glTest

jogl port of the apitest.

Changing Problems / Solutions

The active problem is changed by pressing the left or right arrow keys. The active solution is changed by pressing the up or down arrow keys.

first line is C, second is Java

Null Problem:

• Null Solution -> 15k fps

Dynamic Streaming:

• GLBufferSubData -> 2.1 fps

    CPU time: 444 ms, GPU time: 455 ms, theor. FPS: 2,194
  
    CPU time: 256 ms, GPU time: 458 ms, theor. FPS: 2,18
  

• GLMapUnsynchronized -> 9.0 fps

    CPU time: 115 ms, GPU time: 114 ms, theor. FPS: 8.7
  
    CPU time: 208 ms, GPU time: 206 ms, theor. FPS: 4,8
  

• GLMapPersistent -> 117 fps

    CPU time: 8 ms, GPU time: 6 ms, theor. FPS: 165
  
    CPU time: 16 ms, GPU time: 10,9 ms, theor. FPS: 91,3
  

  Direct3D11:

    D3D11UpdateSubresource -> CPU time: 407 ms, FPS: 2,06
    
    D3D11MapNoOverwrite -> CPU time: 14 ms, FPS: 66
  

Untextured Objects:

• GLUniform -> 26 fps

    CPU time: 35 ms, GPU time: 30 ms, theor. FPS: 33
  
    CPU time: 47 ms, GPU time: 44 ms, theor. FPS: 22.3
  

• GLDrawLoop -> 43 fps

    CPU time: 23 ms, GPU time: 17 ms, theor. FPS: 58
  
    CPU time: 22 ms, GPU time: 20,7 ms, theor. FPS: 48
  

• GLMultiDraw-SDP -> 66 fps

• GLMultiDraw-NoSDP -> 109 fps

• GLMultiDrawBuffer-SDP -> 63 fps

    CPU time: 21 ms, GPU time: 16 ms, theor. FPS: 61
  
    CPU time: 5.8 ms, GPU time: 15.9 ms, theor. FPS: 62.5
  

• GLMultiDrawBuffer-NoSDP -> 115 fps

    CPU time: 13 ms, GPU time: 7.6 ms, theor. FPS: 130
  
    CPU time: 6 ms, GPU time: 7.9 ms, theor. FPS: 126.5
  

• GLBindless -> 25 fps

    CPU time: 37 ms, GPU time: 32 ms, theor. FPS: 31
  
    CPU time: 162 ms, GPU time: 158 ms, theor. FPS: 6,3
  

• GLBindlessIndirect -> 33 fps

    CPU time: 40 ms, GPU time: 28.8 ms, theor. FPS: 34.5
  
    CPU time: 22 ms, GPU time: 31 ms, theor. FPS: 31,5
  

• GLBufferRange -> 15 fps

    CPU time: 60 ms, GPU time: 55 ms, theor. FPS: 18
  
    CPU time: 78,6 ms, GPU time: 76,5 ms, theor. FPS: 13,1
  

• GLBufferStorage-SDP -> 73 fps

    CPU time: 19 ms, GPU time: 13 ms, theor. FPS: 75
  
    CPU time: 5 ms, GPU time: 14,6 ms, theor. FPS: 68,4
  

• GLBufferStorage-NoSDP -> 137 fps

    CPU time: 11 ms, GPU time: 4.7 ms, theor. FPS: 211
  
    CPU time: 5,1 ms, GPU time: 6 ms, theor. FPS: 164,3
  

• GLDynamicBuffer -> 23 fps

    CPU time: 38 ms, GPU time: 33.6 ms, theor. FPS: 29.7
  
    CPU time: 75 ms, GPU time: 71,9 ms, theor. FPS: 13,9
  

• GLMapUnsynchronized -> 1.2 fps

    CPU time: 790 ms, GPU time: 790 ms, theor. FPS: 1.2
  
    CPU time: 747 ms, GPU time: 783 ms, theor. FPS: 1,28
  

• GLMapPersistent -> 48 fps

    CPU time: 20 ms, GPU time: 16 ms, theor. FPS: 62
  
    CPU time: 25,2 ms, GPU time: 22,6 ms, theor. FPS: 44,2
  

• GLTexCoord -> 48 fps

    CPU time: 27 ms, GPU time: 23 ms, theor. FPS: 43
  
    CPU time: 36,9 ms, GPU time: 34,8 ms, theor. FPS: 28,7
  

  Direct3D11:

    D3D11Naive -> CPU time: 20 ms, FPS: 49,5
  

glCalls

Micro benchmark in order to evaluate the relative weights between different GL state changes

I hit 105k framebuffer, 700k Program, 3.1M Texture Bindings, 3.4M Vertex Formats, 4.6M UBO Bindings, 6.3M Vertex Bindings and 15.2M Uniform Updates

This can be also seen as the following.

On my machine a single framebuffer switch is equal to:

• 6.67 Program state changes
• 29 Texture Bindings state changes
• 32 Vertex Formats state changes
• 43 UBO Bindings state changes
• 60 Vertex Bindings state changes
• 144 Uniform Updates state changes

Specs:

• win7 x64
• gtx 770 (353.30)
• jogl 2.3.1

Jogl is 3-5% faster compared to Lwjgl