mirror of
https://github.com/libretro/scummvm.git
synced 2024-12-22 01:39:57 +00:00
3de0b4e3f2
DREAMWEB: Fixes, cleaning and more functions ported to C++
589 lines
12 KiB
C++
589 lines
12 KiB
C++
/* ScummVM - Graphic Adventure Engine
|
|
*
|
|
* ScummVM is the legal property of its developers, whose names
|
|
* are too numerous to list here. Please refer to the COPYRIGHT
|
|
* file distributed with this source distribution.
|
|
*
|
|
* This program is free software; you can redistribute it and/or
|
|
* modify it under the terms of the GNU General Public License
|
|
* as published by the Free Software Foundation; either version 2
|
|
* of the License, or (at your option) any later version.
|
|
|
|
* This program is distributed in the hope that it will be useful,
|
|
* but WITHOUT ANY WARRANTY; without even the implied warranty of
|
|
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
|
* GNU General Public License for more details.
|
|
|
|
* You should have received a copy of the GNU General Public License
|
|
* along with this program; if not, write to the Free Software
|
|
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA.
|
|
*
|
|
*/
|
|
|
|
#ifndef DREAMGEN_RUNTIME_H__
|
|
#define DREAMGEN_RUNTIME_H__
|
|
|
|
#include <assert.h>
|
|
#include "common/scummsys.h"
|
|
#include "common/array.h"
|
|
#include "common/debug.h"
|
|
#include "common/hashmap.h"
|
|
#include "common/list.h"
|
|
#include "common/ptr.h"
|
|
|
|
namespace DreamWeb {
|
|
class DreamWebEngine;
|
|
}
|
|
|
|
namespace DreamGen {
|
|
|
|
//fixme: name clash
|
|
#undef random
|
|
|
|
struct Register {
|
|
union {
|
|
uint16 _value;
|
|
uint8 _part[2];
|
|
};
|
|
inline Register(): _value() {}
|
|
inline Register& operator=(uint16 v) { _value = v; return *this; }
|
|
inline operator uint16&() { return _value; }
|
|
inline void cbw() {
|
|
if (_value & 0x80)
|
|
_value |= 0xff00;
|
|
else
|
|
_value &= 0x7f;
|
|
}
|
|
};
|
|
|
|
template<int kIndex> //from low to high
|
|
struct RegisterPart {
|
|
uint8 &_value;
|
|
|
|
explicit inline RegisterPart(Register ®) : _value(reg._part[kIndex]) {}
|
|
|
|
inline operator uint8&() {
|
|
return _value;
|
|
}
|
|
|
|
inline RegisterPart& operator=(const RegisterPart& o) {
|
|
_value = o._value;
|
|
return *this;
|
|
}
|
|
|
|
inline RegisterPart& operator=(uint8 v) {
|
|
_value = v;
|
|
return *this;
|
|
}
|
|
};
|
|
|
|
#ifdef SCUMM_LITTLE_ENDIAN
|
|
typedef RegisterPart<0> LowPartOfRegister;
|
|
typedef RegisterPart<1> HighPartOfRegister;
|
|
#else
|
|
typedef RegisterPart<1> LowPartOfRegister;
|
|
typedef RegisterPart<0> HighPartOfRegister;
|
|
#endif
|
|
|
|
class WordRef {
|
|
uint8 *_data;
|
|
unsigned _index;
|
|
uint16 _value;
|
|
|
|
public:
|
|
inline WordRef(Common::Array<uint8> &data, unsigned index) : _data(data.begin() + index), _index(index) {
|
|
assert(index + 1 < data.size());
|
|
_value = _data[0] | (_data[1] << 8);
|
|
}
|
|
|
|
inline WordRef& operator=(const WordRef &ref) {
|
|
_value = ref._value;
|
|
return *this;
|
|
}
|
|
|
|
inline WordRef& operator=(uint16 v) {
|
|
_value = v;
|
|
return *this;
|
|
}
|
|
|
|
inline operator uint16&() {
|
|
return _value;
|
|
}
|
|
|
|
inline ~WordRef() {
|
|
_data[0] = _value & 0xff;
|
|
_data[1] = _value >> 8;
|
|
_value = _data[0] | (_data[1] << 8);
|
|
}
|
|
};
|
|
|
|
struct Segment {
|
|
Common::Array<uint8> data;
|
|
|
|
inline void assign(const uint8 *b, const uint8 *e) {
|
|
data.assign(b, e);
|
|
}
|
|
|
|
inline uint8 &byte(unsigned index) {
|
|
assert(index < data.size());
|
|
return data[index];
|
|
}
|
|
|
|
inline WordRef word(unsigned index) {
|
|
return WordRef(data, index);
|
|
}
|
|
|
|
inline uint8 *ptr(unsigned index, unsigned size) {
|
|
assert(index + size <= data.size());
|
|
return data.begin() + index;
|
|
}
|
|
};
|
|
|
|
typedef Common::SharedPtr<Segment> SegmentPtr;
|
|
|
|
class Context;
|
|
|
|
class SegmentRef {
|
|
Context *_context;
|
|
uint16 _value;
|
|
SegmentPtr _segment;
|
|
|
|
public:
|
|
SegmentRef(Context *ctx, uint16 value = 0, SegmentPtr segment = SegmentPtr()): _context(ctx), _value(value), _segment(segment) {
|
|
}
|
|
|
|
inline void reset(uint16 value);
|
|
|
|
inline SegmentRef& operator=(const uint16 id) {
|
|
reset(id);
|
|
return *this;
|
|
}
|
|
|
|
inline SegmentRef& operator=(const SegmentRef &ref) {
|
|
_context = ref._context;
|
|
_value = ref._value;
|
|
_segment = ref._segment;
|
|
return *this;
|
|
}
|
|
|
|
inline uint8 &byte(unsigned index) {
|
|
assert(_segment != 0);
|
|
return _segment->byte(index);
|
|
}
|
|
|
|
inline operator uint16() const {
|
|
return _value;
|
|
}
|
|
|
|
inline WordRef word(unsigned index) {
|
|
//debug(1, "getting word ref for %04x:%d", _value, index);
|
|
assert(_segment != 0);
|
|
return _segment->word(index);
|
|
}
|
|
|
|
inline void assign(const uint8 *b, const uint8 *e) {
|
|
assert(_segment != 0);
|
|
_segment->assign(b, e);
|
|
}
|
|
|
|
inline uint8 *ptr(unsigned index, unsigned size) {
|
|
assert(_segment != 0);
|
|
return _segment->ptr(index, size);
|
|
}
|
|
};
|
|
|
|
struct Flags {
|
|
bool _z, _c, _s, _o;
|
|
inline Flags(): _z(true), _c(false), _s(false), _o(false) {}
|
|
|
|
inline bool z() const { return _z; }
|
|
inline bool c() const { return _c; }
|
|
inline bool s() const { return _s; }
|
|
|
|
inline bool l() const { return _o != _s; }
|
|
inline bool le() const { return _o != _s|| _z; }
|
|
|
|
inline void update_zs(uint8 v) {
|
|
_s = v & 0x80;
|
|
_z = v == 0;
|
|
}
|
|
|
|
inline void update_zs(uint16 v) {
|
|
_s = v & 0x8000;
|
|
_z = v == 0;
|
|
}
|
|
|
|
inline void update_o(uint8 v, uint8 a, uint8 b) {
|
|
uint8 s1 = a & 0x80, s2 = b & 0x80;
|
|
_o = (s1 == s2) && (v & 0x80) != s1;
|
|
}
|
|
|
|
inline void update_o(uint16 v, uint16 a, uint16 b) {
|
|
uint16 s1 = a & 0x8000, s2 = b & 0x8000;
|
|
_o = (s1 == s2) && (v & 0x8000) != s1;
|
|
}
|
|
};
|
|
|
|
class Context {
|
|
typedef Common::HashMap<uint16, SegmentPtr> SegmentMap;
|
|
SegmentMap _segments;
|
|
|
|
typedef Common::List<uint16> FreeSegmentList;
|
|
FreeSegmentList _freeSegments;
|
|
|
|
public:
|
|
DreamWeb::DreamWebEngine *engine;
|
|
|
|
enum { kDefaultDataSegment = 0x1000 };
|
|
|
|
Register ax, dx, bx, cx, si, di;
|
|
LowPartOfRegister al;
|
|
HighPartOfRegister ah;
|
|
LowPartOfRegister bl;
|
|
HighPartOfRegister bh;
|
|
LowPartOfRegister cl;
|
|
HighPartOfRegister ch;
|
|
LowPartOfRegister dl;
|
|
HighPartOfRegister dh;
|
|
|
|
SegmentRef cs, ds, es, data;
|
|
//data == fake segment register always pointing to data segment
|
|
Flags flags;
|
|
|
|
inline Context(): engine(0), al(ax), ah(ax), bl(bx), bh(bx), cl(cx), ch(cx), dl(dx), dh(dx),
|
|
cs(this), ds(this), es(this), data(this) {
|
|
_segments[kDefaultDataSegment] = SegmentPtr(new Segment());
|
|
cs.reset(kDefaultDataSegment);
|
|
ds.reset(kDefaultDataSegment);
|
|
es.reset(kDefaultDataSegment);
|
|
data.reset(kDefaultDataSegment);
|
|
}
|
|
|
|
SegmentRef getSegment(uint16 value) {
|
|
SegmentMap::iterator i = _segments.find(value);
|
|
assert(i != _segments.end());
|
|
return SegmentRef(this, value, i->_value);
|
|
}
|
|
|
|
SegmentRef allocateSegment(uint size) {
|
|
unsigned id;
|
|
if (_freeSegments.empty())
|
|
id = kDefaultDataSegment + _segments.size();
|
|
else {
|
|
id = _freeSegments.front();
|
|
_freeSegments.pop_front();
|
|
}
|
|
assert(!_segments.contains(id));
|
|
SegmentPtr seg(new Segment());
|
|
seg->data.resize(size);
|
|
_segments[id] = seg;
|
|
return SegmentRef(this, id, seg);
|
|
}
|
|
|
|
void deallocateSegment(uint16 id) {
|
|
SegmentMap::iterator i = _segments.find(id);
|
|
assert(i != _segments.end());
|
|
_segments.erase(i);
|
|
_freeSegments.push_back(id);
|
|
}
|
|
|
|
SegmentRef segRef(uint16 seg) {
|
|
SegmentRef result(this);
|
|
result = seg;
|
|
return result;
|
|
}
|
|
|
|
inline void _cmp(uint8 a, uint8 b) {
|
|
_sub(a, b);
|
|
}
|
|
|
|
inline void _cmp(uint16 a, uint16 b) {
|
|
_sub(a, b);
|
|
}
|
|
|
|
inline void _test(uint8 a, uint8 b) {
|
|
_and(a, b);
|
|
}
|
|
|
|
inline void _test(uint16 a, uint16 b) {
|
|
_and(a, b);
|
|
}
|
|
|
|
inline void _add(uint8 &dst, uint8 src) {
|
|
unsigned r = (unsigned)dst + src;
|
|
flags.update_o((uint8)r, dst, src);
|
|
flags._c = r >= 0x100;
|
|
dst = r;
|
|
flags.update_zs(dst);
|
|
}
|
|
|
|
inline void _add(uint16 &dst, uint16 src) {
|
|
unsigned r = (unsigned)dst + src;
|
|
flags.update_o((uint16)r, dst, src);
|
|
flags._c = r >= 0x10000;
|
|
dst = r;
|
|
flags.update_zs(dst);
|
|
}
|
|
|
|
inline void _sub(uint8 &dst, uint8 src) {
|
|
flags.update_o(uint8(dst - src), dst, (uint8)-src);
|
|
flags._c = dst < src;
|
|
dst -= src;
|
|
flags.update_zs(dst);
|
|
}
|
|
|
|
inline void _sub(uint16 &dst, uint16 src) {
|
|
flags.update_o(uint16(dst - src), dst, (uint16)-src);
|
|
flags._c = dst < src;
|
|
dst -= src;
|
|
flags.update_zs(dst);
|
|
}
|
|
|
|
inline void _inc(uint8 &dst) {
|
|
flags.update_o((uint8)(dst + 1), dst, 1);
|
|
++dst;
|
|
flags.update_zs(dst);
|
|
}
|
|
|
|
inline void _inc(uint16 &dst) {
|
|
flags.update_o((uint16)(dst + 1), dst, 1);
|
|
++dst;
|
|
flags.update_zs(dst);
|
|
}
|
|
|
|
inline void _dec(uint8 &dst) {
|
|
flags.update_o(uint8(dst - 1), dst, 1);
|
|
--dst;
|
|
flags.update_zs(dst);
|
|
}
|
|
|
|
inline void _dec(uint16 &dst) {
|
|
flags.update_o(uint16(dst - 1), dst, 1);
|
|
--dst;
|
|
flags.update_zs(dst);
|
|
}
|
|
|
|
inline void _and(uint8 &dst, uint8 src) {
|
|
dst &= src;
|
|
flags.update_zs(dst);
|
|
flags._c = flags._o = false;
|
|
}
|
|
|
|
inline void _and(uint16 &dst, uint16 src) {
|
|
dst &= src;
|
|
flags.update_zs(dst);
|
|
flags._c = flags._o = false;
|
|
}
|
|
|
|
inline void _or(uint8 &dst, uint8 src) {
|
|
dst |= src;
|
|
flags.update_zs(dst);
|
|
flags._c = flags._o = false;
|
|
}
|
|
|
|
inline void _or(uint16 &dst, uint16 src) {
|
|
dst |= src;
|
|
flags.update_zs(dst);
|
|
flags._c = flags._o = false;
|
|
}
|
|
|
|
inline void _xor(uint8 &dst, uint8 src) {
|
|
dst ^= src;
|
|
flags.update_zs(dst);
|
|
flags._c = flags._o = false;
|
|
}
|
|
|
|
inline void _xor(uint16 &dst, uint16 src) {
|
|
dst ^= src;
|
|
flags.update_zs(dst);
|
|
flags._c = flags._o = false;
|
|
}
|
|
|
|
inline void _shr(uint8 &dst, uint8 src) {
|
|
src &= 0x1f;
|
|
if (src > 0) {
|
|
dst >>= (src - 1);
|
|
flags._c = dst & 1;
|
|
dst >>= 1;
|
|
flags.update_zs(dst);
|
|
}
|
|
if (src == 1)
|
|
flags._o = dst & 0x80;
|
|
}
|
|
|
|
inline void _shr(uint16 &dst, uint8 src) {
|
|
src &= 0x1f;
|
|
if (src > 0) {
|
|
dst >>= (src - 1);
|
|
flags._c = dst & 1;
|
|
dst >>= 1;
|
|
flags.update_zs(dst);
|
|
}
|
|
if (src == 1)
|
|
flags._o = dst & 0x8000;
|
|
}
|
|
|
|
inline void _shl(uint8 &dst, uint8 src) {
|
|
src &= 0x1f;
|
|
if (src > 0) {
|
|
dst <<= (src - 1);
|
|
flags._c = dst & 0x80;
|
|
dst <<= 1;
|
|
flags.update_zs(dst);
|
|
}
|
|
if (src == 1)
|
|
flags._o = ((dst & 0x80) != 0) == flags._c;
|
|
}
|
|
inline void _shl(uint16 &dst, uint8 src) {
|
|
src &= 0x1f;
|
|
if (src > 0) {
|
|
dst <<= (src - 1);
|
|
flags._c = dst & 0x8000;
|
|
dst <<= 1;
|
|
flags.update_zs(dst);
|
|
}
|
|
if (src == 1)
|
|
flags._o = ((dst & 0x8000) != 0) == flags._c;
|
|
}
|
|
|
|
inline void _mul(uint8 src) {
|
|
unsigned r = unsigned(al) * src;
|
|
ax = (uint16)r;
|
|
flags._c = r >= 0x10000;
|
|
flags._z = r == 0;
|
|
flags._s = r & 0x8000;
|
|
flags._o = ah != 0;
|
|
}
|
|
|
|
inline void _mul(uint16 src) {
|
|
unsigned r = unsigned(ax) * src; //assuming here that we have at least 32 bits
|
|
dx = (r >> 16) & 0xffff;
|
|
ax = r & 0xffff;
|
|
flags._c = false;
|
|
flags._z = r == 0;
|
|
flags._s = r & 0x80000000;
|
|
flags._o = dx != 0;
|
|
}
|
|
|
|
inline void _neg(uint8 &src) {
|
|
uint8 r = 0;
|
|
_sub(r, src);
|
|
src = r;
|
|
}
|
|
|
|
inline void _neg(uint16 &src) {
|
|
uint16 r = 0;
|
|
_sub(r, src);
|
|
src = r;
|
|
}
|
|
|
|
inline void _lodsb() {
|
|
al = ds.byte(si++);
|
|
}
|
|
|
|
inline void _lodsw() {
|
|
ax = ds.word(si);
|
|
si += 2;
|
|
}
|
|
|
|
inline void _movsb() {
|
|
es.byte(di++) = ds.byte(si++);
|
|
}
|
|
|
|
inline void _movsb(uint size, bool clear_cx = false) {
|
|
assert(size != 0xffff);
|
|
//fixme: add overlap and segment boundary check and rewrite
|
|
while (size--)
|
|
_movsb();
|
|
if (clear_cx)
|
|
cx = 0;
|
|
}
|
|
|
|
inline void _movsw() {
|
|
_movsb();
|
|
_movsb();
|
|
}
|
|
|
|
inline void _movsw(uint size, bool clear_cx = false) {
|
|
assert(size != 0xffff);
|
|
_movsb(size * 2, clear_cx);
|
|
}
|
|
|
|
inline void _stosb() {
|
|
es.byte(di++) = al;
|
|
}
|
|
|
|
inline void _stosb(uint size, bool clear_cx = false) {
|
|
assert(size != 0xffff);
|
|
uint8 *dst = es.ptr(di, size);
|
|
memset(dst, al, size);
|
|
di += size;
|
|
if (clear_cx)
|
|
cx = 0;
|
|
}
|
|
|
|
inline void _stosw() {
|
|
es.byte(di++) = al;
|
|
es.byte(di++) = ah;
|
|
}
|
|
|
|
inline void _stosw(uint size, bool clear_cx = false) {
|
|
assert(size != 0xffff);
|
|
uint8 *dst = es.ptr(di, size * 2);
|
|
di += 2 * size;
|
|
while (size--) {
|
|
*dst++ = al;
|
|
*dst++ = ah;
|
|
}
|
|
if (clear_cx)
|
|
cx = 0;
|
|
}
|
|
|
|
inline void _xchg(uint16 &a, uint16 &b) {
|
|
uint16 x = a;
|
|
a = b;
|
|
b = x;
|
|
}
|
|
|
|
inline void _xchg(uint8 &a, uint8 &b) {
|
|
uint8 t = a;
|
|
a = b;
|
|
b = t;
|
|
}
|
|
|
|
Common::Array<uint16> stack;
|
|
inline void push(uint16 v) {
|
|
stack.push_back(v);
|
|
}
|
|
|
|
inline uint16 pop() {
|
|
assert(!stack.empty());
|
|
uint16 v = stack.back();
|
|
stack.pop_back();
|
|
return v;
|
|
}
|
|
};
|
|
|
|
inline void SegmentRef::reset(uint16 value) {
|
|
*this = _context->getSegment(value);
|
|
}
|
|
|
|
class StackChecker {
|
|
const Context &_context;
|
|
const uint _stackDepth;
|
|
|
|
public:
|
|
StackChecker(const Context &context): _context(context), _stackDepth(context.stack.size()) {}
|
|
~StackChecker() { assert(_context.stack.size() == _stackDepth); }
|
|
};
|
|
|
|
#ifndef NDEBUG
|
|
# define STACK_CHECK StackChecker checker(*this)
|
|
#else
|
|
# define STACK_CHECK do {} while (0)
|
|
#endif
|
|
|
|
}
|
|
|
|
#endif
|