diff options
| author | Kazu Hirata <kazu@google.com> | 2025-03-27 20:34:23 -0700 |
|---|---|---|
| committer | GitHub <noreply@github.com> | 2025-03-27 20:34:23 -0700 |
| commit | c9197b27b484713c312c16cad12f9b518c1323c5 (patch) | |
| tree | 61bc78547b9170cef3518757705193879fecf8b1 /llvm/lib/CodeGen/CodeGenPrepare.cpp | |
| parent | d055e58334a91dcbaee22eb87bcdae85a1f33cd4 (diff) | |
| download | llvm-c9197b27b484713c312c16cad12f9b518c1323c5.zip llvm-c9197b27b484713c312c16cad12f9b518c1323c5.tar.gz llvm-c9197b27b484713c312c16cad12f9b518c1323c5.tar.bz2 | |
[AMDGPU] Use MapVector instead of DenseMap (NFC) (#133356)
This patch combines:
DenseMap<MachineBasicBlock *, bool> ReachableMap;
SmallVector<MachineBasicBlock *, 4> ReachableOrdered;
into:
MapVector<MachineBasicBlock *, bool> ReachableMap;
because we add elements to the two data structures in lockstep, and we
care about preserving the insertion order.
As a side benefit, we get to avoid hash lookups at:
ReachableMap[MBB] = true;
Diffstat (limited to 'llvm/lib/CodeGen/CodeGenPrepare.cpp')
0 files changed, 0 insertions, 0 deletions