_plan/step_over_breakpoint_site/main.c
A lldb/test/API/functionalities/thread_plan/step_over_breakpoint_site/run_to_address_plan.py
A lldb/test/API/functionalities/unwind/outlined-prologue/Makefile
A lldb/test/API/functionalities/unwind/outlined-prologue/TestOutlinedPrologueUnwind.py
A lldb/test/API/functionalities/unwind/outlined-prologue/main.c
M lldb/test/API/functionalities/unwind/sigtramp/TestSigtrampUnwind.py
M lldb/test/API/lang/c/local_variables/TestLocalVariables.py
M lldb/test/API/lang/cpp/class_static/TestStaticVariables.py
M lldb/test/API/lang/cpp/dynamic-value/TestCppValueCast.py
M lldb/test/API/lang/objc/foundation/TestObjCMethods.py
M lldb/test/API/lang/objc/foundation/TestObjectDescriptionAPI.py
M lldb/test/API/lang/objc/objc-ivar-offsets/TestObjCIvarOffsets.py
M lldb/test/API/lang/objc/objc-static-method/TestObjCStaticMethod.py
M lldb/test/API/lang/objc/objc-struct-argument/TestObjCStructArgument.py
M lldb/test/API/lang/objc/objc-struct-return/TestObjCStructReturn.py
M lldb/test/API/lang/objc/objc-super/TestObjCSuper.py
M lldb/test/API/linux/add-symbols/TestTargetSymbolsAddCommand.py
M lldb/test/API/linux/mix-dwo-and-regular-objects/TestMixedDwarfBinary.py
M lldb/test/API/linux/thread/create_during_instruction_step/TestCreateDuringInstructionStep.py
M lldb/test/API/macosx/add-dsym/TestAddDsymDownload.py
M lldb/test/API/macosx/safe-to-func-call/TestSafeFuncCalls.py
M lldb/test/API/python_api/class_members/TestSBTypeClassMembers.py
M lldb/test/API/python_api/findvalue_duplist/TestSBFrameFindValue.py
M lldb/test/API/python_api/frame/TestFrames.py
M lldb/test/API/python_api/frame/inlines/TestInlinedFrame.py
M lldb/test/API/python_api/lldbutil/frame/TestFrameUtils.py
M lldb/test/API/python_api/lldbutil/iter/TestLLDBIterator.py
M lldb/test/API/python_api/lldbutil/iter/TestRegistersIterator.py
M lldb/test/API/python_api/objc_type/TestObjCType.py
M lldb/test/API/python_api/process/TestProcessAPI.py
M lldb/test/API/python_api/symbol-context/TestSymbolContext.py
M lldb/test/API/python_api/target/TestTargetAPI.py
M lldb/test/API/python_api/thread/TestThreadAPI.py
M lldb/test/API/python_api/type/TestTypeList.py
M lldb/test/API/python_api/value/TestValueAPI.py
M lldb/test/API/python_api/value/empty_class/TestValueAPIEmptyClass.py
M lldb/test/API/python_api/value/linked_list/TestValueAPILinkedList.py
M lldb/test/API/python_api/watchpoint/TestSetWatchpoint.py
M lldb/test/API/python_api/watchpoint/TestWatchpointIgnoreCount.py
M lldb/test/API/python_api/watchpoint/TestWatchpointIter.py
M lldb/test/API/python_api/watchpoint/TestWatchpointRead.py
M lldb/test/API/python_api/watchpoint/condition/TestWatchpointConditionAPI.py
M lldb/test/API/python_api/watchpoint/watchlocation/TestSetWatchlocation.py
M lldb/test/API/python_api/watchpoint/watchlocation/TestTargetWatchAddress.py
M lldb/unittests/UnwindAssembly/ARM64/CMakeLists.txt
M lldb/unittests/UnwindAssembly/ARM64/TestArm64InstEmulation.cpp
M llvm/include/llvm/CodeGen/Passes.h
M llvm/lib/CodeGen/BranchFolding.cpp
M llvm/lib/CodeGen/LibcallLoweringInfo.cpp
M llvm/lib/CodeGen/ShadowStackGCLowering.cpp
M llvm/lib/Target/AMDGPU/AMDGPU.td
M llvm/lib/Target/AMDGPU/AMDGPUArgumentUsageInfo.cpp
M llvm/lib/Target/AMDGPU/AMDGPUFeatures.td
M llvm/lib/Target/AMDGPU/AMDGPURegBankLegalizeHelper.cpp
M llvm/lib/Target/AMDGPU/SIISelLowering.cpp
M llvm/lib/Target/AVR/AVRFrameLowering.cpp
M llvm/lib/Target/AVR/AVRISelLowering.cpp
M llvm/lib/Target/AVR/AVRISelLowering.h
M llvm/lib/Target/RISCV/RISCVISelLowering.cpp
M llvm/lib/Target/SPIRV/CMakeLists.txt
M llvm/lib/Target/SPIRV/SPIRV.h
M llvm/lib/Target/SPIRV/SPIRVCodeGenPassBuilder.cpp
R llvm/lib/Target/SPIRV/SPIRVLegalizeImplicitBinding.cpp
A llvm/lib/Target/SPIRV/SPIRVLegalizeResourceBinding.cpp
M llvm/lib/Target/SPIRV/SPIRVPassRegistry.def
M llvm/lib/Target/SPIRV/SPIRVTargetMachine.cpp
M llvm/lib/Transforms/IPO/LowerTypeTests.cpp
M llvm/lib/Transforms/Scalar/ConstraintElimination.cpp
M llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
M llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp
M llvm/test/CodeGen/AMDGPU/GlobalISel/irtranslator-call-abi-attribute-hints.ll
M llvm/test/CodeGen/AMDGPU/GlobalISel/irtranslator-call-implicit-args.ll
M llvm/test/CodeGen/AMDGPU/GlobalISel/irtranslator-call-return-values.ll
M llvm/test/CodeGen/AMDGPU/GlobalISel/irtranslator-call-sret.ll
M llvm/test/CodeGen/AMDGPU/GlobalISel/irtranslator-call.ll
M llvm/test/CodeGen/AMDGPU/GlobalISel/irtranslator-indirect-call.ll
M llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-amdgcn.workitem.id.mir
M llvm/test/CodeGen/AMDGPU/GlobalISel/regbankselect-add.mir
M llvm/test/CodeGen/AMDGPU/GlobalISel/regbankselect-amdgcn.ballot.i64.mir
M llvm/test/CodeGen/AMDGPU/GlobalISel/regbankselect-and-s1.mir
M llvm/test/CodeGen/AMDGPU/GlobalISel/regbankselect-anyext.mir
M llvm/test/CodeGen/AMDGPU/GlobalISel/regbankselect-brcond.mir
M llvm/test/CodeGen/AMDGPU/GlobalISel/regbankselect-freeze.mir
M llvm/test/CodeGen/AMDGPU/GlobalISel/regbankselect-mui-regbanklegalize.mir
M llvm/test/CodeGen/AMDGPU/GlobalISel/regbankselect-mui.mir
M llvm/test/CodeGen/AMDGPU/GlobalISel/regbankselect-sadde.mir
M llvm/test/CodeGen/AMDGPU/GlobalISel/regbankselect-select.mir
M llvm/test/CodeGen/AMDGPU/GlobalISel/regbankselect-sext.mir
M llvm/test/CodeGen/AMDGPU/GlobalISel/regbankselect-ssube.mir
M llvm/test/CodeGen/AMDGPU/GlobalISel/regbankselect-sub.mir
M llvm/test/CodeGen/AMDGPU/GlobalISel/regbankselect-trunc.mir
M llvm/test/CodeGen/AMDGPU/GlobalISel/regbankselect-uadde.mir
M llvm/test/CodeGen/AMDGPU/GlobalISel/regbankselect-usube.mir
M llvm/test/CodeGen/AMDGPU/GlobalISel/regbankselect-zext.mir
M llvm/test/CodeGen/AMDGPU/build_vector.gfx11plus.ll
M llvm/test/CodeGen/AMDGPU/carryout-selection.ll
M llvm/test/CodeGen/AMDGPU/fcopysign.gfx11plus.ll
M llvm/test/CodeGen/AMDGPU/fma-mix.gfx11plus.ll
M llvm/test/CodeGen/AMDGPU/fma.f16.gfx11plus.ll
M llvm/test/CodeGen/AMDGPU/indirect-addressing-term.ll
A llvm/test/CodeGen/AMDGPU/mad-mix-fptrunc-rounding-bf16.ll
A llvm/test/CodeGen/AMDGPU/mad-mix-fptrunc-rounding.ll
M llvm/test/CodeGen/AMDGPU/permlane16_opsel.ll
M llvm/test/CodeGen/AMDGPU/scalar_to_vector.gfx11plus.ll
M llvm/test/CodeGen/AMDGPU/vgpr-liverange-ir.ll
A llvm/test/CodeGen/AVR/frameaddress.ll
A llvm/test/CodeGen/AVR/returnaddress.ll
M llvm/test/CodeGen/RISCV/rvv/vector-interleave-fixed.ll
M llvm/test/CodeGen/RISCV/rvv/vector-interleave.ll
M llvm/test/CodeGen/SPIRV/llc-pipeline.ll
R llvm/test/CodeGen/SPIRV/passes/SPIRVLegalizeImplicitBinding.ll
A llvm/test/CodeGen/SPIRV/passes/SPIRVLegalizeResourceBinding-Implicit.ll
A llvm/test/TableGen/AMDGPUTargetDefGenericClassification.td
M llvm/test/Transforms/ConstraintElimination/decompose-signed-preconditions.ll
M llvm/test/Transforms/ConstraintElimination/induction-exit-condition-or-and.ll
M llvm/test/Transforms/ConstraintElimination/materialize-flags.ll
M llvm/test/Transforms/ConstraintElimination/reproducer-remarks.ll
M llvm/test/Transforms/LoopVectorize/AArch64/force-target-instruction-cost.ll
M llvm/test/Transforms/LoopVectorize/AArch64/interleave-with-gaps.ll
M llvm/test/Transforms/LoopVectorize/AArch64/sve-gather-scatter.ll
M llvm/test/Transforms/LoopVectorize/AArch64/sve-interleaved-access-low-vf.ll
M llvm/test/Transforms/LoopVectorize/AArch64/sve-tail-folding-option.ll
M llvm/test/Transforms/LoopVectorize/AArch64/sve-widen-phi.ll
M llvm/test/Transforms/LoopVectorize/AArch64/transform-narrow-interleave-vscale-x-UF-step.ll
M llvm/test/Transforms/LoopVectorize/RISCV/strided-accesses.ll
M llvm/test/Transforms/LowerTypeTests/cfi-jumptable-hotness-summary.ll
M llvm/test/Transforms/SLPVectorizer/X86/loop-accumulator-reduction.ll
A llvm/test/Transforms/SLPVectorizer/X86/or-disjoint-copyable-self-op.ll
M llvm/test/Transforms/ThinLTOBitcodeWriter/cfi-functions-hotness.ll
M llvm/tools/llvm-gpu-loader/llvm-gpu-loader.cpp
M llvm/tools/llvm-gpu-loader/llvm-gpu-loader.h
M llvm/utils/TableGen/Basic/AMDGPUTargetDefEmitter.cpp
M llvm/utils/profcheck-xfail.txt
M mlir/lib/Conversion/VectorToXeGPU/VectorToXeGPU.cpp
M mlir/test/Conversion/VectorToXeGPU/transfer-read-to-xegpu.mlir
M mlir/test/Conversion/VectorToXeGPU/transfer-write-to-xegpu.mlir
M offload/languages/kernel/src/LanguageRuntime.cpp
M offload/liboffload/API/Memory.td
M offload/liboffload/src/OffloadImpl.cpp
M offload/plugins-nextgen/amdgpu/dynamic_hsa/hsa_ext_amd.h
M offload/plugins-nextgen/amdgpu/src/rtl.cpp
M offload/plugins-nextgen/common/include/PluginInterface.h
M offload/plugins-nextgen/common/src/PluginInterface.cpp
M offload/plugins-nextgen/cuda/dynamic_cuda/cuda.cpp
M offload/plugins-nextgen/cuda/dynamic_cuda/cuda.h
M offload/plugins-nextgen/cuda/src/rtl.cpp
M offload/plugins-nextgen/host/src/rtl.cpp
M offload/plugins-nextgen/level_zero/include/L0Memory.h
M offload/plugins-nextgen/level_zero/include/L0Plugin.h
M offload/plugins-nextgen/level_zero/src/L0Context.cpp
M offload/plugins-nextgen/level_zero/src/L0Device.cpp
M offload/plugins-nextgen/level_zero/src/L0Memory.cpp
M offload/plugins-nextgen/level_zero/src/L0Plugin.cpp
M offload/unittests/Conformance/include/mathtest/DeviceContext.hpp
M offload/unittests/Conformance/include/mathtest/DeviceResources.hpp
M offload/unittests/Conformance/lib/DeviceContext.cpp
M offload/unittests/Conformance/lib/DeviceResources.cpp
M offload/unittests/OffloadAPI/event/olGetEventElapsedTime.cpp
M offload/unittests/OffloadAPI/kernel/olLaunchKernel.cpp
M offload/unittests/OffloadAPI/memory/olGetMemInfo.cpp
M offload/unittests/OffloadAPI/memory/olGetMemInfoSize.cpp
M offload/unittests/OffloadAPI/memory/olMemAlloc.cpp
M offload/unittests/OffloadAPI/memory/olMemAllocAligned.cpp
M offload/unittests/OffloadAPI/memory/olMemFill.cpp
M offload/unittests/OffloadAPI/memory/olMemFree.cpp
M offload/unittests/OffloadAPI/memory/olMemPrefetch.cpp
M offload/unittests/OffloadAPI/memory/olMemcpy.cpp
M offload/unittests/OffloadAPI/queue/olLaunchHostFunction.cpp
M offload/unittests/OffloadAPI/queue/olWaitEvents.cpp
Log Message:
-----------
Rebase, address comments
Created using spr 1.3.7
Compare: https://github.com/llvm/llvm-project/compare/52fd92b190ab...fa98c5013c11
To unsubscribe from these emails, change your notification settings at https://github.com/llvm/llvm-project/settings/notifications
More information about the All-commits
mailing list