diff options
author | Michael Pavone <pavone@retrodev.com> | 2014-03-02 16:34:29 -0800 |
---|---|---|
committer | Michael Pavone <pavone@retrodev.com> | 2014-03-02 16:34:29 -0800 |
commit | 48d6afdd1c9aaafc5b8ce59c584e360d89571a5b (patch) | |
tree | 2974a8819a2c80d7fc0ebc2e71b9a8e1370863fe /m68k_to_x86.c | |
parent | 8664df5a5d2870c67036666b66ce0d2f9fcdd1eb (diff) |
Initial stab at separating the generic parts of the 68K core from the host-cpu specific parts.
Diffstat (limited to 'm68k_to_x86.c')
-rw-r--r-- | m68k_to_x86.c | 4626 |
1 files changed, 0 insertions, 4626 deletions
diff --git a/m68k_to_x86.c b/m68k_to_x86.c deleted file mode 100644 index 295ff8e..0000000 --- a/m68k_to_x86.c +++ /dev/null @@ -1,4626 +0,0 @@ -/* - Copyright 2013 Michael Pavone - This file is part of BlastEm. - BlastEm is free software distributed under the terms of the GNU General Public License version 3 or greater. See COPYING for full license text. -*/ -#include "gen_x86.h" -#include "m68k_core.h" -#include "68kinst.h" -#include "mem.h" -#include "backend.h" -#include <stdio.h> -#include <stddef.h> -#include <stdlib.h> -#include <string.h> - -#define BUS 4 -#define PREDEC_PENALTY 2 - -#define CYCLES RAX -#define LIMIT RBP -#define CONTEXT RSI -#define SCRATCH1 RCX - -#ifdef X86_64 -#define SCRATCH2 RDI -#else -#define SCRATCH2 RBX -#endif - -enum { - FLAG_X, - FLAG_N, - FLAG_Z, - FLAG_V, - FLAG_C -}; - -char disasm_buf[1024]; - -m68k_context * sync_components(m68k_context * context, uint32_t address); - -void m68k_invalid(); -void bcd_add(); -void bcd_sub(); - - -void set_flag(m68k_options * opts, uint8_t val, uint8_t flag) -{ - if (opts->flag_regs[flag] >= 0) { - mov_ir(&opts->gen.code, val, opts->flag_regs[flag], SZ_B); - } else { - int8_t offset = offsetof(m68k_context, flags) + flag; - if (offset) { - mov_irdisp(&opts->gen.code, val, opts->gen.context_reg, offset, SZ_B); - } else { - mov_irind(&opts->gen.code, val, opts->gen.context_reg, SZ_B); - } - } -} - -void set_flag_cond(m68k_options *opts, uint8_t cond, uint8_t flag) -{ - if (opts->flag_regs[flag] >= 0) { - setcc_r(&opts->gen.code, cond, opts->flag_regs[flag]); - } else { - int8_t offset = offsetof(m68k_context, flags) + flag; - if (offset) { - setcc_rdisp(&opts->gen.code, cond, opts->gen.context_reg, offset); - } else { - setcc_rind(&opts->gen.code, cond, opts->gen.context_reg); - } - } -} - -void check_flag(m68k_options *opts, uint8_t flag) -{ - if (opts->flag_regs[flag] >= 0) { - cmp_ir(&opts->gen.code, 0, opts->flag_regs[flag], SZ_B); - } else { - cmp_irdisp(&opts->gen.code, 0, opts->gen.context_reg, offsetof(m68k_context, flags) + flag, SZ_B); - } -} - -void flag_to_reg(m68k_options *opts, uint8_t flag, uint8_t reg) -{ - if (opts->flag_regs[flag] >= 0) { - mov_rr(&opts->gen.code, opts->flag_regs[flag], reg, SZ_B); - } else { - int8_t offset = offsetof(m68k_context, flags) + flag; - if (offset) { - mov_rdispr(&opts->gen.code, opts->gen.context_reg, offset, reg, SZ_B); - } else { - mov_rindr(&opts->gen.code, opts->gen.context_reg, reg, SZ_B); - } - } -} - -void reg_to_flag(m68k_options *opts, uint8_t reg, uint8_t flag) -{ - if (opts->flag_regs[flag] >= 0) { - mov_rr(&opts->gen.code, reg, opts->flag_regs[flag], SZ_B); - } else { - int8_t offset = offsetof(m68k_context, flags) + flag; - if (offset) { - mov_rrdisp(&opts->gen.code, reg, opts->gen.context_reg, offset, SZ_B); - } else { - mov_rrind(&opts->gen.code, reg, opts->gen.context_reg, SZ_B); - } - } -} - -void flag_to_flag(m68k_options *opts, uint8_t flag1, uint8_t flag2) -{ - code_info *code = &opts->gen.code; - if (opts->flag_regs[flag1] >= 0 && opts->flag_regs[flag2] >= 0) { - mov_rr(code, opts->flag_regs[flag1], opts->flag_regs[flag2], SZ_B); - } else if(opts->flag_regs[flag1] >= 0) { - mov_rrdisp(code, opts->flag_regs[flag1], opts->gen.context_reg, offsetof(m68k_context, flags) + flag2, SZ_B); - } else if (opts->flag_regs[flag2] >= 0) { - mov_rdispr(code, opts->gen.context_reg, offsetof(m68k_context, flags) + flag1, opts->flag_regs[flag2], SZ_B); - } else { - push_r(code, opts->gen.scratch1); - mov_rdispr(code, opts->gen.context_reg, offsetof(m68k_context, flags) + flag1, opts->gen.scratch1, SZ_B); - mov_rrdisp(code, opts->gen.scratch1, opts->gen.context_reg, offsetof(m68k_context, flags) + flag2, SZ_B); - pop_r(code, opts->gen.scratch1); - } -} - -void flag_to_carry(m68k_options * opts, uint8_t flag) -{ - if (opts->flag_regs[flag] >= 0) { - bt_ir(&opts->gen.code, 0, opts->flag_regs[flag], SZ_B); - } else { - bt_irdisp(&opts->gen.code, 0, opts->gen.context_reg, offsetof(m68k_context, flags) + flag, SZ_B); - } -} - -void or_flag_to_reg(m68k_options *opts, uint8_t flag, uint8_t reg) -{ - if (opts->flag_regs[flag] >= 0) { - or_rr(&opts->gen.code, opts->flag_regs[flag], reg, SZ_B); - } else { - or_rdispr(&opts->gen.code, opts->gen.context_reg, offsetof(m68k_context, flags) + flag, reg, SZ_B); - } -} - -void xor_flag_to_reg(m68k_options *opts, uint8_t flag, uint8_t reg) -{ - if (opts->flag_regs[flag] >= 0) { - xor_rr(&opts->gen.code, opts->flag_regs[flag], reg, SZ_B); - } else { - xor_rdispr(&opts->gen.code, opts->gen.context_reg, offsetof(m68k_context, flags) + flag, reg, SZ_B); - } -} - -void xor_flag(m68k_options *opts, uint8_t val, uint8_t flag) -{ - if (opts->flag_regs[flag] >= 0) { - xor_ir(&opts->gen.code, val, opts->flag_regs[flag], SZ_B); - } else { - xor_irdisp(&opts->gen.code, val, opts->gen.context_reg, offsetof(m68k_context, flags) + flag, SZ_B); - } -} - -void cmp_flags(m68k_options *opts, uint8_t flag1, uint8_t flag2) -{ - code_info *code = &opts->gen.code; - if (opts->flag_regs[flag1] >= 0 && opts->flag_regs[flag2] >= 0) { - cmp_rr(code, opts->flag_regs[flag1], opts->flag_regs[flag2], SZ_B); - } else if(opts->flag_regs[flag1] >= 0 || opts->flag_regs[flag2] >= 0) { - if (opts->flag_regs[flag2] >= 0) { - uint8_t tmp = flag1; - flag1 = flag2; - flag2 = tmp; - } - cmp_rrdisp(code, opts->flag_regs[flag1], opts->gen.context_reg, offsetof(m68k_context, flags) + flag2, SZ_B); - } else { - mov_rdispr(code, opts->gen.context_reg, offsetof(m68k_context, flags) + flag1, opts->gen.scratch1, SZ_B); - cmp_rrdisp(code, opts->gen.scratch1, opts->gen.context_reg, offsetof(m68k_context, flags) + flag2, SZ_B); - } -} - -int8_t native_reg(m68k_op_info * op, m68k_options * opts) -{ - if (op->addr_mode == MODE_REG) { - return opts->dregs[op->params.regs.pri]; - } - if (op->addr_mode == MODE_AREG) { - return opts->aregs[op->params.regs.pri]; - } - return -1; -} - -//must be called with an m68k_op_info that uses a register -size_t reg_offset(m68k_op_info *op) -{ - if (op->addr_mode == MODE_REG) { - return offsetof(m68k_context, dregs) + sizeof(uint32_t) * op->params.regs.pri; - } - return offsetof(m68k_context, aregs) + sizeof(uint32_t) * op->params.regs.pri; -} - -void print_regs_exit(m68k_context * context) -{ - printf("XNZVC\n%d%d%d%d%d\n", context->flags[0], context->flags[1], context->flags[2], context->flags[3], context->flags[4]); - for (int i = 0; i < 8; i++) { - printf("d%d: %X\n", i, context->dregs[i]); - } - for (int i = 0; i < 8; i++) { - printf("a%d: %X\n", i, context->aregs[i]); - } - exit(0); -} - -void m68k_read_size(m68k_options *opts, uint8_t size) -{ - switch (size) - { - case OPSIZE_BYTE: - call(&opts->gen.code, opts->read_8); - break; - case OPSIZE_WORD: - call(&opts->gen.code, opts->read_16); - break; - case OPSIZE_LONG: - call(&opts->gen.code, opts->read_32); - break; - } -} - -void m68k_write_size(m68k_options *opts, uint8_t size) -{ - switch (size) - { - case OPSIZE_BYTE: - call(&opts->gen.code, opts->write_8); - break; - case OPSIZE_WORD: - call(&opts->gen.code, opts->write_16); - break; - case OPSIZE_LONG: - call(&opts->gen.code, opts->write_32_highfirst); - break; - } -} - -void translate_m68k_src(m68kinst * inst, x86_ea * ea, m68k_options * opts) -{ - code_info *code = &opts->gen.code; - int8_t reg = native_reg(&(inst->src), opts); - uint8_t sec_reg; - int32_t dec_amount,inc_amount; - if (reg >= 0) { - ea->mode = MODE_REG_DIRECT; - if (inst->dst.addr_mode == MODE_AREG && inst->extra.size == OPSIZE_WORD) { - movsx_rr(code, reg, opts->gen.scratch1, SZ_W, SZ_D); - ea->base = opts->gen.scratch1; - } else { - ea->base = reg; - } - return; - } - switch (inst->src.addr_mode) - { - case MODE_REG: - case MODE_AREG: - //We only get one memory parameter, so if the dst operand is a register in memory, - //we need to copy this to a temp register first - reg = native_reg(&(inst->dst), opts); - if (reg >= 0 || inst->dst.addr_mode == MODE_UNUSED || !(inst->dst.addr_mode == MODE_REG || inst->dst.addr_mode == MODE_AREG) - || inst->op == M68K_EXG) { - - ea->mode = MODE_REG_DISPLACE8; - ea->base = opts->gen.context_reg; - ea->disp = reg_offset(&(inst->src)); - } else { - if (inst->dst.addr_mode == MODE_AREG && inst->extra.size == OPSIZE_WORD) { - movsx_rdispr(code, opts->gen.context_reg, reg_offset(&(inst->src)), opts->gen.scratch1, SZ_W, SZ_D); - } else { - mov_rdispr(code, opts->gen.context_reg, reg_offset(&(inst->src)), opts->gen.scratch1, inst->extra.size); - } - ea->mode = MODE_REG_DIRECT; - ea->base = opts->gen.scratch1; - //we're explicitly handling the areg dest here, so we exit immediately - return; - } - break; - case MODE_AREG_PREDEC: - dec_amount = inst->extra.size == OPSIZE_WORD ? 2 : (inst->extra.size == OPSIZE_LONG ? 4 : (inst->src.params.regs.pri == 7 ? 2 :1)); - cycles(&opts->gen, PREDEC_PENALTY); - if (opts->aregs[inst->src.params.regs.pri] >= 0) { - sub_ir(code, dec_amount, opts->aregs[inst->src.params.regs.pri], SZ_D); - } else { - sub_irdisp(code, dec_amount, opts->gen.context_reg, reg_offset(&(inst->src)), SZ_D); - } - case MODE_AREG_INDIRECT: - case MODE_AREG_POSTINC: - if (opts->aregs[inst->src.params.regs.pri] >= 0) { - mov_rr(code, opts->aregs[inst->src.params.regs.pri], opts->gen.scratch1, SZ_D); - } else { - mov_rdispr(code, opts->gen.context_reg, reg_offset(&(inst->src)), opts->gen.scratch1, SZ_D); - } - m68k_read_size(opts, inst->extra.size); - - if (inst->src.addr_mode == MODE_AREG_POSTINC) { - inc_amount = inst->extra.size == OPSIZE_WORD ? 2 : (inst->extra.size == OPSIZE_LONG ? 4 : (inst->src.params.regs.pri == 7 ? 2 : 1)); - if (opts->aregs[inst->src.params.regs.pri] >= 0) { - add_ir(code, inc_amount, opts->aregs[inst->src.params.regs.pri], SZ_D); - } else { - add_irdisp(code, inc_amount, opts->gen.context_reg, reg_offset(&(inst->src)), SZ_D); - } - } - ea->mode = MODE_REG_DIRECT; - ea->base = (inst->dst.addr_mode == MODE_AREG_PREDEC && inst->op != M68K_MOVE) ? opts->gen.scratch2 : opts->gen.scratch1; - break; - case MODE_AREG_DISPLACE: - cycles(&opts->gen, BUS); - if (opts->aregs[inst->src.params.regs.pri] >= 0) { - mov_rr(code, opts->aregs[inst->src.params.regs.pri], opts->gen.scratch1, SZ_D); - } else { - mov_rdispr(code, opts->gen.context_reg, reg_offset(&(inst->src)), opts->gen.scratch1, SZ_D); - } - add_ir(code, inst->src.params.regs.displacement, opts->gen.scratch1, SZ_D); - m68k_read_size(opts, inst->extra.size); - - ea->mode = MODE_REG_DIRECT; - ea->base = opts->gen.scratch1; - break; - case MODE_AREG_INDEX_DISP8: - cycles(&opts->gen, 6); - if (opts->aregs[inst->src.params.regs.pri] >= 0) { - mov_rr(code, opts->aregs[inst->src.params.regs.pri], opts->gen.scratch1, SZ_D); - } else { - mov_rdispr(code, opts->gen.context_reg, reg_offset(&(inst->src)), opts->gen.scratch1, SZ_D); - } - sec_reg = (inst->src.params.regs.sec >> 1) & 0x7; - if (inst->src.params.regs.sec & 1) { - if (inst->src.params.regs.sec & 0x10) { - if (opts->aregs[sec_reg] >= 0) { - add_rr(code, opts->aregs[sec_reg], opts->gen.scratch1, SZ_D); - } else { - add_rdispr(code, opts->gen.context_reg, offsetof(m68k_context, aregs) + sizeof(uint32_t)*sec_reg, opts->gen.scratch1, SZ_D); - } - } else { - if (opts->dregs[sec_reg] >= 0) { - add_rr(code, opts->dregs[sec_reg], opts->gen.scratch1, SZ_D); - } else { - add_rdispr(code, opts->gen.context_reg, offsetof(m68k_context, dregs) + sizeof(uint32_t)*sec_reg, opts->gen.scratch1, SZ_D); - } - } - } else { - if (inst->src.params.regs.sec & 0x10) { - if (opts->aregs[sec_reg] >= 0) { - movsx_rr(code, opts->aregs[sec_reg], opts->gen.scratch2, SZ_W, SZ_D); - } else { - movsx_rdispr(code, opts->gen.context_reg, offsetof(m68k_context, aregs) + sizeof(uint32_t)*sec_reg, opts->gen.scratch2, SZ_W, SZ_D); - } - } else { - if (opts->dregs[sec_reg] >= 0) { - movsx_rr(code, opts->dregs[sec_reg], opts->gen.scratch2, SZ_W, SZ_D); - } else { - movsx_rdispr(code, opts->gen.context_reg, offsetof(m68k_context, dregs) + sizeof(uint32_t)*sec_reg, opts->gen.scratch2, SZ_W, SZ_D); - } - } - add_rr(code, opts->gen.scratch2, opts->gen.scratch1, SZ_D); - } - if (inst->src.params.regs.displacement) { - add_ir(code, inst->src.params.regs.displacement, opts->gen.scratch1, SZ_D); - } - m68k_read_size(opts, inst->extra.size); - - ea->mode = MODE_REG_DIRECT; - ea->base = opts->gen.scratch1; - break; - case MODE_PC_DISPLACE: - cycles(&opts->gen, BUS); - mov_ir(code, inst->src.params.regs.displacement + inst->address+2, opts->gen.scratch1, SZ_D); - m68k_read_size(opts, inst->extra.size); - - ea->mode = MODE_REG_DIRECT; - ea->base = opts->gen.scratch1; - break; - case MODE_PC_INDEX_DISP8: - cycles(&opts->gen, 6); - mov_ir(code, inst->address+2, opts->gen.scratch1, SZ_D); - sec_reg = (inst->src.params.regs.sec >> 1) & 0x7; - if (inst->src.params.regs.sec & 1) { - if (inst->src.params.regs.sec & 0x10) { - if (opts->aregs[sec_reg] >= 0) { - add_rr(code, opts->aregs[sec_reg], opts->gen.scratch1, SZ_D); - } else { - add_rdispr(code, opts->gen.context_reg, offsetof(m68k_context, aregs) + sizeof(uint32_t)*sec_reg, opts->gen.scratch1, SZ_D); - } - } else { - if (opts->dregs[sec_reg] >= 0) { - add_rr(code, opts->dregs[sec_reg], opts->gen.scratch1, SZ_D); - } else { - add_rdispr(code, opts->gen.context_reg, offsetof(m68k_context, dregs) + sizeof(uint32_t)*sec_reg, opts->gen.scratch1, SZ_D); - } - } - } else { - if (inst->src.params.regs.sec & 0x10) { - if (opts->aregs[sec_reg] >= 0) { - movsx_rr(code, opts->aregs[sec_reg], opts->gen.scratch2, SZ_W, SZ_D); - } else { - movsx_rdispr(code, opts->gen.context_reg, offsetof(m68k_context, aregs) + sizeof(uint32_t)*sec_reg, opts->gen.scratch2, SZ_W, SZ_D); - } - } else { - if (opts->dregs[sec_reg] >= 0) { - movsx_rr(code, opts->dregs[sec_reg], opts->gen.scratch2, SZ_W, SZ_D); - } else { - movsx_rdispr(code, opts->gen.context_reg, offsetof(m68k_context, dregs) + sizeof(uint32_t)*sec_reg, opts->gen.scratch2, SZ_W, SZ_D); - } - } - add_rr(code, opts->gen.scratch2, opts->gen.scratch1, SZ_D); - } - if (inst->src.params.regs.displacement) { - add_ir(code, inst->src.params.regs.displacement, opts->gen.scratch1, SZ_D); - } - m68k_read_size(opts, inst->extra.size); - - ea->mode = MODE_REG_DIRECT; - ea->base = opts->gen.scratch1; - break; - case MODE_ABSOLUTE: - case MODE_ABSOLUTE_SHORT: - cycles(&opts->gen, inst->src.addr_mode == MODE_ABSOLUTE ? BUS*2 : BUS); - mov_ir(code, inst->src.params.immed, opts->gen.scratch1, SZ_D); - m68k_read_size(opts, inst->extra.size); - - ea->mode = MODE_REG_DIRECT; - ea->base = opts->gen.scratch1; - break; - case MODE_IMMEDIATE: - case MODE_IMMEDIATE_WORD: - if (inst->variant != VAR_QUICK) { - cycles(&opts->gen, (inst->extra.size == OPSIZE_LONG && inst->src.addr_mode == MODE_IMMEDIATE) ? BUS*2 : BUS); - } - ea->mode = MODE_IMMED; - ea->disp = inst->src.params.immed; - if (inst->dst.addr_mode == MODE_AREG && inst->extra.size == OPSIZE_WORD && ea->disp & 0x8000) { - ea->disp |= 0xFFFF0000; - } - return; - default: - m68k_disasm(inst, disasm_buf); - printf("%X: %s\naddress mode %d not implemented (src)\n", inst->address, disasm_buf, inst->src.addr_mode); - exit(1); - } - if (inst->dst.addr_mode == MODE_AREG && inst->extra.size == OPSIZE_WORD) { - if (ea->mode == MODE_REG_DIRECT) { - movsx_rr(code, ea->base, opts->gen.scratch1, SZ_W, SZ_D); - } else { - movsx_rdispr(code, ea->base, ea->disp, opts->gen.scratch1, SZ_W, SZ_D); - ea->mode = MODE_REG_DIRECT; - } - ea->base = opts->gen.scratch1; - } -} - -void translate_m68k_dst(m68kinst * inst, x86_ea * ea, m68k_options * opts, uint8_t fake_read) -{ - code_info *code = &opts->gen.code; - int8_t reg = native_reg(&(inst->dst), opts), sec_reg; - int32_t dec_amount, inc_amount; - if (reg >= 0) { - ea->mode = MODE_REG_DIRECT; - ea->base = reg; - return; - } - switch (inst->dst.addr_mode) - { - case MODE_REG: - case MODE_AREG: - ea->mode = MODE_REG_DISPLACE8; - ea->base = opts->gen.context_reg; - ea->disp = reg_offset(&(inst->dst)); - break; - case MODE_AREG_PREDEC: - if (inst->src.addr_mode == MODE_AREG_PREDEC) { - push_r(code, opts->gen.scratch1); - } - dec_amount = inst->extra.size == OPSIZE_WORD ? 2 : (inst->extra.size == OPSIZE_LONG ? 4 : (inst->dst.params.regs.pri == 7 ? 2 : 1)); - if (opts->aregs[inst->dst.params.regs.pri] >= 0) { - sub_ir(code, dec_amount, opts->aregs[inst->dst.params.regs.pri], SZ_D); - } else { - sub_irdisp(code, dec_amount, opts->gen.context_reg, reg_offset(&(inst->dst)), SZ_D); - } - case MODE_AREG_INDIRECT: - case MODE_AREG_POSTINC: - if (fake_read) { - cycles(&opts->gen, inst->extra.size == OPSIZE_LONG ? 8 : 4); - } else { - if (opts->aregs[inst->dst.params.regs.pri] >= 0) { - mov_rr(code, opts->aregs[inst->dst.params.regs.pri], opts->gen.scratch1, SZ_D); - } else { - mov_rdispr(code, opts->gen.context_reg, reg_offset(&(inst->dst)), opts->gen.scratch1, SZ_D); - } - m68k_read_size(opts, inst->extra.size); - } - if (inst->src.addr_mode == MODE_AREG_PREDEC) { - //restore src operand to opts->gen.scratch2 - pop_r(code, opts->gen.scratch2); - } else { - //save reg value in opts->gen.scratch2 so we can use it to save the result in memory later - if (opts->aregs[inst->dst.params.regs.pri] >= 0) { - mov_rr(code, opts->aregs[inst->dst.params.regs.pri], opts->gen.scratch2, SZ_D); - } else { - mov_rdispr(code, opts->gen.context_reg, reg_offset(&(inst->dst)), opts->gen.scratch2, SZ_D); - } - } - - if (inst->dst.addr_mode == MODE_AREG_POSTINC) { - inc_amount = inst->extra.size == OPSIZE_WORD ? 2 : (inst->extra.size == OPSIZE_LONG ? 4 : (inst->dst.params.regs.pri == 7 ? 2 : 1)); - if (opts->aregs[inst->dst.params.regs.pri] >= 0) { - add_ir(code, inc_amount, opts->aregs[inst->dst.params.regs.pri], SZ_D); - } else { - add_irdisp(code, inc_amount, opts->gen.context_reg, reg_offset(&(inst->dst)), SZ_D); - } - } - ea->mode = MODE_REG_DIRECT; - ea->base = opts->gen.scratch1; - break; - case MODE_AREG_DISPLACE: - cycles(&opts->gen, fake_read ? BUS+(inst->extra.size == OPSIZE_LONG ? BUS*2 : BUS) : BUS); - reg = fake_read ? opts->gen.scratch2 : opts->gen.scratch1; - if (opts->aregs[inst->dst.params.regs.pri] >= 0) { - mov_rr(code, opts->aregs[inst->dst.params.regs.pri], reg, SZ_D); - } else { - mov_rdispr(code, opts->gen.context_reg, reg_offset(&(inst->dst)), reg, SZ_D); - } - add_ir(code, inst->dst.params.regs.displacement, reg, SZ_D); - if (!fake_read) { - push_r(code, opts->gen.scratch1); - m68k_read_size(opts, inst->extra.size); - pop_r(code, opts->gen.scratch2); - } - ea->mode = MODE_REG_DIRECT; - ea->base = opts->gen.scratch1; - break; - case MODE_AREG_INDEX_DISP8: - cycles(&opts->gen, fake_read ? (6 + inst->extra.size == OPSIZE_LONG ? 8 : 4) : 6); - if (opts->aregs[inst->dst.params.regs.pri] >= 0) { - mov_rr(code, opts->aregs[inst->dst.params.regs.pri], opts->gen.scratch1, SZ_D); - } else { - mov_rdispr(code, opts->gen.context_reg, reg_offset(&(inst->dst)), opts->gen.scratch1, SZ_D); - } - sec_reg = (inst->dst.params.regs.sec >> 1) & 0x7; - if (inst->dst.params.regs.sec & 1) { - if (inst->dst.params.regs.sec & 0x10) { - if (opts->aregs[sec_reg] >= 0) { - add_rr(code, opts->aregs[sec_reg], opts->gen.scratch1, SZ_D); - } else { - add_rdispr(code, opts->gen.context_reg, offsetof(m68k_context, aregs) + sizeof(uint32_t)*sec_reg, opts->gen.scratch1, SZ_D); - } - } else { - if (opts->dregs[sec_reg] >= 0) { - add_rr(code, opts->dregs[sec_reg], opts->gen.scratch1, SZ_D); - } else { - add_rdispr(code, opts->gen.context_reg, offsetof(m68k_context, dregs) + sizeof(uint32_t)*sec_reg, opts->gen.scratch1, SZ_D); - } - } - } else { - if (inst->dst.params.regs.sec & 0x10) { - if (opts->aregs[sec_reg] >= 0) { - movsx_rr(code, opts->aregs[sec_reg], opts->gen.scratch2, SZ_W, SZ_D); - } else { - movsx_rdispr(code, opts->gen.context_reg, offsetof(m68k_context, aregs) + sizeof(uint32_t)*sec_reg, opts->gen.scratch2, SZ_W, SZ_D); - } - } else { - if (opts->dregs[sec_reg] >= 0) { - movsx_rr(code, opts->dregs[sec_reg], opts->gen.scratch2, SZ_W, SZ_D); - } else { - movsx_rdispr(code, opts->gen.context_reg, offsetof(m68k_context, dregs) + sizeof(uint32_t)*sec_reg, opts->gen.scratch2, SZ_W, SZ_D); - } - } - add_rr(code, opts->gen.scratch2, opts->gen.scratch1, SZ_D); - } - if (inst->dst.params.regs.displacement) { - add_ir(code, inst->dst.params.regs.displacement, opts->gen.scratch1, SZ_D); - } - if (fake_read) { - mov_rr(code, opts->gen.scratch1, opts->gen.scratch2, SZ_D); - } else { - push_r(code, opts->gen.scratch1); - m68k_read_size(opts, inst->extra.size); - pop_r(code, opts->gen.scratch2); - } - ea->mode = MODE_REG_DIRECT; - ea->base = opts->gen.scratch1; - break; - case MODE_PC_DISPLACE: - cycles(&opts->gen, fake_read ? BUS+(inst->extra.size == OPSIZE_LONG ? BUS*2 : BUS) : BUS); - mov_ir(code, inst->dst.params.regs.displacement + inst->address+2, fake_read ? opts->gen.scratch2 : opts->gen.scratch1, SZ_D); - if (!fake_read) { - push_r(code, opts->gen.scratch1); - m68k_read_size(opts, inst->extra.size); - pop_r(code, opts->gen.scratch2); - } - ea->mode = MODE_REG_DIRECT; - ea->base = opts->gen.scratch1; - break; - case MODE_PC_INDEX_DISP8: - cycles(&opts->gen, fake_read ? (6 + inst->extra.size == OPSIZE_LONG ? 8 : 4) : 6); - mov_ir(code, inst->address+2, opts->gen.scratch1, SZ_D); - sec_reg = (inst->dst.params.regs.sec >> 1) & 0x7; - if (inst->dst.params.regs.sec & 1) { - if (inst->dst.params.regs.sec & 0x10) { - if (opts->aregs[sec_reg] >= 0) { - add_rr(code, opts->aregs[sec_reg], opts->gen.scratch1, SZ_D); - } else { - add_rdispr(code, opts->gen.context_reg, offsetof(m68k_context, aregs) + sizeof(uint32_t)*sec_reg, opts->gen.scratch1, SZ_D); - } - } else { - if (opts->dregs[sec_reg] >= 0) { - add_rr(code, opts->dregs[sec_reg], opts->gen.scratch1, SZ_D); - } else { - add_rdispr(code, opts->gen.context_reg, offsetof(m68k_context, dregs) + sizeof(uint32_t)*sec_reg, opts->gen.scratch1, SZ_D); - } - } - } else { - if (inst->dst.params.regs.sec & 0x10) { - if (opts->aregs[sec_reg] >= 0) { - movsx_rr(code, opts->aregs[sec_reg], opts->gen.scratch2, SZ_W, SZ_D); - } else { - movsx_rdispr(code, opts->gen.context_reg, offsetof(m68k_context, aregs) + sizeof(uint32_t)*sec_reg, opts->gen.scratch2, SZ_W, SZ_D); - } - } else { - if (opts->dregs[sec_reg] >= 0) { - movsx_rr(code, opts->dregs[sec_reg], opts->gen.scratch2, SZ_W, SZ_D); - } else { - movsx_rdispr(code, opts->gen.context_reg, offsetof(m68k_context, dregs) + sizeof(uint32_t)*sec_reg, opts->gen.scratch2, SZ_W, SZ_D); - } - } - add_rr(code, opts->gen.scratch2, opts->gen.scratch1, SZ_D); - } - if (inst->dst.params.regs.displacement) { - add_ir(code, inst->dst.params.regs.displacement, opts->gen.scratch1, SZ_D); - } - if (fake_read) { - mov_rr(code, opts->gen.scratch1, opts->gen.scratch2, SZ_D); - } else { - push_r(code, opts->gen.scratch1); - m68k_read_size(opts, inst->extra.size); - pop_r(code, opts->gen.scratch2); - } - ea->mode = MODE_REG_DIRECT; - ea->base = opts->gen.scratch1; - break; - case MODE_ABSOLUTE: - case MODE_ABSOLUTE_SHORT: - //Add cycles for reading address from instruction stream - cycles(&opts->gen, (inst->dst.addr_mode == MODE_ABSOLUTE ? BUS*2 : BUS) + (fake_read ? (inst->extra.size == OPSIZE_LONG ? BUS*2 : BUS) : 0)); - mov_ir(code, inst->dst.params.immed, fake_read ? opts->gen.scratch2 : opts->gen.scratch1, SZ_D); - if (!fake_read) { - push_r(code, opts->gen.scratch1); - m68k_read_size(opts, inst->extra.size); - pop_r(code, opts->gen.scratch2); - } - ea->mode = MODE_REG_DIRECT; - ea->base = opts->gen.scratch1; - break; - default: - m68k_disasm(inst, disasm_buf); - printf("%X: %s\naddress mode %d not implemented (dst)\n", inst->address, disasm_buf, inst->dst.addr_mode); - exit(1); - } -} - -void m68k_save_result(m68kinst * inst, m68k_options * opts) -{ - code_info *code = &opts->gen.code; - if (inst->dst.addr_mode != MODE_REG && inst->dst.addr_mode != MODE_AREG) { - if (inst->dst.addr_mode == MODE_AREG_PREDEC && inst->src.addr_mode == MODE_AREG_PREDEC && inst->op != M68K_MOVE) { - if (opts->aregs[inst->dst.params.regs.pri] >= 0) { - mov_rr(code, opts->aregs[inst->dst.params.regs.pri], opts->gen.scratch2, SZ_D); - } else { - mov_rdispr(code, opts->gen.context_reg, reg_offset(&(inst->dst)), opts->gen.scratch2, SZ_D); - } - } - switch (inst->extra.size) - { - case OPSIZE_BYTE: - call(code, opts->write_8); - break; - case OPSIZE_WORD: - call(code, opts->write_16); - break; - case OPSIZE_LONG: - call(code, opts->write_32_lowfirst); - break; - } - } -} - -code_ptr get_native_address(native_map_slot * native_code_map, uint32_t address) -{ - address &= 0xFFFFFF; - address /= 2; - uint32_t chunk = address / NATIVE_CHUNK_SIZE; - if (!native_code_map[chunk].base) { - return NULL; - } - uint32_t offset = address % NATIVE_CHUNK_SIZE; - if (native_code_map[chunk].offsets[offset] == INVALID_OFFSET || native_code_map[chunk].offsets[offset] == EXTENSION_WORD) { - return NULL; - } - return native_code_map[chunk].base + native_code_map[chunk].offsets[offset]; -} - -code_ptr get_native_from_context(m68k_context * context, uint32_t address) -{ - return get_native_address(context->native_code_map, address); -} - -uint32_t get_instruction_start(native_map_slot * native_code_map, uint32_t address) -{ - address &= 0xFFFFFF; - address /= 2; - uint32_t chunk = address / NATIVE_CHUNK_SIZE; - if (!native_code_map[chunk].base) { - return 0; - } - uint32_t offset = address % NATIVE_CHUNK_SIZE; - if (native_code_map[chunk].offsets[offset] == INVALID_OFFSET) { - return 0; - } - while (native_code_map[chunk].offsets[offset] == EXTENSION_WORD) { - --address; - chunk = address / NATIVE_CHUNK_SIZE; - offset = address % NATIVE_CHUNK_SIZE; - } - return address*2; -} - -void map_native_address(m68k_context * context, uint32_t address, code_ptr native_addr, uint8_t size, uint8_t native_size) -{ - native_map_slot * native_code_map = context->native_code_map; - m68k_options * opts = context->options; - address &= 0xFFFFFF; - if (address > 0xE00000) { - context->ram_code_flags[(address & 0xC000) >> 14] |= 1 << ((address & 0x3800) >> 11); - if (((address & 0x3FFF) + size) & 0xC000) { - context->ram_code_flags[((address+size) & 0xC000) >> 14] |= 1 << (((address+size) & 0x3800) >> 11); - } - uint32_t slot = (address & 0xFFFF)/1024; - if (!opts->gen.ram_inst_sizes[slot]) { - opts->gen.ram_inst_sizes[slot] = malloc(sizeof(uint8_t) * 512); - } - opts->gen.ram_inst_sizes[slot][((address & 0xFFFF)/2)%512] = native_size; - } - address/= 2; - uint32_t chunk = address / NATIVE_CHUNK_SIZE; - if (!native_code_map[chunk].base) { - native_code_map[chunk].base = native_addr; - native_code_map[chunk].offsets = malloc(sizeof(int32_t) * NATIVE_CHUNK_SIZE); - memset(native_code_map[chunk].offsets, 0xFF, sizeof(int32_t) * NATIVE_CHUNK_SIZE); - } - uint32_t offset = address % NATIVE_CHUNK_SIZE; - native_code_map[chunk].offsets[offset] = native_addr-native_code_map[chunk].base; - for(address++,size-=2; size; address++,size-=2) { - chunk = address / NATIVE_CHUNK_SIZE; - offset = address % NATIVE_CHUNK_SIZE; - if (!native_code_map[chunk].base) { - native_code_map[chunk].base = native_addr; - native_code_map[chunk].offsets = malloc(sizeof(int32_t) * NATIVE_CHUNK_SIZE); - memset(native_code_map[chunk].offsets, 0xFF, sizeof(int32_t) * NATIVE_CHUNK_SIZE); - } - native_code_map[chunk].offsets[offset] = EXTENSION_WORD; - } -} - -uint8_t get_native_inst_size(m68k_options * opts, uint32_t address) -{ - if (address < 0xE00000) { - return 0; - } - uint32_t slot = (address & 0xFFFF)/1024; - return opts->gen.ram_inst_sizes[slot][((address & 0xFFFF)/2)%512]; -} - -void translate_m68k_move(m68k_options * opts, m68kinst * inst) -{ - code_info *code = &opts->gen.code; - int8_t reg, flags_reg, sec_reg; - uint8_t dir = 0; - int32_t offset; - int32_t inc_amount, dec_amount; - x86_ea src; - translate_m68k_src(inst, &src, opts); - reg = native_reg(&(inst->dst), opts); - if (inst->dst.addr_mode != MODE_AREG) { - //update statically set flags - set_flag(opts, 0, FLAG_V); - set_flag(opts, 0, FLAG_C); - } - - if (inst->dst.addr_mode != MODE_AREG) { - if (src.mode == MODE_REG_DIRECT) { - flags_reg = src.base; - } else { - if (reg >= 0) { - flags_reg = reg; - } else { - if(src.mode == MODE_REG_DISPLACE8) { - mov_rdispr(code, src.base, src.disp, opts->gen.scratch1, inst->extra.size); - } else { - mov_ir(code, src.disp, opts->gen.scratch1, inst->extra.size); - } - src.mode = MODE_REG_DIRECT; - flags_reg = src.base = opts->gen.scratch1; - } - } - } - uint8_t size = inst->extra.size; - switch(inst->dst.addr_mode) - { - case MODE_AREG: - size = OPSIZE_LONG; - case MODE_REG: - if (reg >= 0) { - if (src.mode == MODE_REG_DIRECT) { - mov_rr(code, src.base, reg, size); - } else if (src.mode == MODE_REG_DISPLACE8) { - mov_rdispr(code, src.base, src.disp, reg, size); - } else { - mov_ir(code, src.disp, reg, size); - } - } else if(src.mode == MODE_REG_DIRECT) { - mov_rrdisp(code, src.base, opts->gen.context_reg, reg_offset(&(inst->dst)), size); - } else { - mov_irdisp(code, src.disp, opts->gen.context_reg, reg_offset(&(inst->dst)), size); - } - if (inst->dst.addr_mode != MODE_AREG) { - cmp_ir(code, 0, flags_reg, size); - set_flag_cond(opts, CC_Z, FLAG_Z); - set_flag_cond(opts, CC_S, FLAG_N); - } - break; - case MODE_AREG_PREDEC: - dec_amount = inst->extra.size == OPSIZE_WORD ? 2 : (inst->extra.size == OPSIZE_LONG ? 4 : (inst->dst.params.regs.pri == 7 ? 2 : 1)); - if (opts->aregs[inst->dst.params.regs.pri] >= 0) { - sub_ir(code, dec_amount, opts->aregs[inst->dst.params.regs.pri], SZ_D); - } else { - sub_irdisp(code, dec_amount, opts->gen.context_reg, reg_offset(&(inst->dst)), SZ_D); - } - case MODE_AREG_INDIRECT: - case MODE_AREG_POSTINC: - if (opts->aregs[inst->dst.params.regs.pri] >= 0) { - mov_rr(code, opts->aregs[inst->dst.params.regs.pri], opts->gen.scratch2, SZ_D); - } else { - mov_rdispr(code, opts->gen.context_reg, reg_offset(&(inst->dst)), opts->gen.scratch2, SZ_D); - } - if (src.mode == MODE_REG_DIRECT) { - if (src.base != opts->gen.scratch1) { - mov_rr(code, src.base, opts->gen.scratch1, inst->extra.size); - } - } else if (src.mode == MODE_REG_DISPLACE8) { - mov_rdispr(code, src.base, src.disp, opts->gen.scratch1, inst->extra.size); - } else { - mov_ir(code, src.disp, opts->gen.scratch1, inst->extra.size); - } - if (inst->dst.addr_mode != MODE_AREG) { - cmp_ir(code, 0, flags_reg, inst->extra.size); - set_flag_cond(opts, CC_Z, FLAG_Z); - set_flag_cond(opts, CC_S, FLAG_N); - } - m68k_write_size(opts, inst->extra.size); - - if (inst->dst.addr_mode == MODE_AREG_POSTINC) { - inc_amount = inst->extra.size == OPSIZE_WORD ? 2 : (inst->extra.size == OPSIZE_LONG ? 4 : (inst->dst.params.regs.pri == 7 ? 2 : 1)); - if (opts->aregs[inst->dst.params.regs.pri] >= 0) { - add_ir(code, inc_amount, opts->aregs[inst->dst.params.regs.pri], SZ_D); - } else { - add_irdisp(code, inc_amount, opts->gen.context_reg, reg_offset(&(inst->dst)), SZ_D); - } - } - break; - case MODE_AREG_DISPLACE: - cycles(&opts->gen, BUS); - if (opts->aregs[inst->dst.params.regs.pri] >= 0) { - mov_rr(code, opts->aregs[inst->dst.params.regs.pri], opts->gen.scratch2, SZ_D); - } else { - mov_rdispr(code, opts->gen.context_reg, reg_offset(&(inst->dst)), opts->gen.scratch2, SZ_D); - } - add_ir(code, inst->dst.params.regs.displacement, opts->gen.scratch2, SZ_D); - if (src.mode == MODE_REG_DIRECT) { - if (src.base != opts->gen.scratch1) { - mov_rr(code, src.base, opts->gen.scratch1, inst->extra.size); - } - } else if (src.mode == MODE_REG_DISPLACE8) { - mov_rdispr(code, src.base, src.disp, opts->gen.scratch1, inst->extra.size); - } else { - mov_ir(code, src.disp, opts->gen.scratch1, inst->extra.size); - } - if (inst->dst.addr_mode != MODE_AREG) { - cmp_ir(code, 0, flags_reg, inst->extra.size); - set_flag_cond(opts, CC_Z, FLAG_Z); - set_flag_cond(opts, CC_S, FLAG_N); - } - m68k_write_size(opts, inst->extra.size); - break; - case MODE_AREG_INDEX_DISP8: - cycles(&opts->gen, 6);//TODO: Check to make sure this is correct - if (opts->aregs[inst->dst.params.regs.pri] >= 0) { - mov_rr(code, opts->aregs[inst->dst.params.regs.pri], opts->gen.scratch2, SZ_D); - } else { - mov_rdispr(code, opts->gen.context_reg, reg_offset(&(inst->dst)), opts->gen.scratch2, SZ_D); - } - sec_reg = (inst->dst.params.regs.sec >> 1) & 0x7; - if (inst->dst.params.regs.sec & 1) { - if (inst->dst.params.regs.sec & 0x10) { - if (opts->aregs[sec_reg] >= 0) { - add_rr(code, opts->aregs[sec_reg], opts->gen.scratch2, SZ_D); - } else { - add_rdispr(code, opts->gen.context_reg, offsetof(m68k_context, aregs) + sizeof(uint32_t)*sec_reg, opts->gen.scratch2, SZ_D); - } - } else { - if (opts->dregs[sec_reg] >= 0) { - add_rr(code, opts->dregs[sec_reg], opts->gen.scratch2, SZ_D); - } else { - add_rdispr(code, opts->gen.context_reg, offsetof(m68k_context, dregs) + sizeof(uint32_t)*sec_reg, opts->gen.scratch2, SZ_D); - } - } - } else { - if (src.base == opts->gen.scratch1) { - push_r(code, opts->gen.scratch1); - } - if (inst->dst.params.regs.sec & 0x10) { - if (opts->aregs[sec_reg] >= 0) { - movsx_rr(code, opts->aregs[sec_reg], opts->gen.scratch1, SZ_W, SZ_D); - } else { - movsx_rdispr(code, opts->gen.context_reg, offsetof(m68k_context, aregs) + sizeof(uint32_t)*sec_reg, opts->gen.scratch1, SZ_W, SZ_D); - } - } else { - if (opts->dregs[sec_reg] >= 0) { - movsx_rr(code, opts->dregs[sec_reg], opts->gen.scratch1, SZ_W, SZ_D); - } else { - movsx_rdispr(code, opts->gen.context_reg, offsetof(m68k_context, dregs) + sizeof(uint32_t)*sec_reg, opts->gen.scratch1, SZ_W, SZ_D); - } - } - add_rr(code, opts->gen.scratch1, opts->gen.scratch2, SZ_D); - if (src.base == opts->gen.scratch1) { - pop_r(code, opts->gen.scratch1); - } - } - if (inst->dst.params.regs.displacement) { - add_ir(code, inst->dst.params.regs.displacement, opts->gen.scratch2, SZ_D); - } - if (src.mode == MODE_REG_DIRECT) { - if (src.base != opts->gen.scratch1) { - mov_rr(code, src.base, opts->gen.scratch1, inst->extra.size); - } - } else if (src.mode == MODE_REG_DISPLACE8) { - mov_rdispr(code, src.base, src.disp, opts->gen.scratch1, inst->extra.size); - } else { - mov_ir(code, src.disp, opts->gen.scratch1, inst->extra.size); - } - if (inst->dst.addr_mode != MODE_AREG) { - cmp_ir(code, 0, flags_reg, inst->extra.size); - set_flag_cond(opts, CC_Z, FLAG_Z); - set_flag_cond(opts, CC_S, FLAG_N); - } - m68k_write_size(opts, inst->extra.size); - break; - case MODE_PC_DISPLACE: - cycles(&opts->gen, BUS); - mov_ir(code, inst->dst.params.regs.displacement + inst->address+2, opts->gen.scratch2, SZ_D); - if (src.mode == MODE_REG_DIRECT) { - if (src.base != opts->gen.scratch1) { - mov_rr(code, src.base, opts->gen.scratch1, inst->extra.size); - } - } else if (src.mode == MODE_REG_DISPLACE8) { - mov_rdispr(code, src.base, src.disp, opts->gen.scratch1, inst->extra.size); - } else { - mov_ir(code, src.disp, opts->gen.scratch1, inst->extra.size); - } - if (inst->dst.addr_mode != MODE_AREG) { - cmp_ir(code, 0, flags_reg, inst->extra.size); - set_flag_cond(opts, CC_Z, FLAG_Z); - set_flag_cond(opts, CC_S, FLAG_N); - } - m68k_write_size(opts, inst->extra.size); - break; - case MODE_PC_INDEX_DISP8: - cycles(&opts->gen, 6);//TODO: Check to make sure this is correct - mov_ir(code, inst->address, opts->gen.scratch2, SZ_D); - sec_reg = (inst->dst.params.regs.sec >> 1) & 0x7; - if (inst->dst.params.regs.sec & 1) { - if (inst->dst.params.regs.sec & 0x10) { - if (opts->aregs[sec_reg] >= 0) { - add_rr(code, opts->aregs[sec_reg], opts->gen.scratch2, SZ_D); - } else { - add_rdispr(code, opts->gen.context_reg, offsetof(m68k_context, aregs) + sizeof(uint32_t)*sec_reg, opts->gen.scratch2, SZ_D); - } - } else { - if (opts->dregs[sec_reg] >= 0) { - add_rr(code, opts->dregs[sec_reg], opts->gen.scratch2, SZ_D); - } else { - add_rdispr(code, opts->gen.context_reg, offsetof(m68k_context, dregs) + sizeof(uint32_t)*sec_reg, opts->gen.scratch2, SZ_D); - } - } - } else { - if (src.base == opts->gen.scratch1) { - push_r(code, opts->gen.scratch1); - } - if (inst->dst.params.regs.sec & 0x10) { - if (opts->aregs[sec_reg] >= 0) { - movsx_rr(code, opts->aregs[sec_reg], opts->gen.scratch1, SZ_W, SZ_D); - } else { - movsx_rdispr(code, opts->gen.context_reg, offsetof(m68k_context, aregs) + sizeof(uint32_t)*sec_reg, opts->gen.scratch1, SZ_W, SZ_D); - } - } else { - if (opts->dregs[sec_reg] >= 0) { - movsx_rr(code, opts->dregs[sec_reg], opts->gen.scratch1, SZ_W, SZ_D); - } else { - movsx_rdispr(code, opts->gen.context_reg, offsetof(m68k_context, dregs) + sizeof(uint32_t)*sec_reg, opts->gen.scratch1, SZ_W, SZ_D); - } - } - add_rr(code, opts->gen.scratch1, opts->gen.scratch2, SZ_D); - if (src.base == opts->gen.scratch1) { - pop_r(code, opts->gen.scratch1); - } - } - if (inst->dst.params.regs.displacement) { - add_ir(code, inst->dst.params.regs.displacement, opts->gen.scratch2, SZ_D); - } - if (src.mode == MODE_REG_DIRECT) { - if (src.base != opts->gen.scratch1) { - mov_rr(code, src.base, opts->gen.scratch1, inst->extra.size); - } - } else if (src.mode == MODE_REG_DISPLACE8) { - mov_rdispr(code, src.base, src.disp, opts->gen.scratch1, inst->extra.size); - } else { - mov_ir(code, src.disp, opts->gen.scratch1, inst->extra.size); - } - if (inst->dst.addr_mode != MODE_AREG) { - cmp_ir(code, 0, flags_reg, inst->extra.size); - set_flag_cond(opts, CC_Z, FLAG_Z); - set_flag_cond(opts, CC_S, FLAG_N); - } - m68k_write_size(opts, inst->extra.size); - break; - case MODE_ABSOLUTE: - case MODE_ABSOLUTE_SHORT: - if (src.mode == MODE_REG_DIRECT) { - if (src.base != opts->gen.scratch1) { - mov_rr(code, src.base, opts->gen.scratch1, inst->extra.size); - } - } else if (src.mode == MODE_REG_DISPLACE8) { - mov_rdispr(code, src.base, src.disp, opts->gen.scratch1, inst->extra.size); - } else { - mov_ir(code, src.disp, opts->gen.scratch1, inst->extra.size); - } - if (inst->dst.addr_mode == MODE_ABSOLUTE) { - cycles(&opts->gen, BUS*2); - } else { - cycles(&opts->gen, BUS); - } - mov_ir(code, inst->dst.params.immed, opts->gen.scratch2, SZ_D); - if (inst->dst.addr_mode != MODE_AREG) { - cmp_ir(code, 0, flags_reg, inst->extra.size); - set_flag_cond(opts, CC_Z, FLAG_Z); - set_flag_cond(opts, CC_S, FLAG_N); - } - m68k_write_size(opts, inst->extra.size); - break; - default: - m68k_disasm(inst, disasm_buf); - printf("%X: %s\naddress mode %d not implemented (move dst)\n", inst->address, disasm_buf, inst->dst.addr_mode); - exit(1); - } - - //add cycles for prefetch - cycles(&opts->gen, BUS); -} - -void translate_m68k_movem(m68k_options * opts, m68kinst * inst) -{ - code_info *code = &opts->gen.code; - int8_t bit,reg,sec_reg; - uint8_t early_cycles; - if(inst->src.addr_mode == MODE_REG) { - //reg to mem - early_cycles = 8; - int8_t dir; - switch (inst->dst.addr_mode) - { - case MODE_AREG_INDIRECT: - case MODE_AREG_PREDEC: - if (opts->aregs[inst->dst.params.regs.pri] >= 0) { - mov_rr(code, opts->aregs[inst->dst.params.regs.pri], opts->gen.scratch2, SZ_D); - } else { - mov_rdispr(code, opts->gen.context_reg, reg_offset(&(inst->dst)), opts->gen.scratch2, SZ_D); - } - break; - case MODE_AREG_DISPLACE: - early_cycles += BUS; - reg = opts->gen.scratch2; - if (opts->aregs[inst->dst.params.regs.pri] >= 0) { - mov_rr(code, opts->aregs[inst->dst.params.regs.pri], opts->gen.scratch2, SZ_D); - } else { - mov_rdispr(code, opts->gen.context_reg, reg_offset(&(inst->dst)), opts->gen.scratch2, SZ_D); - } - add_ir(code, inst->dst.params.regs.displacement, opts->gen.scratch2, SZ_D); - break; - case MODE_AREG_INDEX_DISP8: - early_cycles += 6; - if (opts->aregs[inst->dst.params.regs.pri] >= 0) { - mov_rr(code, opts->aregs[inst->dst.params.regs.pri], opts->gen.scratch2, SZ_D); - } else { - mov_rdispr(code, opts->gen.context_reg, reg_offset(&(inst->dst)), opts->gen.scratch2, SZ_D); - } - sec_reg = (inst->dst.params.regs.sec >> 1) & 0x7; - if (inst->dst.params.regs.sec & 1) { - if (inst->dst.params.regs.sec & 0x10) { - if (opts->aregs[sec_reg] >= 0) { - add_rr(code, opts->aregs[sec_reg], opts->gen.scratch2, SZ_D); - } else { - add_rdispr(code, opts->gen.context_reg, offsetof(m68k_context, aregs) + sizeof(uint32_t)*sec_reg, opts->gen.scratch2, SZ_D); - } - } else { - if (opts->dregs[sec_reg] >= 0) { - add_rr(code, opts->dregs[sec_reg], opts->gen.scratch2, SZ_D); - } else { - add_rdispr(code, opts->gen.context_reg, offsetof(m68k_context, dregs) + sizeof(uint32_t)*sec_reg, opts->gen.scratch2, SZ_D); - } - } - } else { - if (inst->dst.params.regs.sec & 0x10) { - if (opts->aregs[sec_reg] >= 0) { - movsx_rr(code, opts->aregs[sec_reg], opts->gen.scratch1, SZ_W, SZ_D); - } else { - movsx_rdispr(code, opts->gen.context_reg, offsetof(m68k_context, aregs) + sizeof(uint32_t)*sec_reg, opts->gen.scratch1, SZ_W, SZ_D); - } - } else { - if (opts->dregs[sec_reg] >= 0) { - movsx_rr(code, opts->dregs[sec_reg], opts->gen.scratch1, SZ_W, SZ_D); - } else { - movsx_rdispr(code, opts->gen.context_reg, offsetof(m68k_context, dregs) + sizeof(uint32_t)*sec_reg, opts->gen.scratch1, SZ_W, SZ_D); - } - } - add_rr(code, opts->gen.scratch1, opts->gen.scratch2, SZ_D); - } - if (inst->dst.params.regs.displacement) { - add_ir(code, inst->dst.params.regs.displacement, opts->gen.scratch2, SZ_D); - } - break; - case MODE_PC_DISPLACE: - early_cycles += BUS; - mov_ir(code, inst->dst.params.regs.displacement + inst->address+2, opts->gen.scratch2, SZ_D); - break; - case MODE_PC_INDEX_DISP8: - early_cycles += 6; - mov_ir(code, inst->address+2, opts->gen.scratch2, SZ_D); - sec_reg = (inst->dst.params.regs.sec >> 1) & 0x7; - if (inst->dst.params.regs.sec & 1) { - if (inst->dst.params.regs.sec & 0x10) { - if (opts->aregs[sec_reg] >= 0) { - add_rr(code, opts->aregs[sec_reg], opts->gen.scratch2, SZ_D); - } else { - add_rdispr(code, opts->gen.context_reg, offsetof(m68k_context, aregs) + sizeof(uint32_t)*sec_reg, opts->gen.scratch2, SZ_D); - } - } else { - if (opts->dregs[sec_reg] >= 0) { - add_rr(code, opts->dregs[sec_reg], opts->gen.scratch2, SZ_D); - } else { - add_rdispr(code, opts->gen.context_reg, offsetof(m68k_context, dregs) + sizeof(uint32_t)*sec_reg, opts->gen.scratch2, SZ_D); - } - } - } else { - if (inst->dst.params.regs.sec & 0x10) { - if (opts->aregs[sec_reg] >= 0) { - movsx_rr(code, opts->aregs[sec_reg], opts->gen.scratch1, SZ_W, SZ_D); - } else { - movsx_rdispr(code, opts->gen.context_reg, offsetof(m68k_context, aregs) + sizeof(uint32_t)*sec_reg, opts->gen.scratch1, SZ_W, SZ_D); - } - } else { - if (opts->dregs[sec_reg] >= 0) { - movsx_rr(code, opts->dregs[sec_reg], opts->gen.scratch1, SZ_W, SZ_D); - } else { - movsx_rdispr(code, opts->gen.context_reg, offsetof(m68k_context, dregs) + sizeof(uint32_t)*sec_reg, opts->gen.scratch1, SZ_W, SZ_D); - } - } - add_rr(code, opts->gen.scratch1, opts->gen.scratch2, SZ_D); - } - if (inst->dst.params.regs.displacement) { - add_ir(code, inst->dst.params.regs.displacement, opts->gen.scratch2, SZ_D); - } - break; - case MODE_ABSOLUTE: - early_cycles += 4; - case MODE_ABSOLUTE_SHORT: - early_cycles += 4; - mov_ir(code, inst->dst.params.immed, opts->gen.scratch2, SZ_D); - break; - default: - m68k_disasm(inst, disasm_buf); - printf("%X: %s\naddress mode %d not implemented (movem dst)\n", inst->address, disasm_buf, inst->dst.addr_mode); - exit(1); - } - if (inst->dst.addr_mode == MODE_AREG_PREDEC) { - reg = 15; - dir = -1; - } else { - reg = 0; - dir = 1; - } - cycles(&opts->gen, early_cycles); - for(bit=0; reg < 16 && reg >= 0; reg += dir, bit++) { - if (inst->src.params.immed & (1 << bit)) { - if (inst->dst.addr_mode == MODE_AREG_PREDEC) { - sub_ir(code, (inst->extra.size == OPSIZE_LONG) ? 4 : 2, opts->gen.scratch2, SZ_D); - } - push_r(code, opts->gen.scratch2); - if (reg > 7) { - if (opts->aregs[reg-8] >= 0) { - mov_rr(code, opts->aregs[reg-8], opts->gen.scratch1, inst->extra.size); - } else { - mov_rdispr(code, opts->gen.context_reg, offsetof(m68k_context, aregs) + sizeof(uint32_t) * (reg-8), opts->gen.scratch1, inst->extra.size); - } - } else { - if (opts->dregs[reg] >= 0) { - mov_rr(code, opts->dregs[reg], opts->gen.scratch1, inst->extra.size); - } else { - mov_rdispr(code, opts->gen.context_reg, offsetof(m68k_context, dregs) + sizeof(uint32_t) * (reg), opts->gen.scratch1, inst->extra.size); - } - } - if (inst->extra.size == OPSIZE_LONG) { - call(code, opts->write_32_lowfirst); - } else { - call(code, opts->write_16); - } - pop_r(code, opts->gen.scratch2); - if (inst->dst.addr_mode != MODE_AREG_PREDEC) { - add_ir(code, (inst->extra.size == OPSIZE_LONG) ? 4 : 2, opts->gen.scratch2, SZ_D); - } - } - } - if (inst->dst.addr_mode == MODE_AREG_PREDEC) { - if (opts->aregs[inst->dst.params.regs.pri] >= 0) { - mov_rr(code, opts->gen.scratch2, opts->aregs[inst->dst.params.regs.pri], SZ_D); - } else { - mov_rrdisp(code, opts->gen.scratch2, opts->gen.context_reg, reg_offset(&(inst->dst)), SZ_D); - } - } - } else { - //mem to reg - early_cycles = 4; - switch (inst->src.addr_mode) - { - case MODE_AREG_INDIRECT: - case MODE_AREG_POSTINC: - if (opts->aregs[inst->src.params.regs.pri] >= 0) { - mov_rr(code, opts->aregs[inst->src.params.regs.pri], opts->gen.scratch1, SZ_D); - } else { - mov_rdispr(code, opts->gen.context_reg, reg_offset(&(inst->src)), opts->gen.scratch1, SZ_D); - } - break; - case MODE_AREG_DISPLACE: - early_cycles += BUS; - reg = opts->gen.scratch2; - if (opts->aregs[inst->src.params.regs.pri] >= 0) { - mov_rr(code, opts->aregs[inst->src.params.regs.pri], opts->gen.scratch1, SZ_D); - } else { - mov_rdispr(code, opts->gen.context_reg, reg_offset(&(inst->src)), opts->gen.scratch1, SZ_D); - } - add_ir(code, inst->src.params.regs.displacement, opts->gen.scratch1, SZ_D); - break; - case MODE_AREG_INDEX_DISP8: - early_cycles += 6; - if (opts->aregs[inst->src.params.regs.pri] >= 0) { - mov_rr(code, opts->aregs[inst->src.params.regs.pri], opts->gen.scratch1, SZ_D); - } else { - mov_rdispr(code, opts->gen.context_reg, reg_offset(&(inst->src)), opts->gen.scratch1, SZ_D); - } - sec_reg = (inst->src.params.regs.sec >> 1) & 0x7; - if (inst->src.params.regs.sec & 1) { - if (inst->src.params.regs.sec & 0x10) { - if (opts->aregs[sec_reg] >= 0) { - add_rr(code, opts->aregs[sec_reg], opts->gen.scratch1, SZ_D); - } else { - add_rdispr(code, opts->gen.context_reg, offsetof(m68k_context, aregs) + sizeof(uint32_t)*sec_reg, opts->gen.scratch1, SZ_D); - } - } else { - if (opts->dregs[sec_reg] >= 0) { - add_rr(code, opts->dregs[sec_reg], opts->gen.scratch1, SZ_D); - } else { - add_rdispr(code, opts->gen.context_reg, offsetof(m68k_context, dregs) + sizeof(uint32_t)*sec_reg, opts->gen.scratch1, SZ_D); - } - } - } else { - if (inst->src.params.regs.sec & 0x10) { - if (opts->aregs[sec_reg] >= 0) { - movsx_rr(code, opts->aregs[sec_reg], opts->gen.scratch2, SZ_W, SZ_D); - } else { - movsx_rdispr(code, opts->gen.context_reg, offsetof(m68k_context, aregs) + sizeof(uint32_t)*sec_reg, opts->gen.scratch2, SZ_W, SZ_D); - } - } else { - if (opts->dregs[sec_reg] >= 0) { - movsx_rr(code, opts->dregs[sec_reg], opts->gen.scratch2, SZ_W, SZ_D); - } else { - movsx_rdispr(code, opts->gen.context_reg, offsetof(m68k_context, dregs) + sizeof(uint32_t)*sec_reg, opts->gen.scratch2, SZ_W, SZ_D); - } - } - add_rr(code, opts->gen.scratch2, opts->gen.scratch1, SZ_D); - } - if (inst->src.params.regs.displacement) { - add_ir(code, inst->src.params.regs.displacement, opts->gen.scratch1, SZ_D); - } - break; - case MODE_PC_DISPLACE: - early_cycles += BUS; - mov_ir(code, inst->src.params.regs.displacement + inst->address+2, opts->gen.scratch1, SZ_D); - break; - case MODE_PC_INDEX_DISP8: - early_cycles += 6; - mov_ir(code, inst->address+2, opts->gen.scratch1, SZ_D); - sec_reg = (inst->src.params.regs.sec >> 1) & 0x7; - if (inst->src.params.regs.sec & 1) { - if (inst->src.params.regs.sec & 0x10) { - if (opts->aregs[sec_reg] >= 0) { - add_rr(code, opts->aregs[sec_reg], opts->gen.scratch1, SZ_D); - } else { - add_rdispr(code, opts->gen.context_reg, offsetof(m68k_context, aregs) + sizeof(uint32_t)*sec_reg, opts->gen.scratch1, SZ_D); - } - } else { - if (opts->dregs[sec_reg] >= 0) { - add_rr(code, opts->dregs[sec_reg], opts->gen.scratch1, SZ_D); - } else { - add_rdispr(code, opts->gen.context_reg, offsetof(m68k_context, dregs) + sizeof(uint32_t)*sec_reg, opts->gen.scratch1, SZ_D); - } - } - } else { - if (inst->src.params.regs.sec & 0x10) { - if (opts->aregs[sec_reg] >= 0) { - movsx_rr(code, opts->aregs[sec_reg], opts->gen.scratch2, SZ_W, SZ_D); - } else { - movsx_rdispr(code, opts->gen.context_reg, offsetof(m68k_context, aregs) + sizeof(uint32_t)*sec_reg, opts->gen.scratch2, SZ_W, SZ_D); - } - } else { - if (opts->dregs[sec_reg] >= 0) { - movsx_rr(code, opts->dregs[sec_reg], opts->gen.scratch2, SZ_W, SZ_D); - } else { - movsx_rdispr(code, opts->gen.context_reg, offsetof(m68k_context, dregs) + sizeof(uint32_t)*sec_reg, opts->gen.scratch2, SZ_W, SZ_D); - } - } - add_rr(code, opts->gen.scratch2, opts->gen.scratch1, SZ_D); - } - if (inst->src.params.regs.displacement) { - add_ir(code, inst->src.params.regs.displacement, opts->gen.scratch1, SZ_D); - } - break; - case MODE_ABSOLUTE: - early_cycles += 4; - case MODE_ABSOLUTE_SHORT: - early_cycles += 4; - mov_ir(code, inst->src.params.immed, opts->gen.scratch1, SZ_D); - break; - default: - m68k_disasm(inst, disasm_buf); - printf("%X: %s\naddress mode %d not implemented (movem src)\n", inst->address, disasm_buf, inst->src.addr_mode); - exit(1); - } - cycles(&opts->gen, early_cycles); - for(reg = 0; reg < 16; reg ++) { - if (inst->dst.params.immed & (1 << reg)) { - push_r(code, opts->gen.scratch1); - if (inst->extra.size == OPSIZE_LONG) { - call(code, opts->read_32); - } else { - call(code, opts->read_16); - } - if (inst->extra.size == OPSIZE_WORD) { - movsx_rr(code, opts->gen.scratch1, opts->gen.scratch1, SZ_W, SZ_D); - } - if (reg > 7) { - if (opts->aregs[reg-8] >= 0) { - mov_rr(code, opts->gen.scratch1, opts->aregs[reg-8], SZ_D); - } else { - mov_rrdisp(code, opts->gen.scratch1, opts->gen.context_reg, offsetof(m68k_context, aregs) + sizeof(uint32_t) * (reg-8), SZ_D); - } - } else { - if (opts->dregs[reg] >= 0) { - mov_rr(code, opts->gen.scratch1, opts->dregs[reg], SZ_D); - } else { - mov_rrdisp(code, opts->gen.scratch1, opts->gen.context_reg, offsetof(m68k_context, dregs) + sizeof(uint32_t) * (reg), SZ_D); - } - } - pop_r(code, opts->gen.scratch1); - add_ir(code, (inst->extra.size == OPSIZE_LONG) ? 4 : 2, opts->gen.scratch1, SZ_D); - } - } - if (inst->src.addr_mode == MODE_AREG_POSTINC) { - if (opts->aregs[inst->src.params.regs.pri] >= 0) { - mov_rr(code, opts->gen.scratch1, opts->aregs[inst->src.params.regs.pri], SZ_D); - } else { - mov_rrdisp(code, opts->gen.scratch1, opts->gen.context_reg, reg_offset(&(inst->src)), SZ_D); - } - } - } - //prefetch - cycles(&opts->gen, 4); -} - -void translate_m68k_clr(m68k_options * opts, m68kinst * inst) -{ - code_info *code = &opts->gen.code; - set_flag(opts, 0, FLAG_N); - set_flag(opts, 0, FLAG_V); - set_flag(opts, 0, FLAG_C); - set_flag(opts, 1, FLAG_Z); - int8_t reg = native_reg(&(inst->dst), opts); - if (reg >= 0) { - cycles(&opts->gen, (inst->extra.size == OPSIZE_LONG ? 6 : 4)); - xor_rr(code, reg, reg, inst->extra.size); - return; - } - x86_ea dst_op; - translate_m68k_dst(inst, &dst_op, opts, 1); - if (dst_op.mode == MODE_REG_DIRECT) { - xor_rr(code, dst_op.base, dst_op.base, inst->extra.size); - } else { - mov_irdisp(code, 0, dst_op.base, dst_op.disp, inst->extra.size); - } - m68k_save_result(inst, opts); -} - -void translate_m68k_ext(m68k_options * opts, m68kinst * inst) -{ - code_info *code = &opts->gen.code; - x86_ea dst_op; - uint8_t dst_size = inst->extra.size; - inst->extra.size--; - translate_m68k_dst(inst, &dst_op, opts, 0); - if (dst_op.mode == MODE_REG_DIRECT) { - movsx_rr(code, dst_op.base, dst_op.base, inst->extra.size, dst_size); - cmp_ir(code, 0, dst_op.base, dst_size); - } else { - movsx_rdispr(code, dst_op.base, dst_op.disp, opts->gen.scratch1, inst->extra.size, dst_size); - cmp_ir(code, 0, opts->gen.scratch1, dst_size); - mov_rrdisp(code, opts->gen.scratch1, dst_op.base, dst_op.disp, dst_size); - } - inst->extra.size = dst_size; - set_flag(opts, 0, FLAG_V); - set_flag(opts, 0, FLAG_C); - set_flag_cond(opts, CC_Z, FLAG_Z); - set_flag_cond(opts, CC_S, FLAG_N); - //M68K EXT only operates on registers so no need for a call to save result here -} - -void translate_m68k_lea(m68k_options * opts, m68kinst * inst) -{ - code_info *code = &opts->gen.code; - int8_t dst_reg = native_reg(&(inst->dst), opts), sec_reg; - switch(inst->src.addr_mode) - { - case MODE_AREG_INDIRECT: - cycles(&opts->gen, BUS); - if (opts->aregs[inst->src.params.regs.pri] >= 0) { - if (dst_reg >= 0) { - mov_rr(code, opts->aregs[inst->src.params.regs.pri], dst_reg, SZ_D); - } else { - mov_rrdisp(code, opts->aregs[inst->src.params.regs.pri], opts->gen.context_reg, offsetof(m68k_context, aregs) + 4 * inst->dst.params.regs.pri, SZ_D); - } - } else { - if (dst_reg >= 0) { - mov_rdispr(code, opts->gen.context_reg, offsetof(m68k_context, aregs) + 4 * inst->src.params.regs.pri, dst_reg, SZ_D); - } else { - mov_rdispr(code, opts->gen.context_reg, offsetof(m68k_context, aregs) + 4 * inst->src.params.regs.pri, opts->gen.scratch1, SZ_D); - mov_rrdisp(code, opts->gen.scratch1, opts->gen.context_reg, offsetof(m68k_context, aregs) + 4 * inst->dst.params.regs.pri, SZ_D); - } - } - break; - case MODE_AREG_DISPLACE: - cycles(&opts->gen, 8); - if (dst_reg >= 0) { - if (inst->src.params.regs.pri != inst->dst.params.regs.pri) { - if (opts->aregs[inst->src.params.regs.pri] >= 0) { - mov_rr(code, opts->aregs[inst->src.params.regs.pri], dst_reg, SZ_D); - } else { - mov_rdispr(code, opts->gen.context_reg, reg_offset(&(inst->src)), dst_reg, SZ_D); - } - } - add_ir(code, inst->src.params.regs.displacement, dst_reg, SZ_D); - } else { - if (inst->src.params.regs.pri != inst->dst.params.regs.pri) { - if (opts->aregs[inst->src.params.regs.pri] >= 0) { - mov_rrdisp(code, opts->aregs[inst->src.params.regs.pri], opts->gen.context_reg, reg_offset(&(inst->dst)), SZ_D); - } else { - mov_rdispr(code, opts->gen.context_reg, reg_offset(&(inst->src)), opts->gen.scratch1, SZ_D); - mov_rrdisp(code, opts->gen.scratch1, opts->gen.context_reg, reg_offset(&(inst->dst)), SZ_D); - } - } - add_irdisp(code, inst->src.params.regs.displacement, opts->gen.context_reg, reg_offset(&(inst->dst)), SZ_D); - } - break; - case MODE_AREG_INDEX_DISP8: - cycles(&opts->gen, 12); - if (opts->aregs[inst->src.params.regs.pri] >= 0) { - mov_rr(code, opts->aregs[inst->src.params.regs.pri], opts->gen.scratch2, SZ_D); - } else { - mov_rdispr(code, opts->gen.context_reg, reg_offset(&(inst->src)), opts->gen.scratch2, SZ_D); - } - sec_reg = (inst->src.params.regs.sec >> 1) & 0x7; - if (inst->src.params.regs.sec & 1) { - if (inst->src.params.regs.sec & 0x10) { - if (opts->aregs[sec_reg] >= 0) { - add_rr(code, opts->aregs[sec_reg], opts->gen.scratch2, SZ_D); - } else { - add_rdispr(code, opts->gen.context_reg, offsetof(m68k_context, aregs) + sizeof(uint32_t)*sec_reg, opts->gen.scratch2, SZ_D); - } - } else { - if (opts->dregs[sec_reg] >= 0) { - add_rr(code, opts->dregs[sec_reg], opts->gen.scratch2, SZ_D); - } else { - add_rdispr(code, opts->gen.context_reg, offsetof(m68k_context, dregs) + sizeof(uint32_t)*sec_reg, opts->gen.scratch2, SZ_D); - } - } - } else { - if (inst->src.params.regs.sec & 0x10) { - if (opts->aregs[sec_reg] >= 0) { - movsx_rr(code, opts->aregs[sec_reg], opts->gen.scratch1, SZ_W, SZ_D); - } else { - movsx_rdispr(code, opts->gen.context_reg, offsetof(m68k_context, aregs) + sizeof(uint32_t)*sec_reg, opts->gen.scratch1, SZ_W, SZ_D); - } - } else { - if (opts->dregs[sec_reg] >= 0) { - movsx_rr(code, opts->dregs[sec_reg], opts->gen.scratch1, SZ_W, SZ_D); - } else { - movsx_rdispr(code, opts->gen.context_reg, offsetof(m68k_context, dregs) + sizeof(uint32_t)*sec_reg, opts->gen.scratch1, SZ_W, SZ_D); - } - } - add_rr(code, opts->gen.scratch1, opts->gen.scratch2, SZ_D); - } - if (inst->src.params.regs.displacement) { - add_ir(code, inst->src.params.regs.displacement, opts->gen.scratch2, SZ_D); - } - if (dst_reg >= 0) { - mov_rr(code, opts->gen.scratch2, dst_reg, SZ_D); - } else { - mov_rrdisp(code, opts->gen.scratch2, opts->gen.context_reg, reg_offset(&(inst->dst)), SZ_D); - } - break; - case MODE_PC_DISPLACE: - cycles(&opts->gen, 8); - if (dst_reg >= 0) { - mov_ir(code, inst->src.params.regs.displacement + inst->address+2, dst_reg, SZ_D); - } else { - mov_irdisp(code, inst->src.params.regs.displacement + inst->address+2, opts->gen.context_reg, offsetof(m68k_context, aregs) + 4 * inst->dst.params.regs.pri, SZ_D); - } - break; - case MODE_PC_INDEX_DISP8: - cycles(&opts->gen, BUS*3); - mov_ir(code, inst->address+2, opts->gen.scratch1, SZ_D); - sec_reg = (inst->src.params.regs.sec >> 1) & 0x7; - if (inst->src.params.regs.sec & 1) { - if (inst->src.params.regs.sec & 0x10) { - if (opts->aregs[sec_reg] >= 0) { - add_rr(code, opts->aregs[sec_reg], opts->gen.scratch1, SZ_D); - } else { - add_rdispr(code, opts->gen.context_reg, offsetof(m68k_context, aregs) + sizeof(uint32_t)*sec_reg, opts->gen.scratch1, SZ_D); - } - } else { - if (opts->dregs[sec_reg] >= 0) { - add_rr(code, opts->dregs[sec_reg], opts->gen.scratch1, SZ_D); - } else { - add_rdispr(code, opts->gen.context_reg, offsetof(m68k_context, dregs) + sizeof(uint32_t)*sec_reg, opts->gen.scratch1, SZ_D); - } - } - } else { - if (inst->src.params.regs.sec & 0x10) { - if (opts->aregs[sec_reg] >= 0) { - movsx_rr(code, opts->aregs[sec_reg], opts->gen.scratch2, SZ_W, SZ_D); - } else { - movsx_rdispr(code, opts->gen.context_reg, offsetof(m68k_context, aregs) + sizeof(uint32_t)*sec_reg, opts->gen.scratch2, SZ_W, SZ_D); - } - } else { - if (opts->dregs[sec_reg] >= 0) { - movsx_rr(code, opts->dregs[sec_reg], opts->gen.scratch2, SZ_W, SZ_D); - } else { - movsx_rdispr(code, opts->gen.context_reg, offsetof(m68k_context, dregs) + sizeof(uint32_t)*sec_reg, opts->gen.scratch2, SZ_W, SZ_D); - } - } - add_rr(code, opts->gen.scratch2, opts->gen.scratch1, SZ_D); - } - if (inst->src.params.regs.displacement) { - add_ir(code, inst->src.params.regs.displacement, opts->gen.scratch1, SZ_D); - } - if (dst_reg >= 0) { - mov_rr(code, opts->gen.scratch1, dst_reg, SZ_D); - } else { - mov_rrdisp(code, opts->gen.scratch1, opts->gen.context_reg, reg_offset(&(inst->dst)), SZ_D); - } - break; - case MODE_ABSOLUTE: - case MODE_ABSOLUTE_SHORT: - cycles(&opts->gen, (inst->src.addr_mode == MODE_ABSOLUTE) ? BUS * 3 : BUS * 2); - if (dst_reg >= 0) { - mov_ir(code, inst->src.params.immed, dst_reg, SZ_D); - } else { - mov_irdisp(code, inst->src.params.immed, opts->gen.context_reg, reg_offset(&(inst->dst)), SZ_D); - } - break; - default: - m68k_disasm(inst, disasm_buf); - printf("%X: %s\naddress mode %d not implemented (lea src)\n", inst->address, disasm_buf, inst->src.addr_mode); - exit(1); - } -} - -void translate_m68k_pea(m68k_options * opts, m68kinst * inst) -{ - code_info *code = &opts->gen.code; - uint8_t sec_reg; - switch(inst->src.addr_mode) - { - case MODE_AREG_INDIRECT: - cycles(&opts->gen, BUS); - if (opts->aregs[inst->src.params.regs.pri] >= 0) { - mov_rr(code, opts->aregs[inst->src.params.regs.pri], opts->gen.scratch1, SZ_D); - } else { - mov_rdispr(code, opts->gen.context_reg, offsetof(m68k_context, aregs) + 4 * inst->src.params.regs.pri, opts->gen.scratch1, SZ_D); - } - break; - case MODE_AREG_DISPLACE: - cycles(&opts->gen, 8); - if (opts->aregs[inst->src.params.regs.pri] >= 0) { - mov_rr(code, opts->aregs[inst->src.params.regs.pri], opts->gen.scratch1, SZ_D); - } else { - mov_rdispr(code, opts->gen.context_reg, reg_offset(&(inst->src)), opts->gen.scratch1, SZ_D); - } - add_ir(code, inst->src.params.regs.displacement, opts->gen.scratch1, SZ_D); - break; - case MODE_AREG_INDEX_DISP8: - cycles(&opts->gen, 6);//TODO: Check to make sure this is correct - if (opts->aregs[inst->src.params.regs.pri] >= 0) { - mov_rr(code, opts->aregs[inst->src.params.regs.pri], opts->gen.scratch1, SZ_D); - } else { - mov_rdispr(code, opts->gen.context_reg, reg_offset(&(inst->src)), opts->gen.scratch1, SZ_D); - } - sec_reg = (inst->src.params.regs.sec >> 1) & 0x7; - if (inst->src.params.regs.sec & 1) { - if (inst->src.params.regs.sec & 0x10) { - if (opts->aregs[sec_reg] >= 0) { - add_rr(code, opts->aregs[sec_reg], opts->gen.scratch1, SZ_D); - } else { - add_rdispr(code, opts->gen.context_reg, offsetof(m68k_context, aregs) + sizeof(uint32_t)*sec_reg, opts->gen.scratch1, SZ_D); - } - } else { - if (opts->dregs[sec_reg] >= 0) { - add_rr(code, opts->dregs[sec_reg], opts->gen.scratch1, SZ_D); - } else { - add_rdispr(code, opts->gen.context_reg, offsetof(m68k_context, dregs) + sizeof(uint32_t)*sec_reg, opts->gen.scratch1, SZ_D); - } - } - } else { - if (inst->src.params.regs.sec & 0x10) { - if (opts->aregs[sec_reg] >= 0) { - movsx_rr(code, opts->aregs[sec_reg], opts->gen.scratch2, SZ_W, SZ_D); - } else { - movsx_rdispr(code, opts->gen.context_reg, offsetof(m68k_context, aregs) + sizeof(uint32_t)*sec_reg, opts->gen.scratch2, SZ_W, SZ_D); - } - } else { - if (opts->dregs[sec_reg] >= 0) { - movsx_rr(code, opts->dregs[sec_reg], opts->gen.scratch2, SZ_W, SZ_D); - } else { - movsx_rdispr(code, opts->gen.context_reg, offsetof(m68k_context, dregs) + sizeof(uint32_t)*sec_reg, opts->gen.scratch2, SZ_W, SZ_D); - } - } - add_rr(code, opts->gen.scratch2, opts->gen.scratch1, SZ_D); - } - if (inst->src.params.regs.displacement) { - add_ir(code, inst->src.params.regs.displacement, opts->gen.scratch1, SZ_D); - } - break; - case MODE_PC_DISPLACE: - cycles(&opts->gen, 8); - mov_ir(code, inst->src.params.regs.displacement + inst->address+2, opts->gen.scratch1, SZ_D); - break; - case MODE_ABSOLUTE: - case MODE_ABSOLUTE_SHORT: - cycles(&opts->gen, (inst->src.addr_mode == MODE_ABSOLUTE) ? BUS * 3 : BUS * 2); - mov_ir(code, inst->src.params.immed, opts->gen.scratch1, SZ_D); - break; - default: - m68k_disasm(inst, disasm_buf); - printf("%X: %s\naddress mode %d not implemented (lea src)\n", inst->address, disasm_buf, inst->src.addr_mode); - exit(1); - } - sub_ir(code, 4, opts->aregs[7], SZ_D); - mov_rr(code, opts->aregs[7], opts->gen.scratch2, SZ_D); - call(code, opts->write_32_lowfirst); -} - -void translate_m68k_bsr(m68k_options * opts, m68kinst * inst) -{ - code_info *code = &opts->gen.code; - int32_t disp = inst->src.params.immed; - uint32_t after = inst->address + (inst->variant == VAR_BYTE ? 2 : 4); - //TODO: Add cycles in the right place relative to pushing the return address on the stack - cycles(&opts->gen, 10); - mov_ir(code, after, opts->gen.scratch1, SZ_D); - sub_ir(code, 4, opts->aregs[7], SZ_D); - mov_rr(code, opts->aregs[7], opts->gen.scratch2, SZ_D); - call(code, opts->write_32_highfirst); - code_ptr dest_addr = get_native_address(opts->gen.native_code_map, (inst->address+2) + disp); - if (!dest_addr) { - opts->gen.deferred = defer_address(opts->gen.deferred, (inst->address+2) + disp, code->cur + 1); - //dummy address to be replaced later - dest_addr = code->cur + 256; - } - jmp(code, dest_addr); -} - -uint8_t m68k_eval_cond(m68k_options * opts, uint8_t cc) -{ - uint8_t cond = CC_NZ; - switch (cc) - { - case COND_HIGH: - cond = CC_Z; - case COND_LOW_SAME: - flag_to_reg(opts, FLAG_Z, opts->gen.scratch1); - or_flag_to_reg(opts, FLAG_C, opts->gen.scratch1); - break; - case COND_CARRY_CLR: - cond = CC_Z; - case COND_CARRY_SET: - check_flag(opts, FLAG_C); - break; - case COND_NOT_EQ: - cond = CC_Z; - case COND_EQ: - check_flag(opts, FLAG_Z); - break; - case COND_OVERF_CLR: - cond = CC_Z; - case COND_OVERF_SET: - check_flag(opts, FLAG_V); - break; - case COND_PLUS: - cond = CC_Z; - case COND_MINUS: - check_flag(opts, FLAG_N); - break; - case COND_GREATER_EQ: - cond = CC_Z; - case COND_LESS: - cmp_flags(opts, FLAG_N, FLAG_V); - break; - case COND_GREATER: - cond = CC_Z; - case COND_LESS_EQ: - flag_to_reg(opts, FLAG_V, opts->gen.scratch1); - xor_flag_to_reg(opts, FLAG_N, opts->gen.scratch1); - or_flag_to_reg(opts, FLAG_Z, opts->gen.scratch1); - break; - } - return cond; -} - -void translate_m68k_bcc(m68k_options * opts, m68kinst * inst) -{ - code_info *code = &opts->gen.code; - cycles(&opts->gen, 10);//TODO: Adjust this for branch not taken case - int32_t disp = inst->src.params.immed; - uint32_t after = inst->address + 2; - code_ptr dest_addr = get_native_address(opts->gen.native_code_map, after + disp); - if (inst->extra.cond == COND_TRUE) { - if (!dest_addr) { - opts->gen.deferred = defer_address(opts->gen.deferred, after + disp, code->cur + 1); - //dummy address to be replaced later, make sure it generates a 4-byte displacement - dest_addr = code->cur + 256; - } - jmp(code, dest_addr); - } else { - uint8_t cond = m68k_eval_cond(opts, inst->extra.cond); - if (!dest_addr) { - opts->gen.deferred = defer_address(opts->gen.deferred, after + disp, code->cur + 2); - //dummy address to be replaced later, make sure it generates a 4-byte displacement - dest_addr = code->cur + 256; - } - jcc(code, cond, dest_addr); - } -} - -void translate_m68k_scc(m68k_options * opts, m68kinst * inst) -{ - code_info *code = &opts->gen.code; - uint8_t cond = inst->extra.cond; - x86_ea dst_op; - inst->extra.size = OPSIZE_BYTE; - translate_m68k_dst(inst, &dst_op, opts, 1); - if (cond == COND_TRUE || cond == COND_FALSE) { - if ((inst->dst.addr_mode == MODE_REG || inst->dst.addr_mode == MODE_AREG) && inst->extra.cond == COND_TRUE) { - cycles(&opts->gen, 6); - } else { - cycles(&opts->gen, BUS); - } - if (dst_op.mode == MODE_REG_DIRECT) { - mov_ir(code, cond == COND_TRUE ? 0xFF : 0, dst_op.base, SZ_B); - } else { - mov_irdisp(code, cond == COND_TRUE ? 0xFF : 0, dst_op.base, dst_op.disp, SZ_B); - } - } else { - uint8_t cc = m68k_eval_cond(opts, cond); - check_alloc_code(code, 6*MAX_INST_LEN); - code_ptr true_off = code->cur + 1; - jcc(code, cc, code->cur+2); - cycles(&opts->gen, BUS); - if (dst_op.mode == MODE_REG_DIRECT) { - mov_ir(code, 0, dst_op.base, SZ_B); - } else { - mov_irdisp(code, 0, dst_op.base, dst_op.disp, SZ_B); - } - code_ptr end_off = code->cur+1; - jmp(code, code->cur+2); - *true_off = code->cur - (true_off+1); - cycles(&opts->gen, 6); - if (dst_op.mode == MODE_REG_DIRECT) { - mov_ir(code, 0xFF, dst_op.base, SZ_B); - } else { - mov_irdisp(code, 0xFF, dst_op.base, dst_op.disp, SZ_B); - } - *end_off = code->cur - (end_off+1); - } - m68k_save_result(inst, opts); -} - -void translate_m68k_jmp_jsr(m68k_options * opts, m68kinst * inst) -{ - uint8_t is_jsr = inst->op == M68K_JSR; - code_info *code = &opts->gen.code; - code_ptr dest_addr; - uint8_t sec_reg; - uint32_t after; - uint32_t m68k_addr; - switch(inst->src.addr_mode) - { - case MODE_AREG_INDIRECT: - cycles(&opts->gen, BUS*2); - if (is_jsr) { - mov_ir(code, inst->address + 2, opts->gen.scratch1, SZ_D); - sub_ir(code, 4, opts->aregs[7], SZ_D); - mov_rr(code, opts->aregs[7], opts->gen.scratch2, SZ_D); - call(code, opts->write_32_highfirst); - } - if (opts->aregs[inst->src.params.regs.pri] >= 0) { - mov_rr(code, opts->aregs[inst->src.params.regs.pri], opts->gen.scratch1, SZ_D); - } else { - mov_rdispr(code, opts->gen.context_reg, offsetof(m68k_context, aregs) + 4 * inst->src.params.regs.pri, opts->gen.scratch1, SZ_D); - } - call(code, opts->native_addr); - jmp_r(code, opts->gen.scratch1); - break; - case MODE_AREG_DISPLACE: - cycles(&opts->gen, BUS*2); - if (is_jsr) { - mov_ir(code, inst->address + 4, opts->gen.scratch1, SZ_D); - sub_ir(code, 4, opts->aregs[7], SZ_D); - mov_rr(code, opts->aregs[7], opts->gen.scratch2, SZ_D); - call(code, opts->write_32_highfirst); - } - if (opts->aregs[inst->src.params.regs.pri] >= 0) { - mov_rr(code, opts->aregs[inst->src.params.regs.pri], opts->gen.scratch1, SZ_D); - } else { - mov_rdispr(code, opts->gen.context_reg, offsetof(m68k_context, aregs) + 4 * inst->src.params.regs.pri, opts->gen.scratch1, SZ_D); - } - add_ir(code, inst->src.params.regs.displacement, opts->gen.scratch1, SZ_D); - call(code, opts->native_addr); - jmp_r(code, opts->gen.scratch1); - break; - case MODE_AREG_INDEX_DISP8: - cycles(&opts->gen, BUS*3);//TODO: CHeck that this is correct - if (is_jsr) { - mov_ir(code, inst->address + 4, opts->gen.scratch1, SZ_D); - sub_ir(code, 4, opts->aregs[7], SZ_D); - mov_rr(code, opts->aregs[7], opts->gen.scratch2, SZ_D); - call(code, opts->write_32_highfirst); - } - if (opts->aregs[inst->src.params.regs.pri] >= 0) { - mov_rr(code, opts->aregs[inst->src.params.regs.pri], opts->gen.scratch1, SZ_D); - } else { - mov_rdispr(code, opts->gen.context_reg, reg_offset(&(inst->src)), opts->gen.scratch1, SZ_D); - } - sec_reg = (inst->src.params.regs.sec >> 1) & 0x7; - if (inst->src.params.regs.sec & 1) { - //32-bit index register - if (inst->src.params.regs.sec & 0x10) { - if (opts->aregs[sec_reg] >= 0) { - add_rr(code, opts->aregs[sec_reg], opts->gen.scratch1, SZ_D); - } else { - add_rdispr(code, opts->gen.context_reg, offsetof(m68k_context, aregs) + sizeof(uint32_t)*sec_reg, opts->gen.scratch1, SZ_D); - } - } else { - if (opts->dregs[sec_reg] >= 0) { - add_rr(code, opts->dregs[sec_reg], opts->gen.scratch1, SZ_D); - } else { - add_rdispr(code, opts->gen.context_reg, offsetof(m68k_context, dregs) + sizeof(uint32_t)*sec_reg, opts->gen.scratch1, SZ_D); - } - } - } else { - //16-bit index register - if (inst->src.params.regs.sec & 0x10) { - if (opts->aregs[sec_reg] >= 0) { - movsx_rr(code, opts->aregs[sec_reg], opts->gen.scratch2, SZ_W, SZ_D); - } else { - movsx_rdispr(code, opts->gen.context_reg, offsetof(m68k_context, aregs) + sizeof(uint32_t)*sec_reg, opts->gen.scratch2, SZ_W, SZ_D); - } - } else { - if (opts->dregs[sec_reg] >= 0) { - movsx_rr(code, opts->dregs[sec_reg], opts->gen.scratch2, SZ_W, SZ_D); - } else { - movsx_rdispr(code, opts->gen.context_reg, offsetof(m68k_context, dregs) + sizeof(uint32_t)*sec_reg, opts->gen.scratch2, SZ_W, SZ_D); - } - } - add_rr(code, opts->gen.scratch2, opts->gen.scratch1, SZ_D); - } - if (inst->src.params.regs.displacement) { - add_ir(code, inst->src.params.regs.displacement, opts->gen.scratch1, SZ_D); - } - call(code, opts->native_addr); - jmp_r(code, opts->gen.scratch1); - break; - case MODE_PC_DISPLACE: - //TODO: Add cycles in the right place relative to pushing the return address on the stack - cycles(&opts->gen, 10); - if (is_jsr) { - mov_ir(code, inst->address + 4, opts->gen.scratch1, SZ_D); - sub_ir(code, 4, opts->aregs[7], SZ_D); - mov_rr(code, opts->aregs[7], opts->gen.scratch2, SZ_D); - call(code, opts->write_32_highfirst); - } - m68k_addr = inst->src.params.regs.displacement + inst->address + 2; - if ((m68k_addr & 0xFFFFFF) < 0x400000) { - dest_addr = get_native_address(opts->gen.native_code_map, m68k_addr); - if (!dest_addr) { - opts->gen.deferred = defer_address(opts->gen.deferred, m68k_addr, code->cur + 1); - //dummy address to be replaced later, make sure it generates a 4-byte displacement - dest_addr = code->cur + 256; - } - jmp(code, dest_addr); - } else { - mov_ir(code, m68k_addr, opts->gen.scratch1, SZ_D); - call(code, opts->native_addr); - jmp_r(code, opts->gen.scratch1); - } - break; - case MODE_PC_INDEX_DISP8: - cycles(&opts->gen, BUS*3);//TODO: CHeck that this is correct - if (is_jsr) { - mov_ir(code, inst->address + 4, opts->gen.scratch1, SZ_D); - sub_ir(code, 4, opts->aregs[7], SZ_D); - mov_rr(code, opts->aregs[7], opts->gen.scratch2, SZ_D); - call(code, opts->write_32_highfirst); - } - mov_ir(code, inst->address+2, opts->gen.scratch1, SZ_D); - sec_reg = (inst->src.params.regs.sec >> 1) & 0x7; - if (inst->src.params.regs.sec & 1) { - if (inst->src.params.regs.sec & 0x10) { - if (opts->aregs[sec_reg] >= 0) { - add_rr(code, opts->aregs[sec_reg], opts->gen.scratch1, SZ_D); - } else { - add_rdispr(code, opts->gen.context_reg, offsetof(m68k_context, aregs) + sizeof(uint32_t)*sec_reg, opts->gen.scratch1, SZ_D); - } - } else { - if (opts->dregs[sec_reg] >= 0) { - add_rr(code, opts->dregs[sec_reg], opts->gen.scratch1, SZ_D); - } else { - add_rdispr(code, opts->gen.context_reg, offsetof(m68k_context, dregs) + sizeof(uint32_t)*sec_reg, opts->gen.scratch1, SZ_D); - } - } - } else { - if (inst->src.params.regs.sec & 0x10) { - if (opts->aregs[sec_reg] >= 0) { - movsx_rr(code, opts->aregs[sec_reg], opts->gen.scratch2, SZ_W, SZ_D); - } else { - movsx_rdispr(code, opts->gen.context_reg, offsetof(m68k_context, aregs) + sizeof(uint32_t)*sec_reg, opts->gen.scratch2, SZ_W, SZ_D); - } - } else { - if (opts->dregs[sec_reg] >= 0) { - movsx_rr(code, opts->dregs[sec_reg], opts->gen.scratch2, SZ_W, SZ_D); - } else { - movsx_rdispr(code, opts->gen.context_reg, offsetof(m68k_context, dregs) + sizeof(uint32_t)*sec_reg, opts->gen.scratch2, SZ_W, SZ_D); - } - } - add_rr(code, opts->gen.scratch2, opts->gen.scratch1, SZ_D); - } - if (inst->src.params.regs.displacement) { - add_ir(code, inst->src.params.regs.displacement, opts->gen.scratch1, SZ_D); - } - call(code, opts->native_addr); - jmp_r(code, opts->gen.scratch1); - break; - case MODE_ABSOLUTE: - case MODE_ABSOLUTE_SHORT: - //TODO: Add cycles in the right place relative to pushing the return address on the stack - cycles(&opts->gen, inst->src.addr_mode == MODE_ABSOLUTE ? 12 : 10); - if (is_jsr) { - mov_ir(code, inst->address + (inst->src.addr_mode == MODE_ABSOLUTE ? 6 : 4), opts->gen.scratch1, SZ_D); - sub_ir(code, 4, opts->aregs[7], SZ_D); - mov_rr(code, opts->aregs[7], opts->gen.scratch2, SZ_D); - call(code, opts->write_32_highfirst); - } - m68k_addr = inst->src.params.immed; - if ((m68k_addr & 0xFFFFFF) < 0x400000) { - dest_addr = get_native_address(opts->gen.native_code_map, m68k_addr); - if (!dest_addr) { - opts->gen.deferred = defer_address(opts->gen.deferred, m68k_addr, code->cur + 1); - //dummy address to be replaced later, make sure it generates a 4-byte displacement - dest_addr = code->cur + 256; - } - jmp(code, dest_addr); - } else { - mov_ir(code, m68k_addr, opts->gen.scratch1, SZ_D); - call(code, opts->native_addr); - jmp_r(code, opts->gen.scratch1); - } - break; - default: - m68k_disasm(inst, disasm_buf); - printf("%s\naddress mode %d not yet supported (%s)\n", disasm_buf, inst->src.addr_mode, is_jsr ? "jsr" : "jmp"); - exit(1); - } -} - -void translate_m68k_rts(m68k_options * opts, m68kinst * inst) -{ - code_info *code = &opts->gen.code; - //TODO: Add cycles - mov_rr(code, opts->aregs[7], opts->gen.scratch1, SZ_D); - add_ir(code, 4, opts->aregs[7], SZ_D); - call(code, opts->read_32); - call(code, opts->native_addr); - jmp_r(code, opts->gen.scratch1); -} - -void translate_m68k_dbcc(m68k_options * opts, m68kinst * inst) -{ - code_info *code = &opts->gen.code; - //best case duration - cycles(&opts->gen, 10); - code_ptr skip_loc = NULL; - //TODO: Check if COND_TRUE technically valid here even though - //it's basically a slow NOP - if (inst->extra.cond != COND_FALSE) { - uint8_t cond = m68k_eval_cond(opts, inst->extra.cond); - check_alloc_code(code, 6*MAX_INST_LEN); - skip_loc = code->cur + 1; - jcc(code, cond, code->cur + 2); - } - if (opts->dregs[inst->dst.params.regs.pri] >= 0) { - sub_ir(code, 1, opts->dregs[inst->dst.params.regs.pri], SZ_W); - cmp_ir(code, -1, opts->dregs[inst->dst.params.regs.pri], SZ_W); - } else { - sub_irdisp(code, 1, opts->gen.context_reg, offsetof(m68k_context, dregs) + 4 * inst->dst.params.regs.pri, SZ_W); - cmp_irdisp(code, -1, opts->gen.context_reg, offsetof(m68k_context, dregs) + 4 * inst->dst.params.regs.pri, SZ_W); - } - code_ptr loop_end_loc = code->cur + 1; - jcc(code, CC_Z, code->cur + 2); - uint32_t after = inst->address + 2; - code_ptr dest_addr = get_native_address(opts->gen.native_code_map, after + inst->src.params.immed); - if (!dest_addr) { - opts->gen.deferred = defer_address(opts->gen.deferred, after + inst->src.params.immed, code->cur + 1); - //dummy address to be replaced later, make sure it generates a 4-byte displacement - dest_addr = code->cur + 256; - } - jmp(code, dest_addr); - *loop_end_loc = code->cur - (loop_end_loc+1); - if (skip_loc) { - cycles(&opts->gen, 2); - *skip_loc = code->cur - (skip_loc+1); - cycles(&opts->gen, 2); - } else { - cycles(&opts->gen, 4); - } -} - -void translate_m68k_link(m68k_options * opts, m68kinst * inst) -{ - code_info *code = &opts->gen.code; - int8_t reg = native_reg(&(inst->src), opts); - //compensate for displacement word - cycles(&opts->gen, BUS); - sub_ir(code, 4, opts->aregs[7], SZ_D); - mov_rr(code, opts->aregs[7], opts->gen.scratch2, SZ_D); - if (reg >= 0) { - mov_rr(code, reg, opts->gen.scratch1, SZ_D); - } else { - mov_rdispr(code, opts->gen.context_reg, reg_offset(&(inst->src)), opts->gen.scratch1, SZ_D); - } - call(code, opts->write_32_highfirst); - if (reg >= 0) { - mov_rr(code, opts->aregs[7], reg, SZ_D); - } else { - mov_rrdisp(code, opts->aregs[7], opts->gen.context_reg, reg_offset(&(inst->src)), SZ_D); - } - add_ir(code, inst->dst.params.immed, opts->aregs[7], SZ_D); - //prefetch - cycles(&opts->gen, BUS); -} - -void translate_m68k_movep(m68k_options * opts, m68kinst * inst) -{ - code_info *code = &opts->gen.code; - int8_t reg; - cycles(&opts->gen, BUS*2); - if (inst->src.addr_mode == MODE_REG) { - if (opts->aregs[inst->dst.params.regs.pri] >= 0) { - mov_rr(code, opts->aregs[inst->dst.params.regs.pri], opts->gen.scratch2, SZ_D); - } else { - mov_rdispr(code, opts->gen.context_reg, reg_offset(&(inst->dst)), opts->gen.scratch2, SZ_D); - } - if (inst->dst.params.regs.displacement) { - add_ir(code, inst->dst.params.regs.displacement, opts->gen.scratch2, SZ_D); - } - reg = native_reg(&(inst->src), opts); - if (inst->extra.size == OPSIZE_LONG) { - if (reg >= 0) { - mov_rr(code, reg, opts->gen.scratch1, SZ_D); - shr_ir(code, 24, opts->gen.scratch1, SZ_D); - push_r(code, opts->gen.scratch2); - call(code, opts->write_8); - pop_r(code, opts->gen.scratch2); - mov_rr(code, reg, opts->gen.scratch1, SZ_D); - shr_ir(code, 16, opts->gen.scratch1, SZ_D); - - } else { - mov_rdispr(code, opts->gen.context_reg, reg_offset(&(inst->src))+3, opts->gen.scratch1, SZ_B); - push_r(code, opts->gen.scratch2); - call(code, opts->write_8); - pop_r(code, opts->gen.scratch2); - mov_rdispr(code, opts->gen.context_reg, reg_offset(&(inst->src))+2, opts->gen.scratch1, SZ_B); - } - add_ir(code, 2, opts->gen.scratch2, SZ_D); - push_r(code, opts->gen.scratch2); - call(code, opts->write_8); - pop_r(code, opts->gen.scratch2); - add_ir(code, 2, opts->gen.scratch2, SZ_D); - } - if (reg >= 0) { - mov_rr(code, reg, opts->gen.scratch1, SZ_W); - shr_ir(code, 8, opts->gen.scratch1, SZ_W); - push_r(code, opts->gen.scratch2); - call(code, opts->write_8); - pop_r(code, opts->gen.scratch2); - mov_rr(code, reg, opts->gen.scratch1, SZ_W); - } else { - mov_rdispr(code, opts->gen.context_reg, reg_offset(&(inst->src))+1, opts->gen.scratch1, SZ_B); - push_r(code, opts->gen.scratch2); - call(code, opts->write_8); - pop_r(code, opts->gen.scratch2); - mov_rdispr(code, opts->gen.context_reg, reg_offset(&(inst->src)), opts->gen.scratch1, SZ_B); - } - add_ir(code, 2, opts->gen.scratch2, SZ_D); - call(code, opts->write_8); - } else { - if (opts->aregs[inst->src.params.regs.pri] >= 0) { - mov_rr(code, opts->aregs[inst->src.params.regs.pri], opts->gen.scratch1, SZ_D); - } else { - mov_rdispr(code, opts->gen.context_reg, reg_offset(&(inst->src)), opts->gen.scratch1, SZ_D); - } - if (inst->src.params.regs.displacement) { - add_ir(code, inst->src.params.regs.displacement, opts->gen.scratch1, SZ_D); - } - reg = native_reg(&(inst->dst), opts); - if (inst->extra.size == OPSIZE_LONG) { - if (reg >= 0) { - push_r(code, opts->gen.scratch1); - call(code, opts->read_8); - shl_ir(code, 24, opts->gen.scratch1, SZ_D); - mov_rr(code, opts->gen.scratch1, reg, SZ_D); - pop_r(code, opts->gen.scratch1); - add_ir(code, 2, opts->gen.scratch1, SZ_D); - push_r(code, opts->gen.scratch1); - call(code, opts->read_8); - shl_ir(code, 16, opts->gen.scratch1, SZ_D); - or_rr(code, opts->gen.scratch1, reg, SZ_D); - } else { - push_r(code, opts->gen.scratch1); - call(code, opts->read_8); - mov_rrdisp(code, opts->gen.scratch1, opts->gen.context_reg, reg_offset(&(inst->dst))+3, SZ_B); - pop_r(code, opts->gen.scratch1); - add_ir(code, 2, opts->gen.scratch1, SZ_D); - push_r(code, opts->gen.scratch1); - call(code, opts->read_8); - mov_rrdisp(code, opts->gen.scratch1, opts->gen.context_reg, reg_offset(&(inst->dst))+2, SZ_B); - } - pop_r(code, opts->gen.scratch1); - add_ir(code, 2, opts->gen.scratch1, SZ_D); - } - push_r(code, opts->gen.scratch1); - call(code, opts->read_8); - if (reg >= 0) { - - shl_ir(code, 8, opts->gen.scratch1, SZ_W); - mov_rr(code, opts->gen.scratch1, reg, SZ_W); - pop_r(code, opts->gen.scratch1); - add_ir(code, 2, opts->gen.scratch1, SZ_D); - call(code, opts->read_8); - mov_rr(code, opts->gen.scratch1, reg, SZ_B); - } else { - mov_rrdisp(code, opts->gen.scratch1, opts->gen.context_reg, reg_offset(&(inst->dst))+1, SZ_B); - pop_r(code, opts->gen.scratch1); - add_ir(code, 2, opts->gen.scratch1, SZ_D); - call(code, opts->read_8); - mov_rrdisp(code, opts->gen.scratch1, opts->gen.context_reg, reg_offset(&(inst->dst)), SZ_B); - } - } -} - -void translate_m68k_cmp(m68k_options * opts, m68kinst * inst) -{ - code_info *code = &opts->gen.code; - uint8_t size = inst->extra.size; - x86_ea src_op, dst_op; - translate_m68k_src(inst, &src_op, opts); - if (inst->dst.addr_mode == MODE_AREG_POSTINC) { - push_r(code, opts->gen.scratch1); - translate_m68k_dst(inst, &dst_op, opts, 0); - pop_r(code, opts->gen.scratch2); - src_op.base = opts->gen.scratch2; - } else { - translate_m68k_dst(inst, &dst_op, opts, 0); - if (inst->dst.addr_mode == MODE_AREG && size == OPSIZE_WORD) { - size = OPSIZE_LONG; - } - } - cycles(&opts->gen, BUS); - if (src_op.mode == MODE_REG_DIRECT) { - if (dst_op.mode == MODE_REG_DIRECT) { - cmp_rr(code, src_op.base, dst_op.base, size); - } else { - cmp_rrdisp(code, src_op.base, dst_op.base, dst_op.disp, size); - } - } else if (src_op.mode == MODE_REG_DISPLACE8) { - cmp_rdispr(code, src_op.base, src_op.disp, dst_op.base, size); - } else { - if (dst_op.mode == MODE_REG_DIRECT) { - cmp_ir(code, src_op.disp, dst_op.base, size); - } else { - cmp_irdisp(code, src_op.disp, dst_op.base, dst_op.disp, size); - } - } - set_flag_cond(opts, CC_C, FLAG_C); - set_flag_cond(opts, CC_Z, FLAG_Z); - set_flag_cond(opts, CC_S, FLAG_N); - set_flag_cond(opts, CC_O, FLAG_V); -} - -typedef void (*shift_ir_t)(code_info *code, uint8_t val, uint8_t dst, uint8_t size); -typedef void (*shift_irdisp_t)(code_info *code, uint8_t val, uint8_t dst_base, int32_t disp, uint8_t size); -typedef void (*shift_clr_t)(code_info *code, uint8_t dst, uint8_t size); -typedef void (*shift_clrdisp_t)(code_info *code, uint8_t dst_base, int32_t disp, uint8_t size); - -void translate_shift(m68k_options * opts, m68kinst * inst, x86_ea *src_op, x86_ea * dst_op, shift_ir_t shift_ir, shift_irdisp_t shift_irdisp, shift_clr_t shift_clr, shift_clrdisp_t shift_clrdisp, shift_ir_t special, shift_irdisp_t special_disp) -{ - code_info *code = &opts->gen.code; - code_ptr end_off = NULL; - code_ptr nz_off = NULL; - code_ptr z_off = NULL; - if (inst->src.addr_mode == MODE_UNUSED) { - cycles(&opts->gen, BUS); - //Memory shift - shift_ir(code, 1, dst_op->base, SZ_W); - } else { - cycles(&opts->gen, inst->extra.size == OPSIZE_LONG ? 8 : 6); - if (src_op->mode == MODE_IMMED) { - if (src_op->disp != 1 && inst->op == M68K_ASL) { - set_flag(opts, 0, FLAG_V); - for (int i = 0; i < src_op->disp; i++) { - if (dst_op->mode == MODE_REG_DIRECT) { - shift_ir(code, 1, dst_op->base, inst->extra.size); - } else { - shift_irdisp(code, 1, dst_op->base, dst_op->disp, inst->extra.size); - } - check_alloc_code(code, 2*MAX_INST_LEN); - code_ptr after_flag_set = code->cur + 1; - jcc(code, CC_NO, code->cur + 2); - set_flag(opts, 1, FLAG_V); - *after_flag_set = code->cur - (after_flag_set+1); - } - } else { - if (dst_op->mode == MODE_REG_DIRECT) { - shift_ir(code, src_op->disp, dst_op->base, inst->extra.size); - } else { - shift_irdisp(code, src_op->disp, dst_op->base, dst_op->disp, inst->extra.size); - } - set_flag_cond(opts, CC_O, FLAG_V); - } - } else { - if (src_op->base != RCX) { - if (src_op->mode == MODE_REG_DIRECT) { - mov_rr(code, src_op->base, RCX, SZ_B); - } else { - mov_rdispr(code, src_op->base, src_op->disp, RCX, SZ_B); - } - - } - and_ir(code, 63, RCX, SZ_D); - check_alloc_code(code, 7*MAX_INST_LEN); - nz_off = code->cur + 1; - jcc(code, CC_NZ, code->cur + 2); - //Flag behavior for shift count of 0 is different for x86 than 68K - if (dst_op->mode == MODE_REG_DIRECT) { - cmp_ir(code, 0, dst_op->base, inst->extra.size); - } else { - cmp_irdisp(code, 0, dst_op->base, dst_op->disp, inst->extra.size); - } - set_flag_cond(opts, CC_Z, FLAG_Z); - set_flag_cond(opts, CC_S, FLAG_N); - set_flag(opts, 0, FLAG_C); - //For other instructions, this flag will be set below - if (inst->op == M68K_ASL) { - set_flag(opts, 0, FLAG_V); - } - z_off = code->cur + 1; - jmp(code, code->cur + 2); - *nz_off = code->cur - (nz_off + 1); - //add 2 cycles for every bit shifted - add_rr(code, RCX, CYCLES, SZ_D); - add_rr(code, RCX, CYCLES, SZ_D); - if (inst->op == M68K_ASL) { - //ASL has Overflow flag behavior that depends on all of the bits shifted through the MSB - //Easiest way to deal with this is to shift one bit at a time - set_flag(opts, 0, FLAG_V); - check_alloc_code(code, 5*MAX_INST_LEN); - code_ptr loop_start = code->cur; - if (dst_op->mode == MODE_REG_DIRECT) { - shift_ir(code, 1, dst_op->base, inst->extra.size); - } else { - shift_irdisp(code, 1, dst_op->base, dst_op->disp, inst->extra.size); - } - code_ptr after_flag_set = code->cur + 1; - jcc(code, CC_NO, code->cur + 2); - set_flag(opts, 1, FLAG_V); - *after_flag_set = code->cur - (after_flag_set+1); - loop(code, loop_start); - } else { - //x86 shifts modulo 32 for operand sizes less than 64-bits - //but M68K shifts modulo 64, so we need to check for large shifts here - cmp_ir(code, 32, RCX, SZ_B); - check_alloc_code(code, 14*MAX_INST_LEN); - code_ptr norm_shift_off = code->cur + 1; - jcc(code, CC_L, code->cur + 2); - if (special) { - code_ptr after_flag_set = NULL; - if (inst->extra.size == OPSIZE_LONG) { - code_ptr neq_32_off = code->cur + 1; - jcc(code, CC_NZ, code->cur + 2); - - //set the carry bit to the lsb - if (dst_op->mode == MODE_REG_DIRECT) { - special(code, 1, dst_op->base, SZ_D); - } else { - special_disp(code, 1, dst_op->base, dst_op->disp, SZ_D); - } - set_flag_cond(opts, CC_C, FLAG_C); - after_flag_set = code->cur + 1; - jmp(code, code->cur + 2); - *neq_32_off = code->cur - (neq_32_off+1); - } - set_flag(opts, 0, FLAG_C); - if (after_flag_set) { - *after_flag_set = code->cur - (after_flag_set+1); - } - set_flag(opts, 1, FLAG_Z); - set_flag(opts, 0, FLAG_N); - if (dst_op->mode == MODE_REG_DIRECT) { - xor_rr(code, dst_op->base, dst_op->base, inst->extra.size); - } else { - mov_irdisp(code, 0, dst_op->base, dst_op->disp, inst->extra.size); - } - } else { - if (dst_op->mode == MODE_REG_DIRECT) { - shift_ir(code, 31, dst_op->base, inst->extra.size); - shift_ir(code, 1, dst_op->base, inst->extra.size); - } else { - shift_irdisp(code, 31, dst_op->base, dst_op->disp, inst->extra.size); - shift_irdisp(code, 1, dst_op->base, dst_op->disp, inst->extra.size); - } - - } - end_off = code->cur + 1; - jmp(code, code->cur + 2); - *norm_shift_off = code->cur - (norm_shift_off+1); - if (dst_op->mode == MODE_REG_DIRECT) { - shift_clr(code, dst_op->base, inst->extra.size); - } else { - shift_clrdisp(code, dst_op->base, dst_op->disp, inst->extra.size); - } - } - } - - } - if (!special && end_off) { - *end_off = code->cur - (end_off + 1); - } - set_flag_cond(opts, CC_C, FLAG_C); - set_flag_cond(opts, CC_Z, FLAG_Z); - set_flag_cond(opts, CC_S, FLAG_N); - if (special && end_off) { - *end_off = code->cur - (end_off + 1); - } - //set X flag to same as C flag - if (opts->flag_regs[FLAG_C] >= 0) { - flag_to_flag(opts, FLAG_C, FLAG_X); - } else { - set_flag_cond(opts, CC_C, FLAG_X); - } - if (z_off) { - *z_off = code->cur - (z_off + 1); - } - if (inst->op != M68K_ASL) { - set_flag(opts, 0, FLAG_V); - } - if (inst->src.addr_mode == MODE_UNUSED) { - m68k_save_result(inst, opts); - } -} - -#define BIT_SUPERVISOR 5 - -void translate_m68k(m68k_options * opts, m68kinst * inst) -{ - code_ptr end_off, zero_off, norm_off; - uint8_t dst_reg; - code_info *code = &opts->gen.code; - check_cycles_int(&opts->gen, inst->address); - if (inst->op == M68K_MOVE) { - return translate_m68k_move(opts, inst); - } else if(inst->op == M68K_LEA) { - return translate_m68k_lea(opts, inst); - } else if(inst->op == M68K_PEA) { - return translate_m68k_pea(opts, inst); - } else if(inst->op == M68K_BSR) { - return translate_m68k_bsr(opts, inst); - } else if(inst->op == M68K_BCC) { - return translate_m68k_bcc(opts, inst); - } else if(inst->op == M68K_JMP) { - return translate_m68k_jmp_jsr(opts, inst); - } else if(inst->op == M68K_JSR) { - return translate_m68k_jmp_jsr(opts, inst); - } else if(inst->op == M68K_RTS) { - return translate_m68k_rts(opts, inst); - } else if(inst->op == M68K_DBCC) { - return translate_m68k_dbcc(opts, inst); - } else if(inst->op == M68K_CLR) { - return translate_m68k_clr(opts, inst); - } else if(inst->op == M68K_MOVEM) { - return translate_m68k_movem(opts, inst); - } else if(inst->op == M68K_LINK) { - return translate_m68k_link(opts, inst); - } else if(inst->op == M68K_EXT) { - return translate_m68k_ext(opts, inst); - } else if(inst->op == M68K_SCC) { - return translate_m68k_scc(opts, inst); - } else if(inst->op == M68K_MOVEP) { - return translate_m68k_movep(opts, inst); - } else if(inst->op == M68K_INVALID) { - if (inst->src.params.immed == 0x7100) { - return retn(code); - } - mov_ir(code, inst->address, opts->gen.scratch1, SZ_D); - return call(code, (code_ptr)m68k_invalid); - } else if(inst->op == M68K_CMP) { - return translate_m68k_cmp(opts, inst); - } - x86_ea src_op, dst_op; - if (inst->src.addr_mode != MODE_UNUSED) { - translate_m68k_src(inst, &src_op, opts); - } - if (inst->dst.addr_mode != MODE_UNUSED) { - translate_m68k_dst(inst, &dst_op, opts, 0); - } - uint8_t size; - switch(inst->op) - { - case M68K_ABCD: - if (src_op.base != opts->gen.scratch2) { - if (src_op.mode == MODE_REG_DIRECT) { - mov_rr(code, src_op.base, opts->gen.scratch2, SZ_B); - } else { - mov_rdispr(code, src_op.base, src_op.disp, opts->gen.scratch2, SZ_B); - } - } - if (dst_op.base != opts->gen.scratch1) { - if (dst_op.mode == MODE_REG_DIRECT) { - mov_rr(code, dst_op.base, opts->gen.scratch1, SZ_B); - } else { - mov_rdispr(code, dst_op.base, dst_op.disp, opts->gen.scratch1, SZ_B); - } - } - flag_to_carry(opts, FLAG_X); - jcc(code, CC_NC, code->cur + 5); - add_ir(code, 1, opts->gen.scratch1, SZ_B); - call(code, (code_ptr)bcd_add); - reg_to_flag(opts, CH, FLAG_C); - reg_to_flag(opts, CH, FLAG_X); - cmp_ir(code, 0, opts->gen.scratch1, SZ_B); - jcc(code, CC_Z, code->cur + 4); - set_flag(opts, 0, FLAG_Z); - if (dst_op.base != opts->gen.scratch1) { - if (dst_op.mode == MODE_REG_DIRECT) { - mov_rr(code, opts->gen.scratch1, dst_op.base, SZ_B); - } else { - mov_rrdisp(code, opts->gen.scratch1, dst_op.base, dst_op.disp, SZ_B); - } - } - m68k_save_result(inst, opts); - break; - case M68K_ADD: - cycles(&opts->gen, BUS); - size = inst->dst.addr_mode == MODE_AREG ? OPSIZE_LONG : inst->extra.size; - if (src_op.mode == MODE_REG_DIRECT) { - if (dst_op.mode == MODE_REG_DIRECT) { - add_rr(code, src_op.base, dst_op.base, size); - } else { - add_rrdisp(code, src_op.base, dst_op.base, dst_op.disp, size); - } - } else if (src_op.mode == MODE_REG_DISPLACE8) { - add_rdispr(code, src_op.base, src_op.disp, dst_op.base, size); - } else { - if (dst_op.mode == MODE_REG_DIRECT) { - add_ir(code, src_op.disp, dst_op.base, size); - } else { - add_irdisp(code, src_op.disp, dst_op.base, dst_op.disp, size); - } - } - if (inst->dst.addr_mode != MODE_AREG) { - set_flag_cond(opts, CC_C, FLAG_C); - set_flag_cond(opts, CC_Z, FLAG_Z); - set_flag_cond(opts, CC_S, FLAG_N); - set_flag_cond(opts, CC_O, FLAG_V); - if (opts->flag_regs[FLAG_C] >= 0) { - flag_to_flag(opts, FLAG_C, FLAG_X); - } else { - set_flag_cond(opts, CC_C, FLAG_X); - } - } - m68k_save_result(inst, opts); - break; - case M68K_ADDX: { - cycles(&opts->gen, BUS); - flag_to_carry(opts, FLAG_X); - if (src_op.mode == MODE_REG_DIRECT) { - if (dst_op.mode == MODE_REG_DIRECT) { - adc_rr(code, src_op.base, dst_op.base, inst->extra.size); - } else { - adc_rrdisp(code, src_op.base, dst_op.base, dst_op.disp, inst->extra.size); - } - } else if (src_op.mode == MODE_REG_DISPLACE8) { - adc_rdispr(code, src_op.base, src_op.disp, dst_op.base, inst->extra.size); - } else { - if (dst_op.mode == MODE_REG_DIRECT) { - adc_ir(code, src_op.disp, dst_op.base, inst->extra.size); - } else { - adc_irdisp(code, src_op.disp, dst_op.base, dst_op.disp, inst->extra.size); - } - } - set_flag_cond(opts, CC_C, FLAG_C); - - check_alloc_code(code, 2*MAX_INST_LEN); - code_ptr after_flag_set = code->cur + 1; - jcc(code, CC_Z, code->cur + 2); - set_flag(opts, 0, FLAG_Z); - *after_flag_set = code->cur - (after_flag_set+1); - set_flag_cond(opts, CC_S, FLAG_N); - set_flag_cond(opts, CC_O, FLAG_V); - if (opts->flag_regs[FLAG_C] >= 0) { - flag_to_flag(opts, FLAG_C, FLAG_X); - } else { - set_flag_cond(opts, CC_C, FLAG_X); - } - m68k_save_result(inst, opts); - break; - } - case M68K_AND: - cycles(&opts->gen, BUS); - if (src_op.mode == MODE_REG_DIRECT) { - if (dst_op.mode == MODE_REG_DIRECT) { - and_rr(code, src_op.base, dst_op.base, inst->extra.size); - } else { - and_rrdisp(code, src_op.base, dst_op.base, dst_op.disp, inst->extra.size); - } - } else if (src_op.mode == MODE_REG_DISPLACE8) { - and_rdispr(code, src_op.base, src_op.disp, dst_op.base, inst->extra.size); - } else { - if (dst_op.mode == MODE_REG_DIRECT) { - and_ir(code, src_op.disp, dst_op.base, inst->extra.size); - } else { - and_irdisp(code, src_op.disp, dst_op.base, dst_op.disp, inst->extra.size); - } - } - set_flag(opts, 0, FLAG_C); - set_flag_cond(opts, CC_Z, FLAG_Z); - set_flag_cond(opts, CC_S, FLAG_N); - set_flag(opts, 0, FLAG_V); - m68k_save_result(inst, opts); - break; - case M68K_ANDI_CCR: - case M68K_ANDI_SR: - cycles(&opts->gen, 20); - //TODO: If ANDI to SR, trap if not in supervisor mode - if (!(inst->src.params.immed & 0x1)) { - set_flag(opts, 0, FLAG_C); - } - if (!(inst->src.params.immed & 0x2)) { - set_flag(opts, 0, FLAG_V); - } - if (!(inst->src.params.immed & 0x4)) { - set_flag(opts, 0, FLAG_Z); - } - if (!(inst->src.params.immed & 0x8)) { - set_flag(opts, 0, FLAG_N); - } - if (!(inst->src.params.immed & 0x10)) { - set_flag(opts, 0, FLAG_X); - } - if (inst->op == M68K_ANDI_SR) { - and_irdisp(code, inst->src.params.immed >> 8, opts->gen.context_reg, offsetof(m68k_context, status), SZ_B); - if (!((inst->src.params.immed >> 8) & (1 << BIT_SUPERVISOR))) { - //leave supervisor mode - mov_rr(code, opts->aregs[7], opts->gen.scratch1, SZ_B); - mov_rdispr(code, opts->gen.context_reg, offsetof(m68k_context, aregs) + sizeof(uint32_t) * 8, opts->aregs[7], SZ_B); - mov_rrdisp(code, opts->gen.scratch1, opts->gen.context_reg, offsetof(m68k_context, aregs) + sizeof(uint32_t) * 8, SZ_B); - } - if (inst->src.params.immed & 0x700) { - call(code, opts->do_sync); - } - } - break; - case M68K_ASL: - case M68K_LSL: - translate_shift(opts, inst, &src_op, &dst_op, shl_ir, shl_irdisp, shl_clr, shl_clrdisp, shr_ir, shr_irdisp); - break; - case M68K_ASR: - translate_shift(opts, inst, &src_op, &dst_op, sar_ir, sar_irdisp, sar_clr, sar_clrdisp, NULL, NULL); - break; - case M68K_LSR: - translate_shift(opts, inst, &src_op, &dst_op, shr_ir, shr_irdisp, shr_clr, shr_clrdisp, shl_ir, shl_irdisp); - break; - case M68K_BCHG: - case M68K_BCLR: - case M68K_BSET: - case M68K_BTST: - cycles(&opts->gen, inst->extra.size == OPSIZE_BYTE ? 4 : ( - inst->op == M68K_BTST ? 6 : (inst->op == M68K_BCLR ? 10 : 8)) - ); - if (src_op.mode == MODE_IMMED) { - if (inst->extra.size == OPSIZE_BYTE) { - src_op.disp &= 0x7; - } - if (inst->op == M68K_BTST) { - if (dst_op.mode == MODE_REG_DIRECT) { - bt_ir(code, src_op.disp, dst_op.base, inst->extra.size); - } else { - bt_irdisp(code, src_op.disp, dst_op.base, dst_op.disp, inst->extra.size); - } - } else if (inst->op == M68K_BSET) { - if (dst_op.mode == MODE_REG_DIRECT) { - bts_ir(code, src_op.disp, dst_op.base, inst->extra.size); - } else { - bts_irdisp(code, src_op.disp, dst_op.base, dst_op.disp, inst->extra.size); - } - } else if (inst->op == M68K_BCLR) { - if (dst_op.mode == MODE_REG_DIRECT) { - btr_ir(code, src_op.disp, dst_op.base, inst->extra.size); - } else { - btr_irdisp(code, src_op.disp, dst_op.base, dst_op.disp, inst->extra.size); - } - } else { - if (dst_op.mode == MODE_REG_DIRECT) { - btc_ir(code, src_op.disp, dst_op.base, inst->extra.size); - } else { - btc_irdisp(code, src_op.disp, dst_op.base, dst_op.disp, inst->extra.size); - } - } - } else { - if (src_op.mode == MODE_REG_DISPLACE8 || (inst->dst.addr_mode != MODE_REG && src_op.base != opts->gen.scratch1 && src_op.base != opts->gen.scratch2)) { - if (dst_op.base == opts->gen.scratch1) { - push_r(code, opts->gen.scratch2); - if (src_op.mode == MODE_REG_DIRECT) { - mov_rr(code, src_op.base, opts->gen.scratch2, SZ_B); - } else { - mov_rdispr(code, src_op.base, src_op.disp, opts->gen.scratch2, SZ_B); - } - src_op.base = opts->gen.scratch2; - } else { - if (src_op.mode == MODE_REG_DIRECT) { - mov_rr(code, src_op.base, opts->gen.scratch1, SZ_B); - } else { - mov_rdispr(code, src_op.base, src_op.disp, opts->gen.scratch1, SZ_B); - } - src_op.base = opts->gen.scratch1; - } - } - uint8_t size = inst->extra.size; - if (dst_op.mode == MODE_REG_DISPLACE8) { - if (src_op.base != opts->gen.scratch1 && src_op.base != opts->gen.scratch2) { - if (src_op.mode == MODE_REG_DIRECT) { - mov_rr(code, src_op.base, opts->gen.scratch1, SZ_D); - } else { - mov_rdispr(code, src_op.base, src_op.disp, opts->gen.scratch1, SZ_D); - src_op.mode = MODE_REG_DIRECT; - } - src_op.base = opts->gen.scratch1; - } - //b### with register destination is modulo 32 - //x86 with a memory destination isn't modulo anything - //so use an and here to force the value to be modulo 32 - and_ir(code, 31, opts->gen.scratch1, SZ_D); - } else if(inst->dst.addr_mode != MODE_REG) { - //b### with memory destination is modulo 8 - //x86-64 doesn't support 8-bit bit operations - //so we fake it by forcing the bit number to be modulo 8 - and_ir(code, 7, src_op.base, SZ_D); - size = SZ_D; - } - if (inst->op == M68K_BTST) { - if (dst_op.mode == MODE_REG_DIRECT) { - bt_rr(code, src_op.base, dst_op.base, size); - } else { - bt_rrdisp(code, src_op.base, dst_op.base, dst_op.disp, size); - } - } else if (inst->op == M68K_BSET) { - if (dst_op.mode == MODE_REG_DIRECT) { - bts_rr(code, src_op.base, dst_op.base, size); - } else { - bts_rrdisp(code, src_op.base, dst_op.base, dst_op.disp, size); - } - } else if (inst->op == M68K_BCLR) { - if (dst_op.mode == MODE_REG_DIRECT) { - btr_rr(code, src_op.base, dst_op.base, size); - } else { - btr_rrdisp(code, src_op.base, dst_op.base, dst_op.disp, size); - } - } else { - if (dst_op.mode == MODE_REG_DIRECT) { - btc_rr(code, src_op.base, dst_op.base, size); - } else { - btc_rrdisp(code, src_op.base, dst_op.base, dst_op.disp, size); - } - } - if (src_op.base == opts->gen.scratch2) { - pop_r(code, opts->gen.scratch2); - } - } - //x86 sets the carry flag to the value of the bit tested - //68K sets the zero flag to the complement of the bit tested - set_flag_cond(opts, CC_NC, FLAG_Z); - if (inst->op != M68K_BTST) { - m68k_save_result(inst, opts); - } - break; - case M68K_CHK: - { - cycles(&opts->gen, 6); - if (dst_op.mode == MODE_REG_DIRECT) { - cmp_ir(code, 0, dst_op.base, inst->extra.size); - } else { - cmp_irdisp(code, 0, dst_op.base, dst_op.disp, inst->extra.size); - } - uint32_t isize; - switch(inst->src.addr_mode) - { - case MODE_AREG_DISPLACE: - case MODE_AREG_INDEX_DISP8: - case MODE_ABSOLUTE_SHORT: - case MODE_PC_INDEX_DISP8: - case MODE_PC_DISPLACE: - case MODE_IMMEDIATE: - isize = 4; - break; - case MODE_ABSOLUTE: - isize = 6; - break; - default: - isize = 2; - } - //make sure we won't start a new chunk in the middle of these branches - check_alloc_code(code, MAX_INST_LEN * 11); - code_ptr passed = code->cur + 1; - jcc(code, CC_GE, code->cur + 2); - set_flag(opts, 1, FLAG_N); - mov_ir(code, VECTOR_CHK, opts->gen.scratch2, SZ_D); - mov_ir(code, inst->address+isize, opts->gen.scratch1, SZ_D); - jmp(code, opts->trap); - *passed = code->cur - (passed+1); - if (dst_op.mode == MODE_REG_DIRECT) { - if (src_op.mode == MODE_REG_DIRECT) { - cmp_rr(code, src_op.base, dst_op.base, inst->extra.size); - } else if(src_op.mode == MODE_REG_DISPLACE8) { - cmp_rdispr(code, src_op.base, src_op.disp, dst_op.base, inst->extra.size); - } else { - cmp_ir(code, src_op.disp, dst_op.base, inst->extra.size); - } - } else if(dst_op.mode == MODE_REG_DISPLACE8) { - if (src_op.mode == MODE_REG_DIRECT) { - cmp_rrdisp(code, src_op.base, dst_op.base, dst_op.disp, inst->extra.size); - } else { - cmp_irdisp(code, src_op.disp, dst_op.base, dst_op.disp, inst->extra.size); - } - } - passed = code->cur + 1; - jcc(code, CC_LE, code->cur + 2); - set_flag(opts, 0, FLAG_N); - mov_ir(code, VECTOR_CHK, opts->gen.scratch2, SZ_D); - mov_ir(code, inst->address+isize, opts->gen.scratch1, SZ_D); - jmp(code, opts->trap); - *passed = code->cur - (passed+1); - cycles(&opts->gen, 4); - break; - } - case M68K_DIVS: - case M68K_DIVU: - { - check_alloc_code(code, MAX_NATIVE_SIZE); - //TODO: cycle exact division - cycles(&opts->gen, inst->op == M68K_DIVS ? 158 : 140); - set_flag(opts, 0, FLAG_C); - push_r(code, RDX); - push_r(code, RAX); - if (dst_op.mode == MODE_REG_DIRECT) { - mov_rr(code, dst_op.base, RAX, SZ_D); - } else { - mov_rdispr(code, dst_op.base, dst_op.disp, RAX, SZ_D); - } - if (src_op.mode == MODE_IMMED) { - mov_ir(code, (src_op.disp & 0x8000) && inst->op == M68K_DIVS ? src_op.disp | 0xFFFF0000 : src_op.disp, opts->gen.scratch2, SZ_D); - } else if (src_op.mode == MODE_REG_DIRECT) { - if (inst->op == M68K_DIVS) { - movsx_rr(code, src_op.base, opts->gen.scratch2, SZ_W, SZ_D); - } else { - movzx_rr(code, src_op.base, opts->gen.scratch2, SZ_W, SZ_D); - } - } else if (src_op.mode == MODE_REG_DISPLACE8) { - if (inst->op == M68K_DIVS) { - movsx_rdispr(code, src_op.base, src_op.disp, opts->gen.scratch2, SZ_W, SZ_D); - } else { - movzx_rdispr(code, src_op.base, src_op.disp, opts->gen.scratch2, SZ_W, SZ_D); - } - } - cmp_ir(code, 0, opts->gen.scratch2, SZ_D); - check_alloc_code(code, 6*MAX_INST_LEN); - code_ptr not_zero = code->cur + 1; - jcc(code, CC_NZ, code->cur + 2); - pop_r(code, RAX); - pop_r(code, RDX); - mov_ir(code, VECTOR_INT_DIV_ZERO, opts->gen.scratch2, SZ_D); - mov_ir(code, inst->address+2, opts->gen.scratch1, SZ_D); - jmp(code, opts->trap); - *not_zero = code->cur - (not_zero+1); - if (inst->op == M68K_DIVS) { - cdq(code); - } else { - xor_rr(code, RDX, RDX, SZ_D); - } - if (inst->op == M68K_DIVS) { - idiv_r(code, opts->gen.scratch2, SZ_D); - } else { - div_r(code, opts->gen.scratch2, SZ_D); - } - code_ptr skip_sec_check; - if (inst->op == M68K_DIVS) { - cmp_ir(code, 0x8000, RAX, SZ_D); - skip_sec_check = code->cur + 1; - jcc(code, CC_GE, code->cur + 2); - cmp_ir(code, -0x8000, RAX, SZ_D); - norm_off = code->cur + 1; - jcc(code, CC_L, code->cur + 2); - } else { - cmp_ir(code, 0x10000, RAX, SZ_D); - norm_off = code->cur + 1; - jcc(code, CC_NC, code->cur + 2); - } - if (dst_op.mode == MODE_REG_DIRECT) { - mov_rr(code, RDX, dst_op.base, SZ_W); - shl_ir(code, 16, dst_op.base, SZ_D); - mov_rr(code, RAX, dst_op.base, SZ_W); - } else { - mov_rrdisp(code, RDX, dst_op.base, dst_op.disp, SZ_W); - shl_irdisp(code, 16, dst_op.base, dst_op.disp, SZ_D); - mov_rrdisp(code, RAX, dst_op.base, dst_op.disp, SZ_W); - } - cmp_ir(code, 0, RAX, SZ_W); - pop_r(code, RAX); - pop_r(code, RDX); - set_flag(opts, 0, FLAG_V); - set_flag_cond(opts, CC_Z, FLAG_Z); - set_flag_cond(opts, CC_S, FLAG_N); - end_off = code->cur + 1; - jmp(code, code->cur + 2); - *norm_off = code->cur - (norm_off + 1); - if (inst->op == M68K_DIVS) { - *skip_sec_check = code->cur - (skip_sec_check+1); - } - pop_r(code, RAX); - pop_r(code, RDX); - set_flag(opts, 1, FLAG_V); - *end_off = code->cur - (end_off + 1); - break; - } - case M68K_EOR: - cycles(&opts->gen, BUS); - if (src_op.mode == MODE_REG_DIRECT) { - if (dst_op.mode == MODE_REG_DIRECT) { - xor_rr(code, src_op.base, dst_op.base, inst->extra.size); - } else { - xor_rrdisp(code, src_op.base, dst_op.base, dst_op.disp, inst->extra.size); - } - } else if (src_op.mode == MODE_REG_DISPLACE8) { - xor_rdispr(code, src_op.base, src_op.disp, dst_op.base, inst->extra.size); - } else { - if (dst_op.mode == MODE_REG_DIRECT) { - xor_ir(code, src_op.disp, dst_op.base, inst->extra.size); - } else { - xor_irdisp(code, src_op.disp, dst_op.base, dst_op.disp, inst->extra.size); - } - } - set_flag(opts, 0, FLAG_C); - set_flag_cond(opts, CC_Z, FLAG_Z); - set_flag_cond(opts, CC_S, FLAG_N); - set_flag(opts, 0, FLAG_V); - m68k_save_result(inst, opts); - break; - case M68K_EORI_CCR: - case M68K_EORI_SR: - cycles(&opts->gen, 20); - //TODO: If ANDI to SR, trap if not in supervisor mode - if (inst->src.params.immed & 0x1) { - xor_flag(opts, 1, FLAG_C); - } - if (inst->src.params.immed & 0x2) { - xor_flag(opts, 1, FLAG_V); - } - if (inst->src.params.immed & 0x4) { - xor_flag(opts, 1, FLAG_Z); - } - if (inst->src.params.immed & 0x8) { - xor_flag(opts, 1, FLAG_N); - } - if (inst->src.params.immed & 0x10) { - xor_flag(opts, 1, FLAG_X); - } - if (inst->op == M68K_ORI_SR) { - xor_irdisp(code, inst->src.params.immed >> 8, opts->gen.context_reg, offsetof(m68k_context, status), SZ_B); - if (inst->src.params.immed & 0x700) { - call(code, opts->do_sync); - } - } - break; - case M68K_EXG: - cycles(&opts->gen, 6); - if (dst_op.mode == MODE_REG_DIRECT) { - mov_rr(code, dst_op.base, opts->gen.scratch2, SZ_D); - if (src_op.mode == MODE_REG_DIRECT) { - mov_rr(code, src_op.base, dst_op.base, SZ_D); - mov_rr(code, opts->gen.scratch2, src_op.base, SZ_D); - } else { - mov_rdispr(code, src_op.base, src_op.disp, dst_op.base, SZ_D); - mov_rrdisp(code, opts->gen.scratch2, src_op.base, src_op.disp, SZ_D); - } - } else { - mov_rdispr(code, dst_op.base, dst_op.disp, opts->gen.scratch2, SZ_D); - if (src_op.mode == MODE_REG_DIRECT) { - mov_rrdisp(code, src_op.base, dst_op.base, dst_op.disp, SZ_D); - mov_rr(code, opts->gen.scratch2, src_op.base, SZ_D); - } else { - mov_rdispr(code, src_op.base, src_op.disp, opts->gen.scratch1, SZ_D); - mov_rrdisp(code, opts->gen.scratch1, dst_op.base, dst_op.disp, SZ_D); - mov_rrdisp(code, opts->gen.scratch2, src_op.base, src_op.disp, SZ_D); - } - } - break; - case M68K_ILLEGAL: - call(code, opts->gen.save_context); -#ifdef X86_64 - mov_rr(code, opts->gen.context_reg, RDI, SZ_PTR); -#else - push_r(code, opts->gen.context_reg); -#endif - call(code, (code_ptr)print_regs_exit); - break; - case M68K_MOVE_FROM_SR: - //TODO: Trap if not in system mode - call(code, opts->get_sr); - if (dst_op.mode == MODE_REG_DIRECT) { - mov_rr(code, opts->gen.scratch1, dst_op.base, SZ_W); - } else { - mov_rrdisp(code, opts->gen.scratch1, dst_op.base, dst_op.disp, SZ_W); - } - m68k_save_result(inst, opts); - break; - case M68K_MOVE_CCR: - case M68K_MOVE_SR: - //TODO: Privilege check for MOVE to SR - if (src_op.mode == MODE_IMMED) { - set_flag(opts, src_op.disp & 0x1, FLAG_C); - set_flag(opts, (src_op.disp >> 1) & 0x1, FLAG_V); - set_flag(opts, (src_op.disp >> 2) & 0x1, FLAG_Z); - set_flag(opts, (src_op.disp >> 3) & 0x1, FLAG_N); - set_flag(opts, (src_op.disp >> 4) & 0x1, FLAG_X); - if (inst->op == M68K_MOVE_SR) { - mov_irdisp(code, (src_op.disp >> 8), opts->gen.context_reg, offsetof(m68k_context, status), SZ_B); - if (!((inst->src.params.immed >> 8) & (1 << BIT_SUPERVISOR))) { - //leave supervisor mode - mov_rr(code, opts->aregs[7], opts->gen.scratch1, SZ_D); - mov_rdispr(code, opts->gen.context_reg, offsetof(m68k_context, aregs) + sizeof(uint32_t) * 8, opts->aregs[7], SZ_D); - mov_rrdisp(code, opts->gen.scratch1, opts->gen.context_reg, offsetof(m68k_context, aregs) + sizeof(uint32_t) * 8, SZ_D); - } - call(code, opts->do_sync); - } - cycles(&opts->gen, 12); - } else { - if (src_op.base != opts->gen.scratch1) { - if (src_op.mode == MODE_REG_DIRECT) { - mov_rr(code, src_op.base, opts->gen.scratch1, SZ_W); - } else { - mov_rdispr(code, src_op.base, src_op.disp, opts->gen.scratch1, SZ_W); - } - } - call(code, inst->op == M68K_MOVE_SR ? opts->set_sr : opts->set_ccr); - cycles(&opts->gen, 12); - - } - break; - case M68K_MOVE_USP: - cycles(&opts->gen, BUS); - //TODO: Trap if not in supervisor mode - //bt_irdisp(code, BIT_SUPERVISOR, opts->gen.context_reg, offsetof(m68k_context, status), SZ_B); - if (inst->src.addr_mode == MODE_UNUSED) { - if (dst_op.mode == MODE_REG_DIRECT) { - mov_rdispr(code, opts->gen.context_reg, offsetof(m68k_context, aregs) + sizeof(uint32_t) * 8, dst_op.base, SZ_D); - } else { - mov_rdispr(code, opts->gen.context_reg, offsetof(m68k_context, aregs) + sizeof(uint32_t) * 8, opts->gen.scratch1, SZ_D); - mov_rrdisp(code, opts->gen.scratch1, dst_op.base, dst_op.disp, SZ_D); - } - } else { - if (src_op.mode == MODE_REG_DIRECT) { - mov_rrdisp(code, src_op.base, opts->gen.context_reg, offsetof(m68k_context, aregs) + sizeof(uint32_t) * 8, SZ_D); - } else { - mov_rdispr(code, src_op.base, src_op.disp, opts->gen.scratch1, SZ_D); - mov_rrdisp(code, opts->gen.scratch1, opts->gen.context_reg, offsetof(m68k_context, aregs) + sizeof(uint32_t) * 8, SZ_D); - } - } - break; - //case M68K_MOVEP: - case M68K_MULS: - case M68K_MULU: - cycles(&opts->gen, 70); //TODO: Calculate the actual value based on the value of the <ea> parameter - if (src_op.mode == MODE_IMMED) { - mov_ir(code, inst->op == M68K_MULU ? (src_op.disp & 0xFFFF) : ((src_op.disp & 0x8000) ? src_op.disp | 0xFFFF0000 : src_op.disp), opts->gen.scratch1, SZ_D); - } else if (src_op.mode == MODE_REG_DIRECT) { - if (inst->op == M68K_MULS) { - movsx_rr(code, src_op.base, opts->gen.scratch1, SZ_W, SZ_D); - } else { - movzx_rr(code, src_op.base, opts->gen.scratch1, SZ_W, SZ_D); - } - } else { - if (inst->op == M68K_MULS) { - movsx_rdispr(code, src_op.base, src_op.disp, opts->gen.scratch1, SZ_W, SZ_D); - } else { - movzx_rdispr(code, src_op.base, src_op.disp, opts->gen.scratch1, SZ_W, SZ_D); - } - } - if (dst_op.mode == MODE_REG_DIRECT) { - dst_reg = dst_op.base; - if (inst->op == M68K_MULS) { - movsx_rr(code, dst_reg, dst_reg, SZ_W, SZ_D); - } else { - movzx_rr(code, dst_reg, dst_reg, SZ_W, SZ_D); - } - } else { - dst_reg = opts->gen.scratch2; - if (inst->op == M68K_MULS) { - movsx_rdispr(code, dst_op.base, dst_op.disp, opts->gen.scratch2, SZ_W, SZ_D); - } else { - movzx_rdispr(code, dst_op.base, dst_op.disp, opts->gen.scratch2, SZ_W, SZ_D); - } - } - imul_rr(code, opts->gen.scratch1, dst_reg, SZ_D); - if (dst_op.mode == MODE_REG_DISPLACE8) { - mov_rrdisp(code, dst_reg, dst_op.base, dst_op.disp, SZ_D); - } - set_flag(opts, 0, FLAG_V); - set_flag(opts, 0, FLAG_C); - cmp_ir(code, 0, dst_reg, SZ_D); - set_flag_cond(opts, CC_Z, FLAG_Z); - set_flag_cond(opts, CC_S, FLAG_N); - break; - //case M68K_NBCD: - case M68K_NEG: - cycles(&opts->gen, BUS); - if (dst_op.mode == MODE_REG_DIRECT) { - neg_r(code, dst_op.base, inst->extra.size); - } else { - neg_rdisp(code, dst_op.base, dst_op.disp, inst->extra.size); - } - set_flag_cond(opts, CC_C, FLAG_C); - set_flag_cond(opts, CC_Z, FLAG_Z); - set_flag_cond(opts, CC_S, FLAG_N); - set_flag_cond(opts, CC_O, FLAG_V); - if (opts->flag_regs[FLAG_C] >= 0) { - flag_to_flag(opts, FLAG_C, FLAG_X); - } else { - set_flag_cond(opts, CC_C, FLAG_X); - } - m68k_save_result(inst, opts); - break; - case M68K_NEGX: { - cycles(&opts->gen, BUS); - if (dst_op.mode == MODE_REG_DIRECT) { - if (dst_op.base == opts->gen.scratch1) { - push_r(code, opts->gen.scratch2); - xor_rr(code, opts->gen.scratch2, opts->gen.scratch2, inst->extra.size); - flag_to_carry(opts, FLAG_X); - sbb_rr(code, dst_op.base, opts->gen.scratch2, inst->extra.size); - mov_rr(code, opts->gen.scratch2, dst_op.base, inst->extra.size); - pop_r(code, opts->gen.scratch2); - } else { - xor_rr(code, opts->gen.scratch1, opts->gen.scratch1, inst->extra.size); - flag_to_carry(opts, FLAG_X); - sbb_rr(code, dst_op.base, opts->gen.scratch1, inst->extra.size); - mov_rr(code, opts->gen.scratch1, dst_op.base, inst->extra.size); - } - } else { - xor_rr(code, opts->gen.scratch1, opts->gen.scratch1, inst->extra.size); - flag_to_carry(opts, FLAG_X); - sbb_rdispr(code, dst_op.base, dst_op.disp, opts->gen.scratch1, inst->extra.size); - mov_rrdisp(code, opts->gen.scratch1, dst_op.base, dst_op.disp, inst->extra.size); - } - set_flag_cond(opts, CC_C, FLAG_C); - code_ptr after_flag_set = code->cur + 1; - jcc(code, CC_Z, code->cur + 2); - set_flag(opts, 0, FLAG_Z); - *after_flag_set = code->cur - (after_flag_set+1); - set_flag_cond(opts, CC_S, FLAG_N); - set_flag_cond(opts, CC_O, FLAG_V); - if (opts->flag_regs[FLAG_C] >= 0) { - flag_to_flag(opts, FLAG_C, FLAG_X); - } else { - set_flag_cond(opts, CC_C, FLAG_X); - } - m68k_save_result(inst, opts); - break; - } - case M68K_NOP: - cycles(&opts->gen, BUS); - break; - case M68K_NOT: - if (dst_op.mode == MODE_REG_DIRECT) { - not_r(code, dst_op.base, inst->extra.size); - cmp_ir(code, 0, dst_op.base, inst->extra.size); - } else { - not_rdisp(code, dst_op.base, dst_op.disp, inst->extra.size); - cmp_irdisp(code, 0, dst_op.base, dst_op.disp, inst->extra.size); - } - - set_flag(opts, 0, FLAG_C); - set_flag_cond(opts, CC_Z, FLAG_Z); - set_flag_cond(opts, CC_S, FLAG_N); - set_flag(opts, 0, FLAG_V); - m68k_save_result(inst, opts); - break; - case M68K_OR: - cycles(&opts->gen, BUS); - if (src_op.mode == MODE_REG_DIRECT) { - if (dst_op.mode == MODE_REG_DIRECT) { - or_rr(code, src_op.base, dst_op.base, inst->extra.size); - } else { - or_rrdisp(code, src_op.base, dst_op.base, dst_op.disp, inst->extra.size); - } - } else if (src_op.mode == MODE_REG_DISPLACE8) { - or_rdispr(code, src_op.base, src_op.disp, dst_op.base, inst->extra.size); - } else { - if (dst_op.mode == MODE_REG_DIRECT) { - or_ir(code, src_op.disp, dst_op.base, inst->extra.size); - } else { - or_irdisp(code, src_op.disp, dst_op.base, dst_op.disp, inst->extra.size); - } - } - set_flag(opts, 0, FLAG_C); - set_flag_cond(opts, CC_Z, FLAG_Z); - set_flag_cond(opts, CC_S, FLAG_N); - set_flag(opts, 0, FLAG_V); - m68k_save_result(inst, opts); - break; - case M68K_ORI_CCR: - case M68K_ORI_SR: - cycles(&opts->gen, 20); - //TODO: If ANDI to SR, trap if not in supervisor mode - if (inst->src.params.immed & 0x1) { - set_flag(opts, 1, FLAG_C); - } - if (inst->src.params.immed & 0x2) { - set_flag(opts, 1, FLAG_V); - } - if (inst->src.params.immed & 0x4) { - set_flag(opts, 1, FLAG_Z); - } - if (inst->src.params.immed & 0x8) { - set_flag(opts, 1, FLAG_N); - } - if (inst->src.params.immed & 0x10) { - set_flag(opts, 1, FLAG_X); - } - if (inst->op == M68K_ORI_SR) { - or_irdisp(code, inst->src.params.immed >> 8, opts->gen.context_reg, offsetof(m68k_context, status), SZ_B); - if (inst->src.params.immed & 0x700) { - call(code, opts->do_sync); - } - } - break; - case M68K_RESET: - call(code, opts->gen.save_context); -#ifdef X86_64 - mov_rr(code, opts->gen.context_reg, RDI, SZ_PTR); -#else - push_r(code, opts->gen.context_reg); -#endif - call(code, (code_ptr)print_regs_exit); - break; - case M68K_ROL: - case M68K_ROR: - set_flag(opts, 0, FLAG_V); - if (inst->src.addr_mode == MODE_UNUSED) { - cycles(&opts->gen, BUS); - //Memory rotate - if (inst->op == M68K_ROL) { - rol_ir(code, 1, dst_op.base, inst->extra.size); - } else { - ror_ir(code, 1, dst_op.base, inst->extra.size); - } - set_flag_cond(opts, CC_C, FLAG_C); - cmp_ir(code, 0, dst_op.base, inst->extra.size); - set_flag_cond(opts, CC_Z, FLAG_Z); - set_flag_cond(opts, CC_S, FLAG_N); - m68k_save_result(inst, opts); - } else { - if (src_op.mode == MODE_IMMED) { - cycles(&opts->gen, (inst->extra.size == OPSIZE_LONG ? 8 : 6) + src_op.disp*2); - if (dst_op.mode == MODE_REG_DIRECT) { - if (inst->op == M68K_ROL) { - rol_ir(code, src_op.disp, dst_op.base, inst->extra.size); - } else { - ror_ir(code, src_op.disp, dst_op.base, inst->extra.size); - } - } else { - if (inst->op == M68K_ROL) { - rol_irdisp(code, src_op.disp, dst_op.base, dst_op.disp, inst->extra.size); - } else { - ror_irdisp(code, src_op.disp, dst_op.base, dst_op.disp, inst->extra.size); - } - } - set_flag_cond(opts, CC_C, FLAG_C); - } else { - if (src_op.mode == MODE_REG_DIRECT) { - if (src_op.base != opts->gen.scratch1) { - mov_rr(code, src_op.base, opts->gen.scratch1, SZ_B); - } - } else { - mov_rdispr(code, src_op.base, src_op.disp, opts->gen.scratch1, SZ_B); - } - and_ir(code, 63, opts->gen.scratch1, SZ_D); - zero_off = code->cur + 1; - jcc(code, CC_Z, code->cur + 2); - add_rr(code, opts->gen.scratch1, CYCLES, SZ_D); - add_rr(code, opts->gen.scratch1, CYCLES, SZ_D); - cmp_ir(code, 32, opts->gen.scratch1, SZ_B); - norm_off = code->cur + 1; - jcc(code, CC_L, code->cur + 2); - sub_ir(code, 32, opts->gen.scratch1, SZ_B); - if (dst_op.mode == MODE_REG_DIRECT) { - if (inst->op == M68K_ROL) { - rol_ir(code, 31, dst_op.base, inst->extra.size); - rol_ir(code, 1, dst_op.base, inst->extra.size); - } else { - ror_ir(code, 31, dst_op.base, inst->extra.size); - ror_ir(code, 1, dst_op.base, inst->extra.size); - } - } else { - if (inst->op == M68K_ROL) { - rol_irdisp(code, 31, dst_op.base, dst_op.disp, inst->extra.size); - rol_irdisp(code, 1, dst_op.base, dst_op.disp, inst->extra.size); - } else { - ror_irdisp(code, 31, dst_op.base, dst_op.disp, inst->extra.size); - ror_irdisp(code, 1, dst_op.base, dst_op.disp, inst->extra.size); - } - } - *norm_off = code->cur - (norm_off+1); - if (dst_op.mode == MODE_REG_DIRECT) { - if (inst->op == M68K_ROL) { - rol_clr(code, dst_op.base, inst->extra.size); - } else { - ror_clr(code, dst_op.base, inst->extra.size); - } - } else { - if (inst->op == M68K_ROL) { - rol_clrdisp(code, dst_op.base, dst_op.disp, inst->extra.size); - } else { - ror_clrdisp(code, dst_op.base, dst_op.disp, inst->extra.size); - } - } - set_flag_cond(opts, CC_C, FLAG_C); - end_off = code->cur + 1; - jmp(code, code->cur + 2); - *zero_off = code->cur - (zero_off+1); - set_flag(opts, 0, FLAG_C); - *end_off = code->cur - (end_off+1); - } - if (dst_op.mode == MODE_REG_DIRECT) { - cmp_ir(code, 0, dst_op.base, inst->extra.size); - } else { - cmp_irdisp(code, 0, dst_op.base, dst_op.disp, inst->extra.size); - } - set_flag_cond(opts, CC_Z, FLAG_Z); - set_flag_cond(opts, CC_S, FLAG_N); - } - break; - case M68K_ROXL: - case M68K_ROXR: - set_flag(opts, 0, FLAG_V); - if (inst->src.addr_mode == MODE_UNUSED) { - cycles(&opts->gen, BUS); - //Memory rotate - flag_to_carry(opts, FLAG_X); - if (inst->op == M68K_ROXL) { - rcl_ir(code, 1, dst_op.base, inst->extra.size); - } else { - rcr_ir(code, 1, dst_op.base, inst->extra.size); - } - set_flag_cond(opts, CC_C, FLAG_C); - if (opts->flag_regs[FLAG_C] < 0) { - set_flag_cond(opts, CC_C, FLAG_X); - } - cmp_ir(code, 0, dst_op.base, inst->extra.size); - set_flag_cond(opts, CC_Z, FLAG_Z); - set_flag_cond(opts, CC_S, FLAG_N); - if (opts->flag_regs[FLAG_C] >= 0) { - flag_to_flag(opts, FLAG_C, FLAG_X); - } - m68k_save_result(inst, opts); - } else { - if (src_op.mode == MODE_IMMED) { - cycles(&opts->gen, (inst->extra.size == OPSIZE_LONG ? 8 : 6) + src_op.disp*2); - flag_to_carry(opts, FLAG_X); - if (dst_op.mode == MODE_REG_DIRECT) { - if (inst->op == M68K_ROXL) { - rcl_ir(code, src_op.disp, dst_op.base, inst->extra.size); - } else { - rcr_ir(code, src_op.disp, dst_op.base, inst->extra.size); - } - } else { - if (inst->op == M68K_ROXL) { - rcl_irdisp(code, src_op.disp, dst_op.base, dst_op.disp, inst->extra.size); - } else { - rcr_irdisp(code, src_op.disp, dst_op.base, dst_op.disp, inst->extra.size); - } - } - set_flag_cond(opts, CC_C, FLAG_C); - if (opts->flag_regs[FLAG_C] >= 0) { - flag_to_flag(opts, FLAG_C, FLAG_X); - } else { - set_flag_cond(opts, CC_C, FLAG_X); - } - } else { - if (src_op.mode == MODE_REG_DIRECT) { - if (src_op.base != opts->gen.scratch1) { - mov_rr(code, src_op.base, opts->gen.scratch1, SZ_B); - } - } else { - mov_rdispr(code, src_op.base, src_op.disp, opts->gen.scratch1, SZ_B); - } - and_ir(code, 63, opts->gen.scratch1, SZ_D); - zero_off = code->cur + 1; - jcc(code, CC_Z, code->cur + 2); - add_rr(code, opts->gen.scratch1, CYCLES, SZ_D); - add_rr(code, opts->gen.scratch1, CYCLES, SZ_D); - cmp_ir(code, 32, opts->gen.scratch1, SZ_B); - norm_off = code->cur + 1; - jcc(code, CC_L, code->cur + 2); - flag_to_carry(opts, FLAG_X); - if (dst_op.mode == MODE_REG_DIRECT) { - if (inst->op == M68K_ROXL) { - rcl_ir(code, 31, dst_op.base, inst->extra.size); - rcl_ir(code, 1, dst_op.base, inst->extra.size); - } else { - rcr_ir(code, 31, dst_op.base, inst->extra.size); - rcr_ir(code, 1, dst_op.base, inst->extra.size); - } - } else { - if (inst->op == M68K_ROXL) { - rcl_irdisp(code, 31, dst_op.base, dst_op.disp, inst->extra.size); - rcl_irdisp(code, 1, dst_op.base, dst_op.disp, inst->extra.size); - } else { - rcr_irdisp(code, 31, dst_op.base, dst_op.disp, inst->extra.size); - rcr_irdisp(code, 1, dst_op.base, dst_op.disp, inst->extra.size); - } - } - set_flag_cond(opts, CC_C, FLAG_X); - sub_ir(code, 32, opts->gen.scratch1, SZ_B); - *norm_off = code->cur - (norm_off+1); - flag_to_carry(opts, FLAG_X); - if (dst_op.mode == MODE_REG_DIRECT) { - if (inst->op == M68K_ROXL) { - rcl_clr(code, dst_op.base, inst->extra.size); - } else { - rcr_clr(code, dst_op.base, inst->extra.size); - } - } else { - if (inst->op == M68K_ROXL) { - rcl_clrdisp(code, dst_op.base, dst_op.disp, inst->extra.size); - } else { - rcr_clrdisp(code, dst_op.base, dst_op.disp, inst->extra.size); - } - } - set_flag_cond(opts, CC_C, FLAG_C); - if (opts->flag_regs[FLAG_C] >= 0) { - flag_to_flag(opts, FLAG_C, FLAG_X); - } else { - set_flag_cond(opts, CC_C, FLAG_X); - } - end_off = code->cur + 1; - jmp(code, code->cur + 2); - *zero_off = code->cur - (zero_off+1); - //Carry flag is set to X flag when count is 0, this is different from ROR/ROL - flag_to_flag(opts, FLAG_X, FLAG_C); - *end_off = code->cur - (end_off+1); - } - if (dst_op.mode == MODE_REG_DIRECT) { - cmp_ir(code, 0, dst_op.base, inst->extra.size); - } else { - cmp_irdisp(code, 0, dst_op.base, dst_op.disp, inst->extra.size); - } - set_flag_cond(opts, CC_Z, FLAG_Z); - set_flag_cond(opts, CC_S, FLAG_N); - } - break; - case M68K_RTE: - //TODO: Trap if not in system mode - //Read saved SR - mov_rr(code, opts->aregs[7], opts->gen.scratch1, SZ_D); - call(code, opts->read_16); - add_ir(code, 2, opts->aregs[7], SZ_D); - call(code, opts->set_sr); - //Read saved PC - mov_rr(code, opts->aregs[7], opts->gen.scratch1, SZ_D); - call(code, opts->read_32); - add_ir(code, 4, opts->aregs[7], SZ_D); - //Check if we've switched to user mode and swap stack pointers if needed - bt_irdisp(code, 5, opts->gen.context_reg, offsetof(m68k_context, status), SZ_B); - end_off = code->cur + 1; - jcc(code, CC_C, code->cur + 2); - mov_rr(code, opts->aregs[7], opts->gen.scratch2, SZ_D); - mov_rdispr(code, opts->gen.context_reg, offsetof(m68k_context, aregs) + sizeof(uint32_t) * 8, opts->aregs[7], SZ_D); - mov_rrdisp(code, opts->gen.scratch2, opts->gen.context_reg, offsetof(m68k_context, aregs) + sizeof(uint32_t) * 8, SZ_D); - *end_off = code->cur - (end_off+1); - //Get native address, sync components, recalculate integer points and jump to returned address - call(code, opts->native_addr_and_sync); - jmp_r(code, opts->gen.scratch1); - break; - case M68K_RTR: - //Read saved CCR - mov_rr(code, opts->aregs[7], opts->gen.scratch1, SZ_D); - call(code, opts->read_16); - add_ir(code, 2, opts->aregs[7], SZ_D); - call(code, opts->set_ccr); - //Read saved PC - mov_rr(code, opts->aregs[7], opts->gen.scratch1, SZ_D); - call(code, opts->read_32); - add_ir(code, 4, opts->aregs[7], SZ_D); - //Get native address and jump to it - call(code, opts->native_addr); - jmp_r(code, opts->gen.scratch1); - break; - case M68K_SBCD: { - if (src_op.base != opts->gen.scratch2) { - if (src_op.mode == MODE_REG_DIRECT) { - mov_rr(code, src_op.base, opts->gen.scratch2, SZ_B); - } else { - mov_rdispr(code, src_op.base, src_op.disp, opts->gen.scratch2, SZ_B); - } - } - if (dst_op.base != opts->gen.scratch1) { - if (dst_op.mode == MODE_REG_DIRECT) { - mov_rr(code, dst_op.base, opts->gen.scratch1, SZ_B); - } else { - mov_rdispr(code, dst_op.base, dst_op.disp, opts->gen.scratch1, SZ_B); - } - } - flag_to_carry(opts, FLAG_X); - jcc(code, CC_NC, code->cur + 5); - sub_ir(code, 1, opts->gen.scratch1, SZ_B); - call(code, (code_ptr)bcd_sub); - reg_to_flag(opts, CH, FLAG_C); - reg_to_flag(opts, CH, FLAG_X); - cmp_ir(code, 0, opts->gen.scratch1, SZ_B); - code_ptr after_flag_set = code->cur+1; - jcc(code, CC_Z, code->cur + 2); - set_flag(opts, 0, FLAG_Z); - *after_flag_set = code->cur - (after_flag_set+1); - if (dst_op.base != opts->gen.scratch1) { - if (dst_op.mode == MODE_REG_DIRECT) { - mov_rr(code, opts->gen.scratch1, dst_op.base, SZ_B); - } else { - mov_rrdisp(code, opts->gen.scratch1, dst_op.base, dst_op.disp, SZ_B); - } - } - m68k_save_result(inst, opts); - break; - } - case M68K_STOP: { - //TODO: Trap if not in system mode - //manual says 4 cycles, but it has to be at least 8 since it's a 2-word instruction - //possibly even 12 since that's how long MOVE to SR takes - cycles(&opts->gen, BUS*2); - set_flag(opts, src_op.disp & 0x1, FLAG_C); - set_flag(opts, (src_op.disp >> 1) & 0x1, FLAG_V); - set_flag(opts, (src_op.disp >> 2) & 0x1, FLAG_Z); - set_flag(opts, (src_op.disp >> 3) & 0x1, FLAG_N); - set_flag(opts, (src_op.disp >> 4) & 0x1, FLAG_X); - mov_irdisp(code, (src_op.disp >> 8), opts->gen.context_reg, offsetof(m68k_context, status), SZ_B); - if (!((inst->src.params.immed >> 8) & (1 << BIT_SUPERVISOR))) { - //leave supervisor mode - mov_rr(code, opts->aregs[7], opts->gen.scratch1, SZ_D); - mov_rdispr(code, opts->gen.context_reg, offsetof(m68k_context, aregs) + sizeof(uint32_t) * 8, opts->aregs[7], SZ_D); - mov_rrdisp(code, opts->gen.scratch1, opts->gen.context_reg, offsetof(m68k_context, aregs) + sizeof(uint32_t) * 8, SZ_D); - } - code_ptr loop_top = code->cur; - call(code, opts->do_sync); - cmp_rr(code, opts->gen.limit, opts->gen.cycles, SZ_D); - code_ptr normal_cycle_up = code->cur + 1; - jcc(code, CC_A, code->cur + 2); - cycles(&opts->gen, BUS); - code_ptr after_cycle_up = code->cur + 1; - jmp(code, code->cur + 2); - *normal_cycle_up = code->cur - (normal_cycle_up + 1); - mov_rr(code, opts->gen.limit, opts->gen.cycles, SZ_D); - *after_cycle_up = code->cur - (after_cycle_up+1); - cmp_rdispr(code, opts->gen.context_reg, offsetof(m68k_context, int_cycle), opts->gen.cycles, SZ_D); - jcc(code, CC_C, loop_top); - break; - } - case M68K_SUB: - size = inst->dst.addr_mode == MODE_AREG ? OPSIZE_LONG : inst->extra.size; - cycles(&opts->gen, BUS); - if (src_op.mode == MODE_REG_DIRECT) { - if (dst_op.mode == MODE_REG_DIRECT) { - sub_rr(code, src_op.base, dst_op.base, size); - } else { - sub_rrdisp(code, src_op.base, dst_op.base, dst_op.disp, size); - } - } else if (src_op.mode == MODE_REG_DISPLACE8) { - sub_rdispr(code, src_op.base, src_op.disp, dst_op.base, size); - } else { - if (dst_op.mode == MODE_REG_DIRECT) { - sub_ir(code, src_op.disp, dst_op.base, size); - } else { - sub_irdisp(code, src_op.disp, dst_op.base, dst_op.disp, size); - } - } - if (inst->dst.addr_mode != MODE_AREG) { - set_flag_cond(opts, CC_C, FLAG_C); - set_flag_cond(opts, CC_Z, FLAG_Z); - set_flag_cond(opts, CC_S, FLAG_N); - set_flag_cond(opts, CC_O, FLAG_V); - if (opts->flag_regs[FLAG_C] >= 0) { - flag_to_flag(opts, FLAG_C, FLAG_X); - } else { - set_flag_cond(opts, CC_C, FLAG_X); - } - } - m68k_save_result(inst, opts); - break; - case M68K_SUBX: { - cycles(&opts->gen, BUS); - flag_to_carry(opts, FLAG_X); - if (src_op.mode == MODE_REG_DIRECT) { - if (dst_op.mode == MODE_REG_DIRECT) { - sbb_rr(code, src_op.base, dst_op.base, inst->extra.size); - } else { - sbb_rrdisp(code, src_op.base, dst_op.base, dst_op.disp, inst->extra.size); - } - } else if (src_op.mode == MODE_REG_DISPLACE8) { - sbb_rdispr(code, src_op.base, src_op.disp, dst_op.base, inst->extra.size); - } else { - if (dst_op.mode == MODE_REG_DIRECT) { - sbb_ir(code, src_op.disp, dst_op.base, inst->extra.size); - } else { - sbb_irdisp(code, src_op.disp, dst_op.base, dst_op.disp, inst->extra.size); - } - } - set_flag_cond(opts, CC_C, FLAG_C); - if (opts->flag_regs[FLAG_C] < 0) { - set_flag_cond(opts, CC_C, FLAG_X); - } - code_ptr after_flag_set = code->cur + 1; - jcc(code, CC_Z, code->cur + 2); - set_flag(opts, 0, FLAG_Z); - *after_flag_set = code->cur - (after_flag_set+1); - set_flag_cond(opts, CC_S, FLAG_N); - set_flag_cond(opts, CC_O, FLAG_V); - if (opts->flag_regs[FLAG_C] >= 0) { - flag_to_flag(opts, FLAG_C, FLAG_X); - } - m68k_save_result(inst, opts); - break; - } - case M68K_SWAP: - cycles(&opts->gen, BUS); - if (src_op.mode == MODE_REG_DIRECT) { - rol_ir(code, 16, src_op.base, SZ_D); - cmp_ir(code, 0, src_op.base, SZ_D); - } else{ - rol_irdisp(code, 16, src_op.base, src_op.disp, SZ_D); - cmp_irdisp(code, 0, src_op.base, src_op.disp, SZ_D); - } - - set_flag(opts, 0, FLAG_C); - set_flag_cond(opts, CC_Z, FLAG_Z); - set_flag_cond(opts, CC_S, FLAG_N); - set_flag(opts, 0, FLAG_V); - break; - //case M68K_TAS: - case M68K_TRAP: - mov_ir(code, src_op.disp + VECTOR_TRAP_0, opts->gen.scratch2, SZ_D); - mov_ir(code, inst->address+2, opts->gen.scratch1, SZ_D); - jmp(code, opts->trap); - break; - //case M68K_TRAPV: - case M68K_TST: - cycles(&opts->gen, BUS); - if (src_op.mode == MODE_REG_DIRECT) { - cmp_ir(code, 0, src_op.base, inst->extra.size); - } else { //M68000 doesn't support immedate operand for tst, so this must be MODE_REG_DISPLACE8 - cmp_irdisp(code, 0, src_op.base, src_op.disp, inst->extra.size); - } - set_flag(opts, 0, FLAG_C); - set_flag_cond(opts, CC_Z, FLAG_Z); - set_flag_cond(opts, CC_S, FLAG_N); - set_flag(opts, 0, FLAG_V); - break; - case M68K_UNLK: - cycles(&opts->gen, BUS); - if (dst_op.mode == MODE_REG_DIRECT) { - mov_rr(code, dst_op.base, opts->aregs[7], SZ_D); - } else { - mov_rdispr(code, dst_op.base, dst_op.disp, opts->aregs[7], SZ_D); - } - mov_rr(code, opts->aregs[7], opts->gen.scratch1, SZ_D); - call(code, opts->read_32); - if (dst_op.mode == MODE_REG_DIRECT) { - mov_rr(code, opts->gen.scratch1, dst_op.base, SZ_D); - } else { - mov_rrdisp(code, opts->gen.scratch1, dst_op.base, dst_op.disp, SZ_D); - } - add_ir(code, 4, opts->aregs[7], SZ_D); - break; - default: - m68k_disasm(inst, disasm_buf); - printf("%X: %s\ninstruction %d not yet implemented\n", inst->address, disasm_buf, inst->op); - exit(1); - } -} - -uint8_t m68k_is_terminal(m68kinst * inst) -{ - return inst->op == M68K_RTS || inst->op == M68K_RTE || inst->op == M68K_RTR || inst->op == M68K_JMP - || inst->op == M68K_TRAP || inst->op == M68K_ILLEGAL || inst->op == M68K_INVALID || inst->op == M68K_RESET - || (inst->op == M68K_BCC && inst->extra.cond == COND_TRUE); -} - -void m68k_handle_deferred(m68k_context * context) -{ - m68k_options * opts = context->options; - process_deferred(&opts->gen.deferred, context, (native_addr_func)get_native_from_context); - if (opts->gen.deferred) { - translate_m68k_stream(opts->gen.deferred->address, context); - } -} - -void translate_m68k_stream(uint32_t address, m68k_context * context) -{ - m68kinst instbuf; - m68k_options * opts = context->options; - code_info *code = &opts->gen.code; - address &= 0xFFFFFF; - if(get_native_address(opts->gen.native_code_map, address)) { - return; - } - char disbuf[1024]; - uint16_t *encoded, *next; - if ((address & 0xFFFFFF) < 0x400000) { - encoded = context->mem_pointers[0] + (address & 0xFFFFFF)/2; - } else if ((address & 0xFFFFFF) > 0xE00000) { - encoded = context->mem_pointers[1] + (address & 0xFFFF)/2; - } else { - printf("attempt to translate non-memory address: %X\n", address); - exit(1); - } - do { - if (opts->address_log) { - fprintf(opts->address_log, "%X\n", address); - } - do { - if (address >= 0x400000 && address < 0xE00000) { - xor_rr(code, RDI, RDI, SZ_D); -#ifdef X86_32 - push_r(code, RDI); -#endif - call(code, (code_ptr)exit); - break; - } - code_ptr existing = get_native_address(opts->gen.native_code_map, address); - if (existing) { - jmp(code, existing); - break; - } - next = m68k_decode(encoded, &instbuf, address); - if (instbuf.op == M68K_INVALID) { - instbuf.src.params.immed = *encoded; - } - uint16_t m68k_size = (next-encoded)*2; - address += m68k_size; - encoded = next; - //m68k_disasm(&instbuf, disbuf); - //printf("%X: %s\n", instbuf.address, disbuf); - - //make sure the beginning of the code for an instruction is contiguous - check_alloc_code(code, MAX_INST_LEN*4); - code_ptr start = code->cur; - translate_m68k(opts, &instbuf); - code_ptr after = code->cur; - map_native_address(context, instbuf.address, start, m68k_size, after-start); - after; - } while(!m68k_is_terminal(&instbuf)); - process_deferred(&opts->gen.deferred, context, (native_addr_func)get_native_from_context); - if (opts->gen.deferred) { - address = opts->gen.deferred->address; - if ((address & 0xFFFFFF) < 0x400000) { - encoded = context->mem_pointers[0] + (address & 0xFFFFFF)/2; - } else if ((address & 0xFFFFFF) > 0xE00000) { - encoded = context->mem_pointers[1] + (address & 0xFFFF)/2; - } else { - printf("attempt to translate non-memory address: %X\n", address); - exit(1); - } - } else { - encoded = NULL; - } - } while(encoded != NULL); -} - -code_ptr get_native_address_trans(m68k_context * context, uint32_t address) -{ - address &= 0xFFFFFF; - code_ptr ret = get_native_address(context->native_code_map, address); - if (!ret) { - translate_m68k_stream(address, context); - ret = get_native_address(context->native_code_map, address); - } - return ret; -} - -void * m68k_retranslate_inst(uint32_t address, m68k_context * context) -{ - m68k_options * opts = context->options; - code_info *code = &opts->gen.code; - uint8_t orig_size = get_native_inst_size(opts, address); - code_ptr orig_start = get_native_address(context->native_code_map, address); - uint32_t orig = address; - code_info orig_code; - orig_code.cur = orig_start; - orig_code.last = orig_start + orig_size + 5; - address &= 0xFFFF; - uint16_t *after, *inst = context->mem_pointers[1] + address/2; - m68kinst instbuf; - after = m68k_decode(inst, &instbuf, orig); - if (orig_size != MAX_NATIVE_SIZE) { - deferred_addr * orig_deferred = opts->gen.deferred; - - //make sure the beginning of the code for an instruction is contiguous - check_alloc_code(code, MAX_INST_LEN*4); - code_ptr native_start = code->cur; - translate_m68k(opts, &instbuf); - code_ptr native_end = code->cur; - uint8_t is_terminal = m68k_is_terminal(&instbuf); - if ((native_end - native_start) <= orig_size) { - code_ptr native_next; - if (!is_terminal) { - native_next = get_native_address(context->native_code_map, orig + (after-inst)*2); - } - if (is_terminal || (native_next && ((native_next == orig_start + orig_size) || (orig_size - (native_end - native_start)) > 5))) { - remove_deferred_until(&opts->gen.deferred, orig_deferred); - code_info tmp; - tmp.cur = code->cur; - tmp.last = code->last; - code->cur = orig_code.cur; - code->last = orig_code.last; - translate_m68k(opts, &instbuf); - native_end = orig_code.cur = code->cur; - code->cur = tmp.cur; - code->last = tmp.last; - if (!is_terminal) { - if (native_next == orig_start + orig_size && (native_next-native_end) < 2) { - while (orig_code.cur < orig_start + orig_size) { - *(orig_code.cur++) = 0x90; //NOP - } - } else { - jmp(&orig_code, native_next); - } - } - m68k_handle_deferred(context); - return orig_start; - } - } - - map_native_address(context, instbuf.address, native_start, (after-inst)*2, MAX_NATIVE_SIZE); - - jmp(&orig_code, native_start); - if (!m68k_is_terminal(&instbuf)) { - code_ptr native_end = code->cur; - code->cur = native_start + MAX_NATIVE_SIZE; - code_ptr rest = get_native_address_trans(context, orig + (after-inst)*2); - code_ptr tmp = code->cur; - code->cur = native_end; - jmp(code, rest); - code->cur = tmp; - } else { - code->cur = native_start + MAX_NATIVE_SIZE; - } - m68k_handle_deferred(context); - return native_start; - } else { - code_info tmp; - tmp.cur = code->cur; - tmp.last = code->last; - code->cur = orig_code.cur; - code->last = orig_code.last; - translate_m68k(opts, &instbuf); - if (!m68k_is_terminal(&instbuf)) { - jmp(code, get_native_address_trans(context, orig + (after-inst)*2)); - } - code->cur = tmp.cur; - code->last = tmp.last; - m68k_handle_deferred(context); - return orig_start; - } -} - -m68k_context * m68k_handle_code_write(uint32_t address, m68k_context * context) -{ - uint32_t inst_start = get_instruction_start(context->native_code_map, address | 0xFF0000); - if (inst_start) { - m68k_options * options = context->options; - code_info *code = &options->gen.code; - code_ptr dst = get_native_address(context->native_code_map, inst_start); - code_info orig; - orig.cur = dst; - orig.last = dst + 128; - mov_ir(&orig, inst_start, options->gen.scratch2, SZ_D); - - if (!options->retrans_stub) { - options->retrans_stub = code->cur; - call(code, options->gen.save_context); - push_r(code, options->gen.context_reg); -#ifdef X86_32 - push_r(code, options->gen.context_reg); - push_r(code, options->gen.scratch2); -#endif - call(code, (code_ptr)m68k_retranslate_inst); -#ifdef X86_32 - add_ir(code, 8, RSP, SZ_D); -#endif - pop_r(code, options->gen.context_reg); - mov_rr(code, RAX, options->gen.scratch1, SZ_PTR); - call(code, options->gen.load_context); - jmp_r(code, options->gen.scratch1); - } - jmp(&orig, options->retrans_stub); - } - return context; -} - -void insert_breakpoint(m68k_context * context, uint32_t address, code_ptr bp_handler) -{ - static code_ptr bp_stub = NULL; - m68k_options * opts = context->options; - code_info native; - native.cur = get_native_address_trans(context, address); - native.last = native.cur + 128; - code_ptr start_native = native.cur; - mov_ir(&native, address, opts->gen.scratch1, SZ_D); - if (!bp_stub) { - code_info *code = &opts->gen.code; - check_alloc_code(code, 5); - bp_stub = code->cur; - call(&native, bp_stub); - - //Calculate length of prologue - check_cycles_int(&opts->gen, address); - int check_int_size = code->cur-bp_stub; - code->cur = bp_stub; - - //Save context and call breakpoint handler - call(code, opts->gen.save_context); - push_r(code, opts->gen.scratch1); -#ifdef X86_64 - mov_rr(code, opts->gen.context_reg, RDI, SZ_PTR); - mov_rr(code, opts->gen.scratch1, RSI, SZ_D); -#else - push_r(code, opts->gen.scratch1); - push_r(code, opts->gen.context_reg); -#endif - call(code, bp_handler); -#ifdef X86_32 - add_ir(code, 8, RSP, SZ_D); -#endif - mov_rr(code, RAX, opts->gen.context_reg, SZ_PTR); - //Restore context - call(code, opts->gen.load_context); - pop_r(code, opts->gen.scratch1); - //do prologue stuff - cmp_rr(code, opts->gen.cycles, opts->gen.limit, SZ_D); - code_ptr jmp_off = code->cur + 1; - jcc(code, CC_NC, code->cur + 7); - call(code, opts->gen.handle_cycle_limit_int); - *jmp_off = code->cur - (jmp_off+1); - //jump back to body of translated instruction - pop_r(code, opts->gen.scratch1); - add_ir(code, check_int_size - (native.cur-start_native), opts->gen.scratch1, SZ_PTR); - jmp_r(code, opts->gen.scratch1); - } else { - call(&native, bp_stub); - } -} - -void remove_breakpoint(m68k_context * context, uint32_t address) -{ - code_ptr native = get_native_address(context->native_code_map, address); - check_cycles_int(context->options, address); -} - -void start_68k_context(m68k_context * context, uint32_t address) -{ - code_ptr addr = get_native_address_trans(context, address); - m68k_options * options = context->options; - options->start_context(addr, context); -} - -void m68k_reset(m68k_context * context) -{ - //TODO: Make this actually use the normal read functions - context->aregs[7] = context->mem_pointers[0][0] << 16 | context->mem_pointers[0][1]; - uint32_t address = context->mem_pointers[0][2] << 16 | context->mem_pointers[0][3]; - start_68k_context(context, address); -} - -code_ptr gen_mem_fun(cpu_options * opts, memmap_chunk * memmap, uint32_t num_chunks, ftype fun_type) -{ - code_info *code = &opts->code; - code_ptr start = code->cur; - check_cycles(opts); - cycles(opts, BUS); - and_ir(code, 0xFFFFFF, opts->scratch1, SZ_D); - code_ptr lb_jcc = NULL, ub_jcc = NULL; - uint8_t is_write = fun_type == WRITE_16 || fun_type == WRITE_8; - uint8_t adr_reg = is_write ? opts->scratch2 : opts->scratch1; - uint16_t access_flag = is_write ? MMAP_WRITE : MMAP_READ; - uint8_t size = (fun_type == READ_16 || fun_type == WRITE_16) ? SZ_W : SZ_B; - for (uint32_t chunk = 0; chunk < num_chunks; chunk++) - { - if (memmap[chunk].start > 0) { - cmp_ir(code, memmap[chunk].start, adr_reg, SZ_D); - lb_jcc = code->cur + 1; - jcc(code, CC_C, code->cur + 2); - } - if (memmap[chunk].end < 0x1000000) { - cmp_ir(code, memmap[chunk].end, adr_reg, SZ_D); - ub_jcc = code->cur + 1; - jcc(code, CC_NC, code->cur + 2); - } - - if (memmap[chunk].mask != 0xFFFFFF) { - and_ir(code, memmap[chunk].mask, adr_reg, SZ_D); - } - void * cfun; - switch (fun_type) - { - case READ_16: - cfun = memmap[chunk].read_16; - break; - case READ_8: - cfun = memmap[chunk].read_8; - break; - case WRITE_16: - cfun = memmap[chunk].write_16; - break; - case WRITE_8: - cfun = memmap[chunk].write_8; - break; - default: - cfun = NULL; - } - if(memmap[chunk].buffer && memmap[chunk].flags & access_flag) { - if (memmap[chunk].flags & MMAP_PTR_IDX) { - if (memmap[chunk].flags & MMAP_FUNC_NULL) { - cmp_irdisp(code, 0, opts->context_reg, offsetof(m68k_context, mem_pointers) + sizeof(void*) * memmap[chunk].ptr_index, SZ_PTR); - code_ptr not_null = code->cur + 1; - jcc(code, CC_NZ, code->cur + 2); - call(code, opts->save_context); -#ifdef X86_64 - if (is_write) { - if (opts->scratch2 != RDI) { - mov_rr(code, opts->scratch2, RDI, SZ_D); - } - mov_rr(code, opts->scratch1, RDX, size); - } else { - push_r(code, opts->context_reg); - mov_rr(code, opts->scratch1, RDI, SZ_D); - } - test_ir(code, 8, RSP, SZ_D); - code_ptr adjust_rsp = code->cur + 1; - jcc(code, CC_NZ, code->cur + 2); - call(code, cfun); - code_ptr no_adjust = code->cur + 1; - jmp(code, code->cur + 2); - *adjust_rsp = code->cur - (adjust_rsp + 1); - sub_ir(code, 8, RSP, SZ_PTR); - call(code, cfun); - add_ir(code, 8, RSP, SZ_PTR); - *no_adjust = code->cur - (no_adjust + 1); -#else - if (is_write) { - push_r(code, opts->scratch1); - } else { - push_r(code, opts->context_reg);//save opts->context_reg for later - } - push_r(code, opts->context_reg); - push_r(code, is_write ? opts->scratch2 : opts->scratch1); - call(code, cfun); - add_ir(code, is_write ? 12 : 8, RSP, SZ_D); -#endif - if (is_write) { - mov_rr(code, RAX, opts->context_reg, SZ_PTR); - } else { - pop_r(code, opts->context_reg); - mov_rr(code, RAX, opts->scratch1, size); - } - jmp(code, opts->load_context); - - *not_null = code->cur - (not_null + 1); - } - if (size == SZ_B) { - xor_ir(code, 1, adr_reg, SZ_D); - } - add_rdispr(code, opts->context_reg, offsetof(m68k_context, mem_pointers) + sizeof(void*) * memmap[chunk].ptr_index, adr_reg, SZ_PTR); - if (is_write) { - mov_rrind(code, opts->scratch1, opts->scratch2, size); - - } else { - mov_rindr(code, opts->scratch1, opts->scratch1, size); - } - } else { - uint8_t tmp_size = size; - if (size == SZ_B) { - if ((memmap[chunk].flags & MMAP_ONLY_ODD) || (memmap[chunk].flags & MMAP_ONLY_EVEN)) { - bt_ir(code, 0, adr_reg, SZ_D); - code_ptr good_addr = code->cur + 1; - jcc(code, (memmap[chunk].flags & MMAP_ONLY_ODD) ? CC_C : CC_NC, code->cur + 2); - if (!is_write) { - mov_ir(code, 0xFF, opts->scratch1, SZ_B); - } - retn(code); - *good_addr = code->cur - (good_addr + 1); - shr_ir(code, 1, adr_reg, SZ_D); - } else { - xor_ir(code, 1, adr_reg, SZ_D); - } - } else if ((memmap[chunk].flags & MMAP_ONLY_ODD) || (memmap[chunk].flags & MMAP_ONLY_EVEN)) { - tmp_size = SZ_B; - shr_ir(code, 1, adr_reg, SZ_D); - if ((memmap[chunk].flags & MMAP_ONLY_EVEN) && is_write) { - shr_ir(code, 8, opts->scratch1, SZ_W); - } - } - if ((intptr_t)memmap[chunk].buffer <= 0x7FFFFFFF && (intptr_t)memmap[chunk].buffer >= -2147483648) { - if (is_write) { - mov_rrdisp(code, opts->scratch1, opts->scratch2, (intptr_t)memmap[chunk].buffer, tmp_size); - } else { - mov_rdispr(code, opts->scratch1, (intptr_t)memmap[chunk].buffer, opts->scratch1, tmp_size); - } - } else { - if (is_write) { - push_r(code, opts->scratch1); - mov_ir(code, (intptr_t)memmap[chunk].buffer, opts->scratch1, SZ_PTR); - add_rr(code, opts->scratch1, opts->scratch2, SZ_PTR); - pop_r(code, opts->scratch1); - mov_rrind(code, opts->scratch1, opts->scratch2, tmp_size); - } else { - mov_ir(code, (intptr_t)memmap[chunk].buffer, opts->scratch2, SZ_PTR); - mov_rindexr(code, opts->scratch2, opts->scratch1, 1, opts->scratch1, tmp_size); - } - } - if (size != tmp_size && !is_write) { - if (memmap[chunk].flags & MMAP_ONLY_EVEN) { - shl_ir(code, 8, opts->scratch1, SZ_W); - mov_ir(code, 0xFF, opts->scratch1, SZ_B); - } else { - or_ir(code, 0xFF00, opts->scratch1, SZ_W); - } - } - } - if (is_write && (memmap[chunk].flags & MMAP_CODE)) { - mov_rr(code, opts->scratch2, opts->scratch1, SZ_D); - shr_ir(code, 11, opts->scratch1, SZ_D); - bt_rrdisp(code, opts->scratch1, opts->context_reg, offsetof(m68k_context, ram_code_flags), SZ_D); - code_ptr not_code = code->cur + 1; - jcc(code, CC_NC, code->cur + 2); - call(code, opts->save_context); -#ifdef X86_32 - push_r(code, opts->context_reg); - push_r(code, opts->scratch2); -#endif - call(code, (code_ptr)m68k_handle_code_write); -#ifdef X86_32 - add_ir(code, 8, RSP, SZ_D); -#endif - mov_rr(code, RAX, opts->context_reg, SZ_PTR); - call(code, opts->load_context); - *not_code = code->cur - (not_code+1); - } - retn(code); - } else if (cfun) { - call(code, opts->save_context); -#ifdef X86_64 - if (is_write) { - if (opts->scratch2 != RDI) { - mov_rr(code, opts->scratch2, RDI, SZ_D); - } - mov_rr(code, opts->scratch1, RDX, size); - } else { - push_r(code, opts->context_reg); - mov_rr(code, opts->scratch1, RDI, SZ_D); - } - test_ir(code, 8, RSP, SZ_D); - code_ptr adjust_rsp = code->cur + 1; - jcc(code, CC_NZ, code->cur + 2); - call(code, cfun); - code_ptr no_adjust = code->cur + 1; - jmp(code, code->cur + 2); - *adjust_rsp = code->cur - (adjust_rsp + 1); - sub_ir(code, 8, RSP, SZ_PTR); - call(code, cfun); - add_ir(code, 8, RSP, SZ_PTR); - *no_adjust = code->cur - (no_adjust+1); -#else - if (is_write) { - push_r(code, opts->scratch1); - } else { - push_r(code, opts->context_reg);//save opts->context_reg for later - } - push_r(code, opts->context_reg); - push_r(code, is_write ? opts->scratch2 : opts->scratch1); - call(code, cfun); - add_ir(code, is_write ? 12 : 8, RSP, SZ_D); -#endif - if (is_write) { - mov_rr(code, RAX, opts->context_reg, SZ_PTR); - } else { - pop_r(code, opts->context_reg); - mov_rr(code, RAX, opts->scratch1, size); - } - jmp(code, opts->load_context); - } else { - //Not sure the best course of action here - if (!is_write) { - mov_ir(code, size == SZ_B ? 0xFF : 0xFFFF, opts->scratch1, size); - } - retn(code); - } - if (lb_jcc) { - *lb_jcc = code->cur - (lb_jcc+1); - lb_jcc = NULL; - } - if (ub_jcc) { - *ub_jcc = code->cur - (ub_jcc+1); - ub_jcc = NULL; - } - } - if (!is_write) { - mov_ir(code, size == SZ_B ? 0xFF : 0xFFFF, opts->scratch1, size); - } - retn(code); - return start; -} - -void init_x86_68k_opts(m68k_options * opts, memmap_chunk * memmap, uint32_t num_chunks) -{ - memset(opts, 0, sizeof(*opts)); - for (int i = 0; i < 8; i++) - { - opts->dregs[i] = opts->aregs[i] = -1; - } -#ifdef X86_64 - opts->dregs[0] = R10; - opts->dregs[1] = R11; - opts->dregs[2] = R12; - opts->dregs[3] = R8; - opts->aregs[0] = R13; - opts->aregs[1] = R14; - opts->aregs[2] = R9; - opts->aregs[7] = R15; - - opts->flag_regs[0] = -1; - opts->flag_regs[1] = RBX; - opts->flag_regs[2] = RDX; - opts->flag_regs[3] = BH; - opts->flag_regs[4] = DH; - - opts->gen.scratch2 = RDI; -#else - opts->dregs[0] = RDX; - opts->aregs[7] = RDI; - - for (int i = 0; i < 5; i++) - { - opts->flag_regs[i] = -1; - } - opts->gen.scratch2 = RBX; -#endif - opts->gen.context_reg = RSI; - opts->gen.cycles = RAX; - opts->gen.limit = RBP; - opts->gen.scratch1 = RCX; - - - opts->gen.native_code_map = malloc(sizeof(native_map_slot) * NATIVE_MAP_CHUNKS); - memset(opts->gen.native_code_map, 0, sizeof(native_map_slot) * NATIVE_MAP_CHUNKS); - opts->gen.deferred = NULL; - opts->gen.ram_inst_sizes = malloc(sizeof(uint8_t *) * 64); - memset(opts->gen.ram_inst_sizes, 0, sizeof(uint8_t *) * 64); - - code_info *code = &opts->gen.code; - init_code_info(code); - - opts->gen.save_context = code->cur; - for (int i = 0; i < 5; i++) - if (opts->flag_regs[i] >= 0) { - mov_rrdisp(code, opts->flag_regs[i], opts->gen.context_reg, offsetof(m68k_context, flags) + i, SZ_B); - } - for (int i = 0; i < 8; i++) - { - if (opts->dregs[i] >= 0) { - mov_rrdisp(code, opts->dregs[i], opts->gen.context_reg, offsetof(m68k_context, dregs) + sizeof(uint32_t) * i, SZ_D); - } - if (opts->aregs[i] >= 0) { - mov_rrdisp(code, opts->aregs[i], opts->gen.context_reg, offsetof(m68k_context, aregs) + sizeof(uint32_t) * i, SZ_D); - } - } - mov_rrdisp(code, opts->gen.cycles, opts->gen.context_reg, offsetof(m68k_context, current_cycle), SZ_D); - retn(code); - - opts->gen.load_context = code->cur; - for (int i = 0; i < 5; i++) - if (opts->flag_regs[i] >= 0) { - mov_rdispr(code, opts->gen.context_reg, offsetof(m68k_context, flags) + i, opts->flag_regs[i], SZ_B); - } - for (int i = 0; i < 8; i++) - { - if (opts->dregs[i] >= 0) { - mov_rdispr(code, opts->gen.context_reg, offsetof(m68k_context, dregs) + sizeof(uint32_t) * i, opts->dregs[i], SZ_D); - } - if (opts->aregs[i] >= 0) { - mov_rdispr(code, opts->gen.context_reg, offsetof(m68k_context, aregs) + sizeof(uint32_t) * i, opts->aregs[i], SZ_D); - } - } - mov_rdispr(code, opts->gen.context_reg, offsetof(m68k_context, current_cycle), CYCLES, SZ_D); - mov_rdispr(code, opts->gen.context_reg, offsetof(m68k_context, target_cycle), LIMIT, SZ_D); - retn(code); - - opts->start_context = (start_fun)code->cur; -#ifdef X86_64 - if (opts->gen.scratch2 != RDI) { - mov_rr(code, RDI, opts->gen.scratch2, SZ_PTR); - } - //save callee save registers - push_r(code, RBP); - push_r(code, R12); - push_r(code, R13); - push_r(code, R14); - push_r(code, R15); -#else - //save callee save registers - push_r(code, RBP); - push_r(code, RBX); - push_r(code, RSI); - push_r(code, RDI); - - mov_rdispr(code, RSP, 20, opts->gen.scratch2, SZ_D); - mov_rdispr(code, RSP, 24, opts->gen.context_reg, SZ_D); -#endif - call(code, opts->gen.load_context); - call_r(code, opts->gen.scratch2); - call(code, opts->gen.save_context); -#ifdef X86_64 - //restore callee save registers - pop_r(code, R15); - pop_r(code, R14); - pop_r(code, R13); - pop_r(code, R12); - pop_r(code, RBP); -#else - pop_r(code, RDI); - pop_r(code, RSI); - pop_r(code, RBX); - pop_r(code, RBP); -#endif - retn(code); - - opts->native_addr = code->cur; - call(code, opts->gen.save_context); - push_r(code, opts->gen.context_reg); -#ifdef X86_64 - mov_rr(code, opts->gen.context_reg, RDI, SZ_PTR); //move context to 1st arg reg - mov_rr(code, opts->gen.scratch1, RSI, SZ_D); //move address to 2nd arg reg -#else - push_r(code, opts->gen.scratch1); - push_r(code, opts->gen.context_reg); -#endif - call(code, (code_ptr)get_native_address_trans); -#ifdef X86_32 - add_ir(code, 8, RSP, SZ_D); -#endif - mov_rr(code, RAX, opts->gen.scratch1, SZ_PTR); //move result to scratch reg - pop_r(code, opts->gen.context_reg); - call(code, opts->gen.load_context); - retn(code); - - opts->native_addr_and_sync = code->cur; - call(code, opts->gen.save_context); - push_r(code, opts->gen.scratch1); -#ifdef X86_64 - mov_rr(code, opts->gen.context_reg, RDI, SZ_PTR); - xor_rr(code, RSI, RSI, SZ_D); - test_ir(code, 8, RSP, SZ_PTR); //check stack alignment - code_ptr do_adjust_rsp = code->cur + 1; - jcc(code, CC_NZ, code->cur + 2); - call(code, (code_ptr)sync_components); - code_ptr no_adjust_rsp = code->cur + 1; - jmp(code, code->cur + 2); - *do_adjust_rsp = code->cur - (do_adjust_rsp+1); - sub_ir(code, 8, RSP, SZ_PTR); - call(code, (code_ptr)sync_components); - add_ir(code, 8, RSP, SZ_PTR); - *no_adjust_rsp = code->cur - (no_adjust_rsp+1); - pop_r(code, RSI); - push_r(code, RAX); - mov_rr(code, RAX, RDI, SZ_PTR); - call(code, (code_ptr)get_native_address_trans); -#else - //TODO: Add support for pushing a constant in gen_x86 - xor_rr(code, RAX, RAX, SZ_D); - push_r(code, RAX); - push_r(code, opts->gen.context_reg); - call(code, (code_ptr)sync_components); - add_ir(code, 8, RSP, SZ_D); - pop_r(code, RSI); //restore saved address from opts->gen.scratch1 - push_r(code, RAX); //save context pointer for later - push_r(code, RSI); //2nd arg -- address - push_r(code, RAX); //1st arg -- context pointer - call(code, (code_ptr)get_native_address_trans); - add_ir(code, 8, RSP, SZ_D); -#endif - - mov_rr(code, RAX, opts->gen.scratch1, SZ_PTR); //move result to scratch reg - pop_r(code, opts->gen.context_reg); - call(code, opts->gen.load_context); - retn(code); - - opts->gen.handle_cycle_limit = code->cur; - cmp_rdispr(code, opts->gen.context_reg, offsetof(m68k_context, sync_cycle), CYCLES, SZ_D); - code_ptr skip_sync = code->cur + 1; - jcc(code, CC_C, code->cur + 2); - opts->do_sync = code->cur; - push_r(code, opts->gen.scratch1); - push_r(code, opts->gen.scratch2); - call(code, opts->gen.save_context); -#ifdef X86_64 - mov_rr(code, opts->gen.context_reg, RDI, SZ_PTR); - xor_rr(code, RSI, RSI, SZ_D); - test_ir(code, 8, RSP, SZ_D); - code_ptr adjust_rsp = code->cur + 1; - jcc(code, CC_NZ, code->cur + 2); - call(code, (code_ptr)sync_components); - code_ptr no_adjust = code->cur + 1; - jmp(code, code->cur + 2); - *adjust_rsp = code->cur - (adjust_rsp + 1); - sub_ir(code, 8, RSP, SZ_PTR); - call(code, (code_ptr)sync_components); - add_ir(code, 8, RSP, SZ_PTR); - *no_adjust = code->cur - (no_adjust+1); -#else - //TODO: Add support for pushing a constant in gen_x86 - xor_rr(code, RAX, RAX, SZ_D); - push_r(code, RAX); - push_r(code, opts->gen.context_reg); - call(code, (code_ptr)sync_components); - add_ir(code, 8, RSP, SZ_D); -#endif - mov_rr(code, RAX, opts->gen.context_reg, SZ_PTR); - call(code, opts->gen.load_context); - pop_r(code, opts->gen.scratch2); - pop_r(code, opts->gen.scratch1); - *skip_sync = code->cur - (skip_sync+1); - retn(code); - - opts->read_16 = gen_mem_fun(&opts->gen, memmap, num_chunks, READ_16); - opts->read_8 = gen_mem_fun(&opts->gen, memmap, num_chunks, READ_8); - opts->write_16 = gen_mem_fun(&opts->gen, memmap, num_chunks, WRITE_16); - opts->write_8 = gen_mem_fun(&opts->gen, memmap, num_chunks, WRITE_8); - - opts->read_32 = code->cur; - push_r(code, opts->gen.scratch1); - call(code, opts->read_16); - mov_rr(code, opts->gen.scratch1, opts->gen.scratch2, SZ_W); - pop_r(code, opts->gen.scratch1); - push_r(code, opts->gen.scratch2); - add_ir(code, 2, opts->gen.scratch1, SZ_D); - call(code, opts->read_16); - pop_r(code, opts->gen.scratch2); - movzx_rr(code, opts->gen.scratch1, opts->gen.scratch1, SZ_W, SZ_D); - shl_ir(code, 16, opts->gen.scratch2, SZ_D); - or_rr(code, opts->gen.scratch2, opts->gen.scratch1, SZ_D); - retn(code); - - opts->write_32_lowfirst = code->cur; - push_r(code, opts->gen.scratch2); - push_r(code, opts->gen.scratch1); - add_ir(code, 2, opts->gen.scratch2, SZ_D); - call(code, opts->write_16); - pop_r(code, opts->gen.scratch1); - pop_r(code, opts->gen.scratch2); - shr_ir(code, 16, opts->gen.scratch1, SZ_D); - jmp(code, opts->write_16); - - opts->write_32_highfirst = code->cur; - push_r(code, opts->gen.scratch1); - push_r(code, opts->gen.scratch2); - shr_ir(code, 16, opts->gen.scratch1, SZ_D); - call(code, opts->write_16); - pop_r(code, opts->gen.scratch2); - pop_r(code, opts->gen.scratch1); - add_ir(code, 2, opts->gen.scratch2, SZ_D); - jmp(code, opts->write_16); - - opts->get_sr = code->cur; - mov_rdispr(code, opts->gen.context_reg, offsetof(m68k_context, status), opts->gen.scratch1, SZ_B); - shl_ir(code, 8, opts->gen.scratch1, SZ_W); - if (opts->flag_regs[FLAG_X] >= 0) { - mov_rr(code, opts->flag_regs[FLAG_X], opts->gen.scratch1, SZ_B); - } else { - int8_t offset = offsetof(m68k_context, flags); - if (offset) { - mov_rdispr(code, opts->gen.context_reg, offset, opts->gen.scratch1, SZ_B); - } else { - mov_rindr(code, opts->gen.context_reg, opts->gen.scratch1, SZ_B); - } - } - for (int flag = FLAG_N; flag <= FLAG_C; flag++) - { - shl_ir(code, 1, opts->gen.scratch1, SZ_B); - if (opts->flag_regs[flag] >= 0) { - or_rr(code, opts->flag_regs[flag], opts->gen.scratch1, SZ_B); - } else { - or_rdispr(code, opts->gen.context_reg, offsetof(m68k_context, flags) + flag, opts->gen.scratch1, SZ_B); - } - } - retn(code); - - opts->set_sr = code->cur; - for (int flag = FLAG_C; flag >= FLAG_X; flag--) - { - rcr_ir(code, 1, opts->gen.scratch1, SZ_B); - if (opts->flag_regs[flag] >= 0) { - setcc_r(code, CC_C, opts->flag_regs[flag]); - } else { - int8_t offset = offsetof(m68k_context, flags) + flag; - if (offset) { - setcc_rdisp(code, CC_C, opts->gen.context_reg, offset); - } else { - setcc_rind(code, CC_C, opts->gen.context_reg); - } - } - } - shr_ir(code, 8, opts->gen.scratch1, SZ_W); - mov_rrdisp(code, opts->gen.scratch1, opts->gen.context_reg, offsetof(m68k_context, status), SZ_B); - retn(code); - - opts->set_ccr = code->cur; - for (int flag = FLAG_C; flag >= FLAG_X; flag--) - { - rcr_ir(code, 1, opts->gen.scratch1, SZ_B); - if (opts->flag_regs[flag] >= 0) { - setcc_r(code, CC_C, opts->flag_regs[flag]); - } else { - int8_t offset = offsetof(m68k_context, flags) + flag; - if (offset) { - setcc_rdisp(code, CC_C, opts->gen.context_reg, offset); - } else { - setcc_rind(code, CC_C, opts->gen.context_reg); - } - } - } - retn(code); - - opts->gen.handle_cycle_limit_int = code->cur; - cmp_rdispr(code, opts->gen.context_reg, offsetof(m68k_context, int_cycle), CYCLES, SZ_D); - code_ptr do_int = code->cur + 1; - jcc(code, CC_NC, code->cur + 2); - cmp_rdispr(code, opts->gen.context_reg, offsetof(m68k_context, sync_cycle), CYCLES, SZ_D); - skip_sync = code->cur + 1; - jcc(code, CC_C, code->cur + 2); - call(code, opts->gen.save_context); -#ifdef X86_64 - mov_rr(code, opts->gen.context_reg, RDI, SZ_PTR); - mov_rr(code, opts->gen.scratch1, RSI, SZ_D); - test_ir(code, 8, RSP, SZ_D); - adjust_rsp = code->cur + 1; - jcc(code, CC_NZ, code->cur + 2); - call(code, (code_ptr)sync_components); - no_adjust = code->cur + 1; - jmp(code, code->cur + 2); - *adjust_rsp = code->cur - (adjust_rsp + 1); - sub_ir(code, 8, RSP, SZ_PTR); - call(code, (code_ptr)sync_components); - add_ir(code, 8, RSP, SZ_PTR); - *no_adjust = code->cur - (no_adjust+1); -#else - push_r(code, opts->gen.scratch1); - push_r(code, opts->gen.context_reg); - call(code, (code_ptr)sync_components); - add_ir(code, 8, RSP, SZ_D); -#endif - mov_rr(code, RAX, opts->gen.context_reg, SZ_PTR); - jmp(code, opts->gen.load_context); - *skip_sync = code->cur - (skip_sync+1); - retn(code); - *do_int = code->cur - (do_int+1); - //set target cycle to sync cycle - mov_rdispr(code, opts->gen.context_reg, offsetof(m68k_context, sync_cycle), LIMIT, SZ_D); - //swap USP and SSP if not already in supervisor mode - bt_irdisp(code, 5, opts->gen.context_reg, offsetof(m68k_context, status), SZ_B); - code_ptr already_supervisor = code->cur + 1; - jcc(code, CC_C, code->cur + 2); - mov_rdispr(code, opts->gen.context_reg, offsetof(m68k_context, aregs) + sizeof(uint32_t) * 8, opts->gen.scratch2, SZ_D); - mov_rrdisp(code, opts->aregs[7], opts->gen.context_reg, offsetof(m68k_context, aregs) + sizeof(uint32_t) * 8, SZ_D); - mov_rr(code, opts->gen.scratch2, opts->aregs[7], SZ_D); - *already_supervisor = code->cur - (already_supervisor+1); - //save PC - sub_ir(code, 4, opts->aregs[7], SZ_D); - mov_rr(code, opts->aregs[7], opts->gen.scratch2, SZ_D); - call(code, opts->write_32_lowfirst); - //save status register - sub_ir(code, 2, opts->aregs[7], SZ_D); - call(code, opts->get_sr); - mov_rr(code, opts->aregs[7], opts->gen.scratch2, SZ_D); - call(code, opts->write_16); - //update status register - and_irdisp(code, 0xF8, opts->gen.context_reg, offsetof(m68k_context, status), SZ_B); - mov_rdispr(code, opts->gen.context_reg, offsetof(m68k_context, int_num), opts->gen.scratch1, SZ_B); - or_ir(code, 0x20, opts->gen.scratch1, SZ_B); - or_rrdisp(code, opts->gen.scratch1, opts->gen.context_reg, offsetof(m68k_context, status), SZ_B); - //calculate interrupt vector address - mov_rdispr(code, opts->gen.context_reg, offsetof(m68k_context, int_num), opts->gen.scratch1, SZ_D); - mov_rrdisp(code, opts->gen.scratch1, opts->gen.context_reg, offsetof(m68k_context, int_ack), SZ_W); - shl_ir(code, 2, opts->gen.scratch1, SZ_D); - add_ir(code, 0x60, opts->gen.scratch1, SZ_D); - call(code, opts->read_32); - call(code, opts->native_addr_and_sync); - cycles(&opts->gen, 24); - //discard function return address - pop_r(code, opts->gen.scratch2); - jmp_r(code, opts->gen.scratch1); - - opts->trap = code->cur; - push_r(code, opts->gen.scratch2); - //swap USP and SSP if not already in supervisor mode - bt_irdisp(code, 5, opts->gen.context_reg, offsetof(m68k_context, status), SZ_B); - already_supervisor = code->cur + 1; - jcc(code, CC_C, code->cur + 2); - mov_rdispr(code, opts->gen.context_reg, offsetof(m68k_context, aregs) + sizeof(uint32_t) * 8, opts->gen.scratch2, SZ_D); - mov_rrdisp(code, opts->aregs[7], opts->gen.context_reg, offsetof(m68k_context, aregs) + sizeof(uint32_t) * 8, SZ_D); - mov_rr(code, opts->gen.scratch2, opts->aregs[7], SZ_D); - *already_supervisor = code->cur - (already_supervisor+1); - //save PC - sub_ir(code, 4, opts->aregs[7], SZ_D); - mov_rr(code, opts->aregs[7], opts->gen.scratch2, SZ_D); - call(code, opts->write_32_lowfirst); - //save status register - sub_ir(code, 2, opts->aregs[7], SZ_D); - call(code, opts->get_sr); - mov_rr(code, opts->aregs[7], opts->gen.scratch2, SZ_D); - call(code, opts->write_16); - //set supervisor bit - or_irdisp(code, 0x20, opts->gen.context_reg, offsetof(m68k_context, status), SZ_B); - //calculate vector address - pop_r(code, opts->gen.scratch1); - shl_ir(code, 2, opts->gen.scratch1, SZ_D); - call(code, opts->read_32); - call(code, opts->native_addr_and_sync); - cycles(&opts->gen, 18); - jmp_r(code, opts->gen.scratch1); -} - -void init_68k_context(m68k_context * context, native_map_slot * native_code_map, void * opts) -{ - memset(context, 0, sizeof(m68k_context)); - context->native_code_map = native_code_map; - context->options = opts; - context->int_cycle = 0xFFFFFFFF; - context->status = 0x27; -} - |