The hardcoded 1024 value for vector_load_bits_max_ in loop_vectorize.cc should be replaced with `GetSunmmioTileProcessorConfig().register_bits` (4096).
The hardcoded 1024 value for vector_load_bits_max_ in loop_vectorize.cc should be replaced with
GetSunmmioTileProcessorConfig().register_bits(4096).