Direct Mapped Cache

A direct-mapped cache is the simplest form of cache mapping, where each memory block is mapped to exactly one cache line. 

Key Concepts of Direct-Mapped Cache

  • Indexing: A portion of the memory address selects the cache line.
  • Tag Comparison: The tag is stored alongside the data and compared to the incoming address.
  • Eviction Policy: If a new block is mapped to an occupied cache line, the old block is evicted (replaced).

Address Breakdown (Typical Structure)

Assuming a 32-bit address and a cache with lines:

  • Offset: Identifies bytes within a block.
  • Index: Selects a cache line.
  • Tag: Identifies which memory block is stored.

Example (Cache with 64 Lines, Block Size = 16B)

A 32-bit address layout:

  • Offset (log₂(16) = 4 bits)
  • Index (log₂(64) = 6 bits)
  • Tag (32 - 6 - 4 = 22 bits)

Comments

Post a Comment

Popular posts from this blog

Cache Basics

A cache is: 1. a small memory 2. Fast Memory between the CPU and main memory that stores frequently accesos data. Key parameters: Cache Line Size : Number of bytes fetched at once (e.g., 64B). Associativity :  1. Direct-mapped,  2. set-associative,  3. or fully associative. Replacement Policy :  1. LRU,  2. PLRU,  3. Round Robin,  4. etc.
Read more

Cache Sizes

Common cache sizes depend on the level of cache (L1, L2, L3, or L4) and the architecture (mobile, desktop, server, or embedded systems). Here’s a general breakdown: 1. L1 Cache (Instruction & Data) Size : 16KB – 128KB per core Latency : ~1-4 cycles Associativity : 4-way to 8-way set-associative Use case : Extremely fast, directly connected to the CPU core for low-latency access 2. L2 Cache Size : 256KB – 8MB per core Latency : ~10-20 cycles Associativity : 8-way to 16-way Use case : Acts as a backup for L1, slightly slower but much larger 3. L3 Cache (Shared across multiple cores) Size : 2MB – 128MB (varies by CPU type) Latency : ~30-50 cycles Associativity : 12-way to 32-way Use case : Shared between cores, reducing memory access bottlenecks 4. L4 Cache (Rare, used in some high-end CPUs & GPUs) Size : 32MB – 256MB Latency : ~100+ cycles Use case : Mostly in enterprise CPUs or GPUs for large-scale data processing Cache Sizes by Platform If we’re wor...
Read more

Cache Memory for RISC V

To become an expert in cache memory for RISC-V, we'll need to focus on: Cache Fundamentals – Understand direct-mapped, set-associative, and fully associative caches, along with policies like LRU, PLRU, and replacement strategies. RISC-V Cache Architecture – Study how RISC-V implementations handle cache hierarchies (L1, L2, L3) and coherence protocols (MESI, MOESI). Cache Verification – Since we're already working with UVM, mastering cache verification (hit/miss testing, replacement validation, coherence checking) will be key. Performance Optimization – Explore prefetching, write policies, and latency reduction techniques in RISC-V. Hardware Implementation – Get hands-on experience designing and optimizing cache controllers in SystemVerilog.
Read more