[LV] Use SmallVector::resize instead of push_back/emplace_back in a loop. NFC

[topperc]

This should be more efficient since the vector can know how much additional space to reserve before creating the new elements.

[llvmbot]

@llvm/pr-subscribers-llvm-transforms

Author: Craig Topper (topperc)

Changes

This should be more efficient since the vector can know how much additional space to reserve before creating the new elements.

---

Full diff: https://github.com/llvm/llvm-project/pull/83696.diff

1 Files Affected:

• (modified) llvm/lib/Transforms/Vectorize/VPlan.h (+4-4)

diff --git a/llvm/lib/Transforms/Vectorize/VPlan.h b/llvm/lib/Transforms/Vectorize/VPlan.h
index 16c09a83e777dd..bc499400bced84 100644
--- a/llvm/lib/Transforms/Vectorize/VPlan.h
+++ b/llvm/lib/Transforms/Vectorize/VPlan.h
@@ -311,12 +311,12 @@ struct VPTransformState {
   void set(VPValue *Def, Value *V, const VPIteration &Instance) {
     auto Iter = Data.PerPartScalars.insert({Def, {}});
     auto &PerPartVec = Iter.first->second;
-    while (PerPartVec.size() <= Instance.Part)
-      PerPartVec.emplace_back();
+    if (PerPartVec.size() <= Instance.Part)
+      PerPartVec.resize(Instance.Part + 1);
     auto &Scalars = PerPartVec[Instance.Part];
     unsigned CacheIdx = Instance.Lane.mapToCacheIndex(VF);
-    while (Scalars.size() <= CacheIdx)
-      Scalars.push_back(nullptr);
+    if (Scalars.size() <= CacheIdx)
+      Scalars.resize(CacheIdx + 1);
     assert(!Scalars[CacheIdx] && "should overwrite existing value");
     Scalars[CacheIdx] = V;
   }

[fhahn]

LGTM, thanks!