[bug] add test case and clean debug info

taichi/analysis/gen_offline_cache_key.cpp

@@ -532,7 +532,6 @@ class ASTSerializer : public IRVisitor, public ExpressionVisitor {
       emit(static_cast<std::size_t>(snode->get_snode_tree_id()));
       emit(static_cast<std::size_t>(snode->id));
       const auto *root = snode->get_root();
-      // snode_tree_roots_.insert(root);
       snode_tree_roots_.push_back(root);
     } else {
       emit(std::numeric_limits<std::size_t>::max());
@@ -655,8 +654,6 @@ class ASTSerializer : public IRVisitor, public ExpressionVisitor {
 #undef DEFINE_EMIT_ENUM
 
   std::ostream *os_{nullptr};
-  // std::unordered_set<const SNode *> snode_tree_roots_;
-  // std::unordered_map<Function *, std::size_t> real_funcs_;
   std::vector<const SNode *> snode_tree_roots_;
   std::map<Function *, std::size_t> real_funcs_;
   std::vector<char> string_pool_;

taichi/analysis/offline_cache_util.cpp

@@ -184,27 +184,6 @@ std::string get_hashed_offline_cache_key(const CompileConfig &config,
   auto device_caps_key = get_offline_cache_key_of_device_caps(caps);
   std::string autodiff_mode =
       std::to_string(static_cast<std::size_t>(kernel->autodiff_mode));
-  // std::cout << "=============== kernel name " << kernel->name <<
-  // "==================="<< std::endl; for (auto elem : kernel->parameter_list)
-  // {
-  //   std::cout << "kernel->parameter_list "<< elem << std::endl;
-  // }
-  // for (auto elem : kernel_rets_string) {
-  //   std::cout << "kernel_rets_string "<< elem << std::endl;
-  // }
-  // for (auto elem : compile_config_key) {
-  //   std::cout << "compile_config_key "<< elem << std::endl;
-  // }
-  // for (auto elem : device_caps_key) {
-  //   std::cout << "device_caps_key "<< elem << std::endl;
-  // }
-  // for (auto elem : kernel_params_string) {
-  //   std::cout << "kernel_params_string "<< elem << std::endl;
-  // }
-  // for (auto elem : kernel_rets_string) {
-  //   std::cout << "kernel_rets_string "<< elem << std::endl;
-  // }
-  // std::cout << "kernel_body_string "<< kernel_body_string << std::endl;
   picosha2::hash256_one_by_one hasher;
   hasher.process(compile_config_key.begin(), compile_config_key.end());
   hasher.process(device_caps_key.begin(), device_caps_key.end());

taichi/compilation_manager/kernel_compilation_manager.cpp

@@ -214,8 +214,6 @@ const CompiledKernelData *KernelCompilationManager::try_load_cached_kernel(
     auto &kernels = cached_data_.kernels;
     auto iter = kernels.find(kernel_key);
 
-    TI_INFO("!!!!!KERNEL KEY {}, in kernels, is exist {}", kernel_key,
-            iter != kernels.end());
     if (iter != kernels.end()) {
       auto &k = iter->second;
       if (k.compiled_kernel_data) {

tests/python/test_offline_cache.py

@@ -384,6 +384,70 @@ def helper():
     assert added_files() == expected_num_cache_files(len(simple_kernels_to_test))
 
 
+@pytest.mark.parametrize("curr_arch", supported_archs_offline_cache)
+@_test_offline_cache_dec
+def test_offline_cache_with_different_snode_trees(curr_arch):
+    count_of_cache_file = cache_files_cnt()
+
+    def added_files():
+        return cache_files_cnt() - count_of_cache_file
+
+    def helper():
+        x = ti.field(float, shape=5)
+        @ti.kernel
+        def trigger_compile():
+            x[0] += 1
+
+        # This case is used for testing SNodeTree storeing order matters (i.e., use a ordered container such as vector instead of unordered_map or unordered_set) when generating kernel offline cache key
+        # The multiple `trigger_compile` equalivant to allocate each field to a different SNodeTree
+        # i.e., 
+        # x = ti.field(float)
+        # fb.dense(ti.i, 5).place(x)
+        # fb.finalize()
+
+        trigger_compile()
+        a = ti.field(float, shape=5)
+        trigger_compile()
+        b = ti.field(float, shape=10)
+        trigger_compile()
+        c = ti.field(float, shape=5)
+        trigger_compile()
+        d = ti.field(float, shape=10)
+        trigger_compile()
+        e = ti.field(float, shape=5)
+        trigger_compile()
+        f = ti.field(float, shape=10)
+        trigger_compile()
+        g = ti.field(float, shape=5)
+        trigger_compile()
+        h = ti.field(float, shape=10)
+
+
+        @ti.kernel
+        def kernel_forward():
+            for i in range(5):
+                a[i] += i
+                b[i] += i
+                c[i] += i
+                d[i] += i
+                e[i] += i
+                f[i] += i
+                g[i] += i
+                h[i] += i
+
+        kernel_forward()
+
+    ti.init(arch=curr_arch, enable_fallback=False, **current_thread_ext_options())
+    helper()
+    assert added_files() == expected_num_cache_files(2)
+
+    # The number of cache file should not change
+    for _ in range(5):
+        ti.init(arch=curr_arch, enable_fallback=False, **current_thread_ext_options())
+        helper()
+        assert added_files() == expected_num_cache_files(2)
+
+
 @pytest.mark.parametrize("curr_arch", supported_archs_offline_cache)
 @_test_offline_cache_dec
 def test_offline_cache_with_changing_compile_config(curr_arch):