Skeleton of a linear scan register allocator.

runtime/vm/flow_graph_compiler_x64.cc

 // Copyright (c) 2012, the Dart project authors.  Please see the AUTHORS file
 // for details. All rights reserved. Use of this source code is governed by a
 // BSD-style license that can be found in the LICENSE file.
 
 #include "vm/globals.h"  // Needed here to get TARGET_ARCH_X64.
 #if defined(TARGET_ARCH_X64)
 
 #include "vm/flow_graph_compiler.h"
 
 #include "lib/error.h"
@@ skipping 10 matching lines @@
 DECLARE_FLAG(bool, print_ast);
 DECLARE_FLAG(bool, print_scopes);
 DECLARE_FLAG(bool, trace_functions);
 
 
 void DeoptimizationStub::GenerateCode(FlowGraphCompiler* compiler) {
   Assembler* assem = compiler->assembler();
 #define __ assem->
   __ Comment("Deopt stub for id %d", deopt_id_);
   __ Bind(entry_label());
-  for (intptr_t i = 0; i < registers_.length(); i++) {
-    if (registers_[i] != kNoRegister) {
-      __ pushq(registers_[i]);
+
+  if (deoptimization_env_ == NULL) {
+    for (intptr_t i = 0; i < registers_.length(); i++) {
+      if (registers_[i] != kNoRegister) {
+        __ pushq(registers_[i]);
+      }
+    }
+  } else {
+    // We have a deoptimization environment, we have to tear down optimized
+    // frame and recreate non-optimized one.
+    __ leaq(RSP,
+            Address(RBP, ParsedFunction::kFirstLocalSlotIndex * kWordSize));

[srdjan, 2012/07/10 23:27:54]

How do you handle incoming parameters and incoming optional arguments which are
copied? I assume the following code handles only non-parameter locals.

[Vyacheslav Egorov (Google), 2012/07/11 13:27:58]

Positional parameters are in the environment as well, so it should work.
(however regalloc will bailout because this parameters do not have definitions
in the graph; I am planning to address this once we land skeleton).

SSA building will bailout if function has optional arguments.

+
+    const ZoneGrowableArray<Value*>& values = deoptimization_env_->values();
+    const GrowableArray<Location>* locations = deoptimization_env_->locations();
+
+    for (intptr_t i = 0; i < values.length(); i++) {
+      Location loc = (*locations)[i];
+      if (loc.IsInvalid()) {
+        ASSERT(values[i]->IsConstant());
+        __ PushObject(values[i]->AsConstant()->value());
+      } else {
+        ASSERT(loc.IsRegister());
+        __ pushq(loc.reg());
+      }
     }
   }
+
   if (compiler->IsLeaf()) {
     Label L;
     __ call(&L);
     const intptr_t offset = assem->CodeSize();
     __ Bind(&L);
     __ popq(RAX);
     __ subq(RAX,
         Immediate(offset - AssemblerMacros::kOffsetOfSavedPCfromEntrypoint));
     __ movq(Address(RBP, -kWordSize), RAX);
   }
@@ skipping 1027 matching lines @@
   __ movsd(result, FieldAddress(reg, Double::value_offset()));
   __ jmp(&done);
   __ Bind(&is_smi);
   __ movq(temp, reg);
   __ SmiUntag(temp);
   __ cvtsi2sd(result, temp);
   __ Bind(&done);
 }
 
 
+ParallelMoveResolver::ParallelMoveResolver(FlowGraphCompiler* compiler)
+    : compiler_(compiler), moves_(32) {}

[srdjan, 2012/07/10 23:27:54]

Some methods should be shared, I guess.

[Vyacheslav Egorov (Google), 2012/07/11 13:27:58]

Yes this is the intent. Done when porting to ia32.

+
+
+void ParallelMoveResolver::EmitNativeCode(ParallelMoveInstr* parallel_move) {
+  ASSERT(moves_.is_empty());
+  // Build up a worklist of moves.
+  BuildInitialMoveList(parallel_move);
+
+  for (int i = 0; i < moves_.length(); ++i) {
+    MoveOperands move = moves_[i];
+    // Skip constants to perform them last.  They don't block other moves
+    // and skipping such moves with register destinations keeps those
+    // registers free for the whole algorithm.
+    if (!move.IsEliminated() && !move.src().IsConstant()) PerformMove(i);
+  }
+
+  // Perform the moves with constant sources.
+  for (int i = 0; i < moves_.length(); ++i) {
+    if (!moves_[i].IsEliminated()) {
+      ASSERT(moves_[i].src().IsConstant());
+      EmitMove(i);
+    }
+  }
+
+  moves_.Clear();
+}
+
+
+void ParallelMoveResolver::BuildInitialMoveList(
+  ParallelMoveInstr* parallel_move) {

[srdjan, 2012/07/10 23:27:54]

indent 4 spaces

[Vyacheslav Egorov (Google), 2012/07/11 13:27:58]

Done.

+  // Perform a linear sweep of the moves to add them to the initial list of
+  // moves to perform, ignoring any move that is redundant (the source is
+  // the same as the destination, the destination is ignored and
+  // unallocated, or the move was already eliminated).
+  const GrowableArray<MoveOperands>& moves = parallel_move->moves();
+  for (int i = 0; i < moves.length(); ++i) {
+    MoveOperands move = moves[i];
+    if (!move.IsRedundant()) moves_.Add(move);
+  }
+}
+
+
+void ParallelMoveResolver::PerformMove(int index) {
+  // Each call to this function performs a move and deletes it from the move
+  // graph.  We first recursively perform any move blocking this one.  We
+  // mark a move as "pending" on entry to PerformMove in order to detect
+  // cycles in the move graph.  We use operand swaps to resolve cycles,
+  // which means that a call to PerformMove could change any source operand
+  // in the move graph.
+
+  ASSERT(!moves_[index].IsPending());
+  ASSERT(!moves_[index].IsRedundant());
+
+  // Clear this move's destination to indicate a pending move.  The actual
+  // destination is saved in a stack-allocated local.  Recursion may allow
+  // multiple moves to be pending.
+  ASSERT(!moves_[index].src().IsInvalid());
+  Location destination = moves_[index].MarkPending();
+
+  // Perform a depth-first traversal of the move graph to resolve
+  // dependencies.  Any unperformed, unpending move with a source the same
+  // as this one's destination blocks this one so recursively perform all
+  // such moves.
+  for (int i = 0; i < moves_.length(); ++i) {
+    MoveOperands other_move = moves_[i];
+    if (other_move.Blocks(destination) && !other_move.IsPending()) {
+      // Though PerformMove can change any source operand in the move graph,
+      // this call cannot create a blocking move via a swap (this loop does
+      // not miss any).  Assume there is a non-blocking move with source A
+      // and this move is blocked on source B and there is a swap of A and
+      // B.  Then A and B must be involved in the same cycle (or they would
+      // not be swapped).  Since this move's destination is B and there is
+      // only a single incoming edge to an operand, this move must also be
+      // involved in the same cycle.  In that case, the blocking move will
+      // be created but will be "pending" when we return from PerformMove.
+      PerformMove(i);
+    }
+  }
+
+  // We are about to resolve this move and don't need it marked as
+  // pending, so restore its destination.
+  moves_[index].ClearPending(destination);
+
+  // This move's source may have changed due to swaps to resolve cycles and
+  // so it may now be the last move in the cycle.  If so remove it.
+  if (moves_[index].src().Equals(destination)) {
+    moves_[index].Eliminate();
+    return;
+  }
+
+  // The move may be blocked on a (at most one) pending move, in which case
+  // we have a cycle.  Search for such a blocking move and perform a swap to
+  // resolve it.
+  for (int i = 0; i < moves_.length(); ++i) {
+    MoveOperands other_move = moves_[i];
+    if (other_move.Blocks(destination)) {
+      ASSERT(other_move.IsPending());
+      EmitSwap(index);
+      return;
+    }
+  }
+
+  // This move is not blocked.
+  EmitMove(index);
+}
+
+#undef __
+
+#undef __

[srdjan, 2012/07/10 23:27:54]

One undef too many.

[Vyacheslav Egorov (Google), 2012/07/11 13:27:58]

Done.

+#define __ compiler_->assembler()->
+
+void ParallelMoveResolver::EmitMove(int index) {
+  Location source = moves_[index].src();
+  Location destination = moves_[index].dest();
+
+  ASSERT(destination.IsRegister());
+  if (source.IsRegister()) {
+     __ movq(destination.reg(), source.reg());

[srdjan, 2012/07/10 23:27:54]

indent

[Vyacheslav Egorov (Google), 2012/07/11 13:27:58]

Done.

+  } else {
+    ASSERT(source.IsConstant());
+    const Object& value = source.constant();
+    if (value.IsSmi()) {

[Vyacheslav Egorov (Google), 2012/07/11 13:27:58]

This is apparently not needed assembler does the same internally.

+      int64_t imm = reinterpret_cast<int64_t>(value.raw());
+      __ movq(destination.reg(), Immediate(imm));
+    } else {
+      __ LoadObject(destination.reg(), value);
+    }
+  }
+  moves_[index].Eliminate();
+}
+
+
+void ParallelMoveResolver::EmitSwap(int index) {
+  Location source = moves_[index].src();
+  Location destination = moves_[index].dest();
+
+  ASSERT(source.IsRegister() && destination.IsRegister());
+  __ xchgq(destination.reg(), source.reg());
+
+  // The swap of source and destination has executed a move from source to
+  // destination.
+  moves_[index].Eliminate();
+
+  // Any unperformed (including pending) move with a source of either
+  // this move's source or destination needs to have their source
+  // changed to reflect the state of affairs after the swap.
+  for (int i = 0; i < moves_.length(); ++i) {
+    MoveOperands other_move = moves_[i];
+    if (other_move.Blocks(source)) {
+      moves_[i].set_src(destination);
+    } else if (other_move.Blocks(destination)) {
+      moves_[i].set_src(source);
+    }
+  }
+}
+
+
 #undef __
 
 }  // namespace dart
 
 #endif  // defined TARGET_ARCH_X64