A method and an apparatus are provided for loading data to a local store of a processor in a computer system having a direct memory access (DMA) mechanism. A transfer of data is performed from a system memory of the computer system to the local store. The data is fetched from the system memory to a cache of the processor. A DMA load request is issued to request data. It is determined whether the requested data is found in the cache. Upon a determination that the requested data is found in the cache, the requested data is loaded directly from the cache to the local store.