createComputePipelineContext: This engine does not support c

Error message

createComputePipelineContext: This engine does not support compute shaders!

What it means

This is the second stub thrown by the compute-shader extension registered on engines without compute support. ThinEngine.prototype.createComputePipelineContext is replaced with a throwing stub so any attempt to build a GPU compute pipeline context on a non-compute engine fails loudly. Only WebGPU-style engines provide a real implementation.

Source

Thrown at packages/dev/core/src/Engines/Extensions/engine.computeShader.pure.ts:84

export type ComputeBindingList = { [key: string]: { type: ComputeBindingType; object: any; indexInGroupEntries?: number } };

let _Registered = false;
/**
 * Register side effects for enginesExtensionsEngineComputeShader.
 * Safe to call multiple times; only the first call has an effect.
 */
export function RegisterEnginesExtensionsEngineComputeShader(): void {
    if (_Registered) {
        return;
    }
    _Registered = true;

    ThinEngine.prototype.createComputeEffect = function (baseName: IComputeShaderPath & { computeToken?: string }, options: IComputeEffectCreationOptions): ComputeEffect {
        throw new Error("createComputeEffect: This engine does not support compute shaders!");
    };

    ThinEngine.prototype.createComputePipelineContext = function (): IComputePipelineContext {
        throw new Error("createComputePipelineContext: This engine does not support compute shaders!");
    };

    ThinEngine.prototype.createComputeContext = function (): IComputeContext | undefined {
        return undefined;
    };

    ThinEngine.prototype.computeDispatch = function (
        effect: ComputeEffect,
        context: IComputeContext,
        bindings: ComputeBindingList,
        x: number,
        y?: number,
        z?: number,
        bindingsMapping?: ComputeBindingMapping
    ): void {
        throw new Error("computeDispatch: This engine does not support compute shaders!");
    };

View on GitHub (pinned to 0592b347b8)

Solutions

Use a WebGPU engine so the real createComputePipelineContext implementation is registered (EngineFactory.CreateAsync with webgpu).
Guard with engine.isWebGPU / supportComputeShaders capability check before touching compute pipeline APIs.
Refactor code to skip compute pipeline setup on non-compute engines and use a rendering-only fallback.
Verify the correct full-engine bundle is loaded so RegisterFullEngineExtensions registered the WebGPU compute path.

Example fix

// before
const ctx = engine.createComputePipelineContext();
// after
if (!engine.isWebGPU) {
    throw new Error('Compute requires WebGPU engine');
}
const ctx = engine.createComputePipelineContext();

Defensive patterns

Strategy: validation

Validate before calling

if (!engine.isWebGPU) {
    throw new Error('createComputePipelineContext requires WebGPU');
}

Type guard

function isComputeEngine(e: AbstractEngine): e is WebGPUEngine {
    return (e as any).isWebGPU === true;
}

Try / catch

try {
    const ctx = engine.createComputePipelineContext();
} catch (e) {
    if (String((e as Error).message).includes('compute shaders')) {
        handleNoCompute();
    } else { throw e; }
}

Prevention

Gate all compute pipeline setup behind a WebGPU capability check
Keep compute code isolated in a module that only loads on WebGPU engines
Verify the correct full-engine bundle registers WebGPU extensions
Add integration tests running under WebGL to catch accidental compute usage

When it happens

Trigger: Calling engine.createComputePipelineContext() on a WebGL-based ThinEngine/Engine, typically indirectly through ComputeEffect creation, pipeline creation, or lower-level compute plumbing when the engine does not support compute shaders.

Common situations: Same family as createComputeEffect failure: WebGPU unavailable at runtime, engine fallback to WebGL2, custom engine wrappers that bypass WebGPUEngine, or tests running against ThinEngine.

Related errors

AI-assisted analysis of BabylonJS/Babylon.js@0592b347b8 (2026-08-30). Data as JSON: /api/errors/193070bf07eb712c. Report an issue: GitHub.