PPC/s390: [wasm] TrapIf and TrapUnless TurboFan operators implemented on arm.

Port ca8d3ba7

Original Commit Message:

    Original commit message:
    [wasm] Introduce the TrapIf and TrapUnless operators to generate trap code.

    Some instructions in WebAssembly trap for some inputs, which means that the
    execution is terminated and (at least at the moment) a JavaScript exception is
    thrown. Examples for traps are out-of-bounds memory accesses, or integer
    divisions by zero.

    Without the TrapIf and TrapUnless operators trap check in WebAssembly introduces 5
    TurboFan nodes (branch, if_true, if_false, trap-reason constant, trap-position
    constant), in addition to the trap condition itself. Additionally, each
    WebAssembly function has four TurboFan nodes (merge, effect_phi, 2 phis) whose
    number of inputs is linear to the number of trap checks in the function.
    Especially for functions with high numbers of trap checks we observe a
    significant slowdown in compilation time, down to 0.22 MiB/s in the sqlite
    benchmark instead of the average of 3 MiB/s in other benchmarks. By introducing
    a TrapIf common operator only a single node is necessary per trap check, in
    addition to the trap condition. Also the nodes which are shared between trap
    checks (merge, effect_phi, 2 phis) would disappear. First measurements suggest a
    speedup of 30-50% on average.

    This CL only implements TrapIf and TrapUnless on x64. The implementation is also
    hidden behind the --wasm-trap-if flag.

    Please take a special look at how the source position is transfered from the
    instruction selector to the code generator, and at the context that is used for
    the runtime call.

R=ahaas@chromium.org, titzer@chromium.org, joransiu@ca.ibm.com, jyan@ca.ibm.com, michael_dawson@ca.ibm.com
BUG=
LOG=N

Review-Url: https://codereview.chromium.org/2666433002
Cr-Commit-Position: refs/heads/master@{#42793}

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

@@ -2018,7 +2018,79 @@ void CodeGenerator::AssembleArchJump(RpoNumber target) {
 
 void CodeGenerator::AssembleArchTrap(Instruction* instr,
                                      FlagsCondition condition) {
-  UNREACHABLE();
+  class OutOfLineTrap final : public OutOfLineCode {
+   public:
+    OutOfLineTrap(CodeGenerator* gen, bool frame_elided, Instruction* instr)
+        : OutOfLineCode(gen),
+          frame_elided_(frame_elided),
+          instr_(instr),
+          gen_(gen) {}
+
+    void Generate() final {
+      PPCOperandConverter i(gen_, instr_);
+
+      Runtime::FunctionId trap_id = static_cast<Runtime::FunctionId>(
+          i.InputInt32(instr_->InputCount() - 1));
+      bool old_has_frame = __ has_frame();
+      if (frame_elided_) {
+        __ set_has_frame(true);
+        __ EnterFrame(StackFrame::WASM_COMPILED, true);
+      }
+      GenerateCallToTrap(trap_id);
+      if (frame_elided_) {
+        __ set_has_frame(old_has_frame);
+      }
+      if (FLAG_debug_code) {
+        __ stop(GetBailoutReason(kUnexpectedReturnFromWasmTrap));
+      }
+    }
+
+   private:
+    void GenerateCallToTrap(Runtime::FunctionId trap_id) {
+      if (trap_id == Runtime::kNumFunctions) {
+        // We cannot test calls to the runtime in cctest/test-run-wasm.
+        // Therefore we emit a call to C here instead of a call to the runtime.
+        // We use the context register as the scratch register, because we do
+        // not have a context here.
+        __ PrepareCallCFunction(0, 0, cp);
+        __ CallCFunction(
+            ExternalReference::wasm_call_trap_callback_for_testing(isolate()),
+            0);
+      } else {
+        __ Move(cp, isolate()->native_context());
+        gen_->AssembleSourcePosition(instr_);
+        __ CallRuntime(trap_id);
+      }
+      ReferenceMap* reference_map =
+          new (gen_->zone()) ReferenceMap(gen_->zone());
+      gen_->RecordSafepoint(reference_map, Safepoint::kSimple, 0,
+                            Safepoint::kNoLazyDeopt);
+    }
+
+    bool frame_elided_;
+    Instruction* instr_;
+    CodeGenerator* gen_;
+  };
+  bool frame_elided = !frame_access_state()->has_frame();
+  auto ool = new (zone()) OutOfLineTrap(this, frame_elided, instr);
+  Label* tlabel = ool->entry();
+  Label end;
+
+  ArchOpcode op = instr->arch_opcode();
+  CRegister cr = cr0;
+  Condition cond = FlagsConditionToCondition(condition, op);
+  if (op == kPPC_CmpDouble) {
+    // check for unordered if necessary
+    if (cond == le) {
+      __ bunordered(&end, cr);
+      // Unnecessary for eq/lt since only FU bit will be set.
+    } else if (cond == gt) {
+      __ bunordered(tlabel, cr);
+      // Unnecessary for ne/ge since only FU bit will be set.
+    }
+  }
+  __ b(cond, tlabel, cr);
+  __ bind(&end);
 }
 
 // Assembles boolean materializations after an instruction.

src/compiler/ppc/instruction-selector-ppc.cc

@@ -155,6 +155,9 @@ void VisitBinop(InstructionSelector* selector, Node* node,
   if (cont->IsDeoptimize()) {
     selector->EmitDeoptimize(opcode, output_count, outputs, input_count, inputs,
                              cont->reason(), cont->frame_state());
+  } else if (cont->IsTrap()) {
+    inputs[input_count++] = g.UseImmediate(cont->trap_id());
+    selector->Emit(opcode, output_count, outputs, input_count, inputs);
   } else {
     selector->Emit(opcode, output_count, outputs, input_count, inputs);
   }
@@ -1535,9 +1538,12 @@ void VisitCompare(InstructionSelector* selector, InstructionCode opcode,
   } else if (cont->IsDeoptimize()) {
     selector->EmitDeoptimize(opcode, g.NoOutput(), left, right, cont->reason(),
                              cont->frame_state());
-  } else {
-    DCHECK(cont->IsSet());
+  } else if (cont->IsSet()) {
     selector->Emit(opcode, g.DefineAsRegister(cont->result()), left, right);
+  } else {
+    DCHECK(cont->IsTrap());
+    selector->Emit(opcode, g.NoOutput(), left, right,
+                   g.UseImmediate(cont->trap_id()));
   }
 }
 
@@ -1788,12 +1794,16 @@ void InstructionSelector::VisitDeoptimizeUnless(Node* node) {
 }
 
 void InstructionSelector::VisitTrapIf(Node* node, Runtime::FunctionId func_id) {
-  UNREACHABLE();
+  FlagsContinuation cont =
+      FlagsContinuation::ForTrap(kNotEqual, func_id, node->InputAt(1));
+  VisitWord32CompareZero(this, node, node->InputAt(0), &cont);
 }
 
 void InstructionSelector::VisitTrapUnless(Node* node,
                                           Runtime::FunctionId func_id) {
-  UNREACHABLE();
+  FlagsContinuation cont =
+      FlagsContinuation::ForTrap(kEqual, func_id, node->InputAt(1));
+  VisitWord32CompareZero(this, node, node->InputAt(0), &cont);
 }
 
 void InstructionSelector::VisitSwitch(Node* node, const SwitchInfo& sw) {

src/compiler/s390/code-generator-s390.cc

@@ -2261,7 +2261,78 @@ void CodeGenerator::AssembleArchJump(RpoNumber target) {
 
 void CodeGenerator::AssembleArchTrap(Instruction* instr,
                                      FlagsCondition condition) {
-  UNREACHABLE();
+  class OutOfLineTrap final : public OutOfLineCode {
+   public:
+    OutOfLineTrap(CodeGenerator* gen, bool frame_elided, Instruction* instr)
+        : OutOfLineCode(gen),
+          frame_elided_(frame_elided),
+          instr_(instr),
+          gen_(gen) {}
+
+    void Generate() final {
+      S390OperandConverter i(gen_, instr_);
+
+      Runtime::FunctionId trap_id = static_cast<Runtime::FunctionId>(
+          i.InputInt32(instr_->InputCount() - 1));
+      bool old_has_frame = __ has_frame();
+      if (frame_elided_) {
+        __ set_has_frame(true);
+        __ EnterFrame(StackFrame::WASM_COMPILED);
+      }
+      GenerateCallToTrap(trap_id);
+      if (frame_elided_) {
+        __ set_has_frame(old_has_frame);
+      }
+      if (FLAG_debug_code) {
+        __ stop(GetBailoutReason(kUnexpectedReturnFromWasmTrap));
+      }
+    }
+
+   private:
+    void GenerateCallToTrap(Runtime::FunctionId trap_id) {
+      if (trap_id == Runtime::kNumFunctions) {
+        // We cannot test calls to the runtime in cctest/test-run-wasm.
+        // Therefore we emit a call to C here instead of a call to the runtime.
+        // We use the context register as the scratch register, because we do
+        // not have a context here.
+        __ PrepareCallCFunction(0, 0, cp);
+        __ CallCFunction(
+            ExternalReference::wasm_call_trap_callback_for_testing(isolate()),
+            0);
+      } else {
+        __ Move(cp, isolate()->native_context());
+        gen_->AssembleSourcePosition(instr_);
+        __ CallRuntime(trap_id);
+      }
+      ReferenceMap* reference_map =
+          new (gen_->zone()) ReferenceMap(gen_->zone());
+      gen_->RecordSafepoint(reference_map, Safepoint::kSimple, 0,
+                            Safepoint::kNoLazyDeopt);
+    }
+
+    bool frame_elided_;
+    Instruction* instr_;
+    CodeGenerator* gen_;
+  };
+  bool frame_elided = !frame_access_state()->has_frame();
+  auto ool = new (zone()) OutOfLineTrap(this, frame_elided, instr);
+  Label* tlabel = ool->entry();
+  Label end;
+
+  ArchOpcode op = instr->arch_opcode();
+  Condition cond = FlagsConditionToCondition(condition, op);
+  if (op == kS390_CmpDouble) {
+    // check for unordered if necessary
+    if (cond == le) {
+      __ bunordered(&end);
+      // Unnecessary for eq/lt since only FU bit will be set.
+    } else if (cond == gt) {
+      __ bunordered(tlabel);
+      // Unnecessary for ne/ge since only FU bit will be set.
+    }
+  }
+  __ b(cond, tlabel);
+  __ bind(&end);
 }
 
 // Assembles boolean materializations after an instruction.

src/compiler/s390/instruction-selector-s390.cc

@@ -421,6 +421,9 @@ void VisitBinop(InstructionSelector* selector, Node* node,
   if (cont->IsDeoptimize()) {
     selector->EmitDeoptimize(opcode, output_count, outputs, input_count, inputs,
                              cont->reason(), cont->frame_state());
+  } else if (cont->IsTrap()) {
+    inputs[input_count++] = g.UseImmediate(cont->trap_id());
+    selector->Emit(opcode, output_count, outputs, input_count, inputs);
   } else {
     selector->Emit(opcode, output_count, outputs, input_count, inputs);
   }
@@ -1719,9 +1722,12 @@ void VisitCompare(InstructionSelector* selector, InstructionCode opcode,
   } else if (cont->IsDeoptimize()) {
     selector->EmitDeoptimize(opcode, g.NoOutput(), left, right, cont->reason(),
                              cont->frame_state());
-  } else {
-    DCHECK(cont->IsSet());
+  } else if (cont->IsSet()) {
     selector->Emit(opcode, g.DefineAsRegister(cont->result()), left, right);
+  } else {
+    DCHECK(cont->IsTrap());
+    selector->Emit(opcode, g.NoOutput(), left, right,
+                   g.UseImmediate(cont->trap_id()));
   }
 }
 
@@ -2002,12 +2008,16 @@ void InstructionSelector::VisitDeoptimizeUnless(Node* node) {
 }
 
 void InstructionSelector::VisitTrapIf(Node* node, Runtime::FunctionId func_id) {
-  UNREACHABLE();
+  FlagsContinuation cont =
+      FlagsContinuation::ForTrap(kNotEqual, func_id, node->InputAt(1));
+  VisitWord32CompareZero(this, node, node->InputAt(0), &cont);
 }
 
 void InstructionSelector::VisitTrapUnless(Node* node,
                                           Runtime::FunctionId func_id) {
-  UNREACHABLE();
+  FlagsContinuation cont =
+      FlagsContinuation::ForTrap(kEqual, func_id, node->InputAt(1));
+  VisitWord32CompareZero(this, node, node->InputAt(0), &cont);
 }
 
 void InstructionSelector::VisitSwitch(Node* node, const SwitchInfo& sw) {

src/compiler/wasm-compiler.cc

@@ -190,8 +190,10 @@ class WasmTrapHelper : public ZoneObject {
     }
   }
 
-#if V8_TARGET_ARCH_X64 || V8_TARGET_ARCH_IA32 || V8_TARGET_ARCH_ARM || \
-    V8_TARGET_ARCH_ARM64 || V8_TARGET_ARCH_MIPS || V8_TARGET_ARCH_MIPS64
+#if V8_TARGET_ARCH_X64 || V8_TARGET_ARCH_IA32 || V8_TARGET_ARCH_ARM ||      \
+    V8_TARGET_ARCH_ARM64 || V8_TARGET_ARCH_MIPS || V8_TARGET_ARCH_MIPS64 || \
+    V8_TARGET_ARCH_PPC || V8_TARGET_ARCH_PPC64 || V8_TARGET_ARCH_S390 ||    \
+    V8_TARGET_ARCH_S390X
 #define WASM_TRAP_IF_SUPPORTED
 #endif