ValidateProgramUniformBlockBindings: Bound buffer is too small (16384 bytes, 65520 bytes expected) - Model rendering failure

[baoJiePro]

Filament Uniform Buffer Size Issue - Model Rendering Failure

Problem Summary

When attempting to load and render 3D models, the application encounters a uniform buffer binding error that prevents successful model rendering. The system allocates a buffer that is too small for the required uniform data.

Error Details

ValidateProgramUniformBlockBindings: Bound buffer is too small (Bound buffer is 16384 bytes, 65520 bytes expected) :0: DrawElements: ValidateState() failed

System Configuration

GPU Information:

• GPU Vendor: Imagination Technologies
• GPU Model: PowerVR Rogue GE8300
• OpenGL ES Version: 3.2 (build 1.11@5516664)
• GLSL ES Version: 3.20 (build 1.11@5516664)
• Feature Level: 2 (Backend), 1 (Engine)

OpenGL ES Capabilities:

• GL_MAX_UNIFORM_BLOCK_SIZE: 134,217,728 bytes (128MB)
• GL_MAX_UNIFORM_BUFFER_BINDINGS: 72
• GL_UNIFORM_BUFFER_OFFSET_ALIGNMENT: 64 bytes
• GL_MAX_TEXTURE_IMAGE_UNITS: 16
• GL_MAX_COMBINED_TEXTURE_IMAGE_UNITS: 96
• GL_MAX_TEXTURE_SIZE: 4096
• GL_MAX_SAMPLES: 4

Active GPU Workarounds:

• disable_glFlush
• allow_read_only_ancillary_feedback_loop
• powervr_shader_workarounds
• delay_fbo_destruction

Platform:

• OS Version: Android API 29

Technical Analysis

The error indicates a significant discrepancy between:

• Allocated Buffer Size: 16,384 bytes (16KB)
• Required Buffer Size: 65,520 bytes (~64KB)

Key Observations

1. Hardware Capability vs. Allocation: While the GPU supports up to 128MB for uniform blocks (GL_MAX_UNIFORM_BLOCK_SIZE), the actual buffer allocated by Filament is only 16KB.
2. PowerVR-Specific Issue: The presence of powervr_shader_workarounds suggests this GPU has known limitations that may affect buffer management.
3. Feature Level Mismatch: There's a discrepancy between backend feature level (2) and engine feature level (1), which may impact buffer allocation strategies.

Expected Behavior

Models should render successfully with properly sized uniform buffers that accommodate all required uniform data for the shaders.

Actual Behavior

Model rendering fails with a buffer size validation error, resulting in no visual output. The rendering pipeline cannot proceed due to insufficient uniform buffer allocation.

This issue prevents the application from displaying 3D models on PowerVR Rogue GE8300 GPU devices running Android API 29.

[pixelflinger]

Please provide the model with the problem. I'm not understanding which uniform block is involved here. Is it because of skinning? Currently filament is limited to 16KB UBOs to accomodate GLES3 minspec.

But regardless, this limitation should not be exposed to end-users, so I'm not sure which UBO is causing the problem here.

[baoJiePro]

skp134.zip

@pixelflinger Thank you for your response! Here are the details about the problem model:

Model Information

Model Name: skp134.glTF File Size: ~32KB (GLTF) + 8.3MB (binary buffer data) Model Type: Smart home 3D interior scene (static architectural model, no skinning/animation)

Problem Details

The model triggers this exact error on PowerVR Rogue GE8300 GPU:

ValidateProgramUniformBlockBindings: Bound buffer is too small (16384 bytes, 65520 bytes expected)

Technical Context

• No Skinning Animation: This is a static architectural model, so it shouldn't require large bone matrices
• • Complex Geometry: The model has detailed interior structure with many mesh entities
• • • Material Count: Multiple materials with different properties (transparency, textures, etc.)

Questions for Optimization

Given the 16KB UBO limitation for GLES3 minspec, what properties and settings can we optimize for low-end GPUs like the PowerVR GE8300:

1. Material Properties: Which material parameters contribute most to UBO size?
2. 2. Geometry Optimization: Should we split the model into smaller chunks?
3. 3. Render Settings: Which Filament settings can reduce uniform buffer usage?
4. 4. Feature Flags: What features can we disable to reduce UBO requirements?
5. 5. Engine Configuration: Are there engine-level settings for uniform buffer optimization? The model files are too large for GitHub attachment, but the GLTF structure shows many mesh entities and materials that might be contributing to the large uniform buffer requirement.

Thank you for your help with this issue!