[Webkit-unassigned] [Bug 159385] ECMAScript 2016: %TypedArray%.prototype.includes implementation

[bugzilla-daemon at webkit.org]

https://bugs.webkit.org/show_bug.cgi?id=159385

Caio Lima <ticaiolima at gmail.com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
 Attachment #282639|Patch                       |RFC Patch
        description|                            |

--- Comment #3 from Caio Lima <ticaiolima at gmail.com> ---
Comment on attachment 282639
  --> https://bugs.webkit.org/attachment.cgi?id=282639
RFC Patch

This Patch contains the implementation of JavaScript version and also the C++ version. I've made a benchmark with this code:

let a = new Float64Array(100);

a.fill(1);

for (let i = 0; i < 100000000; i++) {
   a.includes(0);
}

The idea is reproduce the worst case, since it is going to compare with all array elements. These are the configurations of runs:

----------C++ Implementation----------
Caios-MacBook-Pro:js-tests caiolima$ JSC_useJIT=false time run-jsc includes-benchmark.js 
Running 1 time(s): DYLD_FRAMEWORK_PATH=/Users/caiolima/open_projects/WebKit/WebKitBuild/Release /Users/caiolima/open_projects/WebKit/WebKitBuild/Release/jsc includes-benchmark.js
       16.36 real        16.04 user         0.16 sys
Caios-MacBook-Pro:js-tests caiolima$ JSC_useJIT=true time run-jsc includes-benchmark.js 
Running 1 time(s): DYLD_FRAMEWORK_PATH=/Users/caiolima/open_projects/WebKit/WebKitBuild/Release /Users/caiolima/open_projects/WebKit/WebKitBuild/Release/jsc includes-benchmark.js
       13.71 real        13.40 user         0.16 sys

----------JS Wraper Implementation----------
Caios-MacBook-Pro:js-tests caiolima$ JSC_useJIT=false time run-jsc includes-benchmark.js 
Running 1 time(s): DYLD_FRAMEWORK_PATH=/Users/caiolima/open_projects/WebKit/WebKitBuild/Release /Users/caiolima/open_projects/WebKit/WebKitBuild/Release/jsc includes-benchmark.js
      345.25 real       343.33 user         1.24 sys
Caios-MacBook-Pro:js-tests caiolima$ 
Caios-MacBook-Pro:js-tests caiolima$ JSC_useJIT=true time run-jsc includes-benchmark.js 
Running 1 time(s): DYLD_FRAMEWORK_PATH=/Users/caiolima/open_projects/WebKit/WebKitBuild/Release /Users/caiolima/open_projects/WebKit/WebKitBuild/Release/jsc includes-benchmark.js
       12.54 real        12.22 user         0.23 sys

What I think is the C++ implementation has better LLInt - JIT balance, since it performs well with and without JIT. However, When the javascript wrapper is JITed,it performs better than the C++ version JITed. What do you guys think about it. Also, I am open to new Benchmark ideas.

-- 
You are receiving this mail because:
You are the assignee for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.webkit.org/pipermail/webkit-unassigned/attachments/20160702/f3d5647a/attachment.html>