Thursday, October 5, 2017

Optimizing ES2015 proxies in V8


Proxies have been an integral part of JavaScript since ES2015. They allow intercepting fundamental operations on objects and customizing their behavior. Proxies form a core part of projects like jsdom and the Comlink RPC library. Recently, we put a lot of effort into improving the performance of proxies in V8. This article sheds some light on general performance improvement patterns in V8 and for proxies in particular.

Proxies are “objects used to define custom behavior for fundamental operations (e.g. property lookup, assignment, enumeration, function invocation, etc.)” (definition by MDN). More info can be found in the full specification. For example, the following code snippet adds logging to every property access on the object:

const target = {};
const callTracer = new Proxy(target, {
  get: (target, name, receiver) => {
    console.log(`get was called for: ${name}`);
    return target[name];
}); = 'value';
// get was called for: property
// value

Constructing proxies

The first feature we'll focus on is the construction of proxies. Our original C++ implementation here followed the EcmaScript specification step-by-step, resulting in at least 4 jumps between the C++ and JS runtimes as shown in the following figure. We wanted to port this implementation into the platform-agnostic CodeStubAssembler (CSA), which is executed in the JS runtime as opposed to the C++ runtime.This porting minimizes that number of jumps between the language runtimes. CEntryStub and JSEntryStub represent the runtimes in the figure below. The dotted lines represent the borders between the JS and C++ runtimes. Luckily, lots of helper predicates were already implemented in the assembler, which made the initial version concise and readable.

The figure below shows the execution flow for calling a Proxy with any proxy trap (in this example apply, which is being called when the proxy is used as a function) generated by the following sample code:

function foo(...) {...}
g = new Proxy({...}, {
  apply: foo
g(1, 2);

After porting the trap execution to CSA all of the execution happens in the JS runtime, reducing the number of jumps between languages from 4 to 0.

This change resulted in the following performance improvements::

Our JS performance score shows an improvement between 49% and 74%. This score roughly measures how many times the given microbenchmark can be executed in 1000ms. For some tests the code is run multiple times in order to get an accurate enough measurement given the timer resolution. The code for all of the following benchmarks can be found in our js-perf-test directory.

Call and construct traps

The next section shows the results from optimizing call and construct traps (a.k.a. "apply" and "construct").

The performance improvements when calling proxies are significant — up to 500% faster! Still, the improvement for proxy construction is quite modest, especially in cases where no actual trap is defined — only about 25% gain. We investigated this by running the following command with the d8 shell:

$ out/x64.release/d8 --runtime-call-stats test.js
> run: 120.104000

                      Runtime Function/C++ Builtin        Time             Count
                                         NewObject     59.16ms  48.47%    100000  24.94%
                                      JS_Execution     23.83ms  19.53%         1   0.00%
                              RecompileSynchronous     11.68ms   9.57%        20   0.00%
                        AccessorNameGetterCallback     10.86ms   8.90%    100000  24.94%
      AccessorNameGetterCallback_FunctionPrototype      5.79ms   4.74%    100000  24.94%
                                  Map_SetPrototype      4.46ms   3.65%    100203  25.00%


Where test.js's source is:

function MyClass() {}
MyClass.prototype = {};
const P = new Proxy(MyClass, {});
function run() {
  return new P();
const N = 1e5;
for (let i = 0; i < N; ++i) {

It turned out most of the time is spent in NewObject and the functions called by it, so we started planning how to speed this up in future releases.

Get trap

The next section describes how we optimized the other most common operations — getting and setting properties through proxies. It turned out the get trap is more involved than the previous cases, due to the specific behavior of V8's inline cache. For a detailed explanation of inline caches, you can watch this talk.

Eventually we managed to get a working port to CSA with the following results:

After landing the change, we noticed the size of the Android .apk for Chrome had grown by ~160KB, which is more than expected for a helper function of roughly 20 lines, but fortunately we track such statistics. It turned out this function is called twice from another function, which is called 3 times, from another called 4 times. The cause of the problem turned out to be the aggressive inlining. Eventually we solved the issue by turning the inline function into a separate code stub, thus saving precious KBs - the end version had only ~19KB increase in .apk size.

Has trap

The next section shows the results from optimizing the has trap. Although at first we thought it would be easier (and reuse most of the code of the get trap), it turned out to have its own peculiarities. A particularly hard-to-track-down problem was the prototype chain walking when calling the in operator. The improvement results achieved vary between 71% and 428%. Again the gain is more prominent in cases where the trap is present.

Set trap

The next section talks about porting the set trap. This time we had to differentiate between named and indexed properties (elements). These two main types are not part of the JS language, but are essential for V8's efficient property storage. The initial implementation still bailed out to the runtime for elements, which causes crossing the language boundaries again. Nevertheless we achieved improvements between 27% and 438% for cases when the trap is set, at the cost of a decrease of up to 23% when it's not. This performance regression is due to the overhead of additional check for differentiating between indexed and named properties. For indexed properties, there is no improvement yet. Here are the complete results:

Real-world usage

Results from jsdom-proxy-benchmark

The jsdom-proxy-benchmark project compiles the ECMAScript specification using the Ecmarkup tool. As of v11.2.0, the jsdom project (which underlies Ecmarkup) uses proxies to implement the common data structures NodeList and HTMLCollection. We used this benchmark to get an overview of some more realistic usage than the synthetic micro-benchmarks, and achieved the following results, average of 100 runs:

  • Node v8.4.0 (without Proxy optimizations): 14277 ± 159 ms
  • Node v9.0.0-v8-canary-20170924 (with only half of the traps ported): 11789 ± 308 ms
  • Gain in speed around 2.4 seconds which is ~17% better

(thanks for the results provided by TimothyGu)

Results from Chai.js

Chai.js is a popular assertion library which makes heavy use of proxies. We've created a kind of real-world benchmark by running its tests with different versions of V8 an improvement of roughly 1s out of more than 4s, average of 100 runs:

Optimization approach

We often tackle performance issues using a generic optimization scheme. The main approach that we followed for this particular work included the following steps:

  • Implement performance tests for the particular sub-feature
  • Add more specification conformance tests (or write them from scratch)
  • Investigate the original C++ implementation
  • Port the sub-feature to the platform-agnostic CodeStubAssembler
  • Optimize the code even further by hand-crafting a TurboFan implementation
  • Measure the performance improvement.

This approach can be applied to any general optimization task that you may have.

Written by Maya Lekova (@MayaLekova), Optimizer of Proxies

Wednesday, October 4, 2017

An Internship on Laziness: Lazy Unlinking of Deoptimized Functions

Roughly three months ago, I joined the V8 team (Google Munich) as an intern and since then I’ve been working on the VM’s Deoptimizer — something completely new to me which proved to be an interesting and challenging project. The first part of my internship focused on improving the VM security-wise. The second part focused on performance improvements. Namely, on the removal of a data-structure used for the unlinking of previously deoptimized functions, which was a performance bottleneck during garbage collection. This blog post describes this second part of my internship. I’ll explain how V8 used to unlink deoptimized functions, how we changed this, and what performance improvements were obtained.

Let’s (very) briefly recap the V8 pipeline for a JavaScript function: V8’s interpreter, Ignition, collects profiling information about that function while interpreting it. Once the function becomes hot, this information is passed to V8’s compiler, TurboFan, which generates optimized machine code. When the profiling information is no longer valid — for example because one of the profiled objects gets a different type during runtime — the optimized machine code might become invalid. In that case, V8 needs to deoptimize it.

Source: JavaScript Start-up Performance

Upon optimization, TurboFan generates a code object, i.e. the optimized machine code, for the function under optimization. When this function is invoked the next time, V8 follows the link to optimized code for that function and executes it. Upon deoptimization of this function, we need to unlink the code object in order to make sure that it won’t be executed again. How does that happen?

For example, in the following code, the function f1 will be invoked many times (always passing an integer as argument). TurboFan then generates machine code for that specific case.

function g() {
  return (i) => i;

// Create a closure.
const f1 = g();
// Optimize f1.
for (var i = 0; i < 1000; i++) f1(0);

Each function also has a trampoline to the interpreter — more details in these slides — and will keep a pointer to this trampoline in its SharedFunctionInfo (SFI). This trampoline will be used whenever V8 needs to go back to unoptimized code. Thus, upon deoptimization, triggered by passing an argument of a different type, for example, the Deoptimizer can simply set the code field of the JavaScript function to this trampoline.

Although this seems simple, it forces V8 to keep weak lists of optimized JavaScript functions. This is because it is possible to have different functions pointing to the same optimized code object. We can extend our example as follows, and the functions f1 and f2 both point to the same optimized code.

const f2 = g();

If the function f1 is deoptimized (for example by invoking it with an object of different type {x: 0}) we need to make sure that the invalidated code will not be executed again by invoking f2.

Thus, upon deoptimization, V8 used to iterate over all the optimized JavaScript functions, and would unlink those that pointed to the code object being deoptimized. This iteration in applications with many optimized JavaScript functions became a performance bottleneck. Moreover, other than slowing down deoptimization, V8 used to iterate over these lists upon stop-the-world cycles of garbage collection, making it even worse.

In order to have an idea of the impact of such data-structure in the performance of V8, we wrote a micro-benchmark that stresses its usage, by triggering many scavenge cycles after creating many JavaScript functions.

function g() {
  return (i) => i + 1;

// Create an initial closure and optimize.
var f = g();


// Create 2M closures, those will get the previously optimized code.
var a = [];
for (var i = 0; i < 2000000; i++) {
  var h = g();

// Now cause scavenges, all of them are slow.
for (var i = 0; i < 1000; i++) {
  new Array(50000);

When running this benchmark, we could observe that V8 spent around 98% of its execution time on garbage collection. We then removed this data structure, and instead used an approach for lazy unlinking, and this was what we observed on x64:

Although this is just a micro-benchmark that creates many JavaScript functions and triggers many garbage collection cycles, it gives us an idea of the overhead introduced by this data structure. Other more realistic applications where we saw some overhead, and which motivated this work, were the router benchmark implemented in Node.js and ARES-6 benchmark suite.

Lazy unlinking

Rather than unlinking optimized code from JavaScript functions upon deoptimization, V8 postpones it for the next invocation of such functions. When such functions are invoked, V8 checks whether they have been deoptimized, unlinks them and then continues with their lazy compilation. If these functions are never invoked again, then they will never be unlinked and the deoptimized code objects will not be collected. However, given that during deoptimization, we invalidate all the embedded fields of the code object, we only keep that code object alive.

The commit that removed this list of optimized JavaScript functions required changes in several parts of the VM, but the basic idea is as follows. When assembling the optimized code object, we check if this is the code of a JavaScript function. If so, in its prologue, we assemble machine code to bail out if the code object has been deoptimized. Upon deoptimization we don’t modify the deoptimized code — code patching is gone. Thus, its bit marked_for_deoptimization is still set when invoking the function again. TurboFan generates code to check it, and if it is set, then V8 jumps to a new builtin, CompileLazyDeoptimizedCode, that unlinks the deoptimized code from the JavaScript function and then continues with lazy compilation.

In more detail, the first step is to generate instructions that load the address of the code being currently assembled. We can do that in x64, with the following code:

Label current;
// Load effective address of current instruction into rcx.
__ leaq(rcx, Operand(&current));
__ bind(&current);

After that we need to obtain where in the code object the marked_for_deoptimization bit lives.

int pc = __ pc_offset();
int offset = Code::kKindSpecificFlags1Offset - (Code::kHeaderSize + pc);

We can then test the bit and if it is set, we jump to the CompileLazyDeoptimizedCode built in.

// Test if the bit is set, that is, if the code is marked for deoptimization.
__ testl(Operand(rcx, offset),
         Immediate(1 << Code::kMarkedForDeoptimizationBit));
// Jump to builtin if it is.
__ j(not_zero, /* handle to builtin code here */, RelocInfo::CODE_TARGET);

On the side of this CompileLazyDeoptimizedCode builtin, all that’s left to do is to unlink the code field from the JavaScript function and set it to the trampoline to the Interpreter entry. So, considering that the address of the JavaScript function is in the register rdi, we can obtain the pointer to the SharedFunctionInfo with:

// Field read to obtain the SharedFunctionInfo.
__ movq(rcx, FieldOperand(rdi, JSFunction::kSharedFunctionInfoOffset));

…and similarly the trampoline with:

// Field read to obtain the code object.
__ movq(rcx, FieldOperand(rcx, SharedFunctionInfo::kCodeOffset));

Then we can use it to update the function slot for the code pointer:

// Update the code field of the function with the trampoline.
__ movq(FieldOperand(rdi, JSFunction::kCodeOffset), rcx);
// Write barrier to protect the field.
__ RecordWriteField(rdi, JSFunction::kCodeOffset, rcx, r15,
                    kDontSaveFPRegs, OMIT_REMEMBERED_SET, OMIT_SMI_CHECK);

This produces the same result as before. However, rather than taking care of the unlinking in the Deoptimizer, we need to worry about it during code generation. Hence the handwritten assembly.

The above is how it works in the x64 architecture. We have implemented it for ia32, arm, arm64, mips, and mips64 as well.

This new technique is already integrated in V8 and, as we’ll discuss later, allows for performance improvements. However, it comes with a minor disadvantage: Before, V8 would consider unlinking only upon deoptimization. Now, it has to do so in the activation of all optimized functions. Moreover, the approach to check the marked_for_deoptimization bit is not as efficient as it could be, given that we need to do some work to obtain the address of the code object. Note that this happens when entering every optimized function. A possible solution for this issue is to keep in a code object a pointer to itself. Rather than doing work to find the address of the code object whenever the function is invoked, V8 would do it only once, after its construction.


We now look at the performance gains and regressions obtained with this project.

General Improvements on x64

The following plot shows us some improvements and regressions, relative to the previous commit. Note that the higher, the better.

The promises benchmarks are the ones where we see greater improvements, observing almost 33% gain for the bluebird-parallel benchmark, and 22.40% for wikipedia. We also observed a few regressions in some benchmarks. This is related to the issue explained above, on checking whether the code is marked for deoptimization.

We also see improvements in the ARES-6 benchmark suite. Note that in this chart too, the higher the better. These programs used to spend considerable amount of time in GC-related activities. With lazy unlinking we improve performance by 1.9% overall. The most notable case is the Air steadyState where we get an improvement of around 5.36%.

AreWeFastYet results

The performance results for the Octane and ARES-6 benchmark suites also showed up on the AreWeFastYet tracker. We looked at these performance results on September 5th, 2017, using the provided default machine (macOS 10.10 64-bit, Mac Pro, shell).

Impact on Node.js

We can also see performance improvements in the router-benchmark. The following two plots show the number of operations per second of each tested router. Thus the higher the better. We have performed two kinds of experiments with this benchmark suite. Firstly, we ran each test in isolation, so that we could see the performance improvement, independently from the remaining tests. Secondly, we ran all tests at once, without switching of the VM, thus simulating an environment where each test is integrated with other functionalities.

For the first experiment, we saw that the router and express tests perform about twice as many operations than before, in the same amount of time. For the second experiment, we saw even greater improvement. In some of the cases, such as routr, server-router and router, the benchmark performs approximately 3.80×, 3× and 2× more operations, respectively. This happens because V8 accumulates more optimized JavaScript functions, test after test. Thus, whenever executing a given test, if a garbage collection cycle is triggered, V8 has to visit the optimized functions from the current test and from the previous ones.

Further Optimization

Now that V8 does not keep the linked-list of JavaScript functions in the context, we can remove the field next from the JSFunction class. Although this is a simple modification, it allows us to save the size of a pointer per function, which represent significant savings in several web pages:

Benchmark Kind Memory savings (absolute) Memory savings (relative) Average effective size 170KB 3.7% Average size of allocated objects 284KB 1.2% Average size of allocated objects 788KB 1.53% Average size of allocated objects 129KB 0.79%


Throughout my internship, I had lots of help from several people, who were always available to answer my many questions. Thus I would like to thank the following people: Benedikt Meurer, Jaroslav Sevcik, and Michael Starzinger for discussions on how the Compiler and the Deoptimizer work, Ulan Degenbaev for helping with the Garbage Collector whenever I broke it, and Mathias Bynens, Peter Marshall, Camillo Bruni, and Maya Lekova for proofreading this article.

Finally, this article is my last contribution as a Google intern and I would like to take the opportunity to thank everyone in the V8 team, and especially my host, Benedikt Meurer, for hosting me and for giving me the opportunity to work on such an interesting project — I definitely learned a lot and enjoyed my time at Google!

Juliana Franco, @jupvfranco, Laziness Expert