Cache miss processing using a defer/replay mechanism

A computer-readable storage medium storing instructions that, when executed by a processor including a level-one (L1) cache, cause the L1 cache to process cache misses associated with requests transmitted by a client, by performing the steps of:

• receiving a first request transmitted by a first client of the L1 cache that produces a cache miss for the L1 cache;
• determining whether there is an entry available in a pending request table in which the first request can be stored, wherein the pending request table is configured to store one or more requests that are transmitted by a particular client of the L1 cache and result in cache misses, and wherein the particular client is configured to store information indicating how to retire the one or more requests;
• if the entry is not available, deferring the first request so that the first request is replayed by the first client at a later time, or
• if the entry is available, accepting the first request and storing the first request in the entry, wherein the first client is configured to store information indicating how to retire the first request.

A system for processing cache misses, the system comprising:

• a first client that is coupled to a level-one (L1) cache;
• and
• the L1 cache, wherein the L1 cache is configured to:
• receive a first request transmitted by a first client of the L1 cache that produces a cache miss for the L1 cache;
• determine whether there is an entry available in a pending request table in which the first request can be stored, wherein the pending request table is configured to store one or more requests that are transmitted by a particular client of the L1 cache and result in cache misses, and wherein the particular client is configured to store information indicating how to retire the one or more requests;
• if the entry is not available, defer the first request so that the first request is replayed by the first client at a later time, or
• if the entry is available, accept the first request and storing the first request in the entry, wherein the first client is configured to store information indicating how to retire the first request.