Error Classes#
Every exception raised by brainevent derives from BrainEventError,
so a single except brainevent.BrainEventError clause catches all of them.
The hierarchy is:
BrainEventError
├── MathError
├── UnsupportedOperationError
├── BenchmarkDataFnNotProvidedError
└── KernelError
├── KernelNotAvailableError
├── KernelFallbackExhaustedError
├── KernelExecutionError
├── KernelRegistrationError
├── KernelLoadError
├── CUDANotInstalledError
├── KernelCompilationError
│ └── CompilationError
│ └── HostCompilerIncompatibleError
└── KernelToolchainError
├── NvccNotFoundError
├── HeaderNotFoundError
├── HostCompilerNotFoundError
├── GpuArchDetectionError
└── UnsupportedArchError
Base#
Base exception for brainevent. |
Array and Operation Errors#
Base exception for mathematical errors in brainevent operations. |
|
Raised when an operation is structurally meaningless for a representation. |
|
Raised when |
Kernel Dispatch and Execution Errors#
Raised when a requested kernel backend is not installed or is version-incompatible. |
|
Raised when all fallback kernel backends have been exhausted. |
|
Raised when a compiled kernel fails during execution at runtime. |
|
JAX FFI target registration failed. |
|
A compiled .so failed to load via dlopen (E-LOAD). |
Kernel Compilation Errors#
Raised when a kernel fails to compile on the target backend. |
|
|
CUDA or C++ compilation failed. |
|
Host C++ compiler version is not supported by this CUDA/nvcc (E-CXXVER). |
Toolchain Errors#
Raised when the CUDA / C++ toolchain needed to build a kernel is missing or
unusable. Each carries a stable E- diagnostic code in its message.
Compilation toolchain missing or incompatible. |
|
Raised when a CUDA operation is requested but the package is not installed. |
|
The CUDA compiler (nvcc) could not be located (E-NVCC). |
|
A required header (cuda_runtime.h / XLA FFI / brainevent) is missing (E-HDR). |
|
No host C++ compiler (g++/clang++) could be located (E-CXX). |
|
GPU compute capability could not be detected (E-ARCH). |
|
|
nvcc does not support the requested GPU architecture (E-ARCH-UNSUP). |