Official Build 1583
Mozilla/5.0 (Windows; U; Windows NT 5.1; en-US) AppleWebKit/525.13 (KHTML, like Gecko) Chrome/0.2.149.27 Safari/525.13

* Optimizing Dynamically-Dispatched Calls with Run-Time Type Feedback Proceedings of ACM SIGPLAN `94: Describes the basic principles of type-feedback. This paper is for the Self system, not Strongtalk, but the basic technology is the same.

* Adaptive optimization for Self: Reconciling High Performance with Exploratory Programming Ph.D. thesis, Computer Science Department, Stanford University. This is even more detail on type-feedback for Self

* Optimizing Dynamically-Typed Object-Oriented Programming Languages with Polymorphic Inline Caches ECOOP '91 Conference Proceedings

for(var i = 0; i < 10; ++i) { ... }

The main thing to understand is that Psyco is useful for handling blocks that loop many times, and it knows how to optimize operations involving integers and floating point numbers. For non-looping functions, and for operations on other types of objects, Psyco mostly just adds overhead for its analysis and internal-compilation. Moreover, for applications with large numbers of functions and classes, enabling Psyco application-wide adds a large burden in machine-code compilation and memory-usage for this caching.

.. where the script initially took about 8.5 seconds to run, after Psyco's "speedup" it ran in about 12 seconds.

We've used components from Apple's WebKit and Mozilla's Firefox, among others

Its real goal, embedded in Chrome's open-source code, is to upgrade the capabilities of all browsers so that they can better
support (and eventually disappear behind) the applications.

--// jit-test.lua

local t = {count = 0}

for i=1,1E7 do
    t.count = t.count + 1
end

print(t.count)

bash$ time lua jit-test.lua
10000000

real    0m1.748s
user    0m0.106s
sys    0m0.091s

bash$ time luajit jit-test.lua
10000000

real    0m0.818s
user    0m0.061s
sys    0m0.061s

// jit-test.neko 
var t = {count => 0};

var N = 0;
while (N < 10000000) {
  N += 1
  t.count += 1 
}

$print(t.count)

$ nekoc jit-test.neko

$ time neko jit-test.n
10000000
real    0m0.656s
user    0m0.031s
sys     0m0.000s

// jit-test.tis

var t = {count:0};
for(var i = 0; i < 10000000; ++i){
   t.count += 1;
}

stdout.printf("%d\n",t.count)

$ time ./tis jit-test.tis
10000000

real    0m7.508s
user    0m0.031s
sys     0m0.015s

+----------+-------+--------+
|          |  Jit  | NonJit |
+----------+-------+--------+
| Lua      | 0.82  | 1.75   |
+----------+-------+--------+
| Neko     | 0.66  |   --   |
+----------+-------+--------+
| TIScript |   --  | 7.50   |
+----------+-------+--------+

for(var px in Image.pixels)
{
  px.red *= 0.7;
  px.green *= 0.7;
  px.blue *= 0.7;
}

— каждая закладка будет этаким отдельным браузером. Если что-то сломается, то упавшая закладка не повлечет за собой паение всего браузера

— браузер будет включать в себя Chrome Process Browser, который сможет указать, что именно тормозит/потребляет слишком много ресурсов и позволт это убить

for( var i = 0; i < 10; ++i )
  count += 1;

  var o = {};
  Object.prototype.count = 1;
  
  o.count += 1;
  o["count"] += 1;

  alert(o.count);

Z>--// jit-test.lua

Z>local t = {count = 0}

Z>for i=1,1E7 do
Z>    t.count = t.count + 1
Z>end

Z>print(t.count)
Z>

Z>// jit-test.tis

Z>var t = {count:0};
Z>for(var i = 0; i < 10000000; ++i){
Z>   t.count += 1;
Z>}

Z>stdout.printf("%d\n",t.count)
Z>

Z>$ time ./tis jit-test.tis
Z>10000000

Не эквивалентны в том смысле что local t (LUA) в JS записывается следующим образом:

[code]
{
  var t = {count:0};
  for(var i = 0; i < 10000000; ++i){
    t.count += 1;
  }
  stdout.printf("%d\n",t.count);
}

$global.t = {count:0};
for(var i = 0; i < 10000000; ++i){
   $global.t.count += 1;
}
stdout.printf("%d\n",t.count);

bash$ cat jit-test.tis
// jit-test.tis -- TScript

var t = {count:0};
for(var i = 0; i < 10000000; ++i){
   t.count += 1;
}

stdout.printf("%d\n",t.count)
bash$ time tiscript jit-test.tis
10000000

real    0m7.177s
user    0m0.031s
sys    0m0.000s

var i:int = 0;
i += 1;

Z>bash$ cat jit-test.tis
Z>// jit-test.tis -- TScript

Z>var t = {count:0};
Z>for(var i = 0; i < 10000000; ++i){
Z>   t.count += 1;
Z>}

Z>stdout.printf("%d\n",t.count)
Z>bash$ time tiscript jit-test.tis
Z>10000000

Z>real    0m7.177s
Z>user    0m0.031s
Z>sys    0m0.000s
Z>

var t;

{ var t; }

Z>>bash$ cat jit-test.tis
Z>>// jit-test.tis -- TScript

Z>>var t = {count:0};
Z>>for(var i = 0; i < 10000000; ++i){
Z>>   t.count += 1;
Z>>}

Z>>stdout.printf("%d\n",t.count)
Z>>bash$ time tiscript jit-test.tis
Z>>10000000

Z>>real    0m7.177s
Z>>user    0m0.031s
Z>>sys    0m0.000s
Z>>

CS>var t;
CS>

CS>{ var t; }
CS>

bash$ cat jit-test.tis
// jit-test.tis -- TScript
{
    var t = {count:0};
    for(var i = 0; i < 10000000; ++i){
        t.count += 1;
    }

    stdout.printf("%d\n",t.count);
}

bash$ time tiscript jit-test.tis
10000000

real    [b]0m6.056s[]
user    0m0.015s
sys    0m0.031s
bash$

local t = {count = 0}
for i=1,1E7 do
    t.count = t.count + 1
end
print(t.count)

{
  var t = {count:0};
  for(var i=0; i < 10000000; ++i)
    ++t.count;
  stdout.printf("%d\n",t.count);    
}

The type system is both optional and incremental, and operates completely independently of the compiler technology (which means that normal untyped Smalltalk code runs just as fast as typed code).

(Стр. 32)
* Run-time semantics are independent of type system
* Type annotations are optional
(Стр. 35)
Strongtalk
* An optional type system for Smalltalk
* Fastest Smalltalk ever, but does not rely on types for performance

This level of indirection allows for pre and post operations relative to the abstract operations both immediately and with future unforeseen changes. The NVI pattern can be deployed with very little software production and runtime cost. Many commercial frameworks employ the NVI pattern.

allows for pre and post operations relative to the abstract operations both immediately and with future unforeseen changes