readduringwrite

Read-during-write describes the behavior observed when a read access targets the same memory location that is being written concurrently. The phenomenon occurs in many memory technologies and architectures (SRAM, DRAM, flash, and FPGA block RAMs) and in multi-ported or shared-memory systems. The returned value and system behavior depend on timing, memory implementation, and controller policy.

Common resolution policies are called read-first (return the old data that was resident before the write), write-first (return the new data being written), and no-change or undefined (results are indeterminate during the overlap). In synchronous memories these policies are often explicitly documented for each port; in asynchronous memories the result can be timing-dependent. Dual-port memories must also define what happens when two ports access the same address simultaneously.

Implications include potential data hazards, race conditions in concurrent systems, and subtle bugs in simulation versus silicon if designers assume a particular outcome. In nonvolatile memories (for example, flash) reads during an ongoing erase/program may be blocked, may present stale data, or may be filtered by the controller for integrity reasons.

Mitigation strategies include avoiding concurrent accesses to the same location, using arbitration or locks, employing write buffers and coherent caches, selecting memories with a defined read-during-write policy, and designing protocols that rely on explicit write-complete acknowledgements rather than concurrent reads. Understanding the specific memory primitive semantics is essential for correct system behavior.