Home /Modules/cpu/ARCHITECTURE.md

ARCHITECTURE.md

lib/cpu/powerpc/ARCHITECTURE.md · 43 line(s) · markdown

PowerPC Architecture

Logical processing

The PowerPC logical processor contains branch processing, fixed-point processing, and floating-point processing, with sequencing/control for instruction fetch, execution, and interrupt action. IBM describes this as the principal logical processing organization. citeturn0search1

SLeeLa decomposes it into:

SLPowerPCCPU -> branch processor -> fixed-point processor -> floating-point processor -> load/store -> storage/interconnect -> exception/interrupt control

Branch processor

The branch subsystem owns:

Fixed-point processor

The integer unit operates on the 32-bit GPR set and provides arithmetic, logical, compare, rotate/shift, and address-generation operations.

Floating-point processor

The FPU uses the 32-entry FPR file and FPSCR. It remains a replaceable implementation component because embedded PowerPC processors may omit floating-point support. The MPC860 documentation, for example, explicitly identifies unsupported floating-point instructions. citeturn0search7

Pipeline

SLPipeline represents instruction fetch, decode/dispatch, execute, memory, and completion/writeback phases as a functional model. Actual stage count and out-of-order behavior belong to each concrete PowerPC implementation.

Exceptions

The exception controller couples MSR and supervisor state to vector selection, save/restore state, interrupt recognition, and privileged control.