Home /Modules/cpu/ARCHITECTURE.md

ARCHITECTURE.md

lib/cpu/x86/ARCHITECTURE.md · 48 line(s) · markdown

x86 / x86-64 Architecture

Front end

SLX86CPU -> instruction fetch -> prefix scanner -> variable-length decoder -> micro-operation/operation representation -> execution scheduling

Unlike the fixed-width PowerPC/MIPS profiles, the decoder must determine instruction length before execution.

Execution domains

The functional model exposes:

Not every implementation has every domain.

Memory pipeline

The memory path is:

effective address -> segmentation/linearization where applicable -> paging/MMU -> TLB/cache hierarchy -> system interconnect -> physical memory/device.

Intel documents TLBs and paging-structure caches as mechanisms that accelerate address translation. citeturn0search26

Out-of-order boundary

Modern x86 CPUs may translate architectural instructions into internal operations and execute them out of order. SLeeLa represents this as an implementation-level scheduling/completion layer, not as an architectural requirement.

Compatibility

The same SLeeLa family object can describe legacy and 64-bit modes through explicit mode state, while concrete CPU variants select which instructions/extensions are enabled.