mirror of
https://github.com/st3fan/ewm.git
synced 2025-01-01 11:29:35 +00:00
406 lines
11 KiB
C
406 lines
11 KiB
C
// The MIT License (MIT)
|
|
//
|
|
// Copyright (c) 2015 Stefan Arentz - http://github.com/st3fan/ewm
|
|
//
|
|
// Permission is hereby granted, free of charge, to any person obtaining a copy
|
|
// of this software and associated documentation files (the "Software"), to deal
|
|
// in the Software without restriction, including without limitation the rights
|
|
// to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
|
|
// copies of the Software, and to permit persons to whom the Software is
|
|
// furnished to do so, subject to the following conditions:
|
|
//
|
|
// The above copyright notice and this permission notice shall be included in all
|
|
// copies or substantial portions of the Software.
|
|
//
|
|
// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
|
|
// IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
|
|
// FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
|
|
// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
|
|
// LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
|
|
// OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
|
|
// SOFTWARE.
|
|
|
|
#include <assert.h>
|
|
#include <errno.h>
|
|
#include <fcntl.h>
|
|
#include <stdio.h>
|
|
#include <string.h>
|
|
#include <stdint.h>
|
|
#include <stdbool.h>
|
|
#include <inttypes.h>
|
|
#include <stdlib.h>
|
|
#include <unistd.h>
|
|
#include <sys/stat.h>
|
|
|
|
#include "cpu.h"
|
|
#include "ins.h"
|
|
#include "mem.h"
|
|
#include "fmt.h"
|
|
|
|
/* Private API */
|
|
|
|
typedef void (*cpu_instruction_handler_t)(struct cpu_t *cpu);
|
|
typedef void (*cpu_instruction_handler_byte_t)(struct cpu_t *cpu, uint8_t oper);
|
|
typedef void (*cpu_instruction_handler_word_t)(struct cpu_t *cpu, uint16_t oper);
|
|
|
|
// Stack management.
|
|
|
|
void _cpu_push_byte(struct cpu_t *cpu, uint8_t b) {
|
|
mem_set_byte(cpu, 0x0100 + cpu->state.sp, b);
|
|
cpu->state.sp -= 1;
|
|
}
|
|
|
|
void _cpu_push_word(struct cpu_t *cpu, uint16_t w) {
|
|
_cpu_push_byte(cpu, (uint8_t) (w >> 8));
|
|
_cpu_push_byte(cpu, (uint8_t) w);
|
|
}
|
|
|
|
uint8_t _cpu_pull_byte(struct cpu_t *cpu) {
|
|
cpu->state.sp += 1;
|
|
return mem_get_byte(cpu, 0x0100 + cpu->state.sp);
|
|
}
|
|
|
|
uint16_t _cpu_pull_word(struct cpu_t *cpu) {
|
|
return (uint16_t) _cpu_pull_byte(cpu) | ((uint16_t) _cpu_pull_byte(cpu) << 8);
|
|
}
|
|
|
|
uint8_t _cpu_stack_free(struct cpu_t *cpu) {
|
|
return cpu->state.sp;
|
|
}
|
|
|
|
uint8_t _cpu_stack_used(struct cpu_t *cpu) {
|
|
return 0xff - cpu->state.sp;
|
|
}
|
|
|
|
// Because we keep the processor status bits in separate fields, we
|
|
// need a function to combine them into a single register. This is
|
|
// only used when we need to push the register on the stack for
|
|
// interupt handlers. If this turns out to be inefficient then they
|
|
// can be stored in their native form in a byte.
|
|
|
|
uint8_t _cpu_get_status(struct cpu_t *cpu) {
|
|
return 0x30
|
|
| (((cpu->state.n != 0) & 0x01) << 7)
|
|
| (((cpu->state.v != 0) & 0x01) << 6)
|
|
| (((cpu->state.b != 0) & 0x01) << 4)
|
|
| (((cpu->state.d != 0) & 0x01) << 3)
|
|
| (((cpu->state.i != 0) & 0x01) << 2)
|
|
| (((cpu->state.z != 0) & 0x01) << 1)
|
|
| (((cpu->state.c != 0) & 0x01) << 0);
|
|
}
|
|
|
|
void _cpu_set_status(struct cpu_t *cpu, uint8_t status) {
|
|
cpu->state.n = (status & (1 << 7));
|
|
cpu->state.v = (status & (1 << 6));
|
|
cpu->state.b = (status & (1 << 4));
|
|
cpu->state.d = (status & (1 << 3));
|
|
cpu->state.i = (status & (1 << 2));
|
|
cpu->state.z = (status & (1 << 1));
|
|
cpu->state.c = (status & (1 << 0));
|
|
}
|
|
|
|
static int cpu_execute_instruction(struct cpu_t *cpu) {
|
|
/* Trace code - Refactor into its own function or module */
|
|
char trace_instruction[256];
|
|
char trace_state[256];
|
|
char trace_stack[256];
|
|
|
|
if (cpu->trace) {
|
|
cpu_format_instruction(cpu, trace_instruction);
|
|
}
|
|
|
|
/* Fetch instruction */
|
|
struct cpu_instruction_t *i = &cpu->instructions[mem_get_byte(cpu, cpu->state.pc)];
|
|
if (i->handler == NULL) {
|
|
return EWM_CPU_ERR_UNIMPLEMENTED_INSTRUCTION;
|
|
}
|
|
|
|
// If strict mode and if we need the stack, check if that works out
|
|
if (cpu->strict && i->stack != 0) {
|
|
if (i->stack > 0) {
|
|
if (_cpu_stack_free(cpu) < i->stack) {
|
|
return EWM_CPU_ERR_STACK_OVERFLOW;
|
|
}
|
|
} else {
|
|
if (_cpu_stack_used(cpu) < -(i->stack)) {
|
|
return EWM_CPU_ERR_STACK_UNDERFLOW;
|
|
}
|
|
}
|
|
}
|
|
|
|
/* Remember the PC since some instructions modify it */
|
|
uint16_t pc = cpu->state.pc;
|
|
|
|
/* Advance PC */
|
|
if (pc == cpu->state.pc) {
|
|
cpu->state.pc += i->bytes;
|
|
}
|
|
|
|
/* Execute instruction */
|
|
switch (i->bytes) {
|
|
case 1:
|
|
((cpu_instruction_handler_t) i->handler)(cpu);
|
|
break;
|
|
case 2:
|
|
((cpu_instruction_handler_byte_t) i->handler)(cpu, mem_get_byte(cpu, pc+1));
|
|
break;
|
|
case 3:
|
|
((cpu_instruction_handler_word_t) i->handler)(cpu, mem_get_word(cpu, pc+1));
|
|
break;
|
|
}
|
|
|
|
if (cpu->trace) {
|
|
cpu_format_state(cpu, trace_state);
|
|
cpu_format_stack(cpu, trace_stack);
|
|
|
|
char bytes[10];
|
|
switch (i->bytes) {
|
|
case 1:
|
|
snprintf(bytes, sizeof bytes, "%.2X", mem_get_byte(cpu, pc));
|
|
break;
|
|
case 2:
|
|
snprintf(bytes, sizeof bytes, "%.2X %.2X", mem_get_byte(cpu, pc), mem_get_byte(cpu, pc+1));
|
|
break;
|
|
case 3:
|
|
snprintf(bytes, sizeof bytes, "%.2X %.2X %.2X", mem_get_byte(cpu, pc), mem_get_byte(cpu, pc+1), mem_get_byte(cpu, pc+2));
|
|
break;
|
|
}
|
|
|
|
fprintf(cpu->trace, "%.4X: %-8s %-14s %-20s %s\n",
|
|
pc, bytes, trace_instruction, trace_state, trace_stack);
|
|
}
|
|
|
|
return 0;
|
|
}
|
|
|
|
/* Public API */
|
|
|
|
void cpu_setup() {
|
|
for (int i = 0; i <= 255; i++) {
|
|
if (instructions_65C02[i].handler == NULL) {
|
|
instructions_65C02[i] = instructions[i];
|
|
}
|
|
}
|
|
}
|
|
|
|
void cpu_init(struct cpu_t *cpu, int model) {
|
|
memset(cpu, 0x00, sizeof(struct cpu_t));
|
|
cpu->model = model;
|
|
cpu->instructions = (cpu->model == EWM_CPU_MODEL_6502) ? instructions : instructions_65C02;
|
|
}
|
|
|
|
void cpu_shutdown(struct cpu_t *cpu) {
|
|
if (cpu->trace != NULL) {
|
|
(void) fclose(cpu->trace);
|
|
cpu->trace = NULL;
|
|
}
|
|
}
|
|
|
|
struct mem_t *cpu_add_mem(struct cpu_t *cpu, struct mem_t *mem) {
|
|
if (cpu->mem == NULL) {
|
|
cpu->mem = mem;
|
|
mem->next = NULL;
|
|
} else {
|
|
mem->next = cpu->mem;
|
|
cpu->mem = mem;
|
|
}
|
|
return mem;
|
|
}
|
|
|
|
// RAM Memory
|
|
|
|
static uint8_t _ram_read(struct cpu_t *cpu, struct mem_t *mem, uint16_t addr) {
|
|
return ((uint8_t*) mem->obj)[addr - mem->start];
|
|
}
|
|
|
|
static void _ram_write(struct cpu_t *cpu, struct mem_t *mem, uint16_t addr, uint8_t b) {
|
|
((uint8_t*) mem->obj)[addr - mem->start] = b;
|
|
}
|
|
|
|
struct mem_t *cpu_add_ram(struct cpu_t *cpu, uint16_t start, uint16_t end) {
|
|
return cpu_add_ram_data(cpu, start, end, calloc(end-start+1, 0x01));
|
|
}
|
|
|
|
struct mem_t *cpu_add_ram_data(struct cpu_t *cpu, uint16_t start, uint16_t end, uint8_t *data) {
|
|
struct mem_t *mem = (struct mem_t*) malloc(sizeof(struct mem_t));
|
|
mem->enabled = true;
|
|
mem->flags = MEM_FLAGS_READ | MEM_FLAGS_WRITE;
|
|
mem->obj = data;
|
|
mem->start = start;
|
|
mem->end = end;
|
|
mem->read_handler = _ram_read;
|
|
mem->write_handler = _ram_write;
|
|
mem->next = NULL;
|
|
return cpu_add_mem(cpu, mem);
|
|
}
|
|
|
|
struct mem_t *cpu_add_ram_file(struct cpu_t *cpu, uint16_t start, char *path) {
|
|
int fd = open(path, O_RDONLY);
|
|
if (fd == -1) {
|
|
return NULL;
|
|
}
|
|
|
|
struct stat file_info;
|
|
if (fstat(fd, &file_info) == -1) {
|
|
close(fd);
|
|
return NULL;
|
|
}
|
|
|
|
if (file_info.st_size > (64 * 1024 - start)) {
|
|
close(fd);
|
|
return NULL;
|
|
}
|
|
|
|
char *data = calloc(file_info.st_size, 1);
|
|
if (read(fd, data, file_info.st_size) != file_info.st_size) {
|
|
close(fd);
|
|
return NULL;
|
|
}
|
|
|
|
close(fd);
|
|
|
|
return cpu_add_ram_data(cpu, start, start + file_info.st_size - 1, (uint8_t*) data);
|
|
}
|
|
|
|
// ROM Memory
|
|
|
|
static uint8_t _rom_read(struct cpu_t *cpu, struct mem_t *mem, uint16_t addr) {
|
|
return ((uint8_t*) mem->obj)[addr - mem->start];
|
|
}
|
|
|
|
struct mem_t *cpu_add_rom_data(struct cpu_t *cpu, uint16_t start, uint16_t end, uint8_t *data) {
|
|
struct mem_t *mem = (struct mem_t*) malloc(sizeof(struct mem_t));
|
|
mem->enabled = true;
|
|
mem->flags = MEM_FLAGS_READ;
|
|
mem->obj = data;
|
|
mem->start = start;
|
|
mem->end = end;
|
|
mem->read_handler = _rom_read;
|
|
mem->write_handler = NULL;
|
|
mem->next = NULL;
|
|
return cpu_add_mem(cpu, mem);
|
|
}
|
|
|
|
struct mem_t *cpu_add_rom_file(struct cpu_t *cpu, uint16_t start, char *path) {
|
|
int fd = open(path, O_RDONLY);
|
|
if (fd == -1) {
|
|
return NULL;
|
|
}
|
|
|
|
struct stat file_info;
|
|
if (fstat(fd, &file_info) == -1) {
|
|
close(fd);
|
|
return NULL;
|
|
}
|
|
|
|
if (file_info.st_size > (64 * 1024 - start)) {
|
|
close(fd);
|
|
return NULL;
|
|
}
|
|
|
|
char *data = calloc(file_info.st_size, 1);
|
|
if (read(fd, data, file_info.st_size) != file_info.st_size) {
|
|
close(fd);
|
|
return NULL;
|
|
}
|
|
|
|
close(fd);
|
|
|
|
return cpu_add_rom_data(cpu, start, start + file_info.st_size - 1, (uint8_t*) data);
|
|
}
|
|
|
|
// IO Memory
|
|
|
|
struct mem_t *cpu_add_iom(struct cpu_t *cpu, uint16_t start, uint16_t end, void *obj, mem_read_handler_t read_handler, mem_write_handler_t write_handler) {
|
|
struct mem_t *mem = (struct mem_t*) malloc(sizeof(struct mem_t));
|
|
mem->enabled = true;
|
|
mem->flags = MEM_FLAGS_READ | MEM_FLAGS_WRITE;
|
|
mem->obj = obj;
|
|
mem->start = start;
|
|
mem->end = end;
|
|
mem->read_handler = read_handler;
|
|
mem->write_handler = write_handler;
|
|
mem->next = NULL;
|
|
return cpu_add_mem(cpu, mem);
|
|
}
|
|
|
|
void cpu_strict(struct cpu_t *cpu, bool strict) {
|
|
cpu->strict = strict;
|
|
}
|
|
|
|
int cpu_trace(struct cpu_t *cpu, char *path) {
|
|
if (cpu->trace != NULL) {
|
|
(void) fclose(cpu->trace);
|
|
cpu->trace = NULL;
|
|
}
|
|
|
|
if (path != NULL) {
|
|
cpu->trace = fopen(path, "w");
|
|
if (cpu->trace == NULL) {
|
|
return errno;
|
|
}
|
|
}
|
|
|
|
return 0;
|
|
}
|
|
|
|
void cpu_reset(struct cpu_t *cpu) {
|
|
cpu->state.pc = mem_get_word(cpu, EWM_VECTOR_RES);
|
|
cpu->state.a = 0x00;
|
|
cpu->state.x = 0x00;
|
|
cpu->state.y = 0x00;
|
|
cpu->state.n = 0;
|
|
cpu->state.v = 0;
|
|
cpu->state.b = 0;
|
|
cpu->state.d = 0;
|
|
cpu->state.i = 1;
|
|
cpu->state.z = 0;
|
|
cpu->state.c = 0;
|
|
cpu->state.sp = 0xff;
|
|
}
|
|
|
|
int cpu_irq(struct cpu_t *cpu) {
|
|
if (cpu->strict && _cpu_stack_free(cpu) < 3) {
|
|
return EWM_CPU_ERR_STACK_OVERFLOW;
|
|
}
|
|
|
|
_cpu_push_word(cpu, cpu->state.pc + 1); // TODO +1?? Spec says +2 but test fails then
|
|
_cpu_push_byte(cpu, _cpu_get_status(cpu));
|
|
cpu->state.i = 1;
|
|
cpu->state.pc = mem_get_word(cpu, EWM_VECTOR_IRQ);
|
|
|
|
return 0;
|
|
}
|
|
|
|
int cpu_nmi(struct cpu_t *cpu) {
|
|
if (cpu->strict && _cpu_stack_free(cpu) < 3) {
|
|
return EWM_CPU_ERR_STACK_OVERFLOW;
|
|
}
|
|
|
|
_cpu_push_word(cpu, cpu->state.pc + 1); // TODO +1?? Spec says +2 but test fails then
|
|
_cpu_push_byte(cpu, _cpu_get_status(cpu));
|
|
cpu->state.i = 1;
|
|
cpu->state.pc = mem_get_word(cpu, EWM_VECTOR_NMI);
|
|
|
|
return 0;
|
|
}
|
|
|
|
int cpu_run(struct cpu_t *cpu) {
|
|
uint64_t instruction_count = 0;
|
|
int err = 0;
|
|
while ((err = cpu_execute_instruction(cpu)) == 0) {
|
|
/* TODO: Tick? */
|
|
instruction_count++;
|
|
}
|
|
return err;
|
|
}
|
|
|
|
int cpu_boot(struct cpu_t *cpu) {
|
|
cpu_reset(cpu);
|
|
return cpu_run(cpu);
|
|
}
|
|
|
|
int cpu_step(struct cpu_t *cpu) {
|
|
return cpu_execute_instruction(cpu);
|
|
}
|