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

Git repository change summary for qt/clang/llvm
Pushed by mirror-service into branch 'upstream/users/aengelke/spr/ir-remove-deprecated-branchinst'.
Changed from 0000000000000000000000000000000000000000 to ed1db561ddd8777d31659a587a3667d14e86c5ac
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 bab258ae1eaa56cbdc86b19f8063be612e1e4795 by GitHub (on behalf of Tom Stellard) on 26/07/2026 at 05:14..
workflows/relase-tasks: Fix variable name typo (#211732)
https://invent.kde.org/qt/clang/llvm/-/commit/bab258ae1eaa56cbdc86b19f8063be612e1e4795

Git commit 2ad69983d8b72d8a665a2552aac15517fe0cb351 by GitHub (on behalf of Tom Stellard) on 26/07/2026 at 05:25..
workflows/libclang-abi-tests: Fix missing environment variable (#211959)

This was accidentally removed in
f6af6cedfd2fa731bf323608116d373379838b9a.
https://invent.kde.org/qt/clang/llvm/-/commit/2ad69983d8b72d8a665a2552aac15517fe0cb351

Git commit 6cdc7bd8c97b633f7095a5df32fb8003ee1b49b9 by GitHub (on behalf of Andrew Savonichev) on 26/07/2026 at 05:57..
[AArch64] Add CAS instructions to NeoverseN2 schedule model (#211195)

CAS instructions (B, H, W, X variants) are not described in the ARM
Neoverse N2 Software Optimization Guide. There used to be no schedule
model for them, so `llvm-mca` was unable to analyze code containing
these instructions:

    llvm-mca  -mtriple=aarch64 -mcpu=cortex-a78 -mattr=+lse casb.s

    error: found an unsupported instruction in the input assembly
    sequence, use -skip-unsupported-instructions=lack-sched to ignore
    these on the input.

    note: instruction: 	casb	w0, w1, [sp]

The patch adds a basic schedule model of 14 cycles (as measured on
Cortex-A78 with llvm-exegesis) and L/S utilized pipelines (this is a
guess, the reality is likely more complicated).
https://invent.kde.org/qt/clang/llvm/-/commit/6cdc7bd8c97b633f7095a5df32fb8003ee1b49b9

Git commit 3fd07e32350c4f15a9377466ab4cffb042a1ecb7 by GitHub (on behalf of Timm Baeder) on 26/07/2026 at 06:14..
[clang][bytecode] Only override constant-context state if we have an EvalEmitter (#211475)

This does not make sense when emitting bytecode, as the bytecode would
just contain a `PushCC`/`PopCC` pair with nothing in between.
https://invent.kde.org/qt/clang/llvm/-/commit/3fd07e32350c4f15a9377466ab4cffb042a1ecb7

Git commit 4eeff3e309da3a5d827aadc6fb6915b968b23ad5 by GitHub (on behalf of Lucas Mellone) on 26/07/2026 at 06:30..
[libc++][math] Add `constexpr` to `std::fpclassify` (#210993)

Adds `constexpr` to `std::fpclassify` as required by
[P0533R9](https://www.open-std.org/jtc1/sc22/wg21/docs/papers/2021/p0533r9.pdf).

Towards https://github.com/llvm/llvm-project/issues/105174. Follows-up
8889a3d3b5154d7d5bd4e58ab10d3df503544742.
https://invent.kde.org/qt/clang/llvm/-/commit/4eeff3e309da3a5d827aadc6fb6915b968b23ad5

Git commit 545f9fab9ed84d55a97b3f08d306c0da70b625a6 by GitHub (on behalf of Timm Baeder) on 26/07/2026 at 06:53..
[clang][bytecode] Don't check global variable init size (#212092)

The current interpreter doesn't do this either. If we do, the clang
build fails because AMDGPUGenGlobalISel.inc: contains a global constexpr
array called MatchTable0 with 1'926'005 elements.
https://invent.kde.org/qt/clang/llvm/-/commit/545f9fab9ed84d55a97b3f08d306c0da70b625a6

Git commit 5d582e4003b1cab59900d16fa3db9c3a9d16b462 by GitHub (on behalf of Fangrui Song) on 26/07/2026 at 07:15..
[LoopInfo] Identify loops with a single-pass DFS algorithm. NFC (#212000)

analyze() numbers the dominator tree, scans it in reverse preorder for
header candidates, floods backward through the CFG from each header's
latches, then walks the CFG again to order the blocks.

Do all of it in one forward DFS, with the algorithm of Wei et al.,
"A New Algorithm for Identifying Loops in Decompilation", which
GenericCycleInfo already implements.

On an irreducible CFG that algorithm may return an irreducible superset
of the natural loop subset `discoverAndMapSubloop` would return.
(Depending on the successor visiting order, the algorithm may report a
reducible loop nested in an irreducible loop, where the reducible one
exactly matches `discoverAndMapSubloop`.)

To satisfy verifyLoop and LoopSimplify, reduce each such ireducible loop
to the natural loop of its header's backedges instead: the latches the
header dominates, and the blocks reaching them without passing the
header.

A function whose loops are all reducible needs no dominance query.

1000x require<loops> on sqlite3.bc: instructions -24%, cycles -14%.

Aided by Claude Opus 5
https://invent.kde.org/qt/clang/llvm/-/commit/5d582e4003b1cab59900d16fa3db9c3a9d16b462

Git commit ed1db561ddd8777d31659a587a3667d14e86c5ac by Alexis Engelke on 26/07/2026 at 07:29..
[spr] initial version

Created using spr 1.3.8-wip
https://invent.kde.org/qt/clang/llvm/-/commit/ed1db561ddd8777d31659a587a3667d14e86c5ac