MIPS: [turbofan] Optimize certain chains of Branch into a Switch.

src/compiler/mips/code-generator-mips.cc

 // Copyright 2014 the V8 project authors. All rights reserved.
 // Use of this source code is governed by a BSD-style license that can be
 // found in the LICENSE file.
 
 #include "src/compiler/code-generator.h"
 #include "src/compiler/code-generator-impl.h"
 #include "src/compiler/gap-resolver.h"
 #include "src/compiler/node-matchers.h"
 #include "src/mips/macro-assembler-mips.h"
 #include "src/scopes.h"
@@ skipping 410 matching lines @@
       }
 
       __ lw(at, FieldMemOperand(func, JSFunction::kCodeEntryOffset));
       __ Call(at);
       AddSafepointAndDeopt(instr);
       break;
     }
     case kArchJmp:
       AssembleArchJump(i.InputRpo(0));
       break;
-    case kArchSwitch:
-      AssembleArchSwitch(instr);
+    case kArchLookupSwitch:
+      AssembleArchLookupSwitch(instr);
+      break;
+    case kArchTableSwitch:
+      AssembleArchTableSwitch(instr);
       break;
     case kArchNop:
       // don't emit code for nops.
       break;
     case kArchRet:
       AssembleReturn();
       break;
     case kArchStackPointer:
       __ mov(i.OutputRegister(), sp);
       break;
@@ skipping 355 matching lines @@
     UNIMPLEMENTED();
   }
 }
 
 
 void CodeGenerator::AssembleArchJump(BasicBlock::RpoNumber target) {
   if (!IsNextInAssemblyOrder(target)) __ Branch(GetLabel(target));
 }
 
 
-void CodeGenerator::AssembleArchSwitch(Instruction* instr) {
-  MipsOperandConverter i(this, instr);
-  int const kNumLabels = static_cast<int>(instr->InputCount() - 1);
-  v8::internal::Assembler::BlockTrampolinePoolScope block_trampoline_pool(
-      masm());
-  Label here;
-
-  __ bal(&here);
-  __ nop();  // Branch delay slot nop.
-  __ bind(&here);
-  __ sll(at, i.InputRegister(0), 2);
-  __ addu(at, at, ra);
-  __ lw(at, MemOperand(at, 5 * v8::internal::Assembler::kInstrSize));
-  __ jr(at);
-  __ nop();  // Branch delay slot nop.
-
-  for (int index = 0; index < kNumLabels; ++index) {
-    __ dd(GetLabel(i.InputRpo(index + 1)));
-  }
-}
-
-
 // Assembles boolean materializations after an instruction.
 void CodeGenerator::AssembleArchBoolean(Instruction* instr,
                                         FlagsCondition condition) {
   MipsOperandConverter i(this, instr);
   Label done;
 
   // Materialize a full 32-bit 1 or 0 value. The result register is always the
   // last output of the instruction.
   Label false_value;
   DCHECK_NE(0u, instr->OutputCount());
@@ skipping 57 matching lines @@
     UNIMPLEMENTED();
   }
 
   // Fallthrough case is the false materialization.
   __ bind(&false_value);
   __ li(result, Operand(0));
   __ bind(&done);
 }
 
 
+void CodeGenerator::AssembleArchLookupSwitch(Instruction* instr) {
+  MipsOperandConverter i(this, instr);
+  Register input = i.InputRegister(0);
+  for (size_t index = 2; index < instr->InputCount(); index += 2) {
+    __ Branch(GetLabel(i.InputRpo(index + 1)), eq, input,
+              Operand(i.InputInt32(index + 0)));
+  }
+  AssembleArchJump(i.InputRpo(1));
+}
+
+
+void CodeGenerator::AssembleArchTableSwitch(Instruction* instr) {
+  MipsOperandConverter i(this, instr);
+  Register input = i.InputRegister(0);
+  size_t const case_count = instr->InputCount() - 2;
+  Label here;
+  __ Branch(GetLabel(i.InputRpo(1)), hs, input, Operand(case_count));
+  __ bal(&here);

[paul.l..., 2015/02/17 19:33:59]

We need a BlockTrampolinePoolFor() call before this bal() to cover the table
size, and the instructions to index the table.

+  __ nop();  // Branch delay slot nop.
+  __ bind(&here);
+  __ sll(at, input, 2);
+  __ addu(at, at, ra);
+  __ lw(at, MemOperand(at, 5 * v8::internal::Assembler::kInstrSize));
+  __ jr(at);
+  __ nop();  // Branch delay slot nop.
+  for (size_t index = 0; index < case_count; ++index) {
+    __ dd(GetLabel(i.InputRpo(index + 2)));
+  }
+}
+
+
 void CodeGenerator::AssembleDeoptimizerCall(int deoptimization_id) {
   Address deopt_entry = Deoptimizer::GetDeoptimizationEntry(
       isolate(), deoptimization_id, Deoptimizer::LAZY);
   __ Call(deopt_entry, RelocInfo::RUNTIME_ENTRY);
 }
 
 
 void CodeGenerator::AssemblePrologue() {
   CallDescriptor* descriptor = linkage()->GetIncomingDescriptor();
   int stack_slots = frame()->GetSpillSlotCount();
@@ skipping 268 matching lines @@
     }
   }
   MarkLazyDeoptSite();
 }
 
 #undef __
 
 }  // namespace compiler
 }  // namespace internal
 }  // namespace v8