An architecture for content-aware compression and/or encryption of various segments of a application is disclosed. The architecture advantageously allows decompression and decryption units to be placed various levels of a memory hierarchy.