BiteSizedChunks.comLearn one small thing at a time.
Course contentsShow
← Embedded Systems & Computer Architecture
Lesson 105 of 9362. Instruction Set Architecture and CPU MicroarchitectureFree lesson

Managing Load-Store Architecture Constraints

Design instruction sequences for load-store architectures, ensuring memory access is restricted to specific instructions rather than arithmetic operations.

Managing Load-Store Architecture Constraints

In this lesson, you will learn how to handle data operations in a CPU architecture that strictly separates memory access from arithmetic processing.

In a Load-Store Architecture (a hallmark of RISC designs), the CPU cannot perform arithmetic directly on data stored in main memory. While a CISC processor might allow an instruction like ADD [address], register (which adds memory content directly to a register), a Load-Store architecture forbids this. Instead, memory acts like a warehouse: you must explicitly "load" items from the warehouse into your internal "workbench" (the registers) before you can work on them.

The Two-Step Process

Because the CPU’s Arithmetic Logic Unit (ALU) is only wired to perform calculations on registers, any operation involving memory requires a specific sequence of instructions. If you want to increment a value stored at a specific memory location, you must follow these three steps:

  1. Load: Fetch the data from the memory address into a register.
  2. Operate: Perform the addition using the register.
  3. Store: Write the updated register value back to the memory address.

Example Sequence

Imagine we want to add the constant 1 to a variable stored in memory at a location labeled counter. In a Load-Store architecture, the sequence looks like this:

LOAD  R1, counter   ; Step 1: Bring data from memory into R1
ADD   R1, R1, #1    ; Step 2: Add 1 using the ALU (internal to CPU)
STORE R1, counter   ; Step 3: Write the result back to memory

Why restrict the architecture?

This constraint might seem like extra work, but it simplifies the CPU hardware significantly. Because all instructions have a uniform format, the CPU pipeline becomes much easier to manage. You no longer have to deal with instructions that have wildly different execution times based on whether they are accessing memory or registers. By enforcing this strict separation, we ensure that every instruction follows a predictable path through the pipeline, which is essential for high-performance design.

Key Takeaway: In a load-store architecture, you must explicitly move data into registers before any arithmetic can be performed, ensuring memory remains a distinct stage of the instruction execution cycle.