llvm-project.git/llvm/lib/CodeGen/TailDuplicator.cpp, branch users/nico/python-1

Use early return/continue in TailDuplicator::duplicateInstruction [nfc]

2025-05-16T20:33:11+00:00

[llvm] Use range constructors of *Set (NFC) (#137552)

2025-04-27T22:59:57+00:00

[TailDuplicator] Determine if computed gotos using `blockaddress` (#132536)

2025-03-26T13:27:43+00:00

Using `blockaddress` should be more reliable than determining if an
operand comes from a jump table index.

Alternative: Add the `MachineInstr::MIFlag::ComputedGoto` flag when
lowering `indirectbr`. But I don't think this approach is suitable to
backport.

[llvm] Use *Set::insert_range (NFC) (#132509)

2025-03-22T15:07:33+00:00

DenseSet, SmallPtrSet, SmallSet, SetVector, and StringSet recently
gained C++23-style insert_range.  This patch uses insert_range in
conjunction with llvm::{predecessors,successors} and
MachineBasicBlock::{predecessors,successors}.

[llvm] Use *Set::insert_range (NFC) (#132325)

2025-03-21T05:24:06+00:00

DenseSet, SmallPtrSet, SmallSet, SetVector, and StringSet recently
gained C++23-style insert_range.  This patch replaces:

  Dest.insert(Src.begin(), Src.end());

with:

  Dest.insert_range(Src);

This patch does not touch custom begin like succ_begin for now.

[TailDuplicator] Do not restrict the computed gotos (#114990)

2025-03-10T11:34:07+00:00

Fixes #106846.

This is what I learned from GCC. I found that GCC does not duplicate the
BB that has indirect jumps with the jump table. I believe GCC has
provided a clear explanation here:

> Duplicate the blocks containing computed gotos. This basically
unfactors computed gotos that were factored early on in the compilation
process to speed up edge based data flow. We used to not unfactor them
again, which can seriously pessimize code with many computed jumps in
the source code, such as interpreters.

[TailDuplicator] Use Register. NFC

2025-03-02T20:44:26+00:00

[TailDup] Allow large number of predecessors/successors without phis. (#116072)

2025-01-23T18:24:20+00:00

This adjusts the threshold logic added in #78582 to only trigger for
cases where there are actually phis to duplicate in either TailBB or in
one of the successors.

In cases there are no phis, we only have to pay the cost of extra edges,
but have no explosion in PHI related instructions.

This improves performance of Python on some inputs by 2-3% on Apple
Silicon CPUs.

PR: https://github.com/llvm/llvm-project/pull/116072

[aarch64][win] Update Called Globals info when updating Call Site info (#122762)

2025-01-13T22:00:31+00:00

Fixes the "use after poison" issue introduced by #121516 (see
).

The root cause of this issue is that #121516 introduced "Called Global"
information for call instructions modeling how "Call Site" info is
stored in the machine function, HOWEVER it didn't copy the
copy/move/erase operations for call site information.

The fix is to rename and update the existing copy/move/erase functions
so they also take care of Called Global info.

[CodeGen] Remove unused includes (NFC) (#115996)

2024-11-13T07:15:06+00:00

Identified with misc-include-cleaner.