lower pow(x,1) to x

csrc/codegen.cpp

@@ -990,32 +990,38 @@ class CudaKernelGenerator : private kir::ConstIrVisitor {
       return false;
     }
 
-    // Only **2 and **3 are considered
-    if (!(exponent == 2 || exponent == 3)) {
+    // Only **1, **2 and **3 are considered
+    if (!(exponent == 1 || exponent == 2 || exponent == 3)) {
       return false;
     }
 
     auto lhs = gen(bop->lhs());
 
     if (print_inline_) {
-      code_ << lhs << " * " << lhs;
-      if (exponent == 3) {
-        code_ << " * " << lhs;
+      for (int i = 0; i < exponent; ++i) {
+        if (i != 0) {
+          code_ << " * ";
+        }
+        code_ << lhs;
       }
     } else {
       indent() << gen(bop->out());
       if (bop->out()->isScalar()) {
-        code_ << " = " << lhs << " * " << lhs;
-        if (exponent == 3) {
-          code_ << " * " << lhs;
+        for (int i = 0; i < exponent; ++i) {
+          if (i == 0) {
+            code_ << " = " << lhs;
+          } else {
+            code_ << " * " << lhs;
+          }
         }
       } else {
-        code_ << "\n";
-        indent() << kTab << "= " << lhs << "\n";
-        indent() << kTab << "* " << lhs;
-        if (exponent == 3) {
-          code_ << "\n";
-          indent() << kTab << "* " << lhs;
+        for (int i = 0; i < exponent; ++i) {
+          if (i == 0) {
+            code_ << "\n";
+            indent() << kTab << "= " << lhs;
+          } else {
+            indent() << "\n" << kTab << "* " << lhs;
+          }
         }
       }
     }

tests/cpp/test_gpu3.cpp

@@ -9413,6 +9413,65 @@ TEST_F(NVFuserTest, RegisteredExactMappingWithExtentReplacment) {
   }
 }
 
+
+TEST_F(NVFuserTest, RegisteredExactMappingWithExtentReplacment) {
+  Fusion fusion;
+  FusionGuard fg(&fusion);
+
+  auto tv0 = makeConcreteTensor({16, 32});
+  fusion.addInput(tv0);
+  auto tv1 = makeSymbolicTensor(2);
+  fusion.addInput(tv1);
+  auto tv2 = makeSymbolicTensor(1);
+  fusion.addInput(tv2);
+
+  auto tv3 = set(tv2);
+  auto tv4 = broadcast(tv3, {false, true});
+  auto tv5 = add(tv1, tv4);
+  auto tv6 = add(tv0, tv5);
+  fusion.addOutput(tv6);
+
+  // Make the loop domains of tv3 and tv4 exact mapped with tv1's loop
+  // domain
+  scheduler_tools::scheduleLoopDomainsLike({tv3, tv4}, tv1->getLoopDomain());
+
+  EXPECT_TRUE(fusion.hasRegisteredExactMappings());
+
+  // tv3 and tv4 should have new cloned IDs that are exact mapped with
+  // tv1
+  auto registered_mappings = fusion.registeredExactMappings();
+  auto registered_mappings_it = registered_mappings.find(tv3->axis(1));
+  EXPECT_NE(registered_mappings_it, registered_mappings.end());
+  const auto& registered_ids = registered_mappings_it->second;
+  EXPECT_TRUE(registered_ids->has(tv4->axis(1)));
+  EXPECT_TRUE(registered_ids->has(tv1->axis(1)));
+
+  {
+    IdModel id_model(&fusion, /*build_graphs=*/false);
+    const auto& exact_graph = id_model.buildExactGraph();
+    for (auto tv : {tv3, tv4}) {
+      EXPECT_EQ(
+          exact_graph.toGroups(tv->getLoopDomain()),
+          exact_graph.toGroups(tv1->getLoopDomain()));
+    }
+  }
+
+  // tv0 and tv1 are exact mapped. Since tv0 has static extents,
+  // replaceSymbolicSizes will replace the symbolic extents of tv1 and tv2
+  // with the static extents of tv0.
+  replaceSymbolicSizes(&fusion);
+
+  // Check if the exact mapping is still alive
+  {
+    IdModel id_model(&fusion, /*build_graphs=*/false);
+    const auto& exact_graph = id_model.buildExactGraph();
+    for (auto tv : {tv3, tv4}) {
+      EXPECT_EQ(
+          exact_graph.toGroups(tv->getLoopDomain()),
+          exact_graph.toGroups(tv1->getLoopDomain()));
+    }
+  }
+}
 // Test file size should be up to 10K LoC. Create a new file for more tests.
 
 } // namespace nvfuser