StreamGPU -次元数を高くしたときの性能劣化の問題

http://www.culatools.com/html_guide/index.html#thread-safety
の API を見ると、

typedef enum
{
  culaNoError,                       // No error
  culaNotInitialized,                // CULA has not been initialized
  culaNoHardware,                    // No hardware is available to run
  culaFeatureNotImplemented,         // The requested feature has not been implemented
  culaInsufficientComputeCapability, // Available GPUs do not support the requested operation
  culaInsufficientMemory,            // There is insufficient memory to continue
  culaArgumentError,                 // An invalid argument was passed to a function
  culaDataError,                     // An operation could not complete because of singular data
  culaBlasError,                     // A blas error was encountered
  culaRuntimeError                   // A runtime error has occurred
}culaStatus;

と status が見えるので、NoError 以外のときはこれをデバッグで
まずは表示すべき