[all-commits] [llvm/llvm-project] 5a1c29: [flang] Handle vector operands in logical op codeg...

Steven Wu via All-commits all-commits at lists.llvm.org
Fri Oct 2 10:25:47 PDT 2026


  Branch: refs/heads/users/cachemeifyoucan/spr/cas-support-boot-time-on-windows-and-tolerate-a-moving-boot-time
  Home:   https://github.com/llvm/llvm-project
  Commit: 5a1c296488818dab707d86b4762a96c04f89ea41
      https://github.com/llvm/llvm-project/commit/5a1c296488818dab707d86b4762a96c04f89ea41
  Author: Mattéo Rizza Murgier <matteo.rizza-murgier at sipearl.com>
  Date:   2026-10-02 (Fri, 02 Oct 2026)

  Changed paths:
    M flang/lib/Optimizer/CodeGen/CodeGen.cpp
    M flang/test/Fir/logical-convert.fir

  Log Message:
  -----------
  [flang] Handle vector operands in logical op codegen (#225407)

FIR logical ops take `AnyLogicalOrIntegerLike` operands; this includes
vector types. This PR makes the `normalizeLogicalToI1` and
`extendI1ToType` utils type-agnostic so that they don't assert when
passing a non-scalar type.

Fixes e.g. the crash:
```mlir
; fir-opt --fir-to-llvm-ir="target=aarch64" crash.mlir
func.func @neqv_vector(%a: vector<4xi32>, %b: vector<4xi32>) -> vector<4xi32> {
  %0 = fir.neqv %a, %b : vector<4xi32>
  return %0 : vector<4xi32>
}
```


  Commit: 04cff67bac4c949fea01151599cde96ad2967fe4
      https://github.com/llvm/llvm-project/commit/04cff67bac4c949fea01151599cde96ad2967fe4
  Author: Matt Arsenault <Matthew.Arsenault at amd.com>
  Date:   2026-10-02 (Fri, 02 Oct 2026)

  Changed paths:
    M llvm/lib/Target/X86/X86FastISel.cpp
    A llvm/test/CodeGen/X86/fast-isel-dead-eflags.ll

  Log Message:
  -----------
  X86: Mark the dead EFLAGS clobbers in FastISel i1 masking (#228044)

Co-authored-by: Claude Opus 5 <noreply at anthropic.com>


  Commit: 7533de7a1400ef9a5fd8a8d2e9870a0863eba415
      https://github.com/llvm/llvm-project/commit/7533de7a1400ef9a5fd8a8d2e9870a0863eba415
  Author: Krzysztof Parzyszek <Krzysztof.Parzyszek at amd.com>
  Date:   2026-10-02 (Fri, 02 Oct 2026)

  Changed paths:
    M flang/include/flang/Parser/message.h
    M flang/lib/Parser/message.cpp
    M flang/lib/Semantics/check-acc-structure.cpp
    M flang/lib/Semantics/check-directive-structure.h
    M flang/lib/Semantics/check-omp-loop.cpp
    M flang/lib/Semantics/check-omp-structure.cpp
    M flang/lib/Semantics/check-omp-syntax.cpp
    M flang/lib/Semantics/resolve-directives.cpp

  Log Message:
  -----------
  [flang] Pass llvm::StringRef directly to message formatting (#228449)

Added an overload of MessageFormattedText::Convert for llvm::StringRef.
This removes the need to explicitly convert llvm::StringRef arguments to
std::string in calls to diagnostic functions (Say/Warn/Attach).

Only direct message arguments are changed; uses that feed
ToUpperCaseLetters (which takes std::string_view), llvm::Twine
concatenations, and variable initializers are left as-is. No behavioral
change.

Co-Authored-By: Claude Opus 4 (1M context) <noreply at anthropic.com>


  Commit: 5c38ca153312ea29871fcdd89a0a9ec678c9fb72
      https://github.com/llvm/llvm-project/commit/5c38ca153312ea29871fcdd89a0a9ec678c9fb72
  Author: Matt Arsenault <Matthew.Arsenault at amd.com>
  Date:   2026-10-02 (Fri, 02 Oct 2026)

  Changed paths:
    M llvm/include/llvm/CodeGen/SelectionDAG.h
    M llvm/lib/CodeGen/SelectionDAG/LegalizeDAG.cpp
    M llvm/lib/CodeGen/SelectionDAG/SelectionDAG.cpp
    A llvm/test/CodeGen/NVPTX/fenv-no-libcall-error.ll

  Log Message:
  -----------
  DAG: Gracefully diagnose missing floating-point state libcalls (#228426)

Previously makeStateFunctionCall would hit a fatal error if the target has no 
fegetenv/fesetenv/fegetmode/fesetmode implementation. Report a proper
context error and emit nothing.

Co-authored-by: Claude Opus 5 <noreply at anthropic.com>


  Commit: 070264d5115d3faf3d29228009622e92cf18cdb8
      https://github.com/llvm/llvm-project/commit/070264d5115d3faf3d29228009622e92cf18cdb8
  Author: rdong8 <richarddong08 at gmail.com>
  Date:   2026-10-02 (Fri, 02 Oct 2026)

  Changed paths:
    M utils/bazel/MODULE.bazel
    M utils/bazel/MODULE.bazel.lock

  Log Message:
  -----------
  [bazel] Pin rules_cc_autoconf to 0.24.0 (#227956)

The GNU multiprecision deps (GMP, MPFR, MPC) bring in an old version of
`rules_cc_autoconf` that have some transitive includes pruned by newer
libc++:

```
external/rules_cc_autoconf+/autoconf/private/checker/debug_logger.h:43:33: error: no member named 'back_inserter' in namespace 'std'
   43 |                            std::back_inserter(lower_level), ::tolower);
      |                                 ^~~~~~~~~~~~~
1 error generated.
```

This fix pins it to a newer version.

---------

Co-authored-by: Keith Smiley <keithbsmiley at gmail.com>


  Commit: fc0993001ce85dc5f485ddb26bea29fd8ee15ac9
      https://github.com/llvm/llvm-project/commit/fc0993001ce85dc5f485ddb26bea29fd8ee15ac9
  Author: Arcadiy Ivanov <arcadiy at ivanov.biz>
  Date:   2026-10-02 (Fri, 02 Oct 2026)

  Changed paths:
    M llvm/include/llvm/CodeGen/DIE.h
    M llvm/lib/CodeGen/AsmPrinter/DIE.cpp
    A llvm/test/tools/llvm-dwarfutil/ELF/X86/dwarf5-partial-unit-ref-addr.test
    M llvm/unittests/CodeGen/DIETest.cpp

  Log Message:
  -----------
  [AsmPrinter] Find a DIE's unit by its owner, not by its tag (#219660)

[AsmPrinter] Find a DIE's unit by its owner
Fix DIE::getUnitDie() for DW_TAG_partial_unit. The existing parent walk
recognizes unit roots by tag but omits partial units, so a cross-unit
DW_FORM_ref_addr can reach getDebugSectionOffset() without a unit and
assert.

Have getUnit() walk the parent chain until Owner contains a DIEUnit, then
derive getUnitDie() from that unit. A DIEUnit owns its unit DIE, so this
handles every unit root tag without maintaining another tag list and
still returns null for a DIE tree that no DIEUnit owns. Use
dwarf::isUnitType for the constructor assertion as well. Update the
DIEUnit comments to describe the owner-based walk.

Add unit coverage for all four unit root tags and an unowned DIE tree.
Add a dwarfutil test for resolving DW_FORM_ref_addr into a partial unit
with and without garbage collection. Keep that test on the classic
linker and omit --verify because of #219613 and #219365.

Fixes #219650.

Assisted-by: Claude Code


  Commit: 08e245062b1028d09aadf8344d5d7e9177ce19d6
      https://github.com/llvm/llvm-project/commit/08e245062b1028d09aadf8344d5d7e9177ce19d6
  Author: Hongyu Chen <xxs_chy at outlook.com>
  Date:   2026-10-02 (Fri, 02 Oct 2026)

  Changed paths:
    M clang/lib/Headers/riscv_packed_simd.h
    M clang/test/CodeGen/RISCV/rvp-intrinsics.c
    M cross-project-tests/intrinsic-header-tests/riscv_packed_simd.c
    M llvm/lib/Target/RISCV/RISCVISelLowering.cpp
    M llvm/lib/Target/RISCV/RISCVInstrInfoP.td
    M llvm/test/CodeGen/RISCV/rvp-simd-32.ll
    M llvm/test/CodeGen/RISCV/rvp-simd-64.ll

  Log Message:
  -----------
  [RISCV][P-ext] Support Packed Slide 1 up/down (#228049)

This patch supports packed slide 1 up/down intrinsics and codegen.


  Commit: 2eda6f72ab01914b8aa1a13758e55ff57ad3b3d5
      https://github.com/llvm/llvm-project/commit/2eda6f72ab01914b8aa1a13758e55ff57ad3b3d5
  Author: Matt Arsenault <Matthew.Arsenault at amd.com>
  Date:   2026-10-02 (Fri, 02 Oct 2026)

  Changed paths:
    M llvm/lib/Target/AArch64/GISel/AArch64InstructionSelector.cpp
    A llvm/test/CodeGen/AArch64/mops-dead-nzcv.ll

  Log Message:
  -----------
  AArch64/GlobalISel: Mark the MOPS pseudo NZCV def dead (#227643)

Co-authored-by: Claude Opus 5 <noreply at anthropic.com>


  Commit: 8aee09b1913ebd8f3bb26dd3165f2a645bb2fa80
      https://github.com/llvm/llvm-project/commit/8aee09b1913ebd8f3bb26dd3165f2a645bb2fa80
  Author: Michael Toguchi <michael.d.toguchi at intel.com>
  Date:   2026-10-02 (Fri, 02 Oct 2026)

  Changed paths:
    M clang/lib/Driver/Driver.cpp
    M clang/test/Driver/sycl-offload-jit.cpp

  Log Message:
  -----------
  [SYCL][Driver] Emit textual SPIR-V for device-only -S compilations (#227844)

With -fsycl -fsycl-device-only -S, the SYCL device backend step produced
textual LLVM IR. It now produces assembly, which for spirv64 is textual
SPIR-V from the SPIR-V backend. Adding -emit-llvm produces LLVM IR.


  Commit: 47f4550f76ae4504966770405664c9665b1d861e
      https://github.com/llvm/llvm-project/commit/47f4550f76ae4504966770405664c9665b1d861e
  Author: Razvan Lupusoru <razvan.lupusoru at gmail.com>
  Date:   2026-10-02 (Fri, 02 Oct 2026)

  Changed paths:
    M flang/include/flang/Optimizer/Analysis/AliasAnalysis.h
    M flang/lib/Optimizer/Analysis/AliasAnalysis.cpp
    M flang/test/Analysis/AliasAnalysis/alias-analysis-absent.fir
    M flang/test/Analysis/AliasAnalysis/alias-analysis-regionbranch.mlir
    A flang/test/Analysis/AliasAnalysis/modref-call-null.fir

  Log Message:
  -----------
  [flang] Treat null addresses as their own alias source (#228284)

A null pointer is associated with no address, so classifying it as an
allocation describes an object that does not exist. fir.absent and
address-typed fir.zero_bits both lower to that null. Recognizing the
null lets alias analysis keep it distinct from every object, including
other nulls, while a control-flow join of a null and an allocation is
still an allocation. A join whose every predecessor is a null is still a
null, and a call cannot read or write it.

One pattern this join has to recognize is a private copy that is either
freshly allocated or null:

    %source_box = fir.load %source
    %source_addr = fir.box_addr %source_box
    %is_allocated = arith.cmpi ne, %source_addr, %null

    %private_addr = fir.if %is_allocated -> !fir.heap<f32> {
      %allocation = fir.allocmem f32
      fir.result %allocation : !fir.heap<f32>
    } else {
      %null_allocation = fir.zero_bits !fir.heap<f32>
      fir.result %null_allocation : !fir.heap<f32>
    }

%private_box = fir.embox %private_addr : (!fir.heap<f32>) ->
!fir.box<!fir.heap<f32>>

%allocation is an allocation, %null_allocation is a null, and
%private_addr is an allocation: it is either new storage or no address,
so it cannot alias storage outside the branch.


  Commit: 61914fd3ce4100d61859471f4fbc48db82185c0a
      https://github.com/llvm/llvm-project/commit/61914fd3ce4100d61859471f4fbc48db82185c0a
  Author: Matt Arsenault <Matthew.Arsenault at amd.com>
  Date:   2026-10-02 (Fri, 02 Oct 2026)

  Changed paths:

  Log Message:
  -----------
  AArch64/GlobalISel: Mark the MOPS pseudo NZCV def dead (#227643)

Co-authored-by: Claude Opus 5 <noreply at anthropic.com>


  Commit: d9c741c53f4c9402c93ef163c1d848ab3bec7511
      https://github.com/llvm/llvm-project/commit/d9c741c53f4c9402c93ef163c1d848ab3bec7511
  Author: Simon Pilgrim <llvm-dev at redking.me.uk>
  Date:   2026-10-02 (Fri, 02 Oct 2026)

  Changed paths:
    M llvm/test/CodeGen/X86/vector-shuffle-512-v16.ll
    M llvm/test/CodeGen/X86/vector-shuffle-512-v64.ll

  Log Message:
  -----------
  [X86] Add tests showing failure to convert 512-bit concats of 128-bit chains to X86ISD::SHUF128 nodes (#228512)


  Commit: 0a3ea38db7c53b9134a8f64694f49a6f5c5bda19
      https://github.com/llvm/llvm-project/commit/0a3ea38db7c53b9134a8f64694f49a6f5c5bda19
  Author: Steven Wu <stevenwu at apple.com>
  Date:   2026-10-02 (Fri, 02 Oct 2026)

  Changed paths:
    M clang/lib/Driver/Driver.cpp
    M clang/lib/Headers/riscv_packed_simd.h
    M clang/test/CodeGen/RISCV/rvp-intrinsics.c
    M clang/test/Driver/sycl-offload-jit.cpp
    M cross-project-tests/intrinsic-header-tests/riscv_packed_simd.c
    M flang/include/flang/Optimizer/Analysis/AliasAnalysis.h
    M flang/include/flang/Parser/message.h
    M flang/lib/Optimizer/Analysis/AliasAnalysis.cpp
    M flang/lib/Optimizer/CodeGen/CodeGen.cpp
    M flang/lib/Parser/message.cpp
    M flang/lib/Semantics/check-acc-structure.cpp
    M flang/lib/Semantics/check-directive-structure.h
    M flang/lib/Semantics/check-omp-loop.cpp
    M flang/lib/Semantics/check-omp-structure.cpp
    M flang/lib/Semantics/check-omp-syntax.cpp
    M flang/lib/Semantics/resolve-directives.cpp
    M flang/test/Analysis/AliasAnalysis/alias-analysis-absent.fir
    M flang/test/Analysis/AliasAnalysis/alias-analysis-regionbranch.mlir
    A flang/test/Analysis/AliasAnalysis/modref-call-null.fir
    M flang/test/Fir/logical-convert.fir
    M llvm/include/llvm/CodeGen/DIE.h
    M llvm/include/llvm/CodeGen/SelectionDAG.h
    M llvm/lib/CodeGen/AsmPrinter/DIE.cpp
    M llvm/lib/CodeGen/SelectionDAG/LegalizeDAG.cpp
    M llvm/lib/CodeGen/SelectionDAG/SelectionDAG.cpp
    M llvm/lib/Target/AArch64/GISel/AArch64InstructionSelector.cpp
    M llvm/lib/Target/RISCV/RISCVISelLowering.cpp
    M llvm/lib/Target/RISCV/RISCVInstrInfoP.td
    M llvm/lib/Target/X86/X86FastISel.cpp
    A llvm/test/CodeGen/AArch64/mops-dead-nzcv.ll
    A llvm/test/CodeGen/NVPTX/fenv-no-libcall-error.ll
    M llvm/test/CodeGen/RISCV/rvp-simd-32.ll
    M llvm/test/CodeGen/RISCV/rvp-simd-64.ll
    A llvm/test/CodeGen/X86/fast-isel-dead-eflags.ll
    M llvm/test/CodeGen/X86/vector-shuffle-512-v16.ll
    M llvm/test/CodeGen/X86/vector-shuffle-512-v64.ll
    A llvm/test/tools/llvm-dwarfutil/ELF/X86/dwarf5-partial-unit-ref-addr.test
    M llvm/unittests/CodeGen/DIETest.cpp
    M utils/bazel/MODULE.bazel
    M utils/bazel/MODULE.bazel.lock

  Log Message:
  -----------
  rebase to trigger CI

Created using spr 1.3.7


Compare: https://github.com/llvm/llvm-project/compare/5a428dbff412...0a3ea38db7c5

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