[qt/clang/llvm-project]: Summary of bulk changes made

Git repository change summary for qt/clang/llvm-project
Pushed by mirror-service into branch 'upstream/main'.
Changed from 4e0d78f974499e728de1ee1acbdce1e6491a610c to 4e924a6276ef015e1482b68371bb8229368fe5f7
Acknowledgement was received that this change introduces only existing code that has been pushed to another public open source repository.

This change contains the following new commits:

Git commit 2450b7e17b903fb70560a883d07ad1e6ab537c56 by GitHub (on behalf of Maksim Levental) on 02/08/2026 at 07:08..
[MLIR][Python] Drop the LLVMSupport link dependency from the python extensions (#213509)

Follow-up to #180986, which completed the series started in #178290: it switched the last python bindings off the C++ LLVM APIs and dropped `LLVMSupport` from the support library in `AddMLIRPython.cmake`. But the per extension `PRIVATE_LINK_LIBS LLVMSupport` in `mlir/python/CMakeLists.txt` was missed, so the dependency is still there for every extension module.

Assisted by: Claude
https://invent.kde.org/qt/clang/llvm-project/-/commit/2450b7e17b903fb70560a883d07ad1e6ab537c56

Git commit 6a3d4d935cab115f5a088cea394f606b6c46e991 by GitHub (on behalf of Fangrui Song) on 02/08/2026 at 07:23..
[CSKY] Emit build attributes in assembly output (#213507)

Fixes CodeGen/CSKY/fpu-abi-attribute.ll, failing since #212975.

emitTargetAttributes is implemented only by CSKYTargetELFStreamer, so
llc -filetype=asm drops every .csky_attribute directive. Move it to
CSKYTargetStreamer, as ARM and RISC-V do.
https://invent.kde.org/qt/clang/llvm-project/-/commit/6a3d4d935cab115f5a088cea394f606b6c46e991

Git commit 8d292a7c4b952eb5e9c55ed54a17a25c85bd553c by GitHub (on behalf of Lang Hames) on 02/08/2026 at 07:50..
[ORC] Remove EPC runAsVoidFunction/runAsIntFunction methods (#213265)

Remove runAsVoidFunction and runAsIntFunction from
ExecutorProcessControl and all its implementations, along with the
now-unused RunAs{Void,Int}FunctionWrapperName bootstrap symbols and
SimpleRemoteEPC's corresponding address fields. Their only in-tree
users, COFFPlatform and COFFVCRuntimeSupport, now use
rt::sps::Int32VoidCaller and rt::sps::Int32Int32Caller.

runAsMain is likewise routed through rt::sps::MainCaller's controller-
interface symbol, so it is looked up under the same orc_rt_ci_sps_* name
the target-process bootstrap registers.

This is a step towards decoupling the ExecutorProcessControl interface
from SPS serialization.
https://invent.kde.org/qt/clang/llvm-project/-/commit/8d292a7c4b952eb5e9c55ed54a17a25c85bd553c

Git commit e18c3ea024846a60c43e0066af3be953745cc34e by GitHub (on behalf of Fangrui Song) on 02/08/2026 at 08:12..
[test] Add irreducible CFGs with closed-form block frequencies (#213492)

The functions here have irreducible control flow, but none of them pins
down how mass is divided among the entries of an irreducible region.

Add four cases whose exact frequencies follow from the branch weights:

- equalrows: all blocks share one successor distribution; 5:3:2.
- selfloops: self edges of differing probability; ignoring them, each
block
  splits evenly between the other two; 8:5:5.
- unequalrows: symmetric non-header successors, differing header row;
8:3:3.
- nonentry: a member of the region that is not an entry, so its mass is
never
  adjusted; 6:4:3.

BFI computes the first two exactly and the last two not. #213488 will
show up as a diff.
https://invent.kde.org/qt/clang/llvm-project/-/commit/e18c3ea024846a60c43e0066af3be953745cc34e

Git commit ffc32ca95939e131c87072fdbfb36e08e0a495cc by GitHub (on behalf of Harrison Hao) on 02/08/2026 at 08:54..
[InstCombine] Don't change the type of elementwise atomic loads (#213414)

Skip load type canonicalization for elementwise atomic loads, which
would
otherwise drop the vector type and produce an invalid scalar elementwise
load.

Reference: https://github.com/llvm/llvm-project/pull/204556
https://invent.kde.org/qt/clang/llvm-project/-/commit/ffc32ca95939e131c87072fdbfb36e08e0a495cc

Git commit e1d649a644a5a18585aa7324fb35e7792314eb4f by GitHub (on behalf of Harrison Hao) on 02/08/2026 at 08:54..
[LowerAtomic] Clear elementwise flag when lowering atomic load (#213401)

The lower atomic pass demotes atomic loads to non atomic by
calling setAtomic(NotAtomic), but left the elementwise flag untouched.
Since elementwise is only valid on atomic operations, this produced a
non-atomic elementwise load, which the verifier rejects with "non-atomic
load cannot be elementwise".

Reference: https://github.com/llvm/llvm-project/pull/204556
https://invent.kde.org/qt/clang/llvm-project/-/commit/e1d649a644a5a18585aa7324fb35e7792314eb4f

Git commit 6c838170ac5d612902dcf9858aed2f26a7fbd8e2 by GitHub (on behalf of Florian Hahn) on 02/08/2026 at 09:42..
[ConstraintElim] Add facts for SREM. (#213453)

Add facts for SRem, if operands are known to be non-negative.

Add signed bounds for `srem x, n`:

  * `x s>= 0`  =>  result s>= 0  and  result s<= x
  * `n s>  0`  =>  result s<=  n

Alive2 Proofs: https://alive2.llvm.org/ce/z/e-zoAP

Compile-time is in the noise

https://llvm-compile-time-tracker.com/compare.php?from=60f965b1f62c0c77bcdb2997ea9bb6603aa0d002&to=ebc652af5700d569884e27812d33735d604990a1&stat=instructions:u

InstCombine already has a similar fold, but with more limited reasoning.
It does not trigger any changes on
https://github.com/dtcxzyw/llvm-opt-benchmark-nightly/pull/841

It simplifies a few times on other C/C++ workloads, including ffmpeg and
OpenColorIO.

Extracted end-to-end examples simplified with the change:
https://clang.godbolt.org/z/1co6rvjKs

This is part of an effort to improve ConstraintElimination support for
IR generated by the Swift compiler, where such patterns are more common
due to a number of signed runtime checks.

PR: https://github.com/llvm/llvm-project/pull/213453
https://invent.kde.org/qt/clang/llvm-project/-/commit/6c838170ac5d612902dcf9858aed2f26a7fbd8e2

Git commit 45e4b0e9b41c3ea685349462ac6c6f831de0c3a1 by GitHub (on behalf of David Green) on 02/08/2026 at 09:47..
[AArch64][GlobalISel] Update extended scalar type to integer. (#213457)

This updates a number of scalar types in CallLowering to use integer.
https://invent.kde.org/qt/clang/llvm-project/-/commit/45e4b0e9b41c3ea685349462ac6c6f831de0c3a1

Git commit a3da10236c52f4cf05936790a7ebc6e8f12b34ba by GitHub (on behalf of jinge90) on 02/08/2026 at 09:52..
[libc] Use llvm-link approach to create libc bitcode library for GPU targets (#210662)
https://invent.kde.org/qt/clang/llvm-project/-/commit/a3da10236c52f4cf05936790a7ebc6e8f12b34ba

Git commit 012e06b43f7345492b9f5792a74c56c04663ae7d by GitHub (on behalf of David Green) on 02/08/2026 at 10:04..
[AArch64][GlobalISel] Update a number of combiner tests to concrete types. NFC (#213522)
https://invent.kde.org/qt/clang/llvm-project/-/commit/012e06b43f7345492b9f5792a74c56c04663ae7d

Git commit e096d2f60dbc6cab991d5c02a5f7125a6dc694dc by GitHub (on behalf of Jacek Caban) on 02/08/2026 at 10:39..
[LLD][COFF] Replace ARM64EC TLS directory chunks with native chunks when available (#212845)

On ARM64X targets, CRT provides separate TLS directory chunks, expecting
the linker to sort it out. TLS directory uses _tls_start and _tls_end
symbols to reference .tls section. Those symbols use section sorting to
ensure that they are emitted at the start and end of .tls section, but that's
not enough when we have two separate chunks for views: only one of them
can really be the first one. Following MSVC, merge those chunks instead so
that both symbol tables point to the same chunk.

Additionally apply the same logic to _tls_used and _tls_index. This
allows entire TLS directory to be shared between EC and native views. To
achieve that, CRT additionally needs to mark each TLS callback with
-arm64xsameaddress. This matches how MSVC linker and libraries work, but
it requires EC and native views to use the same set of TLS callbacks. We
may emit separate TLS directories in the future to make it more robust.
https://invent.kde.org/qt/clang/llvm-project/-/commit/e096d2f60dbc6cab991d5c02a5f7125a6dc694dc

Git commit 4e924a6276ef015e1482b68371bb8229368fe5f7 by GitHub (on behalf of Lang Hames) on 02/08/2026 at 11:04..
[lldb] std::move unique_ptrs, rather than calling .release. (#213525)

These .release() calls are legacy from the std::auto_ptr to
std::unique_ptr transition.
https://invent.kde.org/qt/clang/llvm-project/-/commit/4e924a6276ef015e1482b68371bb8229368fe5f7