
Chapter 6. Instruction Pipeline and Timing
For More Information On This Product,
Go to: www.freescale.com
6-1
Chapter 6
Instruction Pipeline and Timing
This chapter describes performance features of the V4 ColdFire processor pipeline
structure. These features are common for all V4 standard products as well as custom
products using the CF4e core. Relevant information on the CF4e EMAC, FPU, and MMU
are also included. It is intended as a guide for developing compilers or optimizing assembly
language application code.
This chapter describes the basic V4 pipeline strategy, contrasting it with Version 2 and 3
designs. It provides performance-related details of the instruction fetch and operand
execution pipelines (IFP and OEP). Appendixes describe performance parameters for the
complete ColdFire instruction set.
6.1 Basic V4 Pipeline Strategy
All ColdFire generations include two independent, decoupled pipelines. The instruction
fetch pipeline (IFP) prefetches the instruction stream, examines it to predict changes of
flow, partially decodes instructions, and packages fetched data into instructions for the
OEP. The instruction stream is then gated into the operand execution pipeline (OEP), which
decodes instructions, fetches required operands, and executes required functions. The IFP
and OEP are decoupled by a FIFO instruction buffer. The IFP can prefetch instructions
before the OEP needs them, minimizing the wait for instructions. This organization has
proven very efficient implementation as dictated by the variable-length ColdFire instruction
set.
Version 2 ColdFire processor design minimizes overall size while maintaining a reasonable
performance level. As a result, the V2 OEP is a simple two-stage design, where instructions
requiring operand memory read references are sent through both stages twice. With this and
the significant K-Bus enhancements in the V3 design, V4 minimizes instruction latency by
unrolling the OEP to support single-cycle latency for most opcodes.
Table 6-1 highlights these differences and presents a subset of execution times for the basic
instructions across the ColdFire architecture versions.
F
Freescale Semiconductor, Inc.
n
.