Method and system for clearing dependent speculations from a request queue

A memory hierarchy for a computer system, said memory hierarchy comprising:

• at least one cache;
• means for receiving from a processor and buffering within the memory hierarchy a first load request for loading a first prefetch value from lower level memory into the cache and a second load request for loading a second prefetch value from lower level memory into the cache, wherein said second load request is dependent on said first load request;
• at least a first flag for indicating whether said first load request is canceled and a second flag for indicating whether said second load request is canceled;
• and
• means for canceling said first load request before any provision of said first prefetch value to said cache and for setting said second flag to indicate cancellation of said second load request in response to cancellation of said first load request.

A data processing system, comprising:

• a processor;
• and
• a memory hierarchy, including:
• lower level memory;
• and
• at least one cache having a cache controller that receives from the processor and buffers within the memory hierarchy a first load request for loading a first prefetch value from said lower level memory into the cache and a second load request for loading a second prefetch value from said lower level memory into the cache, wherein said second load request is dependent on said first load request, and wherein said cache controller includes means for canceling said first load request before any provision of said first prefetch value to the cache and means for setting a flag to indicate cancellation of said second load request in response to cancellation of said first load request.