Merge branch 'dev' of github.com:strongoier/taichi into dev

Author: strongoier

taichi/codegen/codegen_llvm.cpp

@@ -1676,13 +1676,14 @@ void CodeGenLLVM::create_offload_struct_for(OffloadedStmt *stmt, bool spmd) {
     if (!prog->config.packed) {
       for (int i = 0; i < snode->num_active_indices; i++) {
         auto j = snode->physical_index_position[i];
-        if (!bit::is_power_of_two(snode->extractors[j].num_elements_from_root)) {
+        if (!bit::is_power_of_two(
+                snode->extractors[j].num_elements_from_root)) {
           auto coord = coord_object.get("val", tlctx->get_constant(j));
           exec_cond = builder->CreateAnd(
-              exec_cond,
-              builder->CreateICmp(
-                  llvm::CmpInst::ICMP_SLT, coord,
-                  tlctx->get_constant(snode->extractors[j].num_elements_from_root)));
+              exec_cond, builder->CreateICmp(
+                             llvm::CmpInst::ICMP_SLT, coord,
+                             tlctx->get_constant(
+                                 snode->extractors[j].num_elements_from_root)));
         }
       }
     }

taichi/struct/struct.cpp

@@ -14,7 +14,8 @@ void infer_snode_properties(SNode &snode, bool packed) {
     auto &ch = snode.ch[ch_id];
     ch->parent = &snode;
     for (int i = 0; i < taichi_max_num_indices; i++) {
-      ch->extractors[i].num_elements_from_root *= snode.extractors[i].num_elements_from_root;
+      ch->extractors[i].num_elements_from_root *=
+          snode.extractors[i].num_elements_from_root;
       bool found = false;
       for (int k = 0; k < taichi_max_num_indices; k++) {
         if (snode.physical_index_position[k] == i) {

taichi/transforms/lower_access.cpp

@@ -49,9 +49,11 @@ class LowerAccess : public IRVisitor {
   bool packed;
 
   LowerAccess(const std::vector<SNode *> &kernel_forces_no_activate,
-              bool lower_atomic_ptr, bool packed)
+              bool lower_atomic_ptr,
+              bool packed)
       : kernel_forces_no_activate(kernel_forces_no_activate),
-        lower_atomic_ptr(lower_atomic_ptr), packed(packed) {
+        lower_atomic_ptr(lower_atomic_ptr),
+        packed(packed) {
     // TODO: change this to false
     allow_undefined_visitor = true;
     current_struct_for = nullptr;
@@ -100,8 +102,8 @@ class LowerAccess : public IRVisitor {
       TI_ASSERT(!pointer_needs_activation);
     }
 
-    PtrLowererImpl lowerer{leaf_snode, indices, snode_op, is_bit_vectorized,
-                           lowered, packed};
+    PtrLowererImpl lowerer{leaf_snode,        indices, snode_op,
+                           is_bit_vectorized, lowered, packed};
     lowerer.set_pointer_needs_activation(pointer_needs_activation);
     lowerer.set_lower_access(this);
     lowerer.run();
@@ -305,8 +307,8 @@ namespace irpass {
 bool lower_access(IRNode *root,
                   const CompileConfig &config,
                   const LowerAccessPass::Args &args) {
-  bool modified =
-      LowerAccess::run(root, args.kernel_forces_no_activate, args.lower_atomic, config.packed);
+  bool modified = LowerAccess::run(root, args.kernel_forces_no_activate,
+                                   args.lower_atomic, config.packed);
   type_check(root, config);
   return modified;
 }

tests/cpp/transforms/scalar_pointer_lowerer_test.cpp

@@ -62,7 +62,9 @@ TEST_F(ScalarPointerLowererTest, Basic) {
       LowererImpl lowerer{leaf_snode_,
                           std::vector<Stmt *>{builder.get_int32(loop_index)},
                           SNodeOpType::undefined,
-                          /*is_bit_vectorized=*/false, &lowered, /*packed=*/false};
+                          /*is_bit_vectorized=*/false,
+                          &lowered,
+                          /*packed=*/false};
       lowerer.run();
       // There are three linearized stmts:
       // 0: for root

tests/python/test_packed_size.py

@@ -5,4 +5,4 @@
 def test_packed_size():
     x = ti.field(ti.i32)
     ti.root.dense(ti.i, 20).dense(ti.ijk, 334).place(x)
-    assert x.snode.parent().parent().cell_size_bytes == 4 * 334 ** 3
+    assert x.snode.parent().parent().cell_size_bytes == 4 * 334**3