authors: Andrew Bardsley
last edited: 2024-07-09 22:50:19 +0000

Minor CPU Model

This document contains a description of the structure and function of the Minor gem5 in-order processor model.

It is recommended reading for anyone who wants to understand Minor’s internal organisation, design decisions, C++ implementation and Python configuration. A familiarity with gem5 and some of its internal structures is assumed. This document is meant to be read alongside the Minor source code and to explain its general structure without being too slavish about naming every function and data type.

What is Minor?

Minor is an in-order processor model with a fixed pipeline but configurable data structures and execute behaviour. It is intended to be used to model processors with strict in-order execution behaviour and allows visualisation of an instruction’s position in the pipeline through the MinorTrace/ format/tool. The intention is to provide a framework for micro-architecturally correlating the model with a particular, chosen processor with similar capabilities.

Design Philosophy


The model isn’t currently capable of multithreading but there are THREAD comments in key places where stage data needs to be arrayed to support multithreading.

Data structures

Decorating data structures with large amounts of life-cycle information is avoided. Only instructions (MinorDynInst) contain a significant proportion of their data content whose values are not set at construction.

All internal structures have fixed sizes on construction. Data held in queues and FIFOs (MinorBuffer, FUPipeline) should have a BubbleIF interface to allow a distinct ‘bubble’/no data value option for each type.

Inter-stage ‘struct’ data is packaged in structures which are passed by value. Only MinorDynInst, the line data in ForwardLineData and the memory-interfacing objects Fetch1::FetchRequest and LSQ::LSQRequest are ::new allocated while running the model.

Model structure

Objects of class MinorCPU are provided by the model to gem5. MinorCPU implements the interfaces of (cpu.hh) and can provide data and instruction interfaces for connection to a cache system. The model is configured in a similar way to other gem5 models through Python. That configuration is passed on to MinorCPU::pipeline (of class Pipeline) which actually implements the processor pipeline.

The hierarchy of major unit ownership from MinorCPU down looks like this:

--- Pipeline - container for the pipeline, owns the cyclic 'tick' event mechanism and the idling (cycle skipping) mechanism.
--- --- Fetch1 - instruction fetch unit responsible for fetching cache lines (or parts of lines from the I-cache interface).
--- --- --- Fetch1::IcachePort - interface to the I-cache from Fetch1.
--- --- Fetch2 - line to instruction decomposition.
--- --- Decode - instruction to micro-op decomposition.
--- --- Execute - instruction execution and data memory interface.
--- --- --- LSQ - load store queue for memory ref. instructions.
--- --- --- LSQ::DcachePort - interface to the D-ache from Execute.

Key data structures

Instruction and line identity: Instld (dyn_inst.hh)

- T/S.P/L - for fetched cache lines
- T/S.P/L/F - for instructions before Decode
- T/S.P/L/F.E - for instructions from Decode onwards

for example:

- 0/10.12/5/6.7

InstId fields are:

Field Symbol Generated by Checked by Function
InstId::threadId T Fetch1 Everywhere the thread number is needed Thread number (currently always 0).
InstId::streamSeqNum S Execute Fetch1, Fetch2, Execute (to discard lines/insts) Stream sequence number as chosen by Execute. Stream sequence numbers change after changes of PC (branches, exceptions) in Execue and are used to separate pre and post brnach instrucion streams.
InstId::predictionSeqNum Fetch2 Fetch2 (while discarding lines after prediction) Prediction sequence numbers represent branch prediction decisions. This is used by Fetch2 to mark lines/instructions/ according to the last followed branch prediction made by Fetch2. Fetch2 can signal to Fetch1 that it should change its fetch address and mark lines with a new prediction sequence number (which it will only do if the stream sequence number Fetch1 expects matches that of the request).  
InstId::lineSeqNum Fetch1 (just for debugging) Line fetch sequence number of this cache line or the line this instruction was extracted from.  
InstId::fetchSeqNum Fetch2 Fetch2 (as the inst. sequence number for branches) Instruction fetch order assigned by Fetch2 when lines are decomposed into instructions.  
InstId::execSeqNum Decode Execute (to check instruction identify in queues/FUs/LSQ Instruction order after micro-op decomposition  

The sequence number fields are all independent of each other and although, for instance, InstId::execSeqNum for an instruction will always be >= InstId::fetchSeqNum, the comparison is not useful.

The originating stage of each sequence number field keeps a counter for that field which can be incremented in order to generate new, unique numbers.

Instructi ns: MinorDynInst (dyn_inst.hh)

MinorDynInst represents an instruction’s progression through the pipeline. An instruction can be three things:

Things Predicate Explanation
A bubble MinorDynInst::isBubble() no instruction at all, just a space-filler
A fault MinorDynInst::isFault() a fault to pass down the pipeline in an insturction’s clothing
A decoded instruction MinorDynInst::isInst() instructions are actually passed to the gem5 decoder in Fetch2 and so are created fully decoded. MinorDynInst::staticInst is the decoded instruction form.

Instructions are reference counted using the gem5 RefCountingPtr (base/refcnt.hh) wrapper. They therefore usually appear as MinorDynInstPtr in code. Note that as RefCountingPtr initialises as nullptr rather than an object that supports BubbleIF::isBubble passing raw MinorDynInstPtrs to Queues and other similar structures from stage.hh without boxing is dangerous.

ForwardLineData (pipe_data.hh)

ForwardLineData is used to pass cache lines from Fetch1 to Fetch2. Like MinorDynInsts, they can be bubbles (ForwardLineData::isBubble()), fault-carrying or can contain a line (partial line) fetched by Fetch1. The data carried by ForwardLineData is owned by a Packet object returned from memory and is explicitly memory managed and do must be deleted once processed (by Fetch2 deleting the Packet).

ForwardInstData (pipe_data.hh)

ForwardInstData can contain up to ForwardInstData::width() instructions in its ForwardInstData::insts vector. This structure is used to carry instructions between Fetch2, Decode and Execute and to store input buffer vectors in Decode and Execute.

Fetch1::FetchRequest (fetch1.hh)

FetchRequests represent I-cache line fetch requests. The are used in the memory queues of Fetch1 and are pushed into/popped from Packet::senderState while traversing the memory system.

FetchRequests contain a memory system Request (mem/request.hh) for that fetch access, a packet (Packet, mem/packet.hh), if the request gets to memory, and a fault field that can be populated with a TLB-sourced prefetch fault (if any).

LSQ::LSQRequest (execute.hh)

LSQRequests are similar to FetchRequests but for D-cache accesses. They carry the instruction associated with a memory access.

The pipeline


    [] : inter-stage BufferBuffer
    |  | : pipeline stage
    ---> : forward communication
    <--- : backward communication

    rv : reservation information for input buffers

                ,------.     ,------.     ,------.     ,-------.
 (from  --[]-v->|Fetch1|-[]->|Fetch2|-[]->|Decode|-[]->|Execute|--> (to Fetch1
 Execute)    |  |      |<-[]-|      |<-rv-|      |<-rv-|       |     & Fetch2)
             |  `------'<-rv-|      |     |      |     |       |
             `-------------->|      |     |      |     |       |
                             `------'     `------'     `-------'

The four pipeline stages are connected together by MinorBuffer FIFO (stage.hh, derived ultimately from TimeBuffer) structures which allow inter-stage delays to be modelled. There is a MinorBuffers between adjacent stages in the forward direction (for example: passing lines from Fetch1 to Fetch2) and, between Fetch2 and Fetch1, a buffer in the backwards direction carrying branch predictions.

Stages Fetch2, Decode and Execute have input buffers which, each cycle, can accept input data from the previous stage and can hold that data if the stage is not ready to process it. Input buffers store data in the same form as it is received and so Decode and Execute’s input buffers contain the output instruction vector (ForwardInstData (pipe_data.hh)) from their previous stages with the instructions and bubbles in the same positions as a single buffer entry.

Stage input buffers provide a Reservable (stage.hh) interface to their previous stages, to allow slots to be reserved in their input buffers, and communicate their input buffer occupancy backwards to allow the previous stage to plan whether it should make an output in a given cycle.

Event handling: MinorActivityRecorder (activity.hh, pipeline.hh)

Minor is essentially a cycle-callable model with some ability to skip cycles based on pipeline activity. External events are mostly received by callbacks (e.g. Fetch1::IcachePort::recvTimingResp) and cause the pipeline to be woken up to service advancing request queues.

Ticked (sim/ticked.hh) is a base class bringing together an evaluate member function and a provided SimObject. It provides a Ticked::start/stop interface to start and pause clock events from being periodically issued. Pipeline is a derived class of Ticked.

During evaluate calls, stages can signal that they still have work to do in the next cycle by calling either MinorCPU::activityRecorder->activity() (for non-callable related activity) or MinorCPU::wakeupOnEvent() (for stage callback-related 'wakeup' activity).

Pipeline::evaluate contains calls to evaluate for each unit and a test for pipeline idling which can turns off the clock tick if no unit has signalled that it may become active next cycle.

Within Pipeline (pipeline.hh), the stages are evaluated in reverse order (and so will ::evaluate in reverse order) and their backwards data can be read immediately after being written in each cycle allowing output decisions to be ‘perfect’ (allowing synchronous stalling of the whole pipeline). Branch predictions from Fetch2 to Fetch1 can also be transported in 0 cycles making fetch1ToFetch2BackwardDelay the only configurable delay which can be set as low as 0 cycles.

The MinorCPU::activateContext and MinorCPU::suspendContext interface can be called to start and pause threads (threads in the MT sense) and to start and pause the pipeline. Executing instructions can call this interface (indirectly through the ThreadContext) to idle the CPU/their threads.

Each pipeline stage

In general, the behaviour of a stage (each cycle) is:

        push input to inputBuffer
        setup references to input/output data slots

        do 'every cycle' 'step' tasks

        if there is input and there is space in the next stage:
            process and generate a new output
            maybe re-activate the stage

        send backwards data

        if the stage generated output to the following FIFO:
            signal pipe activity

        if the stage has more processable input and space in the next stage:
            re-activate the stage for the next cycle

        commit the push to the inputBuffer if that data hasn't all been used

The Execute stage differs from this model as its forward output (branch) data is unconditionally sent to Fetch1 and Fetch2. To allow this behaviour, Fetch1 and Fetch2 must be unconditionally receptive to that data.

Fetch1 stage

Fetch1 is responsible for fetching cache lines or partial cache lines from the I-cache and passing them on to Fetch2 to be decomposed into instructions. It can receive ‘change of stream’ indications from both Execute and Fetch2 to signal that it should change its internal fetch address and tag newly fetched lines with new stream or prediction sequence numbers. When both Execute and Fetch2 signal changes of stream at the same time, Fetch1 takes Execute’s change.

Every line issued by Fetch1 will bear a unique line sequence number which can be used for debugging stream changes.

When fetching from the I-cache, Fetch1 will ask for data from the current fetch address (Fetch1::pc) up to the end of the ‘data snap’ size set in the parameter fetch1LineSnapWidth. Subsequent autonomous line fetches will fetch whole lines at a snap boundary and of size fetch1LineWidth.

Fetch1 will only initiate a memory fetch if it can reserve space in Fetch2 input buffer. That input buffer serves an the fetch queue/LFL for the system.

Fetch1 contains two queues: requests and transfers to handle the stages of translating the address of a line fetch (via the TLB) and accommodating the request/response of fetches to/from memory.

Fetch requests from Fetch1 are pushed into the requests queue as newly allocated FetchRequest objects once they have been sent to the ITLB with a call to itb->translateTiming.

A response from the TLB moves the request from the requests queue to the transfers queue. If there is more than one entry in each queue, it is possible to get a TLB response for request which is not at the head of the requests queue. In that case, the TLB response is marked up as a state change to Translated in the request object, and advancing the request to transfers (and the memory system) is left to calls to Fetch1::stepQueues which is called in the cycle following any event is received.

Fetch1::tryToSendToTransfers — layout: documentation title: Execution Basics doc: gem5 documentation parent: cpu_models permalink: /documentation/general_docs/cpu_models/execution_basics —

is responsible for moving requests between the two queues and issuing requests to memory. Failed TLB lookups (prefetch aborts) continue to occupy space in the queues until they are recovered at the head of transfers.

Responses from memory change the request object state to Complete and Fetch1::evaluate can pick up response data, package it in the ForwardLineData object, and forward it to Fetch2’s input buffer.

As space is always reserved in Fetch2::inputBuffer, setting the input buffer’s size to 1 results in non-prefetching behaviour.

When a change of stream occurs, translated requests queue members and completed transfers queue members can be unconditionally discarded to make way for new transfers.

Fetch2 stage

Fetch2 receives a line from Fetch1 into its input buffer. The data in the head line in that buffer is iterated over and separated into individual instructions which are packed into a vector of instructions which can be passed to Decode. Packing instructions can be aborted early if a fault is found in either the input line as a whole or a decomposed instruction.

Branch prediction

Fetch2 contains the branch prediction mechanism. This is a wrapper around the branch predictor interface provided by gem5 (cpu/pred/…).

Branches are predicted for any control instructions found. If prediction is attempted for an instruction, the MinorDynInst::triedToPredict flag is set on that instruction.

When a branch is predicted to take, the MinorDynInst::predictedTaken flag is set and MinorDynInst::predictedTarget is set to the predicted target PC value. The predicted branch instruction is then packed into Fetch2’s output vector, the prediction sequence number is incremented, and the branch is communicated to Fetch1.

After signalling a prediction, Fetch2 will discard its input buffer contents and will reject any new lines which have the same stream sequence number as that branch but have a different prediction sequence number. This allows following sequentially fetched lines to be rejected without ignoring new lines generated by a change of stream indicated from a ‘real’ branch from Execute (which will have a new stream sequence number).

The program counter value provided to Fetch2 by Fetch1 packets is only updated when there is a change of stream. Fetch2::havePC indicates whether the PC will be picked up from the next processed input line. Fetch2::havePC is necessary to allow line-wrapping instructions to be tracked through decode.

Branches (and instructions predicted to branch) which are processed by Execute will generate BranchData (pipe_data.hh) data explaining the outcome of the branch which is sent forwards to Fetch1 and Fetch2. Fetch1 uses this data to change stream (and update its stream sequence number and address for new lines). Fetch2 uses it to update the branch predictor. Minor does not communicate branch data to the branch predictor for instructions which are discarded on the way to commit.

BranchData::BranchReason (pipe_data.hh) encodes the possible branch scenarios:

Branch enum val. In Execute Fetch1 reaction Fetch2 reaction
No Branch (output bubble data) - -
CorrectlyPredictedBranch Predicted, taken - Update BP as taken branch
UnpredictedBranch Not predicted, taken and was taken New stream Update BP as taken branch
BadlyPredictedBranch Predicted, not taken New stream to restore to old Inst. source Update BP as not taken branch
BadlyPredictedBranchTarget Predicted, taken, but to a different target than predicted one New stream Update BTB to new target
SuspendThread Hint to suspend fetch Suspend fetch for this thread (branch to next inst. as wakeup fetch addr -
Interrupt Interrupt detected New stream -

layout: documentation title: Execution Basics doc: gem5 documentation parent: cpu_models permalink: /documentation/general_docs/cpu_models/execution_basics —

Decode Stage

Decode takes a vector of instructions from Fetch2 (via its input buffer) and decomposes those instructions into micro-ops (if necessary) and packs them into its output instruction vector.

The parameter executeInputWidth sets the number of instructions which can be packed into the output per cycle. If the parameter decodeCycleInput is true, Decode can try to take instructions from more than one entry in its input buffer per cycle.

Execute Stage

Execute provides all the instruction execution and memory access mechanisms. An instructions passage through Execute can take multiple cycles with its precise timing modelled by a functional unit pipeline FIFO.

A vector of instructions (possibly including fault ‘instructions’) is provided to Execute by Decode and can be queued in the Execute input buffer before being issued. Setting the parameter executeCycleInput allows execute to examine more than one input buffer entry (more than one instruction vector). The number of instructions in the input vector can be set with executeInputWidth and the depth of the input buffer can be set with parameter executeInputBufferSize.

Functional units

The Execute stage contains pipelines for each functional unit comprising the computational core of the CPU. Functional units are configured via the executeFuncUnits parameter. Each functional unit has a number of instruction classes it supports, a stated delay between instruction issues, and a delay from instruction issue to (possible) commit and an optional timing annotation capable of more complicated timing.

Each active cycle, Execute::evaluate performs this action:

        push input to inputBuffer
        setup references to input/output data slots and branch output slot

        step D-cache interface queues (similar to Fetch1)

        if interrupt posted:
            take interrupt (signalling branch to Fetch1/Fetch2)
            commit instructions
            issue new instructions

        advance functional unit pipelines

        reactivate Execute if the unit is still active

        commit the push to the inputBuffer if that data hasn't all been used

Functional unit FIFOs

Functional units are implemented as SelfStallingPipelines (stage.hh). These are TimeBuffer FIFOs with two distinct ‘push’ and ‘pop’ wires. They respond to SelfStallingPipeline::advance in the same way as TimeBuffers unless there is data at the far, ‘pop’, end of the FIFO. A ‘stalled’ flag is provided for signalling stalling and to allow a stall to be cleared. The intention is to provide a pipeline for each functional unit which will never advance an instruction out of that pipeline until it has been processed and the pipeline is explicitly unstalled.

The actions ‘issue’, ‘commit’, and ‘advance’ act on the functional units.


Issuing instructions involves iterating over both the input buffer instructions and the heads of the functional units to try and issue instructions in order. The number of instructions which can be issued each cycle is limited by the parameter executeIssueLimit, how executeCycleInput is set, the availability of — layout: documentation title: Execution Basics doc: gem5 documentation parent: cpu_models permalink: /documentation/general_docs/cpu_models/execution_basics —

pipeline space and the policy used to choose a pipeline in which the instruction can be issued.

At present, the only issue policy is strict round-robin visiting of each pipeline with the given instructions in sequence. For greater flexibility, better (and more specific policies) will need to be possible.

Memory operation instructions traverse their functional units to perform their EA calculations. On ‘commit’, the ExecContext::initiateAcc execution phase is performed and any memory access is issued (via. ExecContext::{read,write}Mem calling LSQ::pushRequest) to the LSQ.

Note that faults are issued as if they are instructions and can (currently) be issued to any functional unit.

Every issued instruction is also pushed into the Execute::inFlightInsts queue. Memory ref. instructions are pushing into Execute::inFUMemInsts queue.


Instructions are committed by examining the head of the Execute::inFlightInsts queue (which is decorated with the functional unit number to which the instruction was issued). Instructions which can then be found in their functional units are executed and popped from Execute::inFlightInsts.

Memory operation instructions are committed into the memory queues (as described above) and exit their functional unit pipeline but are not popped from the Execute::inFlightInsts queue. The Execute::inFUMemInsts queue provides ordering to memory operations as they pass through the functional units (maintaining issue order). On entering the LSQ, instructions are popped from Execute::inFUMemInsts.

If the parameter executeAllowEarlyMemoryIssue is set, memory operations can be sent from their FU to the LSQ before reaching the head of Execute::inFlightInsts but after their dependencies are met. MinorDynInst::instToWaitFor is marked up with the latest dependent instruction execSeqNum required to be committed for a memory operation to progress to the LSQ.

Once a memory response is available (by testing the head of Execute::inFlightInsts against LSQ::findResponse), commit will process that response (ExecContext::completeAcc) and pop the instruction from Execute::inFlightInsts.

Any branch, fault or interrupt will cause a stream sequence number change and signal a branch to Fetch1/Fetch2. Only instructions with the current stream sequence number will be issued and/or committed.


All non-stalled pipeline are advanced and may, thereafter, become stalled. Potential activity in the next cycle is signalled if there are any instructions remaining in any pipeline.


The scoreboard (Scoreboard) is used to control instruction issue. It contains a count of the number of in flight instructions which will write each general purpose CPU integer or float register. Instructions will only be issued where the scoreboard contains a count of 0 instructions which will write to one of the instructions source registers.

Once an instruction is issued, the scoreboard counts for each destination register for an instruction will be incremented.

The estimated delivery time of the instruction’s result is marked up in the scoreboard by adding the length of the issued-to FU to the current time. The timings parameter on each FU provides a list of additional rules for calculating the delivery time. These are documented in the parameter comments in

On commit, (for memory operations, memory response commit) the scoreboard counters for an instruction’s source registers are decremented. will be decremented.


The Execute::inFlightInsts queue will always contain all instructions in flight in Execute in the correct issue order. Execute::issue is the only process which will push an instruction into the queue. Execute::commit is the only process that can pop an instruction.


The LSQ can support multiple outstanding transactions to memory in a number of conservative cases.

There are three queues to contain requests: requests, transfers and the store buffer. The requests and transfers queue operate in a similar manner to the queues in Fetch1. The store buffer is used to decouple the delay of completing store operations from following loads.

Requests are issued to the DTLB as their instructions leave their functional unit. At the head of requests, cacheable load requests can be sent to memory and on to the transfers queue. Cacheable stores will be passed to transfers unprocessed and progress that queue maintaining order with other transactions.

The conditions in LSQ::tryToSendToTransfers dictate when requests can be sent to memory.

All uncacheable transactions, split transactions and locked transactions are processed in order at the head of requests. Additionally, store results residing in the store buffer can have their data forwarded to cacheable loads (removing the need to perform a read from memory) but no cacheable load can be issue to the transfers queue until that queue’s stores have drained into the store buffer.

At the end of transfers, requests which are LSQ::LSQRequest::Complete (are faulting, are cacheable stores, or have been sent to memory and received a response) can be picked off by Execute and either committed (ExecContext::completeAcc) and, for stores, be sent to the store buffer.

Barrier instructions do not prevent cacheable loads from progressing to memory but do cause a stream change which will discard that load. Stores will not be committed to the store buffer if they are in the shadow of the barrier but before the new instruction stream has arrived at Execute. As all other memory transactions are delayed at the end of the requests queue until they are at the head of Execute::inFlightInsts, they will be discarded by any barrier stream change.

After commit, LSQ::BarrierDataRequest requests are inserted into the store buffer to track each barrier until all preceding memory transactions have drained from the store buffer. No further memory transactions will be issued from the ends of FUs until after the barrier has drained.


Draining is mostly handled by the Execute stage. When initiated by calling MinorCPU::drain, Pipeline::evaluate checks the draining status of each unit each cycle and keeps the pipeline active until draining is complete. It is Pipeline that signals the completion of draining. Execute is triggered by MinorCPU::drain and starts stepping through its Execute::DrainState state machine, starting from state Execute::NotDraining, in this order:

State Meaning
Execute::NotDraining Not trying to drain, normal execution
Execute::DrainCurrentInst Draining micro-ops to complete inst.
Execute::DrainHaltFetch Halt fetching instructions
Execute::DrainAllInsts Discarding all instructions presented

When complete, a drained Execute unit will be in the Execute::DrainAllInsts state where it will continue to discard instructions but has no knowledge of the drained state of the rest of the model.

Debug options

The model provides a number of debug flags which can be passed to gem5 with the –debug-flags option.

The available flags are:

Debug flag Unit which will generate debugging output
Activity Debug ActivityMonitor actions
Branch Fetch2 and Execute branch prediction decisions
MinorCPU CPU global actions such as wakeup/thread suspension
Decode Decode
MinorExec Execute behaviour
Fetch Fetch1 and Fetch2
MinorInterrupt Execute interrupt handling
MinorMem Execute memory interactions
MinorScoreboard Execute scoreboard activity
MinorTrace Generate MinorTrace cyclic state trace output (see below)
MinorTiming MinorTiming instruction timing modification operations

The group flag Minor enables all the flags beginning with Minor.

MinorTrace and

The debug flag MinorTrace causes cycle-by-cycle state data to be printed which can then be processed and viewed by the tool. This output is very verbose and so it is recommended it only be used for small examples.

MinorTrace format

There are three types of line outputted by MinorTrace:

MinorTrace - Ticked unit cycle state

For example:

 110000: system.cpu.dcachePort: MinorTrace: state=MemoryRunning in_tlb_mem=0/0

For each time step, the MinorTrace flag will cause one MinorTrace line to be printed for every named element in the model.

MinorInst - summaries of instructions issued by Decode


For example:

 140000: system.cpu.execute: MinorInst: id=0/1.1/1/1.1 addr=0x5c \
                             inst="  mov r0, #0" class=IntAlu

MinorInst lines are currently only generated for instructions which are committed.

MinorLine - summaries of line fetches issued by Fetch1


For example:

  92000: system.cpu.icachePort: MinorLine: id=0/1.1/1 size=36 \
                                vaddr=0x5c paddr=0x5c

Minorview (util/ can be used to visualise the data created by MinorTrace.

usage: [-h] [--picture picture-file] [--prefix name]
                   [--start-time time] [--end-time time] [--mini-views]

Minor visualiser

positional arguments:

optional arguments:
  -h, --help            show this help message and exit
  --picture picture-file
                        markup file containing blob information (default:
  --prefix name         name prefix in trace for CPU to be visualised
                        (default: system.cpu)
  --start-time time     time of first event to load from file
  --end-time time       time of last event to load from file
  --mini-views          show tiny views of the next 10 time steps

Raw debugging output can be passed to as the event-file. It will pick out the MinorTrace lines and use other lines where units in the simulation are named (such as system.cpu.dcachePort in the above example) will appear as ‘comments’ when units are clicked on the visualiser.

Clicking on a unit which contains instructions or lines will bring up a speech bubble giving extra information derived from the MinorInst/MinorLine lines.

–start-time and –end-time allow only sections of debug files to be loaded.

–prefix allows the name prefix of the CPU to be inspected to be supplied. This defaults to system.cpu.

In the visualiser, The buttons Start, End, Back, Forward, Play and Stop can be used to control the displayed simulation time.

The diagonally striped coloured blocks are showing the InstId of the instruction or line they represent. Note that lines in Fetch1 and f1ToF2.F only show the id fields of a line and that instructions in Fetch2, f2ToD, and decode.inputBuffer do not yet have execute sequence numbers. The T/S.P/L/F.E buttons can be used to toggle parts of InstId on and off to make it easier to understand the display. Useful combinations are:

Combination Reason
E just show the final execute sequence number
F/E show the instruction-related numbers
S/P show just the stream-related numbers (watch the stream sequence change with branches and not change with predicted branches)
S/E show instructions and their stream

The key to the right shows all the displayable colours (some of the colour choices are quite bad!):

Symbol Meaning
U Uknown data
B Blocked stage
- Bubble
E Empty queue slot
R Reserved queue slot
F Fault
r Read (used as the leftmost stripe on data in the dcachePort)
w Write “ “
0 to 9 last decimal digit of the corresponding data
    ,---------------.         .--------------.  *U
    | |=|->|=|->|=| |         ||=|||->||->|| |  *-  <- Fetch queues/LSQ
    `---------------'         `--------------'  *R
    === ======                                  *w  <- Activity/Stage activity
                              ,--------------.  *1
    ,--.      ,.      ,.      | ============ |  *3  <- Scoreboard
    |  |-\[]-\||-\[]-\||-\[]-\| ============ |  *5  <- Execute::inFlightInsts
    |  | :[] :||-/[]-/||-/[]-/| -. --------  |  *7
    |  |-/[]-/||  ^   ||      |  | --------- |  *9
    |  |      ||  |   ||      |  | ------    |
[]->|  |    ->||  |   ||      |  | ----      |
    |  |<-[]<-||<-+-<-||<-[]<-|  | ------    |->[] <- Execute to Fetch1,
    '--`      `'  ^   `'      | -' ------    |        Fetch2 branch data
             ---. |  ---.     `--------------'
             ---' |  ---'       ^       ^
                  |   ^         |       `------------ Execute
  MinorBuffer ----' input       `-------------------- Execute input buffer

Stages show the colours of the instructions currently being generated/processed.

Forward FIFOs between stages show the data being pushed into them at the current tick (to the left), the data in transit, and the data available at their outputs (to the right).

The backwards FIFO between Fetch2 and Fetch1 shows branch prediction data.

In general, all displayed data is correct at the end of a cycle’s activity at the time indicated but before the inter-stage FIFOs are ticked. Each FIFO has, therefore an extra slot to show the asserted new input data, and all the data currently within the FIFO.

Input buffers for each stage are shown below the corresponding stage and show the contents of those buffers as horizontal strips. Strips marked as reserved (cyan by default) are reserved to be filled by the previous stage. An input buffer with all reserved or occupied slots will, therefore, block the previous stage from generating output.

Fetch queues and LSQ show the lines/instructions in the queues of each interface and show the number of lines/instructions in TLB and memory in the two striped colours of the top of their frames.

Inside Execute, the horizontal bars represent the individual FU pipelines. The vertical bar to the left is the input buffer and the bar to the right, the instructions committed this cycle. The background of Execute shows instructions which are being committed this cycle in their original FU pipeline positions.

The strip at the top of the Execute block shows the current streamSeqNum that Execute is committing. A similar stripe at the top of Fetch1 shows that stage’s expected streamSeqNum and the stripe at the top of Fetch2 shows its issuing predictionSeqNum.

The scoreboard shows the number of instructions in flight which will commit a result to the register in the position shown. The scoreboard contains slots for each integer and floating point register.

The Execute::inFlightInsts queue shows all the instructions in flight in Execute with the oldest instruction (the next instruction to be committed) to the right.

Stage activity shows the signalled activity (as E/1) for each stage (with CPU miscellaneous activity to the left)

Activity show a count of stage and pipe activity.

minor.pic format

The minor.pic file (src/minor/minor.pic) describes the layout of the models blocks on the visualiser. Its format is described in the supplied minor.pic file.