[llvm] Use llvm::less_first and llvm::less_second (NFC) (#136272)

Author: kazutakahirata

llvm/lib/Target/AMDGPU/AMDGPUIGroupLP.cpp

@@ -588,12 +588,8 @@ void PipelineSolver::populateReadyList(
       ReadyList.push_back(std::pair(*I, -1));
   }
 
-  if (UseCostHeur) {
-    std::sort(ReadyList.begin(), ReadyList.end(),
-              [](std::pair<int, int> A, std::pair<int, int> B) {
-                return A.second < B.second;
-              });
-  }
+  if (UseCostHeur)
+    llvm::sort(ReadyList, llvm::less_second());
 
   assert(ReadyList.size() == CurrSU.second.size());
 }

llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp

@@ -6466,10 +6466,7 @@ static bool clusterSortPtrAccesses(ArrayRef<Value *> VL,
   for (auto &Base : Bases) {
     for (auto &Vec : Base.second) {
       if (Vec.size() > 1) {
-        stable_sort(Vec, [](const std::tuple<Value *, int, unsigned> &X,
-                            const std::tuple<Value *, int, unsigned> &Y) {
-          return std::get<1>(X) < std::get<1>(Y);
-        });
+        stable_sort(Vec, llvm::less_second());
         int InitialOffset = std::get<1>(Vec[0]);
         bool AnyConsecutive =
             all_of(enumerate(Vec), [InitialOffset](const auto &P) {
@@ -7986,9 +7983,7 @@ bool BoUpSLP::canFormVector(ArrayRef<StoreInst *> StoresVec,
   // Check if the stores are consecutive by checking if their difference is 1.
   if (StoreOffsetVec.size() != StoresVec.size())
     return false;
-  sort(StoreOffsetVec,
-       [](const std::pair<int, unsigned> &L,
-          const std::pair<int, unsigned> &R) { return L.first < R.first; });
+  sort(StoreOffsetVec, llvm::less_first());
   unsigned Idx = 0;
   int PrevDist = 0;
   for (const auto &P : StoreOffsetVec) {

llvm/tools/llvm-gpu-loader/nvptx.cpp

@@ -84,8 +84,8 @@ Expected<void *> get_ctor_dtor_array(const void *image, const size_t size,
   }
   // Lower priority constructors are run before higher ones. The reverse is true
   // for destructors.
-  llvm::sort(ctors, [](auto x, auto y) { return x.second < y.second; });
-  llvm::sort(dtors, [](auto x, auto y) { return x.second < y.second; });
+  llvm::sort(ctors, llvm::less_second());
+  llvm::sort(dtors, llvm::less_second());
 
   // Allocate host pinned memory to make these arrays visible to the GPU.
   CUdeviceptr *dev_memory = reinterpret_cast<CUdeviceptr *>(allocator(