Files
SplitBit-Emulator/Source/Emulator/cpu.h
T
Anachronaut c3188ed657 Seventy becomes seventy one: a machine that can wait
HALT is terminal - stepCPU returns at once when the Halt Flag is up, so a
halted machine does not execute, service devices, or take an interrupt -
and that has to stay true, because every test ends with a halt and "halted"
is how a program says it has finished. The consequence was that SplitBit
had no way to wait at all. Every wait was a spin, and a spin is bus
traffic: 11.5% of Type over a 14K file on a disk of ten thousand cycles,
after read-ahead had already hidden three quarters of the latency.

WAIT is 0xFE, one byte, no operands, sitting under HALT where the
instruction that almost stops the machine belongs. Three decisions in it:

- A line already standing means there is nothing to wait for, so WAIT does
  nothing. That is what makes test-then-wait race-free.
- Any line ends the wait, masked or not, so a program can sleep on a device
  it has no handler for and read its status afterwards. Masking says who
  answers a request, not whether it happened.
- A line that wakes the CPU without being dispatched is taken down by the
  WAIT. Left standing it would be found by the next WAIT, which would
  return at once - the program would spin exactly as before while looking
  as though it slept.

Waiting is NOT a Status bit, and that is the trap avoided rather than a
gap: Status rides into the interrupt frame and comes back out, so a machine
interrupted mid-wait would return from its handler still waiting, and wait
again for what it had already been given. An internal field instead.

Idle cycles are counted apart from bus cycles and the halt line says so
when there are any, which is what makes the difference observable at all -
with the line-clearing removed the total moves by ONE cycle, 20,100 against
20,099, and only the idle half changes, halving to 9,976. A test on
totals could never have seen it. Tests/terminal.sh asks that question,
being the file for things a recorded output cannot see, and fails with the
clear removed while "both reads finished" still passes.

Three collisions, all found by building it:

- 0xFE was the assembler's "not an instruction" sentinel. getOpcode now
  answers a negative NOT_AN_OPCODE, which is outside the range of every
  possible answer instead of inside the unused part of it.
- 0xFE was also what faultTest and faultResumeTest executed to provoke a
  fault. They now use 0xFD and say why, because they did not fail when it
  became an instruction - they HUNG, having started sleeping instead.
- Keys.asm has had a label called "wait" for a year, and mnemonics are
  matched uppercased. What that reported was "Branch without label" at the
  BRQ thirty lines away. The assembler now refuses a label that is already
  an instruction, at the label, by name; every instruction added takes a
  word out of the space of label names, so this will happen again.
2026-08-26 11:11:25 -04:00

103 lines
4.9 KiB
C

// cpu.h
// SplitBit CPU Emulator Core
// Written by Anachronaut
// 10/16/2024
#ifndef CPU_H
#define CPU_H
#include <stdint.h>
// How many Data Pointers the CPU has. The instructions that name one take a full
// byte to do it, so the encoding would allow up to 256. The limit here is the size
// of the register file and the cost of saving pointers across a CALL, not the
// instruction format. Must be a power of two, so that the selector can be masked
// down to a valid pointer.
#define DATA_POINTERS 4
// How many Data Pointers survive a CALL. The low numbered pointers are saved and
// restored around a subroutine; the rest are left alone, so a subroutine can use
// one to hand a pointer back to its caller the way Q hands back a byte. This is
// deliberately independent of DATA_POINTERS: adding more pointers should not make
// every CALL more expensive.
#define PRESERVED_DATA_POINTERS 3
#if PRESERVED_DATA_POINTERS > DATA_POINTERS
#error "Cannot preserve more Data Pointers than the CPU has."
#endif
// The bits of the Status register that mean something.
#define STATUS_CARRY 0x01 // An arithmetic result carried out of, or borrowed into, a byte.
#define STATUS_FAULT 0x02 // The CPU met a byte it could not decode, and stopped.
#define STATUS_INTERRUPT 0x04 // Hardware interrupts are enabled. Nothing reads this yet.
#define STATUS_HALT 0x80 // Execution has stopped, either from HALT or from a fault.
// What an interrupt puts on the Stack: the resume address, every Data Pointer, and
// every register the CPU has. The CALL frame leaves Q and DP3 alone, but that is a
// convention between a caller and the subroutine it called. An interrupt arrives in
// code that never agreed to give anything up, so it saves the lot.
#define INTERRUPT_FRAME_BYTES (2 + DATA_POINTERS * 2 + 4)
// Why the CPU stopped, when the Fault Flag is set. This is not something a program can
// read, and it is deliberately not a register: when a handler is installed, the vector
// it arrived through already says what happened, which is why the ISA has no fault
// cause. This exists for the case where nothing is installed and the machine is dead,
// so that whatever examines the wreckage can say something better than "it stopped".
typedef enum {
FAULT_NONE = 0,
FAULT_BAD_OPCODE, // A byte that does not decode to an instruction.
FAULT_NO_HANDLER, // Dispatched through a software vector with nothing in it.
FAULT_NO_DEVICE_HANDLER, // A device interrupted, and its vector was empty.
FAULT_DEVICE_REFUSED // A device refused, and nothing was installed to catch it.
} FaultCause;
// The struct containing the CPU registers.
typedef struct {
uint8_t A;
uint8_t B;
uint8_t Q;
uint8_t Status;
uint16_t ProgramCounter;
uint16_t DataPointer[DATA_POINTERS];
uint16_t StackPointer;
uint8_t *Program;
uint8_t *Data;
// ---- What the machine has cost so far ----
//
// ONE BUS ACCESS IS ONE CYCLE, and every access goes through it: fetching an opcode,
// fetching the bytes after it, reading or writing Data Memory, pushing or popping the
// Stack, and reaching a device port. Nothing is overlapped - no fetching the next
// instruction while this one finishes - because that is a thing hardware may or may
// not do and this is the model to design against before deciding.
//
// It replaces counting instructions. Counting instructions said an RSTA and a SETD
// cost the same, and that a CALL moving ten bytes of Stack cost what a branch costs,
// which is not true of any machine anybody could build.
unsigned long busCycles;
// ---- And what it has cost while doing nothing ----
//
// Clocks spent inside WAIT, where the CPU is stopped and the bus is idle. They are
// counted because time still has to pass - a device that takes a while has to be able
// to finish - and they are counted SEPARATELY because they are not the same thing as
// work. A machine waiting on a disk is not using memory, and charging it as though it
// were is exactly the sort of dishonest number the bus count was built to replace.
unsigned long idleCycles;
// Whether the CPU is stopped in a WAIT, which is not a Status bit and must not become
// one: the Status register rides into the interrupt frame and comes back out of it, so
// a machine interrupted while waiting would return from the handler still waiting, and
// wait again for the thing it had already been given.
uint8_t Waiting;
// Set alongside the Fault Flag, and read only by whatever reports the stop.
uint8_t Fault; // A FaultCause.
uint8_t FaultVector; // Which vector was empty, when Fault is FAULT_NO_HANDLER.
} CPURegisters;
uint8_t executeOperation(uint8_t instruction, CPURegisters *cpu);
void initializeCPU(CPURegisters *cpu, uint8_t *programMemory, uint8_t *dataMemory);
void stepCPU(CPURegisters *cpu);
#endif // CPU_H