Skip to content

Latest commit

 

History

History
92 lines (71 loc) · 5.42 KB

instruction_fetch.rst

File metadata and controls

92 lines (71 loc) · 5.42 KB

Instruction Fetch

The Instruction Fetch (IF) stage of the CV32E40P is able to supply one instruction to the Instruction Decode (ID ) stage per cycle if the external bus interface is able to serve one instruction per cycle. In case of executing compressed instructions, on average less than one 32-bit instruction fetch will we needed per instruction in the ID stage.

For optimal performance and timing closure reasons, a prefetcher is used which fetches instructions via the external bus interface from for example an externally connected instruction memory or instruction cache.

The prefetch unit performs word-aligned 32-bit prefetches and stores the fetched words in a FIFO with four entries. As a result of this (speculative) prefetch, CV32E40P can fetch up to four words outside of the code region and care should therefore be taken that no unwanted read side effects occur for such prefetches outside of the actual code region.

:numref:`Instruction Fetch interface signals` describes the signals that are used to fetch instructions. This interface is a simplified version of the interface that is used by the LSU, which is described in :ref:`load-store-unit`. The difference is that no writes are possible and thus it needs fewer signals.

Instruction Fetch interface signals
Signal Direction Description
instr_req_o output Request valid, will stay high until instr_gnt_i is high for one cycle
instr_addr_o[31:0] output Address, word aligned
instr_rdata_i[31:0] input Data read from memory
instr_rvalid_i input instr_rdata_i holds valid data when instr_rvalid_i is high. This signal will be high for exactly one cycle per request.
instr_gnt_i input The other side accepted the request. instr_addr_o may change in the next cycle.

Misaligned Accesses

Externally, the IF interface performs word-aligned instruction fetches only. Misaligned instruction fetches are handled by performing two separate word-aligned instruction fetches. Internally, the core can deal with both word- and half-word-aligned instruction addresses to support compressed instructions. The LSB of the instruction address is ignored internally.

Protocol

The CV32E40P instruction fetch interface does not implement the following optional OBI signals: we, be, wdata, auser, wuser, aid, rready, err, ruser, rid. These signals can be thought of as being tied off as specified in the OBI specification. The CV32E40P instruction fetch interface can cause up to two outstanding transactions.

:numref:`obi-instruction-basic` and :numref:`obi-instruction-multiple-outstanding` show example timing diagrams of the protocol.

Back-to-back Memory Transactions

Multiple Outstanding Memory Transactions