Skip to content

Commit

Permalink
Reland: [TypeFeedbackVector] Store optimized code in the vector
Browse files Browse the repository at this point in the history 
Since the feedback vector is itself a native context structure, why
not store optimized code for a function in there rather than in
a map from native context to code? This allows us to get rid of
the optimized code map in the SharedFunctionInfo, saving a pointer,
and making lookup of any optimized code quicker.

Original patch by Michael Stanton <mvstanton@chromium.org>

BUG=v8:6246
TBR=yangguo@chromium.org,ulan@chromium.org

Change-Id: Ic83e4011148164ef080c63215a0c77f1dfb7f327
Reviewed-on: https://chromium-review.googlesource.com/494487
Reviewed-by: Jaroslav Sevcik <jarin@chromium.org>
Commit-Queue: Ross McIlroy <rmcilroy@chromium.org>
Cr-Commit-Position: refs/heads/master@{#45084}
  • Loading branch information
@rmcilroy
rmcilroy authored and Commit Bot committed May 4, 2017
1 parent 24d7890 commit 662aa42
Show file tree
Hide file tree
Showing 97 changed files with 2,332 additions and 2,689 deletions.
81 changes: 15 additions & 66 deletions src/builtins/arm/builtins-arm.cc
View file
Original file line number Diff line number Diff line change
Expand Up @@ -1319,76 +1319,39 @@ void Builtins::Generate_CompileLazy(MacroAssembler* masm) {
// -- r1 : target function (preserved for callee)
// -----------------------------------
// First lookup code, maybe we don't need to compile!
Label gotta_call_runtime, gotta_call_runtime_no_stack;
Label gotta_call_runtime;
Label try_shared;
Label loop_top, loop_bottom;

Register argument_count = r0;
Register closure = r1;
Register new_target = r3;
Register map = argument_count;
Register index = r2;

// Do we have a valid feedback vector?
__ ldr(index, FieldMemOperand(closure, JSFunction::kFeedbackVectorOffset));
__ ldr(index, FieldMemOperand(index, Cell::kValueOffset));
__ JumpIfRoot(index, Heap::kUndefinedValueRootIndex,
&gotta_call_runtime_no_stack);

__ push(argument_count);
__ push(new_target);
__ push(closure);

__ ldr(map, FieldMemOperand(closure, JSFunction::kSharedFunctionInfoOffset));
__ ldr(map,
FieldMemOperand(map, SharedFunctionInfo::kOptimizedCodeMapOffset));
__ ldr(index, FieldMemOperand(map, FixedArray::kLengthOffset));
__ cmp(index, Operand(Smi::FromInt(2)));
__ b(lt, &try_shared);

// r3 : native context
// r2 : length / index
// r0 : optimized code map
// stack[0] : new target
// stack[4] : closure
Register native_context = r3;
__ ldr(native_context, NativeContextMemOperand());

__ bind(&loop_top);
Register temp = r1;
Register array_pointer = r5;

// Does the native context match?
__ add(array_pointer, map, Operand::PointerOffsetFromSmiKey(index));
__ ldr(temp, FieldMemOperand(array_pointer,
SharedFunctionInfo::kOffsetToPreviousContext));
__ ldr(temp, FieldMemOperand(temp, WeakCell::kValueOffset));
__ cmp(temp, native_context);
__ b(ne, &loop_bottom);
__ JumpIfRoot(index, Heap::kUndefinedValueRootIndex, &gotta_call_runtime);

// Code available?
// Is optimized code available in the feedback vector?
Register entry = r4;
__ ldr(entry,
FieldMemOperand(array_pointer,
SharedFunctionInfo::kOffsetToPreviousCachedCode));
__ ldr(entry, FieldMemOperand(
index, FeedbackVector::kOptimizedCodeIndex * kPointerSize +
FeedbackVector::kHeaderSize));
__ ldr(entry, FieldMemOperand(entry, WeakCell::kValueOffset));
__ JumpIfSmi(entry, &try_shared);

// Found code. Get it into the closure and return.
__ pop(closure);
// Store code entry in the closure.
__ add(entry, entry, Operand(Code::kHeaderSize - kHeapObjectTag));
__ str(entry, FieldMemOperand(closure, JSFunction::kCodeEntryOffset));
__ RecordWriteCodeEntryField(closure, entry, r5);

// Load native context into r6.
Register native_context = r6;
__ ldr(native_context, NativeContextMemOperand());

// Link the closure into the optimized function list.
// r4 : code entry
// r3 : native context
// r1 : closure
__ ldr(r5,
ContextMemOperand(native_context, Context::OPTIMIZED_FUNCTIONS_LIST));
__ str(r5, FieldMemOperand(closure, JSFunction::kNextFunctionLinkOffset));
__ RecordWriteField(closure, JSFunction::kNextFunctionLinkOffset, r5, r0,
__ RecordWriteField(closure, JSFunction::kNextFunctionLinkOffset, r5, r2,
kLRHasNotBeenSaved, kDontSaveFPRegs, EMIT_REMEMBERED_SET,
OMIT_SMI_CHECK);
const int function_list_offset =
Expand All @@ -1397,36 +1360,26 @@ void Builtins::Generate_CompileLazy(MacroAssembler* masm) {
ContextMemOperand(native_context, Context::OPTIMIZED_FUNCTIONS_LIST));
// Save closure before the write barrier.
__ mov(r5, closure);
__ RecordWriteContextSlot(native_context, function_list_offset, closure, r0,
__ RecordWriteContextSlot(native_context, function_list_offset, closure, r2,
kLRHasNotBeenSaved, kDontSaveFPRegs);
__ mov(closure, r5);
__ pop(new_target);
__ pop(argument_count);
__ Jump(entry);

__ bind(&loop_bottom);
__ sub(index, index, Operand(Smi::FromInt(SharedFunctionInfo::kEntryLength)));
__ cmp(index, Operand(Smi::FromInt(1)));
__ b(gt, &loop_top);

// We found no code.
// We found no optimized code.
__ bind(&try_shared);
__ pop(closure);
__ pop(new_target);
__ pop(argument_count);
__ ldr(entry,
FieldMemOperand(closure, JSFunction::kSharedFunctionInfoOffset));
// Is the shared function marked for tier up?
__ ldrb(r5, FieldMemOperand(entry,
SharedFunctionInfo::kMarkedForTierUpByteOffset));
__ tst(r5, Operand(1 << SharedFunctionInfo::kMarkedForTierUpBitWithinByte));
__ b(ne, &gotta_call_runtime_no_stack);
__ b(ne, &gotta_call_runtime);

// If SFI points to anything other than CompileLazy, install that.
__ ldr(entry, FieldMemOperand(entry, SharedFunctionInfo::kCodeOffset));
__ Move(r5, masm->CodeObject());
__ cmp(entry, r5);
__ b(eq, &gotta_call_runtime_no_stack);
__ b(eq, &gotta_call_runtime);

// Install the SFI's code entry.
__ add(entry, entry, Operand(Code::kHeaderSize - kHeapObjectTag));
Expand All @@ -1435,10 +1388,6 @@ void Builtins::Generate_CompileLazy(MacroAssembler* masm) {
__ Jump(entry);

__ bind(&gotta_call_runtime);
__ pop(closure);
__ pop(new_target);
__ pop(argument_count);
__ bind(&gotta_call_runtime_no_stack);
GenerateTailCallToReturnedCode(masm, Runtime::kCompileLazy);
}

Expand Down
52 changes: 10 additions & 42 deletions src/builtins/arm64/builtins-arm64.cc
View file
Original file line number Diff line number Diff line change
Expand Up @@ -1346,49 +1346,20 @@ void Builtins::Generate_CompileLazy(MacroAssembler* masm) {
// First lookup code, maybe we don't need to compile!
Label gotta_call_runtime;
Label try_shared;
Label loop_top, loop_bottom;

Register closure = x1;
Register map = x13;
Register index = x2;

// Do we have a valid feedback vector?
__ Ldr(index, FieldMemOperand(closure, JSFunction::kFeedbackVectorOffset));
__ Ldr(index, FieldMemOperand(index, Cell::kValueOffset));
__ JumpIfRoot(index, Heap::kUndefinedValueRootIndex, &gotta_call_runtime);

__ Ldr(map, FieldMemOperand(closure, JSFunction::kSharedFunctionInfoOffset));
__ Ldr(map,
FieldMemOperand(map, SharedFunctionInfo::kOptimizedCodeMapOffset));
__ Ldrsw(index, UntagSmiFieldMemOperand(map, FixedArray::kLengthOffset));
__ Cmp(index, Operand(2));
__ B(lt, &try_shared);

// x3 : native context
// x2 : length / index
// x13 : optimized code map
// stack[0] : new target
// stack[4] : closure
Register native_context = x4;
__ Ldr(native_context, NativeContextMemOperand());

__ Bind(&loop_top);
Register temp = x5;
Register array_pointer = x6;

// Does the native context match?
__ Add(array_pointer, map, Operand(index, LSL, kPointerSizeLog2));
__ Ldr(temp, FieldMemOperand(array_pointer,
SharedFunctionInfo::kOffsetToPreviousContext));
__ Ldr(temp, FieldMemOperand(temp, WeakCell::kValueOffset));
__ Cmp(temp, native_context);
__ B(ne, &loop_bottom);

// Code available?
// Is optimized code available in the feedback vector?
Register entry = x7;
__ Ldr(entry,
FieldMemOperand(array_pointer,
SharedFunctionInfo::kOffsetToPreviousCachedCode));
__ Ldr(entry, FieldMemOperand(
index, FeedbackVector::kOptimizedCodeIndex * kPointerSize +
FeedbackVector::kHeaderSize));
__ Ldr(entry, FieldMemOperand(entry, WeakCell::kValueOffset));
__ JumpIfSmi(entry, &try_shared);

Expand All @@ -1397,10 +1368,11 @@ void Builtins::Generate_CompileLazy(MacroAssembler* masm) {
__ Str(entry, FieldMemOperand(closure, JSFunction::kCodeEntryOffset));
__ RecordWriteCodeEntryField(closure, entry, x5);

// Load native context into x4.
Register native_context = x4;
__ Ldr(native_context, NativeContextMemOperand());

// Link the closure into the optimized function list.
// x7 : code entry
// x4 : native context
// x1 : closure
__ Ldr(x8,
ContextMemOperand(native_context, Context::OPTIMIZED_FUNCTIONS_LIST));
__ Str(x8, FieldMemOperand(closure, JSFunction::kNextFunctionLinkOffset));
Expand All @@ -1416,12 +1388,8 @@ void Builtins::Generate_CompileLazy(MacroAssembler* masm) {
kLRHasNotBeenSaved, kDontSaveFPRegs);
__ Jump(entry);

__ Bind(&loop_bottom);
__ Sub(index, index, Operand(SharedFunctionInfo::kEntryLength));
__ Cmp(index, Operand(1));
__ B(gt, &loop_top);

// We found no code.
// We found no optimized code.
Register temp = x5;
__ Bind(&try_shared);
__ Ldr(entry,
FieldMemOperand(closure, JSFunction::kSharedFunctionInfoOffset));
Expand Down
23 changes: 23 additions & 0 deletions src/builtins/builtins-constructor-gen.cc
View file
Original file line number Diff line number Diff line change
Expand Up @@ -153,6 +153,29 @@ Node* ConstructorBuiltinsAssembler::EmitFastNewClosure(Node* shared_info,

BIND(&cell_done);
}
{
// If the feedback vector has optimized code, check whether it is marked
// for deopt and, if so, clear it.
Label optimized_code_ok(this);
Node* literals = LoadObjectField(literals_cell, Cell::kValueOffset);
GotoIfNot(IsFeedbackVector(literals), &optimized_code_ok);
Node* optimized_code_cell =
LoadFixedArrayElement(literals, FeedbackVector::kOptimizedCodeIndex);
Node* optimized_code =
LoadWeakCellValue(optimized_code_cell, &optimized_code_ok);
Node* code_flags = LoadObjectField(
optimized_code, Code::kKindSpecificFlags1Offset, MachineType::Uint32());
Node* marked_for_deopt =
DecodeWord32<Code::MarkedForDeoptimizationField>(code_flags);
GotoIf(Word32Equal(marked_for_deopt, Int32Constant(0)), &optimized_code_ok);

// Code is marked for deopt, clear the optimized code slot.
StoreFixedArrayElement(literals, FeedbackVector::kOptimizedCodeIndex,
EmptyWeakCellConstant(), SKIP_WRITE_BARRIER);
Goto(&optimized_code_ok);

BIND(&optimized_code_ok);
}
StoreObjectFieldNoWriteBarrier(result, JSFunction::kFeedbackVectorOffset,
literals_cell);
StoreObjectFieldNoWriteBarrier(
Expand Down
72 changes: 15 additions & 57 deletions src/builtins/ia32/builtins-ia32.cc
View file
Original file line number Diff line number Diff line change
Expand Up @@ -1101,9 +1101,8 @@ void Builtins::Generate_CompileLazy(MacroAssembler* masm) {
// -- edi : target function (preserved for callee)
// -----------------------------------
// First lookup code, maybe we don't need to compile!
Label gotta_call_runtime, gotta_call_runtime_no_stack;
Label gotta_call_runtime;
Label try_shared;
Label loop_top, loop_bottom;

Register closure = edi;
Register new_target = edx;
Expand All @@ -1112,57 +1111,28 @@ void Builtins::Generate_CompileLazy(MacroAssembler* masm) {
// Do we have a valid feedback vector?
__ mov(ebx, FieldOperand(closure, JSFunction::kFeedbackVectorOffset));
__ mov(ebx, FieldOperand(ebx, Cell::kValueOffset));
__ JumpIfRoot(ebx, Heap::kUndefinedValueRootIndex,
&gotta_call_runtime_no_stack);
__ JumpIfRoot(ebx, Heap::kUndefinedValueRootIndex, &gotta_call_runtime);

__ push(argument_count);
__ push(new_target);
__ push(closure);

Register map = argument_count;
Register index = ebx;
__ mov(map, FieldOperand(closure, JSFunction::kSharedFunctionInfoOffset));
__ mov(map, FieldOperand(map, SharedFunctionInfo::kOptimizedCodeMapOffset));
__ mov(index, FieldOperand(map, FixedArray::kLengthOffset));
__ cmp(index, Immediate(Smi::FromInt(2)));
__ j(less, &try_shared);

// edx : native context
// ebx : length / index
// eax : optimized code map
// stack[0] : new target
// stack[4] : closure
Register native_context = edx;
__ mov(native_context, NativeContextOperand());

__ bind(&loop_top);
Register temp = edi;

// Does the native context match?
__ mov(temp, FieldOperand(map, index, times_half_pointer_size,
SharedFunctionInfo::kOffsetToPreviousContext));
__ mov(temp, FieldOperand(temp, WeakCell::kValueOffset));
__ cmp(temp, native_context);
__ j(not_equal, &loop_bottom);

// Code available?
// Is optimized code available in the feedback vector?
Register entry = ecx;
__ mov(entry, FieldOperand(map, index, times_half_pointer_size,
SharedFunctionInfo::kOffsetToPreviousCachedCode));
__ mov(entry,
FieldOperand(ebx, FeedbackVector::kOptimizedCodeIndex * kPointerSize +
FeedbackVector::kHeaderSize));
__ mov(entry, FieldOperand(entry, WeakCell::kValueOffset));
__ JumpIfSmi(entry, &try_shared);

// Found code. Get it into the closure and return.
__ pop(closure);
// Store code entry in the closure.
__ lea(entry, FieldOperand(entry, Code::kHeaderSize));
__ mov(FieldOperand(closure, JSFunction::kCodeEntryOffset), entry);
__ push(argument_count);
__ push(new_target);
__ RecordWriteCodeEntryField(closure, entry, eax);

// Load native context into edx.
Register native_context = edx;
__ mov(native_context, NativeContextOperand());

// Link the closure into the optimized function list.
// ecx : code entry
// edx : native context
// edi : closure
__ mov(ebx,
ContextOperand(native_context, Context::OPTIMIZED_FUNCTIONS_LIST));
__ mov(FieldOperand(closure, JSFunction::kNextFunctionLinkOffset), ebx);
Expand All @@ -1181,27 +1151,19 @@ void Builtins::Generate_CompileLazy(MacroAssembler* masm) {
__ pop(argument_count);
__ jmp(entry);

__ bind(&loop_bottom);
__ sub(index, Immediate(Smi::FromInt(SharedFunctionInfo::kEntryLength)));
__ cmp(index, Immediate(Smi::FromInt(1)));
__ j(greater, &loop_top);

// We found no code.
// We found no optimized code.
__ bind(&try_shared);
__ pop(closure);
__ pop(new_target);
__ pop(argument_count);
__ mov(entry, FieldOperand(closure, JSFunction::kSharedFunctionInfoOffset));
// Is the shared function marked for tier up?
__ test_b(FieldOperand(entry, SharedFunctionInfo::kMarkedForTierUpByteOffset),
Immediate(1 << SharedFunctionInfo::kMarkedForTierUpBitWithinByte));
__ j(not_zero, &gotta_call_runtime_no_stack);
__ j(not_zero, &gotta_call_runtime);

// If SFI points to anything other than CompileLazy, install that.
__ mov(entry, FieldOperand(entry, SharedFunctionInfo::kCodeOffset));
__ Move(ebx, masm->CodeObject());
__ cmp(entry, ebx);
__ j(equal, &gotta_call_runtime_no_stack);
__ j(equal, &gotta_call_runtime);

// Install the SFI's code entry.
__ lea(entry, FieldOperand(entry, Code::kHeaderSize));
Expand All @@ -1210,10 +1172,6 @@ void Builtins::Generate_CompileLazy(MacroAssembler* masm) {
__ jmp(entry);

__ bind(&gotta_call_runtime);
__ pop(closure);
__ pop(new_target);
__ pop(argument_count);
__ bind(&gotta_call_runtime_no_stack);

GenerateTailCallToReturnedCode(masm, Runtime::kCompileLazy);
}
Expand Down
Loading

0 comments on commit 662aa42

Please sign in to comment.