I have 2 patch of particles I want to run together on 5 screens and 1 projector

Have you ever seen this kind of message ?

That happens when I load your 2 particles patchs together.

The weird thing is that it works when I set the jit.window to cover 3 of 4 monitors.

Below the messages I get on the Max Console

I'm not familiar with that and don't know how to repare that.

OSC-route ("OpenSoundControl route") object version 1.17.2 by Matt Wright, Michael Zbyszynski.

jit.gl.fbo: unsupported framebuffer format!

jit.gl.fbo: error binding fbo: GL Error: Out of memory

jit.gl.fbo: error unbinding fbo: GL Error: Stack underflow

jit.qt.movie: could not create temporary GWorld for open()

I have 2 patch of particles I want to run together on 5 screens and 1 projector
for an exhibition and I have problems.

Have you ever seen this kind of message ?
That happens when I load your 2 particles patchs together.

Below the messages I get on the Max Console
Doest it seem like a memory leak ?

I am on Max 6.1.9 and MacPro
Yosemite 10.10.2
Processor 3.5 GHz 6-Core Intel Xeon E5
Memory 16 GB 1866 MHz DDR3 ECC
Graphic card AMD FirePro D700 6144 MB

max-crash-2-particles-patch-with-5-monitors

I was going to say it looks like your gfx card is running out of memory but then I was your system specs...

Under menu > options > opengl status, you can change opengl readback modes. Do the others work better? Is the AMD Firepro listed as renderer? I don't think your system has integrated gfx so it should be but check to be sure.

Wow ! I just tried RTT and "ouf" it doesn't crash.

Where do I list AMD Firepro as renderer? I dont see that.

Wow ! I just tried RTT and "ouf" it doesn't crash.
What I have listed is FBO, RTT and  CTT

Where do I list AMD Firepro as renderer? I dont see that.


"ATI Radeon HD - FirePro D700 OpenGL Engine" : {

"MIN_PBUFFER_VIEWPORT_DIMS_APPLE" : [ 32, 32 ],

"ALIASED_POINT_SIZE_RANGE" : [ 1, 8191 ],

"SMOOTH_LINE_WIDTH_GRANULARITY" : [ 0.125 ],

"SMOOTH_POINT_SIZE_GRANULARITY" : [ 0.125 ],

"MIN_PBUFFER_VIEWPORT_DIMS_APPLE" : [ 1, 1 ],

"SMOOTH_LINE_WIDTH_GRANULARITY" : [ 0.0001 ],

Oh yes it is liste with RTT
Here's the infos

{
"ATI Radeon HD - FirePro D700 OpenGL Engine" : {
"Vendor Name" : "ATI Technologies Inc.",
"Version" : "2.1 ATI-1.30.5",
"GLSL Version" : "1.20",
"Renderer Properties" : {
"OffScreen" : 0,
"FullScreen" : 1,
"RendererID" : 16915457,
"Accelerated" : 1,
"Robust" : 0,
"BackingStore" : 1,
"Safe" : 1,
"Window" : 1,
"MultiScreen" : 0,
"Compliant" : 1,
"DisplayMask" : 127,
"BufferModes" : 31,
"ColorModes" : 176456704,
"AccumModes" : 134217728,
"DepthModes" : 7169,
"StencilModes" : 129,
"MaxAuxBuffers" : 2,
"MaxSampleBuffers" : 1,
"MaxSamples" : 8,
"SampleModes" : 3,
"SampleAlpha" : 1,
"VideoMemory" : 2147483647,
"TextureMemory" : 2147483647,
"GPUVertProcCapable" : 1,
"GPUFragProcCapable" : 1,
"Online" : 1
}
,
"OpenGL Limits" : {
"Framebuffers" : {
"MAX_COLOR_ATTACHMENTS_EXT" : [ 8 ],
"MAX_RENDERBUFFER_SIZE_EXT" : [ 16384 ],
"MAX_VIEWPORT_DIMS" : [ 16384, 16384 ],
"MAX_DRAW_BUFFERS_ARB" : [ 8 ],
"MIN_PBUFFER_VIEWPORT_DIMS_APPLE" : [ 32, 32 ],
"SUBPIXEL_BITS" : [ 8 ]
}
,
"Points and Lines" : {
"ALIASED_LINE_WIDTH_RANGE" : [ 1, 64 ],
"ALIASED_POINT_SIZE_RANGE" : [ 1, 8191 ],
"SMOOTH_LINE_WIDTH_GRANULARITY" : [ 0.125 ],
"SMOOTH_LINE_WIDTH_RANGE" : [ 1, 64 ],
"SMOOTH_POINT_SIZE_GRANULARITY" : [ 0.125 ],
"SMOOTH_POINT_SIZE_RANGE" : [ 1, 64 ]
}
,
"Stacks" : {
"MAX_ATTRIB_STACK_DEPTH" : [ 16 ],
"MAX_CLIENT_ATTRIB_STACK_DEPTH" : [ 16 ],
"MAX_COLOR_MATRIX_STACK_DEPTH" : [ 10 ],
"MAX_MODELVIEW_STACK_DEPTH" : [ 32 ],
"MAX_NAME_STACK_DEPTH" : [ 128 ],
"MAX_PROGRAM_MATRIX_STACK_DEPTH_ARB" : [ 2 ],
"MAX_PROJECTION_STACK_DEPTH" : [ 10 ],
"MAX_TEXTURE_STACK_DEPTH" : [ 10 ]
}
,
"Textures" : {
"MAX_3D_TEXTURE_SIZE" : [ 16384 ],
"MAX_CUBE_MAP_TEXTURE_SIZE" : [ 16384 ],
"MAX_RECTANGLE_TEXTURE_SIZE_EXT" : [ 16384 ],
"MAX_TEXTURE_SIZE" : [ 16384 ],
"MAX_TEXTURE_LOD_BIAS" : [ 16 ],
"MAX_TEXTURE_MAX_ANISOTROPY_EXT" : [ 16 ],
"MAX_TEXTURE_UNITS" : [ 8 ]
}
,
"Shaders" : {
"MAX_COMBINED_TEXTURE_IMAGE_UNITS_ARB" : [ 16 ],
"MAX_FRAGMENT_UNIFORM_COMPONENTS_ARB" : [ 4096 ],
"MAX_TEXTURE_COORDS_ARB" : [ 8 ],
"MAX_TEXTURE_IMAGE_UNITS_ARB" : [ 16 ],
"MAX_VERTEX_ATTRIBS_ARB" : [ 16 ],
"MAX_VERTEX_TEXTURE_IMAGE_UNITS_ARB" : [ 16 ],
"MAX_VERTEX_UNIFORM_COMPONENTS_ARB" : [ 4096 ]
}

}
,
"Apple Software Renderer" : {
"Vendor Name" : "Apple Inc.",
"Version" : "2.1 APPLE-11.1.1",
"GLSL Version" : "1.20",
"Renderer Properties" : {
"OffScreen" : 1,
"FullScreen" : 0,
"RendererID" : 16909312,
"Accelerated" : 0,
"Robust" : 1,
"BackingStore" : 1,
"Safe" : 1,
"Window" : 1,
"MultiScreen" : 1,
"Compliant" : 1,
"DisplayMask" : 255,
"BufferModes" : 13,
"ColorModes" : 134250496,
"AccumModes" : 134217728,
"DepthModes" : 4097,
"StencilModes" : 129,
"MaxAuxBuffers" : 4,
"MaxSampleBuffers" : 1,
"MaxSamples" : 16,
"SampleModes" : 1,
"SampleAlpha" : 1,
"VideoMemory" : 0,
"TextureMemory" : 0,
"GPUVertProcCapable" : 0,
"GPUFragProcCapable" : 0,
"Online" : 1
}
,
"OpenGL Limits" : {
"Framebuffers" : {
"MAX_COLOR_ATTACHMENTS_EXT" : [ 8 ],
"MAX_RENDERBUFFER_SIZE_EXT" : [ 16384 ],
"MAX_VIEWPORT_DIMS" : [ 16384, 16384 ],
"MAX_DRAW_BUFFERS_ARB" : [ 8 ],
"MIN_PBUFFER_VIEWPORT_DIMS_APPLE" : [ 1, 1 ],
"SUBPIXEL_BITS" : [ 10 ]
}
,
"Points and Lines" : {
"ALIASED_LINE_WIDTH_RANGE" : [ 1, 64 ],
"ALIASED_POINT_SIZE_RANGE" : [ 1, 8191 ],
"SMOOTH_LINE_WIDTH_GRANULARITY" : [ 0.0001 ],
"SMOOTH_LINE_WIDTH_RANGE" : [ 0, 64 ],
"SMOOTH_POINT_SIZE_GRANULARITY" : [ 0.0001 ],
"SMOOTH_POINT_SIZE_RANGE" : [ 0, 256 ]
}
,
"Stacks" : {
"MAX_ATTRIB_STACK_DEPTH" : [ 16 ],
"MAX_CLIENT_ATTRIB_STACK_DEPTH" : [ 16 ],
"MAX_COLOR_MATRIX_STACK_DEPTH" : [ 10 ],
"MAX_MODELVIEW_STACK_DEPTH" : [ 32 ],
"MAX_NAME_STACK_DEPTH" : [ 128 ],
"MAX_PROGRAM_MATRIX_STACK_DEPTH_ARB" : [ 2 ],
"MAX_PROJECTION_STACK_DEPTH" : [ 10 ],
"MAX_TEXTURE_STACK_DEPTH" : [ 10 ]
}
,
"Textures" : {
"MAX_3D_TEXTURE_SIZE" : [ 16384 ],
"MAX_CUBE_MAP_TEXTURE_SIZE" : [ 16384 ],
"MAX_RECTANGLE_TEXTURE_SIZE_EXT" : [ 16384 ],
"MAX_TEXTURE_SIZE" : [ 16384 ],
"MAX_TEXTURE_LOD_BIAS" : [ 16 ],
"MAX_TEXTURE_MAX_ANISOTROPY_EXT" : [ 16 ],
"MAX_TEXTURE_UNITS" : [ 8 ]
}
,
"Shaders" : {
"MAX_COMBINED_TEXTURE_IMAGE_UNITS_ARB" : [ 16 ],
"MAX_FRAGMENT_UNIFORM_COMPONENTS_ARB" : [ 4096 ],
"MAX_TEXTURE_COORDS_ARB" : [ 8 ],
"MAX_TEXTURE_IMAGE_UNITS_ARB" : [ 16 ],
"MAX_VERTEX_ATTRIBS_ARB" : [ 16 ],
"MAX_VERTEX_TEXTURE_IMAGE_UNITS_ARB" : [ 16 ],
"MAX_VERTEX_UNIFORM_COMPONENTS_ARB" : [ 4096 ]
}

i'm not too well versed in the readback mode intricacies but a search for opengl + these modes should shed some light.

Now the resolution of the texture 1920 1080 on my patch doesn't extend to 7680 1080.

The coordinates are not applied properly and the patch doesn't respond well.

Now the resolution of the texture 1920 1080 on my patch doesn't extend to 7680 1080.
The coordinates are not applied properly and the patch doesn't respond well.


not much i can say without seeing a patch

I understand. I just can't send it ... I have to keep this work confidential at the moment.

Strip out the secret stuff and just keep the basic rendering and window. This will also tell you if it's a conflict between your hardware and Jitter openGl rendering or rather something funny in your content generation.

are you using a single GPU or multiple GPUs?

Well I dont know if that is an answer but we have 3 max running.

At the moment I'm only testing with the Kinect on one max and a second Max for the particles patches.

Otherwise I have no idea how to set things to different GPUs

When I resize the jit.window to cover 3 screens everything works (well almost but no crash of jit.gl.fbo

Here's what my programmer did for the purpose of testing (after we want to create

https://cycling74.com/forums/launching-and-monitoring-multiple-max-app-instances-with-lunch/

Well I dont know if that is an answer but we have 3 max running.
At the moment I'm only testing with the Kinect on one max and a second Max for the particles patches. 

Here's what my programmer did for the purpose of testing (after we want to create
standalones).

https://cycling74.com/forums/launching-and-monitoring-multiple-max-app-instances-with-lunch/


- one Kinect Patch with OpenNi on Max#1 (it works)

- 2 particles patch mixing with a 4 videos patch on Max#2 and displaying on 4 HD monitors

- another 2 particles patch mixing with video on Max#3 and displaying on one HD monitor

I can not move on Max7 even 7.2 because the video reader and test I did are playing at 16 fps but on Max6 showing an fps of 110.

Anyway ... It is difficult for me as well to know exactly how to branch things (with the proper parameters ... transform_reset isn't acting as planned with fbo or rtt) and mix from patches ending to videoplanes and or gl.nodes in regards to the jit.window.

- one Kinect Patch with OpenNi on Max#1 (it works)
- 2 particles patch mixing with a 4 videos patch on Max#2 and displaying on 4 HD monitors
- another 2 particles patch mixing with video on Max#3 and displaying on one HD monitor

Anyway ... It is difficult for me as well to know exactly how to branch things (with the proper parameters ... transform_reset isn't acting as planned with fbo or rtt) and mix from patches ending to videoplanes and or gl.nodes in regards to the jit.window.


Rob: it seems to me that we can see only a single GPU from within Max. Here are our specs as reported from the System Information app. (on OS X 10.10) All the monitors seem to be attached to the seconds GPU. So far, I think that these two GPUs are linked in SLI, hence they share their resources. (VRAM, monitors, etc.) It's as if they were a single GPU.

Resolution:    1920 x 1080 @ 60Hz (1080p)

Pixel Depth:    32-Bit Color (ARGB8888)

Display Serial Number:    C8E00958SL0

Adapter Type:    Apple Mini DisplayPort To Dual-Link DVI Adapter

Display Serial Number:    C8E00651SL0

Connection Type:    Analog VGA or Analog Over DVI-I

Display Serial Number:    C8E01066SL0

Hello everyone,
Rob: it seems to me that we can see only a single GPU from within Max. Here are our specs as reported from the System Information app. (on OS X 10.10) All the monitors seem to be attached to the seconds GPU. So far, I think that these two GPUs are linked in SLI, hence they share their resources. (VRAM, monitors, etc.) It's as if they were a single GPU.

  Chipset Model:	AMD FirePro D700
  Type:	GPU
  Bus:	PCIe
  Slot:	Slot-1
  PCIe Lane Width:	x16
  VRAM (Total):	6144 MB
  Vendor:	ATI (0x1002)
  Device ID:	0x6798
  Revision ID:	0x0000
  ROM Revision:	113-C3861J-687
  gMux Version:	4.0.11 [3.2.8]
  EFI Driver Version:	01.00.687

  Chipset Model:	AMD FirePro D700
  Type:	GPU
  Bus:	PCIe
  Slot:	Slot-2
  PCIe Lane Width:	x16
  VRAM (Total):	6144 MB
  Vendor:	ATI (0x1002)
  Device ID:	0x6798
  Revision ID:	0x0000
  ROM Revision:	113-C3861J-687
  gMux Version:	4.0.11 [3.2.8]
  EFI Driver Version:	01.00.687
  Displays:
BenQ GL2460:
  Resolution:	1920 x 1080 @ 60Hz (1080p)
  Pixel Depth:	32-Bit Color (ARGB8888)
  Display Serial Number:	C8E00958SL0
  Mirror:	Off
  Online:	Yes
  Rotation:	Supported
  Adapter Type:	Apple Mini DisplayPort To Dual-Link DVI Adapter
  Adapter Firmware Version:	1.03
  Television:	Yes
BenQ GL2460:
  Resolution:	1920 x 1080 @ 60Hz (1080p)
  Pixel Depth:	32-Bit Color (ARGB8888)
  Display Serial Number:	C8E00651SL0
  Mirror:	Off
  Online:	Yes
  Rotation:	Supported
  Adapter Type:	Apple Mini DisplayPort To Dual-Link DVI Adapter
  Adapter Firmware Version:	1.03
  Television:	Yes
MP515 ST:
  Resolution:	800 x 600 @ 60 Hz
  Pixel Depth:	32-Bit Color (ARGB8888)
  Mirror:	Off
  Online:	Yes
  Rotation:	Supported
  Connection Type:	Analog VGA or Analog Over DVI-I
  Adapter Firmware Version:	0.00
BenQ GL2460:
  Resolution:	1920 x 1080 @ 60Hz (1080p)
  Pixel Depth:	32-Bit Color (ARGB8888)
  Display Serial Number:	C8E01066SL0
  Mirror:	Off
  Online:	Yes
  Rotation:	Supported
  Adapter Type:	Apple Mini DisplayPort To Dual-Link DVI Adapter
  Adapter Firmware Version:	1.03
  Television:	Yes
BenQ GL2460:
  Resolution:	1920 x 1080 @ 60Hz (1080p)
  Pixel Depth:	32-Bit Color (ARGB8888)
  Display Serial Number:	H2E05606SL0
  Mirror:	Off
  Online:	Yes
  Rotation:	Supported
  Adapter Type:	Apple Mini DisplayPort To Dual-Link DVI Adapter
  Adapter Firmware Version:	1.03
  Television:	Yes
Apple Cinema HD Display:
  Display Type:	LCD
  Resolution:	1280 x 800
  Pixel Depth:	32-Bit Color (ARGB8888)
  Main Display:	Yes
  Mirror:	Off
  Online:	Yes
  Rotation:	Supported


Another question regarding the same setup:

Our resolution is 7680x1080. We have several jit.gl.node offscreen rendering destinations that have the same size. (7680x1080) If each pixel uses 32 bits, that means each offscreen-rendered texture use about 31.6 Mb of VRAM. Hence, we could create many jit.gl.node objects without any major performance hit. (we have 16 Gb of VRAM!) Is this correct?

Hi again,
Another question regarding the same setup: 

Our resolution is 7680x1080. We have several jit.gl.node offscreen rendering destinations that have the same size. (7680x1080) If each pixel uses 32 bits, that means each offscreen-rendered texture use about 31.6 Mb of VRAM. Hence, we could create many jit.gl.node objects without any major performance hit. (we have 16 Gb of VRAM!) Is this correct?


7680x1080x32 bits = 265420800 bits = 33177600 bytes = 33.2 megabytes or MB (Mb is megabits)

That applies for a single output texture but the card is going to have a lot more going on in its memory. How much is really dependent on the actual patch, the geometry objects, any slabs/shaders, etc.

if you're using two gpus, you should create a unique window and context for each gpu. i'm not sure if the SLI linking thing works in opengl. in any case, i would try creating the two contexts and see if that fixes the crash. you *should* be able to set @shared 1 on the jit.windows and share texture resources between them.

make sure you set the window @rect attribute to the correct display on the correct card before enabling rendering. jit.displays can help with this, but requires that you know which displays are connected to which gpu.

Also we noticed that loadbang and loadmess don't send bang to our patches even trigerbang in 6.1.9

This morning I experienced a crash because of that. We have replaced by sending the command via OSC.

Also we noticed that loadbang and loadmess don't send bang to our patches even trigerbang in 6.1.9
This morning I experienced a crash because of that. We have replaced by sending the command via OSC.


Do you mean I could split the resulting image of this set of patchs that spreads into 4 screens to 2 gl.render and jit.window ?

by spliting the textures and sending them to a second gl.render ?

Do you mean I could split the resulting image of this set of patchs that spreads into 4 screens to 2 gl.render and jit.window ?
by spliting the textures and sending them to a second gl.render ?


Further more ... what software I can use to read the VRAM usage over the 2 GPU's

I tried IStat but that doesn't give me details on the 2 GPU.

Further more ... what software I can use to read the VRAM usage over the 2 GPU's
I tried IStat but that doesn't give me details on the 2 GPU.


Max Crash - 2 particles patch with 5 monitors