fix: fix typos

Author: XuehaiPan

src/layout/gemm_layouts.cc

@@ -588,7 +588,7 @@ Layout makeGemmVoltaABLayout(int stride, int continuous, bool is_a,
 
 // ref:
 // https://github.com/nvidia/cutlass/blob/ad7b2f5e84fcfa124cb02b91d5bd26d238c0459e/include/cutlass/layout/tensor_op_multiplicand_sm75.h#L54
-// Althought the four settings (T or NT) used distinct layouts in CUTLASS, they
+// Although the four settings (T or NT) used distinct layouts in CUTLASS, they
 // appeared to result in the same mem layout
 Layout makeTensorOpMultiplicand(int mat_stride, int mat_continuous,
                                 int elementsize, int crosswise) {

src/op/parallel.cc

@@ -215,9 +215,9 @@ LayoutMap ParallelOpNode::InferLayout(const LayoutInferArgs &T,
     return {};
   if (level == InferLevel::kStrict) {
     LayoutMap results;
-    // Deduce buffers that shoule be complicated replicated.
+    // Deduce buffers that should be complicated replicated.
     // For example:
-    // for i in T.Parllel(m):
+    // for i in T.Parallel(m):
     //   fragment[0] = x[i]
     // then fragment[0] must be replicated on all threads.
     for (const auto &[buffer, indices] : indice_map_) {

src/target/codegen_cuda.cc

@@ -2088,7 +2088,7 @@ void CodeGenTileLangCUDA::VisitExpr_(const BufferLoadNode *op,
   DataType element_dtype = op->buffer->dtype;
 
   int lanes = op->dtype.lanes();
-  // delcare type.
+  // declare type.
   if (value_dtype.lanes() == element_dtype.lanes()) {
     std::string ref = GetBufferRef(op->dtype, op->buffer.get(), index);
     HandleVolatileLoads(ref, op, os);

src/target/ptx.h

@@ -150,7 +150,7 @@ std::string PrintArriveBarrierAsm(const std::string &barrier);
  * \brief Print ptx barrier arrival with expect tx operation using
  * mbarrier.arrive.expect_tx \param barrier: The name of the barrier in shared
  * memory. \param byte_count: Increases the tx count of the mbarrier object to
- * track completion of addtional async transactions.
+ * track completion of additional async transactions.
  */
 std::string PrintArriveBarrierExpectTxAsm(const std::string &barrier,
                                           const std::string &byte_count);

src/transform/inject_assumes.cc

@@ -33,8 +33,8 @@ class AssumeInjector : public tvm::tir::StmtExprMutator {
     };
     tvm::StructuralHash sh;
     tvm::StructuralEqual se;
-    // grouped by expr, since the amount of varidic shape symbols is usualy much
-    // smaller than buffer
+    // grouped by expr, since the amount of variadic shape symbols is usually
+    // much smaller than buffer
     std::vector<Item> items;
     // hash => index in items
     std::unordered_map<size_t, std::vector<size_t>> buckets;

src/transform/loop_vectorize_dynamic.cc

@@ -243,9 +243,9 @@ class VectorizedBodyMutator : public StmtExprMutator {
   std::vector<PrimExpr> conditions_;
 };
 
-class VectorizedConditionExtracter : public StmtExprVisitor {
+class VectorizedConditionExtractor : public StmtExprVisitor {
 public:
-  VectorizedConditionExtracter() = default;
+  VectorizedConditionExtractor() = default;
   std::vector<PrimExpr> GetConditions(const Stmt &body) {
     this->VisitStmt(body);
     return conditions_;
@@ -268,6 +268,9 @@ class VectorizedConditionExtracter : public StmtExprVisitor {
   std::vector<PrimExpr> conditions_;
 };
 
+// backward-compatibility: extracter -> extractor
+using VectorizedConditionExtracter = VectorizedConditionExtractor;
+
 class NestedLoopChecker : public StmtExprVisitor {
 public:
   NestedLoopChecker() : loop_num_(0) {}
@@ -391,8 +394,8 @@ class VectorizeRewriterDynamic : public StmtExprMutator {
     vmap.Set(fnode->loop_var, outer_var * vector_size_ + inner_var);
     Stmt body = Substitute(fnode->body, vmap);
 
-    VectorizedConditionExtracter extracter;
-    std::vector<PrimExpr> conditions = extracter.GetConditions(body);
+    VectorizedConditionExtractor extractor;
+    std::vector<PrimExpr> conditions = extractor.GetConditions(body);
 
     VectorizedConditionMutator condition_mutator(inner_var, vector_size_);