{
  "context": "This criteria evaluates how well the engineer uses PyCUDA to implement GPU-accelerated image brightness adjustment, specifically focusing on kernel compilation, proper kernel launch configuration with grid/block dimensions, and efficient memory management.",
  "type": "weighted_checklist",
  "checklist": [
    {
      "name": "Kernel Compilation",
      "description": "Uses pycuda.compiler.SourceModule to compile a CUDA kernel that performs the brightness adjustment operation",
      "max_score": 20
    },
    {
      "name": "Grid/Block Configuration",
      "description": "Correctly configures grid and block dimensions when launching the kernel, using appropriate 2D thread organization (e.g., block=(16, 16, 1) and grid dimensions calculated based on image size)",
      "max_score": 30
    },
    {
      "name": "Kernel Invocation",
      "description": "Properly invokes the compiled kernel function by calling it with the configured grid, block, and appropriate arguments",
      "max_score": 20
    },
    {
      "name": "Memory Transfer",
      "description": "Uses pycuda.driver memory allocation (mem_alloc) and transfer functions (memcpy_htod, memcpy_dtoh or gpuarray.to_gpu/get) to move data between host and device",
      "max_score": 20
    },
    {
      "name": "Context Initialization",
      "description": "Properly initializes PyCUDA with pycuda.autoinit or manually creates and manages a CUDA context using pycuda.driver.Device and Context",
      "max_score": 10
    }
  ]
}