Benchmark: (ColorfulCakeChen) CNN Block (3b6721ce94752313d64aac7b21523a601e325f18) 1505

(ColorfulCakeChen) CNN Block (3b6721ce94752313d64aac7b21523a601e325f18) 1505 (version: 1505) Benchmark.js

Comparing performance of: Async wait for init (WASM) vs (WASM) DConv_2_bias_COS_AddInputToOutput vs (WASM) P32_bias_COS_P32_bias vs (WASM) DConv_16_bias_COS_P64_bias vs (WASM) DConv_2_COS_AddInputToOutput vs (WASM) DConv_2_COS vs Async wait for init (WebGL) vs (WebGL) DConv_1_bias_COS_AddInputToOutput vs (WebGL) DConv_16_bias_COS_P64_bias vs (WebGL) P32_bias_COS_P32_bias vs (WebGL) DConv_2_bias_COS_AddInputToOutput vs (WebGL) DConv_2_COS_AddInputToOutput vs (WebGL) DConv_2_COS vs Async wait for init (CPU) vs (CPU) DConv_1_bias_COS_AddInputToOutput vs (CPU) DConv_16_bias_COS_P64_bias vs (CPU) P32_bias_COS_P32_bias vs (CPU) DConv_2_bias_COS_AddInputToOutput vs (CPU) DConv_2_COS_AddInputToOutput vs (CPU) DConv_2_COS

Created: 5 years ago by: Registered User

Go to the latest result

HTML Preparation code:

<script src="https://cdn.jsdelivr.net/npm/@tensorflow/tfjs@4.22.0/dist/tf.min.js"></script>
<script src="https://cdn.jsdelivr.net/npm/@tensorflow/tfjs-backend-wasm/dist/tf-backend-wasm.js"></script>

Script Preparation code:

globalThis.backendName_testing = undefined;
globalThis.jsPerf_Module = undefined;
globalThis.testCorrectnessIterator = undefined;

async function init( backendName ) {
  if ( !globalThis.jsPerf_Module ) {
    console.log( "import jsPerf_Block..." );
    globalThis.jsPerf_Module = await import(
      "https://cdn.jsdelivr.net/gh/colorfulcakechen/query-submit-canvas@3b6721ce94752313d64aac7b21523a601e325f18/CNN/x_jsPerf/jsPerf_Block.js" );
  }

  if ( backendName_testing != backendName ) {
    console.log( `Change backend to "${backendName}"...` );
    globalThis.backendName_testing = backendName;

    globalThis.jsPerf_Module.disposeResources();

    await tf.ready();
    await tf.setBackend( backendName );

    globalThis.jsPerf_Module.init();
  	globalThis.testCorrectnessIterator = jsPerf_Module.testCorrectness();
  }

  if ( !globalThis.testCorrectnessIterator )
    return true;

  let testCorrectnessYield = globalThis.testCorrectnessIterator.next();
  if ( testCorrectnessYield.done ) {
    console.log( `${backendName}, initialization done.` );
    globalThis.testCorrectnessIterator = null;
    return true;
  } else {
	return false;
  }
}

Tests:

Async wait for init (WASM)
if ( ( globalThis.initPromiseWASM ) && ( globalThis.initPromiseWebGL ) && ( globalThis.initPromiseCPU ) ) { // For benchmark re-run. globalThis.initPromiseWASM = globalThis.initPromiseWebGL = globalThis.initPromiseCPU = null; } //const backendName = "wasm"; //const backendName = "webgl"; const backendName = "cpu"; if ( !globalThis.initPromiseWASM ) { globalThis.initPromiseWASM = init( backendName ); } else { globalThis.initPromiseWASM.then( ( done ) => { if ( !done ) { globalThis.initPromiseWASM = init( backendName ); } else { } }); }
(WASM) DConv_2_bias_COS_AddInputToOutput
globalThis.testSet_108x192x4.testBlock_ByName( "DConv_2_bias_COS_AddInputToOutput" );
(WASM) P32_bias_COS_P32_bias
globalThis.testSet_108x192x4.testBlock_ByName( "P32_bias_COS_P32_bias" );
(WASM) DConv_16_bias_COS_P64_bias
globalThis.testSet_108x192x4.testBlock_ByName( "DConv_16_bias_COS_P64_bias" );
(WASM) DConv_2_COS_AddInputToOutput
globalThis.testSet_108x192x4.testBlock_ByName( "DConv_2_COS_AddInputToOutput" );
(WASM) DConv_2_COS
globalThis.testSet_108x192x4.testBlock_ByName( "DConv_2_COS" );
Async wait for init (WebGL)
const backendName = "webgl"; //const backendName = "cpu"; if ( !globalThis.initPromiseWebGL ) { globalThis.initPromiseWebGL = init( backendName ); } else { globalThis.initPromiseWebGL.then( ( done ) => { if ( !done ) { globalThis.initPromiseWebGL = init( backendName ); } else { } }); }
(WebGL) DConv_1_bias_COS_AddInputToOutput
globalThis.testSet_108x192x4.testBlock_ByName( "DConv_1_bias_COS_AddInputToOutput" );
(WebGL) DConv_16_bias_COS_P64_bias
globalThis.testSet_108x192x4.testBlock_ByName( "DConv_16_bias_COS_P64_bias" );
(WebGL) P32_bias_COS_P32_bias
globalThis.testSet_108x192x4.testBlock_ByName( "P32_bias_COS_P32_bias" );
(WebGL) DConv_2_bias_COS_AddInputToOutput
globalThis.testSet_108x192x4.testBlock_ByName( "DConv_2_bias_COS_AddInputToOutput" );
(WebGL) DConv_2_COS_AddInputToOutput
globalThis.testSet_108x192x4.testBlock_ByName( "DConv_2_COS_AddInputToOutput" );
(WebGL) DConv_2_COS
globalThis.testSet_108x192x4.testBlock_ByName( "DConv_2_COS" );
Async wait for init (CPU)
const backendName = "cpu"; //const backendName = "wasm"; if ( !globalThis.initPromiseCPU ) { globalThis.initPromiseCPU = init( backendName ); } else { globalThis.initPromiseCPU.then( ( done ) => { if ( !done ) { globalThis.initPromiseCPU = init( backendName ); } else { } }); }
(CPU) DConv_1_bias_COS_AddInputToOutput
globalThis.testSet_108x192x4.testBlock_ByName( "DConv_1_bias_COS_AddInputToOutput" );
(CPU) DConv_16_bias_COS_P64_bias
globalThis.testSet_108x192x4.testBlock_ByName( "DConv_16_bias_COS_P64_bias" );
(CPU) P32_bias_COS_P32_bias
globalThis.testSet_108x192x4.testBlock_ByName( "P32_bias_COS_P32_bias" );
(CPU) DConv_2_bias_COS_AddInputToOutput
globalThis.testSet_108x192x4.testBlock_ByName( "DConv_2_bias_COS_AddInputToOutput" );
(CPU) DConv_2_COS_AddInputToOutput
globalThis.testSet_108x192x4.testBlock_ByName( "DConv_2_COS_AddInputToOutput" );
(CPU) DConv_2_COS
globalThis.testSet_108x192x4.testBlock_ByName( "DConv_2_COS" );

Rendered benchmark preparation results:

Suite status: <idle, ready to run>

Previous results

Test case name	Result
Async wait for init (WASM)
(WASM) DConv_2_bias_COS_AddInputToOutput
(WASM) P32_bias_COS_P32_bias
(WASM) DConv_16_bias_COS_P64_bias
(WASM) DConv_2_COS_AddInputToOutput
(WASM) DConv_2_COS
Async wait for init (WebGL)
(WebGL) DConv_1_bias_COS_AddInputToOutput
(WebGL) DConv_16_bias_COS_P64_bias
(WebGL) P32_bias_COS_P32_bias
(WebGL) DConv_2_bias_COS_AddInputToOutput
(WebGL) DConv_2_COS_AddInputToOutput
(WebGL) DConv_2_COS
Async wait for init (CPU)
(CPU) DConv_1_bias_COS_AddInputToOutput
(CPU) DConv_16_bias_COS_P64_bias
(CPU) P32_bias_COS_P32_bias
(CPU) DConv_2_bias_COS_AddInputToOutput
(CPU) DConv_2_COS_AddInputToOutput
(CPU) DConv_2_COS

Fastest: N/A

Slowest: N/A

Latest run results:

Run details: (Test run date: one year ago)

User agent: Mozilla/5.0 (X11; CrOS x86_64 14541.0.0) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/137.0.0.0 Safari/537.36

Browser/OS: Chrome 137 on Chrome OS 14541.0.0

View result in a separate tab

Async wait for init (WebGL)

2.6M/s

Async wait for init (CPU)

1.9M/s

Async wait for init (WASM)

856K/s

(WebGL) DConv_1_bias_COS_AddInputToOutput

504/s

(WebGL) DConv_2_COS

426/s

(WebGL) DConv_2_COS_AddInputToOutput

390/s

(WebGL) DConv_16_bias_COS_P64_bias

187/s

(WebGL) DConv_2_bias_COS_AddInputToOutput

133/s

(WebGL) P32_bias_COS_P32_bias

81/s

(CPU) DConv_1_bias_COS_AddInputToOutput

6/s

(WASM) DConv_2_COS

5/s

(WASM) DConv_2_COS_AddInputToOutput

4/s

(CPU) DConv_2_COS

4/s

(CPU) DConv_2_bias_COS_AddInputToOutput

4/s

(CPU) DConv_2_COS_AddInputToOutput

4/s

(WASM) DConv_2_bias_COS_AddInputToOutput

2/s

(CPU) P32_bias_COS_P32_bias

1/s

(WASM) P32_bias_COS_P32_bias

1/s

(WASM) DConv_16_bias_COS_P64_bias

0/s

(CPU) DConv_16_bias_COS_P64_bias

0/s

View exact numbers

Test name	Executions per second
✓ Async wait for init (WebGL)	2,557,442 Ops/sec
Async wait for init (CPU)	1,865,384 Ops/sec
Async wait for init (WASM)	856,100 Ops/sec
(WebGL) DConv_1_bias_COS_AddInputToOutput	504 Ops/sec
(WebGL) DConv_2_COS	426 Ops/sec
(WebGL) DConv_2_COS_AddInputToOutput	390 Ops/sec
(WebGL) DConv_16_bias_COS_P64_bias	187 Ops/sec
(WebGL) DConv_2_bias_COS_AddInputToOutput	133 Ops/sec
(WebGL) P32_bias_COS_P32_bias	81 Ops/sec
(CPU) DConv_1_bias_COS_AddInputToOutput	6 Ops/sec
(WASM) DConv_2_COS	5 Ops/sec
(WASM) DConv_2_COS_AddInputToOutput	4 Ops/sec
(CPU) DConv_2_COS	4 Ops/sec
(CPU) DConv_2_bias_COS_AddInputToOutput	4 Ops/sec
(CPU) DConv_2_COS_AddInputToOutput	4 Ops/sec
(WASM) DConv_2_bias_COS_AddInputToOutput	2 Ops/sec
(CPU) P32_bias_COS_P32_bias	1 Ops/sec
(WASM) P32_bias_COS_P32_bias	1 Ops/sec
(WASM) DConv_16_bias_COS_P64_bias	0 Ops/sec
(CPU) DConv_16_bias_COS_P64_bias	0 Ops/sec

Autogenerated LLM Summary (model llama3.2:3b, generated one year ago):

LLMs can make mistakes. Check important info.