// license:BSD-3-Clause // copyright-holders:Aaron Giles,Olivier Galibert /*************************************************************************** emumem.c Functions which handle device memory access. **************************************************************************** Basic theory of memory handling: An address with up to 32 bits is passed to a memory handler. First, an address mask is applied to the address, removing unused bits. Next, the address is broken into two halves, an upper half and a lower half. The number of bits in each half can be controlled via the macros in LEVEL1_BITS and LEVEL2_BITS, but they default to the upper 18 bits and the lower 14 bits. The upper half is then used as an index into a lookup table of bytes. If the value pulled from the table is between SUBTABLE_BASE and 255, then the lower half of the address is needed to resolve the final handler. In this case, the value from the table is combined with the lower address bits to form an index into a subtable. The final result of the lookup is a value from 0 to SUBTABLE_BASE - 1. These values correspond to memory handlers. The lower numbered handlers (from 0 through STATIC_COUNT - 1) are fixed handlers and refer to either memory banks or other special cases. The remaining handlers (from STATIC_COUNT through SUBTABLE_BASE - 1) are dynamically allocated to driver-specified handlers. Thus, table entries fall into these categories: 0 .. STATIC_COUNT - 1 = fixed handlers STATIC_COUNT .. SUBTABLE_BASE - 1 = driver-specific handlers SUBTABLE_BASE .. TOTAL_MEMORY_BANKS - 1 = need to look up lower bits in subtable Caveats: * If your driver executes an opcode which crosses a bank-switched boundary, it will pull the wrong data out of memory. Although not a common case, you may need to revert to memcpy to work around this. See machine/tnzs.c for an example. To do: - Add local banks for RAM/ROM to reduce pressure on banking - Always mirror everything out to 32 bits so we don't have to mask the address? - Add the ability to start with another memory map and modify it - Add fourth memory space for encrypted opcodes - Automatically mirror program space into data space if no data space - Get rid of opcode/data separation by using address spaces? - Add support for internal addressing (maybe just accessors - see TMS3202x) **************************************************************************** Address map fields and restrictions: AM_RANGE(start, end) Specifies a range of consecutive addresses beginning with 'start' and ending with 'end' inclusive. An address hits in this bucket if the 'address' >= 'start' and 'address' <= 'end'. AM_MASK(mask) Specifies a mask for the addresses in the current bucket. This mask is applied after a positive hit in the bucket specified by AM_RANGE or AM_SPACE, and is computed before accessing the RAM or calling through to the read/write handler. If you use AM_MIRROR, below, there should not be any bits in common between mask and mirror. Same for select. AM_MIRROR(mirror) Specifies mirror addresses for the given bucket. The current bucket is mapped repeatedly according to the mirror mask, once where each mirror bit is 0, and once where it is 1. For example, a 'mirror' value of 0x14000 would map the bucket at 0x00000, 0x04000, 0x10000, and 0x14000. AM_SELECT(select) Mirrors the addresses for the given bucket and pass the corresponding address bits to the handler. Very useful for devices with multiple slots/channels/etc were each slot is on a series of consecutive addresses regrouping all its registers. AM_ROM Specifies that this bucket contains ROM data by attaching an internal read handler. If this address space describes the first address space for a device, and if there is a region whose name matches the device's name, and if the bucket start/end range is within the bounds of that region, then this bucket will automatically map to the memory contained in that region. AM_RAM AM_READONLY AM_WRITEONLY Specifies that this bucket contains RAM data by attaching internal read and/or write handlers. Memory is automatically allocated to back this area. AM_RAM maps both reads and writes, while AM_READONLY only maps reads and AM_WRITEONLY only maps writes. AM_NOP AM_READNOP AM_WRITENOP Specifies that reads and/or writes in this bucket are unmapped, but that accesses to them should not be logged. AM_NOP unmaps both reads and writes, while AM_READNOP only unmaps reads, and AM_WRITENOP only unmaps writes. AM_UNMAP Specifies that both reads and writes in thus bucket are unmapeed, and that accesses to them should be logged. There is rarely a need for this, as the entire address space is initialized to behave this way by default. AM_READ_BANK(tag) AM_WRITE_BANK(tag) AM_READWRITE_BANK(tag) Specifies that reads and/or writes in this bucket map to a memory bank with the provided 'tag'. The actual memory this bank points to can be later controlled via the same tag. AM_READ(read) AM_WRITE(write) AM_READWRITE(read, write) Specifies read and/or write handler callbacks for this bucket. All reads and writes in this bucket will trigger a call to the provided functions. AM_DEVREAD(tag, read) AM_DEVWRITE(tag, read) AM_DEVREADWRITE(tag, read) Specifies a device-specific read and/or write handler for this bucket, automatically bound to the device specified by the provided 'tag'. AM_READ_PORT(tag) AM_WRITE_PORT(tag) AM_READWRITE_PORT(tag) Specifies that read and/or write accesses in this bucket will map to the I/O port with the provided 'tag'. An internal read/write handler is set up to handle this mapping. AM_REGION(class, tag, offs) Only useful if used in conjunction with AM_ROM, AM_RAM, or AM_READ/WRITE_BANK. By default, memory is allocated to back each bucket. By specifying AM_REGION, you can tell the memory system to point the base of the memory backing this bucket to a given memory 'region' at the specified 'offs' instead of allocating it. AM_SHARE(tag) Similar to AM_REGION, this specifies that the memory backing the current bucket is shared with other buckets. The first bucket to specify the share 'tag' will use its memory as backing for all future buckets that specify AM_SHARE with the same 'tag'. AM_SETOFFSET(setoffset) Specifies a handler for a 'set address' operation. The intended use case for this operation is to emulate a split-phase memory access: The caller (usually a CPU) sets the address bus lines using set_address. Some component may then react, for instance, by asserting a control line like WAIT before delivering the data on the data bus. The data bits are then sampled on the read operation or delivered on the write operation that must be called subsequently. It is not checked whether the address of the set_address operation matches the address of the subsequent read/write operation. The address map translates the address to a bucket and an offset, hence the name of the macro. If no handler is specified for a bucket, a set_address operation hitting that bucket returns silently. AM_DEVSETOFFSET(tag, setoffset) Specifies a handler for a set_address operation, bound to the device specified by 'tag'. ***************************************************************************/ #include #include #include "emu.h" #include "emuopts.h" #include "debug/debugcpu.h" //************************************************************************** // DEBUGGING //************************************************************************** #define MEM_DUMP (0) #define VERBOSE (0) #define TEST_HANDLER (0) #define VPRINTF(x) do { if (VERBOSE) printf x; } while (0) /*------------------------------------------------- core_i64_hex_format - i64 format printf helper -------------------------------------------------*/ static char *core_i64_hex_format(u64 value, u8 mindigits) { static char buffer[16][64]; // TODO: this can overflow - e.g. when a lot of unmapped writes are logged static int index; char *bufbase = &buffer[index++ % 16][0]; char *bufptr = bufbase; s8 curdigit; for (curdigit = 15; curdigit >= 0; curdigit--) { int nibble = (value >> (curdigit * 4)) & 0xf; if (nibble != 0 || curdigit < mindigits) { mindigits = curdigit; *bufptr++ = "0123456789ABCDEF"[nibble]; } } if (bufptr == bufbase) *bufptr++ = '0'; *bufptr = 0; return bufbase; } //************************************************************************** // CONSTANTS //************************************************************************** // banking constants const int BANK_ENTRY_UNSPECIFIED = -1; // other address map constants const int MEMORY_BLOCK_CHUNK = 65536; // minimum chunk size of allocated memory blocks // static data access handler constants enum { STATIC_INVALID = 0, // invalid - should never be used STATIC_BANK1 = 1, // first memory bank STATIC_BANKMAX = 0xfb, // last memory bank STATIC_NOP, // NOP - reads = unmapped value; writes = no-op STATIC_UNMAP, // unmapped - same as NOP except we log errors STATIC_WATCHPOINT, // watchpoint - used internally STATIC_COUNT // total number of static handlers }; //************************************************************************** // TYPE DEFINITIONS //************************************************************************** // ======================> handler_entry // a handler entry contains information about a memory handler class handler_entry { DISABLE_COPYING(handler_entry); protected: // construction/destruction handler_entry(u8 width, endianness_t endianness, u8 **rambaseptr); virtual ~handler_entry(); public: // getters bool populated() const { return m_populated; } offs_t addrstart() const { return m_addrstart; } offs_t addrend() const { return m_addrend; } offs_t addrmask() const { return m_addrmask; } offs_t bytemask() const { return m_bytemask; } virtual const char *name() const = 0; virtual const char *subunit_name(int entry) const = 0; void description(char *buffer) const; virtual void copy(handler_entry *entry); // return offset within the range referenced by this handler offs_t offset(offs_t address) const { return (address - m_addrstart) & m_addrmask; } // return a pointer to the backing RAM at the given offset u8 *ramptr(offs_t offset = 0) const { return *m_rambaseptr + offset; } // see if we are an exact match to the given parameters bool matches_exactly(offs_t addrstart, offs_t addrend, offs_t addrmask) const { return (m_populated && m_addrstart == addrstart && m_addrend == addrend && m_addrmask == addrmask); } // get the start/end address with the given mirror void mirrored_start_end(offs_t address, offs_t &start, offs_t &end) const { offs_t mirrorbits = (address - m_addrstart) & ~m_addrmask; start = m_addrstart | mirrorbits; end = m_addrend | mirrorbits; } // configure the handler addresses, and mark as populated void configure(offs_t addrstart, offs_t addrend, offs_t addrmask, offs_t bytemask) { if (m_populated && m_subunits) reconfigure_subunits(addrstart); m_populated = true; m_addrstart = addrstart; m_addrend = addrend; m_addrmask = addrmask; m_bytemask = bytemask; } // Re-expand the addrmask after subunit fun void expand_bytemask(offs_t previous_mask) { m_bytemask |= previous_mask; } // reconfigure the subunits on a base address change void reconfigure_subunits(offs_t addrstart); // depopulate an handler void deconfigure() { m_populated = false; m_subunits = 0; } // apply a global mask void apply_mask(offs_t addrmask) { m_addrmask &= addrmask; } void clear_conflicting_subunits(u64 handlermask); bool overriden_by_mask(u64 handlermask); protected: // Subunit description information struct subunit_info { offs_t m_addrmask; // addrmask for this subunit u32 m_mask; // mask (ff, ffff or ffffffff) s32 m_offset; // offset to add to the address u32 m_multiplier; // multiplier to the pre-split address u8 m_size; // size (8, 16 or 32) u8 m_shift; // shift of the subunit }; // internal helpers void configure_subunits(u64 handlermask, int handlerbits, int &start_slot, int &end_slot); virtual void remove_subunit(int entry) = 0; // internal state bool m_populated; // populated? u8 m_datawidth; endianness_t m_endianness; offs_t m_addrstart; // start address for handler offs_t m_addrend; // end address for handler offs_t m_addrmask; // mask against the final address offs_t m_bytemask; // mask against the final address, byte resolution u8 ** m_rambaseptr; // pointer to the bank base u8 m_subunits; // for width stubs, the number of subunits subunit_info m_subunit_infos[8]; // for width stubs, the associated subunit info u64 m_invsubmask; // inverted mask of the populated subunits }; // ======================> handler_entry_read // a read-access-specific extension of handler_entry class handler_entry_read : public handler_entry { public: struct access_handler { // Constructors mean you can't union them read8_delegate r8; read16_delegate r16; read32_delegate r32; read64_delegate r64; }; // construction/destruction handler_entry_read(u8 width, endianness_t endianness, u8 **rambaseptr) : handler_entry(width, endianness, rambaseptr), m_ioport(nullptr) { } virtual void copy(handler_entry *entry) override; // getters virtual const char *name() const override; virtual const char *subunit_name(int entry) const override; // configure delegate callbacks void set_delegate(read8_delegate delegate, u64 mask = 0); void set_delegate(read16_delegate delegate, u64 mask = 0); void set_delegate(read32_delegate delegate, u64 mask = 0); void set_delegate(read64_delegate delegate, u64 mask = 0); // configure I/O port access void set_ioport(ioport_port &ioport); // read via the underlying delegates u8 read8(address_space &space, offs_t offset, u8 mask) const { return m_read.r8(space, offset, mask); } u16 read16(address_space &space, offs_t offset, u16 mask) const { return m_read.r16(space, offset, mask); } u32 read32(address_space &space, offs_t offset, u32 mask) const { return m_read.r32(space, offset, mask); } u64 read64(address_space &space, offs_t offset, u64 mask) const { return m_read.r64(space, offset, mask); } private: // stubs for converting between address sizes u16 read_stub_16(address_space &space, offs_t offset, u16 mask); u32 read_stub_32(address_space &space, offs_t offset, u32 mask); u64 read_stub_64(address_space &space, offs_t offset, u64 mask); // stubs for reading I/O ports template UintType read_stub_ioport(address_space &space, offs_t offset, UintType mask) { return m_ioport->read(); } // internal helper virtual void remove_subunit(int entry) override; // internal state access_handler m_read; access_handler m_subread[8]; ioport_port * m_ioport; }; // ======================> handler_entry_write // a write-access-specific extension of handler_entry class handler_entry_write : public handler_entry { public: struct access_handler { // Constructors mean you can't union them write8_delegate w8; write16_delegate w16; write32_delegate w32; write64_delegate w64; }; // construction/destruction handler_entry_write(u8 width, endianness_t endianness, u8 **rambaseptr) : handler_entry(width, endianness, rambaseptr), m_ioport(nullptr) { } virtual void copy(handler_entry *entry) override; // getters virtual const char *name() const override; virtual const char *subunit_name(int entry) const override; // configure delegate callbacks void set_delegate(write8_delegate delegate, u64 mask = 0); void set_delegate(write16_delegate delegate, u64 mask = 0); void set_delegate(write32_delegate delegate, u64 mask = 0); void set_delegate(write64_delegate delegate, u64 mask = 0); // configure I/O port access void set_ioport(ioport_port &ioport); // write via the underlying delegates void write8(address_space &space, offs_t offset, u8 data, u8 mask) const { m_write.w8(space, offset, data, mask); } void write16(address_space &space, offs_t offset, u16 data, u16 mask) const { m_write.w16(space, offset, data, mask); } void write32(address_space &space, offs_t offset, u32 data, u32 mask) const { m_write.w32(space, offset, data, mask); } void write64(address_space &space, offs_t offset, u64 data, u64 mask) const { m_write.w64(space, offset, data, mask); } private: // stubs for converting between address sizes void write_stub_16(address_space &space, offs_t offset, u16 data, u16 mask); void write_stub_32(address_space &space, offs_t offset, u32 data, u32 mask); void write_stub_64(address_space &space, offs_t offset, u64 data, u64 mask); // stubs for writing I/O ports template void write_stub_ioport(address_space &space, offs_t offset, UintType data, UintType mask) { m_ioport->write(data, mask); } // internal helper virtual void remove_subunit(int entry) override; // internal state access_handler m_write; access_handler m_subwrite[8]; ioport_port * m_ioport; }; // ======================> handler_entry_setoffset // a setoffset-access-specific extension of handler_entry class handler_entry_setoffset : public handler_entry { public: // construction/destruction handler_entry_setoffset() : handler_entry(0, ENDIANNESS_LITTLE, nullptr) { } const char *name() const override { return m_setoffset.name(); } const char *subunit_name(int entry) const override { return "no subunit"; } // Call through only if the setoffset handler has been late-bound before // (i.e. if it was declared in the address map) void setoffset(address_space &space, offs_t offset) const { if (m_setoffset.has_object()) m_setoffset(space, offset); } // configure delegate callbacks void set_delegate(setoffset_delegate delegate, u64 mask = 0) { m_setoffset = delegate; } private: setoffset_delegate m_setoffset; // We do not have subunits for setoffset // Accordingly, we need not implement unused functions. void remove_subunit(int entry) override { } }; // ======================> handler_entry_proxy // A proxy class that contains an handler_entry_read or _write and forwards the setter calls template class handler_entry_proxy { public: handler_entry_proxy(std::list _handlers, u64 _mask) : handlers(std::move(_handlers)), mask(_mask) {} handler_entry_proxy(const handler_entry_proxy &hep) : handlers(hep.handlers), mask(hep.mask) {} // forward delegate callbacks configuration template void set_delegate(Delegate delegate) const { for (const auto & elem : handlers) (elem)->set_delegate(delegate, mask); } // forward I/O port access configuration void set_ioport(ioport_port &ioport) const { for (const auto & elem : handlers) (elem)->set_ioport(ioport); } private: std::list handlers; u64 mask; }; // ======================> address_table // address_table contains information about read/write accesses within an address space class address_table { // address map lookup table definitions static const int LEVEL1_BITS = 18; // number of address bits in the level 1 table static const int LEVEL2_BITS = 32 - LEVEL1_BITS; // number of address bits in the level 2 table static const int SUBTABLE_COUNT = 64; // number of slots reserved for subtables static const int SUBTABLE_BASE = TOTAL_MEMORY_BANKS - SUBTABLE_COUNT; // first index of a subtable static const int ENTRY_COUNT = SUBTABLE_BASE; // number of legitimate (non-subtable) entries static const int SUBTABLE_ALLOC = 8; // number of subtables to allocate at a time inline int level2_bits() const { return m_large ? LEVEL2_BITS : 0; } public: // construction/destruction address_table(address_space &space, bool large); virtual ~address_table(); // getters virtual handler_entry &handler(u32 index) const = 0; bool watchpoints_enabled() const { return (m_live_lookup == s_watchpoint_table); } // address lookups u32 lookup_live(offs_t address) const { return m_large ? lookup_live_large(address) : lookup_live_small(address); } u32 lookup_live_small(offs_t address) const { return m_live_lookup[address]; } u32 lookup_live_large(offs_t address) const { u32 entry = m_live_lookup[level1_index_large(address)]; if (entry >= SUBTABLE_BASE) entry = m_live_lookup[level2_index_large(entry, address)]; return entry; } u32 lookup_live_nowp(offs_t address) const { return m_large ? lookup_live_large_nowp(address) : lookup_live_small_nowp(address); } u32 lookup_live_small_nowp(offs_t address) const { return m_table[address]; } u32 lookup_live_large_nowp(offs_t address) const { u32 entry = m_table[level1_index_large(address)]; if (entry >= SUBTABLE_BASE) entry = m_table[level2_index_large(entry, address)]; return entry; } u32 lookup(offs_t address) const { u32 entry = m_live_lookup[level1_index(address)]; if (entry >= SUBTABLE_BASE) entry = m_live_lookup[level2_index(entry, address)]; return entry; } // enable watchpoints by swapping in the watchpoint table void enable_watchpoints(bool enable = true) { m_live_lookup = enable ? s_watchpoint_table : &m_table[0]; } // table mapping helpers void map_range(offs_t addrstart, offs_t addrend, offs_t addrmask, offs_t addrmirror, u16 staticentry); void setup_range(offs_t addrstart, offs_t addrend, offs_t addrmask, offs_t addrmirror, u64 mask, std::list &entries); u16 derive_range(offs_t address, offs_t &addrstart, offs_t &addrend) const; // misc helpers void mask_all_handlers(offs_t mask); const char *handler_name(u16 entry) const; protected: // determine table indexes based on the address u32 level1_index_large(offs_t address) const { return address >> LEVEL2_BITS; } u32 level2_index_large(u16 l1entry, offs_t address) const { return (1 << LEVEL1_BITS) + ((l1entry - SUBTABLE_BASE) << LEVEL2_BITS) + (address & ((1 << LEVEL2_BITS) - 1)); } u32 level1_index(offs_t address) const { return m_large ? level1_index_large(address) : address; } u32 level2_index(u16 l1entry, offs_t address) const { return m_large ? level2_index_large(l1entry, address) : 0; } // table population/depopulation void populate_range_mirrored(offs_t addrstart, offs_t addrend, offs_t addrmirror, u16 handler); void populate_range(offs_t addrstart, offs_t addrend, u16 handler); // subtable management u16 subtable_alloc(); void subtable_realloc(u16 subentry); int subtable_merge(); void subtable_release(u16 subentry); u16 *subtable_open(offs_t l1index); void subtable_close(offs_t l1index); u16 *subtable_ptr(u16 entry) { return &m_table[level2_index(entry, 0)]; } // internal state std::vector m_table; // pointer to base of table u16 * m_live_lookup; // current lookup address_space & m_space; // pointer back to the space bool m_large; // large memory model? // subtable_data is an internal class with information about each subtable class subtable_data { public: subtable_data() : m_checksum_valid(false), m_checksum(0), m_usecount(0) { } bool m_checksum_valid; // is the checksum valid u32 m_checksum; // checksum over all the bytes u32 m_usecount; // number of times this has been used }; std::vector m_subtable; // info about each subtable u16 m_subtable_alloc; // number of subtables allocated // static global read-only watchpoint table static u16 s_watchpoint_table[1 << LEVEL1_BITS]; private: int handler_refcount[SUBTABLE_BASE-STATIC_COUNT]; u16 handler_next_free[SUBTABLE_BASE-STATIC_COUNT]; u16 handler_free; u16 get_free_handler(); void verify_reference_counts(); void setup_range_solid(offs_t addrstart, offs_t addrend, offs_t addrmask, offs_t addrmirror, std::list &entries); void setup_range_masked(offs_t addrstart, offs_t addrend, offs_t addrmask, offs_t addrmirror, u64 mask, std::list &entries); void handler_ref(u16 entry, int count) { assert(entry < SUBTABLE_BASE); if (entry >= STATIC_COUNT) handler_refcount[entry - STATIC_COUNT] += count; } void handler_unref(u16 entry) { assert(entry < SUBTABLE_BASE); if (entry >= STATIC_COUNT) if (! --handler_refcount[entry - STATIC_COUNT]) { handler(entry).deconfigure(); handler_next_free[entry - STATIC_COUNT] = handler_free; handler_free = entry; } } }; // ======================> address_table_read // read access-specific version of an address table class address_table_read : public address_table { public: // construction/destruction address_table_read(address_space &space, bool large); virtual ~address_table_read(); // getters virtual handler_entry &handler(u32 index) const override; handler_entry_read &handler_read(u32 index) const { assert(index < ARRAY_LENGTH(m_handlers)); return *m_handlers[index]; } // range getter handler_entry_proxy handler_map_range(offs_t addrstart, offs_t addrend, offs_t addrmask, offs_t addrmirror, u64 mask = 0) { std::list entries; setup_range(addrstart, addrend, addrmask, addrmirror, mask, entries); std::list handlers; for (std::list::const_iterator i = entries.begin(); i != entries.end(); ++i) handlers.push_back(&handler_read(*i)); return handler_entry_proxy(handlers, mask); } private: // internal unmapped handler template UintType unmap_r(address_space &space, offs_t offset, UintType mask) { if (m_space.log_unmap() && !m_space.machine().side_effect_disabled()) { m_space.device().logerror( m_space.is_octal() ? "%s: unmapped %s memory read from %0*o & %0*o\n" : "%s: unmapped %s memory read from %0*X & %0*X\n", m_space.machine().describe_context(), m_space.name(), m_space.addrchars(), m_space.byte_to_address(offset * sizeof(UintType)), 2 * sizeof(UintType), mask); } return m_space.unmap(); } // internal no-op handler template UintType nop_r(address_space &space, offs_t offset, UintType mask) { return m_space.unmap(); } // internal watchpoint handler template UintType watchpoint_r(address_space &space, offs_t offset, UintType mask) { m_space.device().debug()->memory_read_hook(m_space, offset * sizeof(UintType), mask); u16 *oldtable = m_live_lookup; m_live_lookup = &m_table[0]; UintType result; if (sizeof(UintType) == 1) result = m_space.read_byte(offset); if (sizeof(UintType) == 2) result = m_space.read_word(offset << 1, mask); if (sizeof(UintType) == 4) result = m_space.read_dword(offset << 2, mask); if (sizeof(UintType) == 8) result = m_space.read_qword(offset << 3, mask); m_live_lookup = oldtable; return result; } // internal state std::unique_ptr m_handlers[TOTAL_MEMORY_BANKS]; // array of user-installed handlers }; // ======================> address_table_write // write access-specific version of an address table class address_table_write : public address_table { public: // construction/destruction address_table_write(address_space &space, bool large); virtual ~address_table_write(); // getters virtual handler_entry &handler(u32 index) const override; handler_entry_write &handler_write(u32 index) const { assert(index < ARRAY_LENGTH(m_handlers)); return *m_handlers[index]; } // range getter handler_entry_proxy handler_map_range(offs_t addrstart, offs_t addrend, offs_t addrmask, offs_t addrmirror, u64 mask = 0) { std::list entries; setup_range(addrstart, addrend, addrmask, addrmirror, mask, entries); std::list handlers; for (std::list::const_iterator i = entries.begin(); i != entries.end(); ++i) handlers.push_back(&handler_write(*i)); return handler_entry_proxy(handlers, mask); } private: // internal handlers template void unmap_w(address_space &space, offs_t offset, UintType data, UintType mask) { if (m_space.log_unmap() && !m_space.machine().side_effect_disabled()) { m_space.device().logerror( m_space.is_octal() ? "%s: unmapped %s memory write to %0*o = %0*o & %0*o\n" : "%s: unmapped %s memory write to %0*X = %0*X & %0*X\n", m_space.machine().describe_context(), m_space.name(), m_space.addrchars(), m_space.byte_to_address(offset * sizeof(UintType)), 2 * sizeof(UintType), data, 2 * sizeof(UintType), mask); } } template void nop_w(address_space &space, offs_t offset, UintType data, UintType mask) { } template void watchpoint_w(address_space &space, offs_t offset, UintType data, UintType mask) { m_space.device().debug()->memory_write_hook(m_space, offset * sizeof(UintType), data, mask); u16 *oldtable = m_live_lookup; m_live_lookup = &m_table[0]; if (sizeof(UintType) == 1) m_space.write_byte(offset, data); if (sizeof(UintType) == 2) m_space.write_word(offset << 1, data, mask); if (sizeof(UintType) == 4) m_space.write_dword(offset << 2, data, mask); if (sizeof(UintType) == 8) m_space.write_qword(offset << 3, data, mask); m_live_lookup = oldtable; } // internal state std::unique_ptr m_handlers[TOTAL_MEMORY_BANKS]; // array of user-installed handlers }; // ======================> address_table_setoffset // setoffset access-specific version of an address table class address_table_setoffset : public address_table { public: // construction/destruction address_table_setoffset(address_space &space, bool large) : address_table(space, large) { // allocate handlers for each entry, prepopulating the bankptrs for banks for (auto & elem : m_handlers) elem = std::make_unique(); // Watchpoints and unmap states do not make sense for setoffset m_handlers[STATIC_NOP]->set_delegate(setoffset_delegate(FUNC(address_table_setoffset::nop_so), this)); m_handlers[STATIC_NOP]->configure(0, space.addrmask(), ~0, ~0); } ~address_table_setoffset() { } handler_entry &handler(u32 index) const override { assert(index < ARRAY_LENGTH(m_handlers)); return *m_handlers[index]; } handler_entry_setoffset &handler_setoffset(u32 index) const { assert(index < ARRAY_LENGTH(m_handlers)); return *m_handlers[index]; } // range getter handler_entry_proxy handler_map_range(offs_t addrstart, offs_t addrend, offs_t addrmask, offs_t addrmirror, u64 mask = 0) { std::list entries; setup_range(addrstart, addrend, addrmask, addrmirror, mask, entries); std::list handlers; for (std::list::const_iterator i = entries.begin(); i != entries.end(); ++i) handlers.push_back(&handler_setoffset(*i)); return handler_entry_proxy(handlers, mask); } private: // internal handlers // Setoffset does not allow for watchpoints, since we assume that a // corresponding read/write operation will follow, and the watchpoint will // apply for that operation // For the same reason it does not make sense to put a warning into the log // for unmapped locations, as this will be done by the read/write operation void nop_so(address_space &space, offs_t offset) { } // internal state std::unique_ptr m_handlers[TOTAL_MEMORY_BANKS]; // array of user-installed handlers }; // ======================> address_space_specific // this is a derived class of address_space with specific width, endianness, and table size template class address_space_specific : public address_space { typedef address_space_specific this_type; // constants describing the native size static constexpr u32 NATIVE_BYTES = sizeof(NativeType); static constexpr u32 NATIVE_STEP = AddrShift >= 0 ? NATIVE_BYTES << iabs(AddrShift) : NATIVE_BYTES >> iabs(AddrShift); static constexpr u32 NATIVE_MASK = NATIVE_STEP - 1; static constexpr u32 NATIVE_BITS = 8 * NATIVE_BYTES; // helpers to simplify core code u32 read_lookup(offs_t address) const { return Large ? m_read.lookup_live_large(address) : m_read.lookup_live_small(address); } u32 write_lookup(offs_t address) const { return Large ? m_write.lookup_live_large(address) : m_write.lookup_live_small(address); } u32 setoffset_lookup(offs_t address) const { return Large ? m_setoffset.lookup_live_large(address) : m_setoffset.lookup_live_small(address); } static constexpr offs_t offset_to_byte(offs_t offset) { return AddrShift < 0 ? offset << iabs(AddrShift) : offset >> iabs(AddrShift); } public: // construction/destruction address_space_specific(memory_manager &manager, device_memory_interface &memory, int spacenum) : address_space(manager, memory, spacenum, Large), m_read(*this, Large), m_write(*this, Large), m_setoffset(*this, Large) { #if (TEST_HANDLER) // test code to verify the read/write handlers are touching the correct bits // and returning the correct results // install some dummy RAM for the first 16 bytes with well-known values u8 buffer[16]; for (int index = 0; index < 16; index++) buffer[index ^ ((Endian == ENDIANNESS_NATIVE) ? 0 : (data_width()/8 - 1))] = index * 0x11; install_ram_generic(0x00, 0x0f, 0x0f, 0, read_or_write::READWRITE, buffer); printf("\n\naddress_space(%d, %s, %s)\n", NATIVE_BITS, (Endian == ENDIANNESS_LITTLE) ? "little" : "big", Large ? "large" : "small"); // walk through the first 8 addresses for (int address = 0; address < 8; address++) { // determine expected values u64 expected64 = (u64((address + ((Endian == ENDIANNESS_LITTLE) ? 7 : 0)) * 0x11) << 56) | (u64((address + ((Endian == ENDIANNESS_LITTLE) ? 6 : 1)) * 0x11) << 48) | (u64((address + ((Endian == ENDIANNESS_LITTLE) ? 5 : 2)) * 0x11) << 40) | (u64((address + ((Endian == ENDIANNESS_LITTLE) ? 4 : 3)) * 0x11) << 32) | (u64((address + ((Endian == ENDIANNESS_LITTLE) ? 3 : 4)) * 0x11) << 24) | (u64((address + ((Endian == ENDIANNESS_LITTLE) ? 2 : 5)) * 0x11) << 16) | (u64((address + ((Endian == ENDIANNESS_LITTLE) ? 1 : 6)) * 0x11) << 8) | (u64((address + ((Endian == ENDIANNESS_LITTLE) ? 0 : 7)) * 0x11) << 0); u32 expected32 = (Endian == ENDIANNESS_LITTLE) ? expected64 : (expected64 >> 32); u16 expected16 = (Endian == ENDIANNESS_LITTLE) ? expected32 : (expected32 >> 16); u8 expected8 = (Endian == ENDIANNESS_LITTLE) ? expected16 : (expected16 >> 8); u64 result64; u32 result32; u16 result16; u8 result8; // validate byte accesses printf("\nAddress %d\n", address); printf(" read_byte = "); printf("%02X\n", result8 = read_byte(address)); assert(result8 == expected8); // validate word accesses (if aligned) if (WORD_ALIGNED(address)) { printf(" read_word = "); printf("%04X\n", result16 = read_word(address)); assert(result16 == expected16); } if (WORD_ALIGNED(address)) { printf(" read_word (0xff00) = "); printf("%04X\n", result16 = read_word(address, 0xff00)); assert((result16 & 0xff00) == (expected16 & 0xff00)); } if (WORD_ALIGNED(address)) { printf(" (0x00ff) = "); printf("%04X\n", result16 = read_word(address, 0x00ff)); assert((result16 & 0x00ff) == (expected16 & 0x00ff)); } // validate unaligned word accesses printf(" read_word_unaligned = "); printf("%04X\n", result16 = read_word_unaligned(address)); assert(result16 == expected16); printf(" read_word_unaligned (0xff00) = "); printf("%04X\n", result16 = read_word_unaligned(address, 0xff00)); assert((result16 & 0xff00) == (expected16 & 0xff00)); printf(" (0x00ff) = "); printf("%04X\n", result16 = read_word_unaligned(address, 0x00ff)); assert((result16 & 0x00ff) == (expected16 & 0x00ff)); // validate dword acceses (if aligned) if (DWORD_ALIGNED(address)) { printf(" read_dword = "); printf("%08X\n", result32 = read_dword(address)); assert(result32 == expected32); } if (DWORD_ALIGNED(address)) { printf(" read_dword (0xff000000) = "); printf("%08X\n", result32 = read_dword(address, 0xff000000)); assert((result32 & 0xff000000) == (expected32 & 0xff000000)); } if (DWORD_ALIGNED(address)) { printf(" (0x00ff0000) = "); printf("%08X\n", result32 = read_dword(address, 0x00ff0000)); assert((result32 & 0x00ff0000) == (expected32 & 0x00ff0000)); } if (DWORD_ALIGNED(address)) { printf(" (0x0000ff00) = "); printf("%08X\n", result32 = read_dword(address, 0x0000ff00)); assert((result32 & 0x0000ff00) == (expected32 & 0x0000ff00)); } if (DWORD_ALIGNED(address)) { printf(" (0x000000ff) = "); printf("%08X\n", result32 = read_dword(address, 0x000000ff)); assert((result32 & 0x000000ff) == (expected32 & 0x000000ff)); } if (DWORD_ALIGNED(address)) { printf(" (0xffff0000) = "); printf("%08X\n", result32 = read_dword(address, 0xffff0000)); assert((result32 & 0xffff0000) == (expected32 & 0xffff0000)); } if (DWORD_ALIGNED(address)) { printf(" (0x0000ffff) = "); printf("%08X\n", result32 = read_dword(address, 0x0000ffff)); assert((result32 & 0x0000ffff) == (expected32 & 0x0000ffff)); } if (DWORD_ALIGNED(address)) { printf(" (0xffffff00) = "); printf("%08X\n", result32 = read_dword(address, 0xffffff00)); assert((result32 & 0xffffff00) == (expected32 & 0xffffff00)); } if (DWORD_ALIGNED(address)) { printf(" (0x00ffffff) = "); printf("%08X\n", result32 = read_dword(address, 0x00ffffff)); assert((result32 & 0x00ffffff) == (expected32 & 0x00ffffff)); } // validate unaligned dword accesses printf(" read_dword_unaligned = "); printf("%08X\n", result32 = read_dword_unaligned(address)); assert(result32 == expected32); printf(" read_dword_unaligned (0xff000000) = "); printf("%08X\n", result32 = read_dword_unaligned(address, 0xff000000)); assert((result32 & 0xff000000) == (expected32 & 0xff000000)); printf(" (0x00ff0000) = "); printf("%08X\n", result32 = read_dword_unaligned(address, 0x00ff0000)); assert((result32 & 0x00ff0000) == (expected32 & 0x00ff0000)); printf(" (0x0000ff00) = "); printf("%08X\n", result32 = read_dword_unaligned(address, 0x0000ff00)); assert((result32 & 0x0000ff00) == (expected32 & 0x0000ff00)); printf(" (0x000000ff) = "); printf("%08X\n", result32 = read_dword_unaligned(address, 0x000000ff)); assert((result32 & 0x000000ff) == (expected32 & 0x000000ff)); printf(" (0xffff0000) = "); printf("%08X\n", result32 = read_dword_unaligned(address, 0xffff0000)); assert((result32 & 0xffff0000) == (expected32 & 0xffff0000)); printf(" (0x0000ffff) = "); printf("%08X\n", result32 = read_dword_unaligned(address, 0x0000ffff)); assert((result32 & 0x0000ffff) == (expected32 & 0x0000ffff)); printf(" (0xffffff00) = "); printf("%08X\n", result32 = read_dword_unaligned(address, 0xffffff00)); assert((result32 & 0xffffff00) == (expected32 & 0xffffff00)); printf(" (0x00ffffff) = "); printf("%08X\n", result32 = read_dword_unaligned(address, 0x00ffffff)); assert((result32 & 0x00ffffff) == (expected32 & 0x00ffffff)); // validate qword acceses (if aligned) if (QWORD_ALIGNED(address)) { printf(" read_qword = "); printf("%s\n", core_i64_hex_format(result64 = read_qword(address), 16)); assert(result64 == expected64); } if (QWORD_ALIGNED(address)) { printf(" read_qword (0xff00000000000000) = "); printf("%s\n", core_i64_hex_format(result64 = read_qword(address, 0xff00000000000000U), 16)); assert((result64 & 0xff00000000000000U) == (expected64 & 0xff00000000000000U)); } if (QWORD_ALIGNED(address)) { printf(" (0x00ff000000000000) = "); printf("%s\n", core_i64_hex_format(result64 = read_qword(address, 0x00ff000000000000U), 16)); assert((result64 & 0x00ff000000000000U) == (expected64 & 0x00ff000000000000U)); } if (QWORD_ALIGNED(address)) { printf(" (0x0000ff0000000000) = "); printf("%s\n", core_i64_hex_format(result64 = read_qword(address, 0x0000ff0000000000U), 16)); assert((result64 & 0x0000ff0000000000U) == (expected64 & 0x0000ff0000000000U)); } if (QWORD_ALIGNED(address)) { printf(" (0x000000ff00000000) = "); printf("%s\n", core_i64_hex_format(result64 = read_qword(address, 0x000000ff00000000U), 16)); assert((result64 & 0x000000ff00000000U) == (expected64 & 0x000000ff00000000U)); } if (QWORD_ALIGNED(address)) { printf(" (0x00000000ff000000) = "); printf("%s\n", core_i64_hex_format(result64 = read_qword(address, 0x00000000ff000000U), 16)); assert((result64 & 0x00000000ff000000U) == (expected64 & 0x00000000ff000000U)); } if (QWORD_ALIGNED(address)) { printf(" (0x0000000000ff0000) = "); printf("%s\n", core_i64_hex_format(result64 = read_qword(address, 0x0000000000ff0000U), 16)); assert((result64 & 0x0000000000ff0000U) == (expected64 & 0x0000000000ff0000U)); } if (QWORD_ALIGNED(address)) { printf(" (0x000000000000ff00) = "); printf("%s\n", core_i64_hex_format(result64 = read_qword(address, 0x000000000000ff00U), 16)); assert((result64 & 0x000000000000ff00U) == (expected64 & 0x000000000000ff00U)); } if (QWORD_ALIGNED(address)) { printf(" (0x00000000000000ff) = "); printf("%s\n", core_i64_hex_format(result64 = read_qword(address, 0x00000000000000ffU), 16)); assert((result64 & 0x00000000000000ffU) == (expected64 & 0x00000000000000ffU)); } if (QWORD_ALIGNED(address)) { printf(" (0xffff000000000000) = "); printf("%s\n", core_i64_hex_format(result64 = read_qword(address, 0xffff000000000000U), 16)); assert((result64 & 0xffff000000000000U) == (expected64 & 0xffff000000000000U)); } if (QWORD_ALIGNED(address)) { printf(" (0x0000ffff00000000) = "); printf("%s\n", core_i64_hex_format(result64 = read_qword(address, 0x0000ffff00000000U), 16)); assert((result64 & 0x0000ffff00000000U) == (expected64 & 0x0000ffff00000000U)); } if (QWORD_ALIGNED(address)) { printf(" (0x00000000ffff0000) = "); printf("%s\n", core_i64_hex_format(result64 = read_qword(address, 0x00000000ffff0000U), 16)); assert((result64 & 0x00000000ffff0000U) == (expected64 & 0x00000000ffff0000U)); } if (QWORD_ALIGNED(address)) { printf(" (0x000000000000ffff) = "); printf("%s\n", core_i64_hex_format(result64 = read_qword(address, 0x000000000000ffffU), 16)); assert((result64 & 0x000000000000ffffU) == (expected64 & 0x000000000000ffffU)); } if (QWORD_ALIGNED(address)) { printf(" (0xffffff0000000000) = "); printf("%s\n", core_i64_hex_format(result64 = read_qword(address, 0xffffff0000000000U), 16)); assert((result64 & 0xffffff0000000000U) == (expected64 & 0xffffff0000000000U)); } if (QWORD_ALIGNED(address)) { printf(" (0x0000ffffff000000) = "); printf("%s\n", core_i64_hex_format(result64 = read_qword(address, 0x0000ffffff000000U), 16)); assert((result64 & 0x0000ffffff000000U) == (expected64 & 0x0000ffffff000000U)); } if (QWORD_ALIGNED(address)) { printf(" (0x000000ffffff0000) = "); printf("%s\n", core_i64_hex_format(result64 = read_qword(address, 0x000000ffffff0000U), 16)); assert((result64 & 0x000000ffffff0000U) == (expected64 & 0x000000ffffff0000U)); } if (QWORD_ALIGNED(address)) { printf(" (0x0000000000ffffff) = "); printf("%s\n", core_i64_hex_format(result64 = read_qword(address, 0x0000000000ffffffU), 16)); assert((result64 & 0x0000000000ffffffU) == (expected64 & 0x0000000000ffffffU)); } if (QWORD_ALIGNED(address)) { printf(" (0xffffffff00000000) = "); printf("%s\n", core_i64_hex_format(result64 = read_qword(address, 0xffffffff00000000U), 16)); assert((result64 & 0xffffffff00000000U) == (expected64 & 0xffffffff00000000U)); } if (QWORD_ALIGNED(address)) { printf(" (0x00ffffffff000000) = "); printf("%s\n", core_i64_hex_format(result64 = read_qword(address, 0x00ffffffff000000U), 16)); assert((result64 & 0x00ffffffff000000U) == (expected64 & 0x00ffffffff000000U)); } if (QWORD_ALIGNED(address)) { printf(" (0x0000ffffffff0000) = "); printf("%s\n", core_i64_hex_format(result64 = read_qword(address, 0x0000ffffffff0000U), 16)); assert((result64 & 0x0000ffffffff0000U) == (expected64 & 0x0000ffffffff0000U)); } if (QWORD_ALIGNED(address)) { printf(" (0x000000ffffffff00) = "); printf("%s\n", core_i64_hex_format(result64 = read_qword(address, 0x000000ffffffff00U), 16)); assert((result64 & 0x000000ffffffff00U) == (expected64 & 0x000000ffffffff00U)); } if (QWORD_ALIGNED(address)) { printf(" (0x00000000ffffffff) = "); printf("%s\n", core_i64_hex_format(result64 = read_qword(address, 0x00000000ffffffffU), 16)); assert((result64 & 0x00000000ffffffffU) == (expected64 & 0x00000000ffffffffU)); } if (QWORD_ALIGNED(address)) { printf(" (0xffffffffff000000) = "); printf("%s\n", core_i64_hex_format(result64 = read_qword(address, 0xffffffffff000000U), 16)); assert((result64 & 0xffffffffff000000U) == (expected64 & 0xffffffffff000000U)); } if (QWORD_ALIGNED(address)) { printf(" (0x00ffffffffff0000) = "); printf("%s\n", core_i64_hex_format(result64 = read_qword(address, 0x00ffffffffff0000U), 16)); assert((result64 & 0x00ffffffffff0000U) == (expected64 & 0x00ffffffffff0000U)); } if (QWORD_ALIGNED(address)) { printf(" (0x0000ffffffffff00) = "); printf("%s\n", core_i64_hex_format(result64 = read_qword(address, 0x0000ffffffffff00U), 16)); assert((result64 & 0x0000ffffffffff00U) == (expected64 & 0x0000ffffffffff00U)); } if (QWORD_ALIGNED(address)) { printf(" (0x000000ffffffffff) = "); printf("%s\n", core_i64_hex_format(result64 = read_qword(address, 0x000000ffffffffffU), 16)); assert((result64 & 0x000000ffffffffffU) == (expected64 & 0x000000ffffffffffU)); } if (QWORD_ALIGNED(address)) { printf(" (0xffffffffffff0000) = "); printf("%s\n", core_i64_hex_format(result64 = read_qword(address, 0xffffffffffff0000U), 16)); assert((result64 & 0xffffffffffff0000U) == (expected64 & 0xffffffffffff0000U)); } if (QWORD_ALIGNED(address)) { printf(" (0x00ffffffffffff00) = "); printf("%s\n", core_i64_hex_format(result64 = read_qword(address, 0x00ffffffffffff00U), 16)); assert((result64 & 0x00ffffffffffff00U) == (expected64 & 0x00ffffffffffff00U)); } if (QWORD_ALIGNED(address)) { printf(" (0x0000ffffffffffff) = "); printf("%s\n", core_i64_hex_format(result64 = read_qword(address, 0x0000ffffffffffffU), 16)); assert((result64 & 0x0000ffffffffffffU) == (expected64 & 0x0000ffffffffffffU)); } if (QWORD_ALIGNED(address)) { printf(" (0xffffffffffffff00) = "); printf("%s\n", core_i64_hex_format(result64 = read_qword(address, 0xffffffffffffff00U), 16)); assert((result64 & 0xffffffffffffff00U) == (expected64 & 0xffffffffffffff00U)); } if (QWORD_ALIGNED(address)) { printf(" (0x00ffffffffffffff) = "); printf("%s\n", core_i64_hex_format(result64 = read_qword(address, 0x00ffffffffffffffU), 16)); assert((result64 & 0x00ffffffffffffffU) == (expected64 & 0x00ffffffffffffffU)); } // validate unaligned qword accesses printf(" read_qword_unaligned = "); printf("%s\n", core_i64_hex_format(result64 = read_qword_unaligned(address), 16)); assert(result64 == expected64); printf(" read_qword_unaligned (0xff00000000000000) = "); printf("%s\n", core_i64_hex_format(result64 = read_qword_unaligned(address, 0xff00000000000000U), 16)); assert((result64 & 0xff00000000000000U) == (expected64 & 0xff00000000000000U)); printf(" (0x00ff000000000000) = "); printf("%s\n", core_i64_hex_format(result64 = read_qword_unaligned(address, 0x00ff000000000000U), 16)); assert((result64 & 0x00ff000000000000U) == (expected64 & 0x00ff000000000000U)); printf(" (0x0000ff0000000000) = "); printf("%s\n", core_i64_hex_format(result64 = read_qword_unaligned(address, 0x0000ff0000000000U), 16)); assert((result64 & 0x0000ff0000000000U) == (expected64 & 0x0000ff0000000000U)); printf(" (0x000000ff00000000) = "); printf("%s\n", core_i64_hex_format(result64 = read_qword_unaligned(address, 0x000000ff00000000U), 16)); assert((result64 & 0x000000ff00000000U) == (expected64 & 0x000000ff00000000U)); printf(" (0x00000000ff000000) = "); printf("%s\n", core_i64_hex_format(result64 = read_qword_unaligned(address, 0x00000000ff000000U), 16)); assert((result64 & 0x00000000ff000000U) == (expected64 & 0x00000000ff000000U)); printf(" (0x0000000000ff0000) = "); printf("%s\n", core_i64_hex_format(result64 = read_qword_unaligned(address, 0x0000000000ff0000U), 16)); assert((result64 & 0x0000000000ff0000U) == (expected64 & 0x0000000000ff0000U)); printf(" (0x000000000000ff00) = "); printf("%s\n", core_i64_hex_format(result64 = read_qword_unaligned(address, 0x000000000000ff00U), 16)); assert((result64 & 0x000000000000ff00U) == (expected64 & 0x000000000000ff00U)); printf(" (0x00000000000000ff) = "); printf("%s\n", core_i64_hex_format(result64 = read_qword_unaligned(address, 0x00000000000000ffU), 16)); assert((result64 & 0x00000000000000ffU) == (expected64 & 0x00000000000000ffU)); printf(" (0xffff000000000000) = "); printf("%s\n", core_i64_hex_format(result64 = read_qword_unaligned(address, 0xffff000000000000U), 16)); assert((result64 & 0xffff000000000000U) == (expected64 & 0xffff000000000000U)); printf(" (0x0000ffff00000000) = "); printf("%s\n", core_i64_hex_format(result64 = read_qword_unaligned(address, 0x0000ffff00000000U), 16)); assert((result64 & 0x0000ffff00000000U) == (expected64 & 0x0000ffff00000000U)); printf(" (0x00000000ffff0000) = "); printf("%s\n", core_i64_hex_format(result64 = read_qword_unaligned(address, 0x00000000ffff0000U), 16)); assert((result64 & 0x00000000ffff0000U) == (expected64 & 0x00000000ffff0000U)); printf(" (0x000000000000ffff) = "); printf("%s\n", core_i64_hex_format(result64 = read_qword_unaligned(address, 0x000000000000ffffU), 16)); assert((result64 & 0x000000000000ffffU) == (expected64 & 0x000000000000ffffU)); printf(" (0xffffff0000000000) = "); printf("%s\n", core_i64_hex_format(result64 = read_qword_unaligned(address, 0xffffff0000000000U), 16)); assert((result64 & 0xffffff0000000000U) == (expected64 & 0xffffff0000000000U)); printf(" (0x0000ffffff000000) = "); printf("%s\n", core_i64_hex_format(result64 = read_qword_unaligned(address, 0x0000ffffff000000U), 16)); assert((result64 & 0x0000ffffff000000U) == (expected64 & 0x0000ffffff000000U)); printf(" (0x000000ffffff0000) = "); printf("%s\n", core_i64_hex_format(result64 = read_qword_unaligned(address, 0x000000ffffff0000U), 16)); assert((result64 & 0x000000ffffff0000U) == (expected64 & 0x000000ffffff0000U)); printf(" (0x0000000000ffffff) = "); printf("%s\n", core_i64_hex_format(result64 = read_qword_unaligned(address, 0x0000000000ffffffU), 16)); assert((result64 & 0x0000000000ffffffU) == (expected64 & 0x0000000000ffffffU)); printf(" (0xffffffff00000000) = "); printf("%s\n", core_i64_hex_format(result64 = read_qword_unaligned(address, 0xffffffff00000000U), 16)); assert((result64 & 0xffffffff00000000U) == (expected64 & 0xffffffff00000000U)); printf(" (0x00ffffffff000000) = "); printf("%s\n", core_i64_hex_format(result64 = read_qword_unaligned(address, 0x00ffffffff000000U), 16)); assert((result64 & 0x00ffffffff000000U) == (expected64 & 0x00ffffffff000000U)); printf(" (0x0000ffffffff0000) = "); printf("%s\n", core_i64_hex_format(result64 = read_qword_unaligned(address, 0x0000ffffffff0000U), 16)); assert((result64 & 0x0000ffffffff0000U) == (expected64 & 0x0000ffffffff0000U)); printf(" (0x000000ffffffff00) = "); printf("%s\n", core_i64_hex_format(result64 = read_qword_unaligned(address, 0x000000ffffffff00U), 16)); assert((result64 & 0x000000ffffffff00U) == (expected64 & 0x000000ffffffff00U)); printf(" (0x00000000ffffffff) = "); printf("%s\n", core_i64_hex_format(result64 = read_qword_unaligned(address, 0x00000000ffffffffU), 16)); assert((result64 & 0x00000000ffffffffU) == (expected64 & 0x00000000ffffffffU)); printf(" (0xffffffffff000000) = "); printf("%s\n", core_i64_hex_format(result64 = read_qword_unaligned(address, 0xffffffffff000000U), 16)); assert((result64 & 0xffffffffff000000U) == (expected64 & 0xffffffffff000000U)); printf(" (0x00ffffffffff0000) = "); printf("%s\n", core_i64_hex_format(result64 = read_qword_unaligned(address, 0x00ffffffffff0000U), 16)); assert((result64 & 0x00ffffffffff0000U) == (expected64 & 0x00ffffffffff0000U)); printf(" (0x0000ffffffffff00) = "); printf("%s\n", core_i64_hex_format(result64 = read_qword_unaligned(address, 0x0000ffffffffff00U), 16)); assert((result64 & 0x0000ffffffffff00U) == (expected64 & 0x0000ffffffffff00U)); printf(" (0x000000ffffffffff) = "); printf("%s\n", core_i64_hex_format(result64 = read_qword_unaligned(address, 0x000000ffffffffffU), 16)); assert((result64 & 0x000000ffffffffffU) == (expected64 & 0x000000ffffffffffU)); printf(" (0xffffffffffff0000) = "); printf("%s\n", core_i64_hex_format(result64 = read_qword_unaligned(address, 0xffffffffffff0000U), 16)); assert((result64 & 0xffffffffffff0000U) == (expected64 & 0xffffffffffff0000U)); printf(" (0x00ffffffffffff00) = "); printf("%s\n", core_i64_hex_format(result64 = read_qword_unaligned(address, 0x00ffffffffffff00U), 16)); assert((result64 & 0x00ffffffffffff00U) == (expected64 & 0x00ffffffffffff00U)); printf(" (0x0000ffffffffffff) = "); printf("%s\n", core_i64_hex_format(result64 = read_qword_unaligned(address, 0x0000ffffffffffffU), 16)); assert((result64 & 0x0000ffffffffffffU) == (expected64 & 0x0000ffffffffffffU)); printf(" (0xffffffffffffff00) = "); printf("%s\n", core_i64_hex_format(result64 = read_qword_unaligned(address, 0xffffffffffffff00U), 16)); assert((result64 & 0xffffffffffffff00U) == (expected64 & 0xffffffffffffff00U)); printf(" (0x00ffffffffffffff) = "); printf("%s\n", core_i64_hex_format(result64 = read_qword_unaligned(address, 0x00ffffffffffffffU), 16)); assert((result64 & 0x00ffffffffffffffU) == (expected64 & 0x00ffffffffffffffU)); } #endif } // accessors virtual address_table_read &read() override { return m_read; } virtual address_table_write &write() override { return m_write; } virtual address_table_setoffset &setoffset() override { return m_setoffset; } // watchpoint control virtual void enable_read_watchpoints(bool enable = true) override { m_read.enable_watchpoints(enable); } virtual void enable_write_watchpoints(bool enable = true) override { m_write.enable_watchpoints(enable); } // generate accessor table virtual void accessors(data_accessors &accessors) const override { accessors.read_byte = reinterpret_cast(&read_byte_static); accessors.read_word = reinterpret_cast(&read_word_static); accessors.read_word_masked = reinterpret_cast(&read_word_masked_static); accessors.read_dword = reinterpret_cast(&read_dword_static); accessors.read_dword_masked = reinterpret_cast(&read_dword_masked_static); accessors.read_qword = reinterpret_cast(&read_qword_static); accessors.read_qword_masked = reinterpret_cast(&read_qword_masked_static); accessors.write_byte = reinterpret_cast(&write_byte_static); accessors.write_word = reinterpret_cast(&write_word_static); accessors.write_word_masked = reinterpret_cast(&write_word_masked_static); accessors.write_dword = reinterpret_cast(&write_dword_static); accessors.write_dword_masked = reinterpret_cast(&write_dword_masked_static); accessors.write_qword = reinterpret_cast(&write_qword_static); accessors.write_qword_masked = reinterpret_cast(&write_qword_masked_static); } // return a pointer to the read bank, or nullptr if none virtual void *get_read_ptr(offs_t address) override { // perform the lookup address &= m_addrmask; u32 entry = read_lookup(address); const handler_entry_read &handler = m_read.handler_read(entry); // 8-bit case: RAM/ROM if (entry > STATIC_BANKMAX) return nullptr; return handler.ramptr(handler.offset(address)); } // return a pointer to the write bank, or nullptr if none virtual void *get_write_ptr(offs_t address) override { // perform the lookup address &= m_addrmask; u32 entry = write_lookup(address); const handler_entry_write &handler = m_write.handler_write(entry); // 8-bit case: RAM/ROM if (entry > STATIC_BANKMAX) return nullptr; return handler.ramptr(handler.offset(address)); } // native read NativeType read_native(offs_t offset, NativeType mask) { g_profiler.start(PROFILER_MEMREAD); if (TEST_HANDLER) printf("[r%X,%s]", offset, core_i64_hex_format(mask, sizeof(NativeType) * 2)); // look up the handler offs_t address = offset & m_addrmask; u32 entry = read_lookup(address); const handler_entry_read &handler = m_read.handler_read(entry); // either read directly from RAM, or call the delegate offset = offset_to_byte(handler.offset(address)); NativeType result; if (entry <= STATIC_BANKMAX) result = *reinterpret_cast(handler.ramptr(offset)); else if (sizeof(NativeType) == 1) result = handler.read8(*this, offset, mask); else if (sizeof(NativeType) == 2) result = handler.read16(*this, offset >> 1, mask); else if (sizeof(NativeType) == 4) result = handler.read32(*this, offset >> 2, mask); else if (sizeof(NativeType) == 8) result = handler.read64(*this, offset >> 3, mask); g_profiler.stop(); return result; } // mask-less native read NativeType read_native(offs_t offset) { g_profiler.start(PROFILER_MEMREAD); if (TEST_HANDLER) printf("[r%X]", offset); // look up the handler offs_t address = offset & m_addrmask; u32 entry = read_lookup(address); const handler_entry_read &handler = m_read.handler_read(entry); // either read directly from RAM, or call the delegate offset = offset_to_byte(handler.offset(address)); NativeType result; if (entry <= STATIC_BANKMAX) result = *reinterpret_cast(handler.ramptr(offset)); else if (sizeof(NativeType) == 1) result = handler.read8(*this, offset, 0xff); else if (sizeof(NativeType) == 2) result = handler.read16(*this, offset >> 1, 0xffff); else if (sizeof(NativeType) == 4) result = handler.read32(*this, offset >> 2, 0xffffffff); else if (sizeof(NativeType) == 8) result = handler.read64(*this, offset >> 3, 0xffffffffffffffffU); g_profiler.stop(); return result; } // native write void write_native(offs_t offset, NativeType data, NativeType mask) { g_profiler.start(PROFILER_MEMWRITE); // look up the handler offs_t address = offset & m_addrmask; u32 entry = write_lookup(address); const handler_entry_write &handler = m_write.handler_write(entry); // either write directly to RAM, or call the delegate offset = offset_to_byte(handler.offset(address)); if (entry <= STATIC_BANKMAX) { NativeType *dest = reinterpret_cast(handler.ramptr(offset)); *dest = (*dest & ~mask) | (data & mask); } else if (sizeof(NativeType) == 1) handler.write8(*this, offset, data, mask); else if (sizeof(NativeType) == 2) handler.write16(*this, offset >> 1, data, mask); else if (sizeof(NativeType) == 4) handler.write32(*this, offset >> 2, data, mask); else if (sizeof(NativeType) == 8) handler.write64(*this, offset >> 3, data, mask); g_profiler.stop(); } // mask-less native write void write_native(offs_t offset, NativeType data) { g_profiler.start(PROFILER_MEMWRITE); // look up the handler offs_t address = offset & m_addrmask; u32 entry = write_lookup(address); const handler_entry_write &handler = m_write.handler_write(entry); // either write directly to RAM, or call the delegate offset = offset_to_byte(handler.offset(address)); if (entry <= STATIC_BANKMAX) *reinterpret_cast(handler.ramptr(offset)) = data; else if (sizeof(NativeType) == 1) handler.write8(*this, offset, data, 0xff); else if (sizeof(NativeType) == 2) handler.write16(*this, offset >> 1, data, 0xffff); else if (sizeof(NativeType) == 4) handler.write32(*this, offset >> 2, data, 0xffffffff); else if (sizeof(NativeType) == 8) handler.write64(*this, offset >> 3, data, 0xffffffffffffffffU); g_profiler.stop(); } // generic direct read template TargetType read_direct(offs_t address, TargetType mask) { const u32 TARGET_BYTES = sizeof(TargetType); const u32 TARGET_BITS = 8 * TARGET_BYTES; // equal to native size and aligned; simple pass-through to the native reader if (NATIVE_BYTES == TARGET_BYTES && (Aligned || (address & NATIVE_MASK) == 0)) return read_native(address & ~NATIVE_MASK, mask); // if native size is larger, see if we can do a single masked read (guaranteed if we're aligned) if (NATIVE_BYTES > TARGET_BYTES) { u32 offsbits = 8 * (offset_to_byte(address) & (NATIVE_BYTES - (Aligned ? TARGET_BYTES : 1))); if (Aligned || (offsbits + TARGET_BITS <= NATIVE_BITS)) { if (Endian != ENDIANNESS_LITTLE) offsbits = NATIVE_BITS - TARGET_BITS - offsbits; return read_native(address & ~NATIVE_MASK, (NativeType)mask << offsbits) >> offsbits; } } // determine our alignment against the native boundaries, and mask the address u32 offsbits = 8 * (offset_to_byte(address) & (NATIVE_BYTES - 1)); address &= ~NATIVE_MASK; // if we're here, and native size is larger or equal to the target, we need exactly 2 reads if (NATIVE_BYTES >= TARGET_BYTES) { // little-endian case if (Endian == ENDIANNESS_LITTLE) { // read lower bits from lower address TargetType result = 0; NativeType curmask = (NativeType)mask << offsbits; if (curmask != 0) result = read_native(address, curmask) >> offsbits; // read upper bits from upper address offsbits = NATIVE_BITS - offsbits; curmask = mask >> offsbits; if (curmask != 0) result |= read_native(address + NATIVE_STEP, curmask) << offsbits; return result; } // big-endian case else { // left-justify the mask to the target type const u32 LEFT_JUSTIFY_TARGET_TO_NATIVE_SHIFT = ((NATIVE_BITS >= TARGET_BITS) ? (NATIVE_BITS - TARGET_BITS) : 0); NativeType result = 0; NativeType ljmask = (NativeType)mask << LEFT_JUSTIFY_TARGET_TO_NATIVE_SHIFT; NativeType curmask = ljmask >> offsbits; // read upper bits from lower address if (curmask != 0) result = read_native(address, curmask) << offsbits; offsbits = NATIVE_BITS - offsbits; // read lower bits from upper address curmask = ljmask << offsbits; if (curmask != 0) result |= read_native(address + NATIVE_STEP, curmask) >> offsbits; // return the un-justified result return result >> LEFT_JUSTIFY_TARGET_TO_NATIVE_SHIFT; } } // if we're here, then we have 2 or more reads needed to get our final result else { // compute the maximum number of loops; we do it this way so that there are // a fixed number of loops for the compiler to unroll if it desires const u32 MAX_SPLITS_MINUS_ONE = TARGET_BYTES / NATIVE_BYTES - 1; TargetType result = 0; // little-endian case if (Endian == ENDIANNESS_LITTLE) { // read lowest bits from first address NativeType curmask = mask << offsbits; if (curmask != 0) result = read_native(address, curmask) >> offsbits; // read middle bits from subsequent addresses offsbits = NATIVE_BITS - offsbits; for (u32 index = 0; index < MAX_SPLITS_MINUS_ONE; index++) { address += NATIVE_STEP; curmask = mask >> offsbits; if (curmask != 0) result |= (TargetType)read_native(address, curmask) << offsbits; offsbits += NATIVE_BITS; } // if we're not aligned and we still have bits left, read uppermost bits from last address if (!Aligned && offsbits < TARGET_BITS) { curmask = mask >> offsbits; if (curmask != 0) result |= (TargetType)read_native(address + NATIVE_STEP, curmask) << offsbits; } } // big-endian case else { // read highest bits from first address offsbits = TARGET_BITS - (NATIVE_BITS - offsbits); NativeType curmask = mask >> offsbits; if (curmask != 0) result = (TargetType)read_native(address, curmask) << offsbits; // read middle bits from subsequent addresses for (u32 index = 0; index < MAX_SPLITS_MINUS_ONE; index++) { offsbits -= NATIVE_BITS; address += NATIVE_STEP; curmask = mask >> offsbits; if (curmask != 0) result |= (TargetType)read_native(address, curmask) << offsbits; } // if we're not aligned and we still have bits left, read lowermost bits from the last address if (!Aligned && offsbits != 0) { offsbits = NATIVE_BITS - offsbits; curmask = mask << offsbits; if (curmask != 0) result |= read_native(address + NATIVE_STEP, curmask) >> offsbits; } } return result; } } // generic direct write template void write_direct(offs_t address, TargetType data, TargetType mask) { const u32 TARGET_BYTES = sizeof(TargetType); const u32 TARGET_BITS = 8 * TARGET_BYTES; // equal to native size and aligned; simple pass-through to the native writer if (NATIVE_BYTES == TARGET_BYTES && (Aligned || (address & NATIVE_MASK) == 0)) return write_native(address & ~NATIVE_MASK, data, mask); // if native size is larger, see if we can do a single masked write (guaranteed if we're aligned) if (NATIVE_BYTES > TARGET_BYTES) { u32 offsbits = 8 * (offset_to_byte(address) & (NATIVE_BYTES - (Aligned ? TARGET_BYTES : 1))); if (Aligned || (offsbits + TARGET_BITS <= NATIVE_BITS)) { if (Endian != ENDIANNESS_LITTLE) offsbits = NATIVE_BITS - TARGET_BITS - offsbits; return write_native(address & ~NATIVE_MASK, (NativeType)data << offsbits, (NativeType)mask << offsbits); } } // determine our alignment against the native boundaries, and mask the address u32 offsbits = 8 * (offset_to_byte(address) & (NATIVE_BYTES - 1)); address &= ~NATIVE_MASK; // if we're here, and native size is larger or equal to the target, we need exactly 2 writes if (NATIVE_BYTES >= TARGET_BYTES) { // little-endian case if (Endian == ENDIANNESS_LITTLE) { // write lower bits to lower address NativeType curmask = (NativeType)mask << offsbits; if (curmask != 0) write_native(address, (NativeType)data << offsbits, curmask); // write upper bits to upper address offsbits = NATIVE_BITS - offsbits; curmask = mask >> offsbits; if (curmask != 0) write_native(address + NATIVE_STEP, data >> offsbits, curmask); } // big-endian case else { // left-justify the mask and data to the target type const u32 LEFT_JUSTIFY_TARGET_TO_NATIVE_SHIFT = ((NATIVE_BITS >= TARGET_BITS) ? (NATIVE_BITS - TARGET_BITS) : 0); NativeType ljdata = (NativeType)data << LEFT_JUSTIFY_TARGET_TO_NATIVE_SHIFT; NativeType ljmask = (NativeType)mask << LEFT_JUSTIFY_TARGET_TO_NATIVE_SHIFT; // write upper bits to lower address NativeType curmask = ljmask >> offsbits; if (curmask != 0) write_native(address, ljdata >> offsbits, curmask); // write lower bits to upper address offsbits = NATIVE_BITS - offsbits; curmask = ljmask << offsbits; if (curmask != 0) write_native(address + NATIVE_STEP, ljdata << offsbits, curmask); } } // if we're here, then we have 2 or more writes needed to get our final result else { // compute the maximum number of loops; we do it this way so that there are // a fixed number of loops for the compiler to unroll if it desires const u32 MAX_SPLITS_MINUS_ONE = TARGET_BYTES / NATIVE_BYTES - 1; // little-endian case if (Endian == ENDIANNESS_LITTLE) { // write lowest bits to first address NativeType curmask = mask << offsbits; if (curmask != 0) write_native(address, data << offsbits, curmask); // write middle bits to subsequent addresses offsbits = NATIVE_BITS - offsbits; for (u32 index = 0; index < MAX_SPLITS_MINUS_ONE; index++) { address += NATIVE_STEP; curmask = mask >> offsbits; if (curmask != 0) write_native(address, data >> offsbits, curmask); offsbits += NATIVE_BITS; } // if we're not aligned and we still have bits left, write uppermost bits to last address if (!Aligned && offsbits < TARGET_BITS) { curmask = mask >> offsbits; if (curmask != 0) write_native(address + NATIVE_STEP, data >> offsbits, curmask); } } // big-endian case else { // write highest bits to first address offsbits = TARGET_BITS - (NATIVE_BITS - offsbits); NativeType curmask = mask >> offsbits; if (curmask != 0) write_native(address, data >> offsbits, curmask); // write middle bits to subsequent addresses for (u32 index = 0; index < MAX_SPLITS_MINUS_ONE; index++) { offsbits -= NATIVE_BITS; address += NATIVE_STEP; curmask = mask >> offsbits; if (curmask != 0) write_native(address, data >> offsbits, curmask); } // if we're not aligned and we still have bits left, write lowermost bits to the last address if (!Aligned && offsbits != 0) { offsbits = NATIVE_BITS - offsbits; curmask = mask << offsbits; if (curmask != 0) write_native(address + NATIVE_STEP, data << offsbits, curmask); } } } } // Allows to announce a pending read or write operation on this address. // The user of the address_space calls a set_address operation which leads // to some particular set_offset operation for an entry in the address map. void set_address(offs_t address) override { address &= m_addrmask; u32 entry = setoffset_lookup(address); const handler_entry_setoffset &handler = m_setoffset.handler_setoffset(entry); offs_t offset = handler.offset(address); handler.setoffset(*this, offset / sizeof(NativeType)); } // virtual access to these functions u8 read_byte(offs_t address) override { return (NATIVE_BITS == 8) ? read_native(address & ~NATIVE_MASK) : read_direct(address, 0xff); } u16 read_word(offs_t address) override { return (NATIVE_BITS == 16) ? read_native(address & ~NATIVE_MASK) : read_direct(address, 0xffff); } u16 read_word(offs_t address, u16 mask) override { return read_direct(address, mask); } u16 read_word_unaligned(offs_t address) override { return read_direct(address, 0xffff); } u16 read_word_unaligned(offs_t address, u16 mask) override { return read_direct(address, mask); } u32 read_dword(offs_t address) override { return (NATIVE_BITS == 32) ? read_native(address & ~NATIVE_MASK) : read_direct(address, 0xffffffff); } u32 read_dword(offs_t address, u32 mask) override { return read_direct(address, mask); } u32 read_dword_unaligned(offs_t address) override { return read_direct(address, 0xffffffff); } u32 read_dword_unaligned(offs_t address, u32 mask) override { return read_direct(address, mask); } u64 read_qword(offs_t address) override { return (NATIVE_BITS == 64) ? read_native(address & ~NATIVE_MASK) : read_direct(address, 0xffffffffffffffffU); } u64 read_qword(offs_t address, u64 mask) override { return read_direct(address, mask); } u64 read_qword_unaligned(offs_t address) override { return read_direct(address, 0xffffffffffffffffU); } u64 read_qword_unaligned(offs_t address, u64 mask) override { return read_direct(address, mask); } void write_byte(offs_t address, u8 data) override { if (NATIVE_BITS == 8) write_native(address & ~NATIVE_MASK, data); else write_direct(address, data, 0xff); } void write_word(offs_t address, u16 data) override { if (NATIVE_BITS == 16) write_native(address & ~NATIVE_MASK, data); else write_direct(address, data, 0xffff); } void write_word(offs_t address, u16 data, u16 mask) override { write_direct(address, data, mask); } void write_word_unaligned(offs_t address, u16 data) override { write_direct(address, data, 0xffff); } void write_word_unaligned(offs_t address, u16 data, u16 mask) override { write_direct(address, data, mask); } void write_dword(offs_t address, u32 data) override { if (NATIVE_BITS == 32) write_native(address & ~NATIVE_MASK, data); else write_direct(address, data, 0xffffffff); } void write_dword(offs_t address, u32 data, u32 mask) override { write_direct(address, data, mask); } void write_dword_unaligned(offs_t address, u32 data) override { write_direct(address, data, 0xffffffff); } void write_dword_unaligned(offs_t address, u32 data, u32 mask) override { write_direct(address, data, mask); } void write_qword(offs_t address, u64 data) override { if (NATIVE_BITS == 64) write_native(address & ~NATIVE_MASK, data); else write_direct(address, data, 0xffffffffffffffffU); } void write_qword(offs_t address, u64 data, u64 mask) override { write_direct(address, data, mask); } void write_qword_unaligned(offs_t address, u64 data) override { write_direct(address, data, 0xffffffffffffffffU); } void write_qword_unaligned(offs_t address, u64 data, u64 mask) override { write_direct(address, data, mask); } // static access to these functions static u8 read_byte_static(this_type &space, offs_t address) { return (NATIVE_BITS == 8) ? space.read_native(address & ~NATIVE_MASK) : space.read_direct(address, 0xff); } static u16 read_word_static(this_type &space, offs_t address) { return (NATIVE_BITS == 16) ? space.read_native(address & ~NATIVE_MASK) : space.read_direct(address, 0xffff); } static u16 read_word_masked_static(this_type &space, offs_t address, u16 mask) { return space.read_direct(address, mask); } static u32 read_dword_static(this_type &space, offs_t address) { return (NATIVE_BITS == 32) ? space.read_native(address & ~NATIVE_MASK) : space.read_direct(address, 0xffffffff); } static u32 read_dword_masked_static(this_type &space, offs_t address, u32 mask) { return space.read_direct(address, mask); } static u64 read_qword_static(this_type &space, offs_t address) { return (NATIVE_BITS == 64) ? space.read_native(address & ~NATIVE_MASK) : space.read_direct(address, 0xffffffffffffffffU); } static u64 read_qword_masked_static(this_type &space, offs_t address, u64 mask) { return space.read_direct(address, mask); } static void write_byte_static(this_type &space, offs_t address, u8 data) { if (NATIVE_BITS == 8) space.write_native(address & ~NATIVE_MASK, data); else space.write_direct(address, data, 0xff); } static void write_word_static(this_type &space, offs_t address, u16 data) { if (NATIVE_BITS == 16) space.write_native(address & ~NATIVE_MASK, data); else space.write_direct(address, data, 0xffff); } static void write_word_masked_static(this_type &space, offs_t address, u16 data, u16 mask) { space.write_direct(address, data, mask); } static void write_dword_static(this_type &space, offs_t address, u32 data) { if (NATIVE_BITS == 32) space.write_native(address & ~NATIVE_MASK, data); else space.write_direct(address, data, 0xffffffff); } static void write_dword_masked_static(this_type &space, offs_t address, u32 data, u32 mask) { space.write_direct(address, data, mask); } static void write_qword_static(this_type &space, offs_t address, u64 data) { if (NATIVE_BITS == 64) space.write_native(address & ~NATIVE_MASK, data); else space.write_direct(address, data, 0xffffffffffffffffU); } static void write_qword_masked_static(this_type &space, offs_t address, u64 data, u64 mask) { space.write_direct(address, data, mask); } address_table_read m_read; // memory read lookup table address_table_write m_write; // memory write lookup table address_table_setoffset m_setoffset; // memory setoffset lookup table }; typedef address_space_specific address_space_8_8le_small; typedef address_space_specific address_space_8_8be_small; typedef address_space_specific address_space_16_1le_small; typedef address_space_specific address_space_16_1be_small; typedef address_space_specific address_space_16_8le_small; typedef address_space_specific address_space_16_8be_small; typedef address_space_specific address_space_16_16le_small; typedef address_space_specific address_space_16_16be_small; typedef address_space_specific address_space_32_8le_small; typedef address_space_specific address_space_32_8be_small; typedef address_space_specific address_space_32_16le_small; typedef address_space_specific address_space_32_16be_small; typedef address_space_specific address_space_32_32le_small; typedef address_space_specific address_space_32_32be_small; typedef address_space_specific address_space_64_8le_small; typedef address_space_specific address_space_64_8be_small; typedef address_space_specific address_space_64_16le_small; typedef address_space_specific address_space_64_16be_small; typedef address_space_specific address_space_64_32le_small; typedef address_space_specific address_space_64_32be_small; typedef address_space_specific address_space_64_64le_small; typedef address_space_specific address_space_64_64be_small; typedef address_space_specific address_space_8_8le_large; typedef address_space_specific address_space_8_8be_large; typedef address_space_specific address_space_16_1le_large; typedef address_space_specific address_space_16_1be_large; typedef address_space_specific address_space_16_8le_large; typedef address_space_specific address_space_16_8be_large; typedef address_space_specific address_space_16_16le_large; typedef address_space_specific address_space_16_16be_large; typedef address_space_specific address_space_32_8le_large; typedef address_space_specific address_space_32_8be_large; typedef address_space_specific address_space_32_16le_large; typedef address_space_specific address_space_32_16be_large; typedef address_space_specific address_space_32_32le_large; typedef address_space_specific address_space_32_32be_large; typedef address_space_specific address_space_64_8le_large; typedef address_space_specific address_space_64_8be_large; typedef address_space_specific address_space_64_16le_large; typedef address_space_specific address_space_64_16be_large; typedef address_space_specific address_space_64_32le_large; typedef address_space_specific address_space_64_32be_large; typedef address_space_specific address_space_64_64le_large; typedef address_space_specific address_space_64_64be_large; //************************************************************************** // GLOBAL VARIABLES //************************************************************************** // global watchpoint table u16 address_table::s_watchpoint_table[1 << LEVEL1_BITS]; //************************************************************************** // FUNCTION PROTOTYPES //************************************************************************** // debugging static void generate_memdump(running_machine &machine); //************************************************************************** // MEMORY MANAGER //************************************************************************** //------------------------------------------------- // memory_manager - constructor //------------------------------------------------- memory_manager::memory_manager(running_machine &machine) : m_machine(machine), m_initialized(false), m_banknext(STATIC_BANK1) { memset(m_bank_ptr, 0, sizeof(m_bank_ptr)); } //------------------------------------------------- // allocate - allocate memory spaces //------------------------------------------------- void memory_manager::allocate(device_memory_interface &memory) { for (int spacenum = 0; spacenum < memory.max_space_count(); ++spacenum) { // if there is a configuration for this space, we need an address space address_space_config const *const spaceconfig = memory.space_config(spacenum); if (spaceconfig) { // allocate one of the appropriate type bool const large(spaceconfig->addr2byte_end(0xffffffffUL >> (32 - spaceconfig->m_addr_width)) >= (1 << 18)); switch (spaceconfig->data_width()) { case 8: if (spaceconfig->endianness() == ENDIANNESS_LITTLE) { if (large) memory.allocate(*this, spacenum); else memory.allocate(*this, spacenum); } else { if (large) memory.allocate(*this, spacenum); else memory.allocate(*this, spacenum); } break; case 16: switch (spaceconfig->addr_shift()) { case 3: if (spaceconfig->endianness() == ENDIANNESS_LITTLE) { if (large) memory.allocate(*this, spacenum); else memory.allocate(*this, spacenum); } else { if (large) memory.allocate(*this, spacenum); else memory.allocate(*this, spacenum); } break; case 0: if (spaceconfig->endianness() == ENDIANNESS_LITTLE) { if (large) memory.allocate(*this, spacenum); else memory.allocate(*this, spacenum); } else { if (large) memory.allocate(*this, spacenum); else memory.allocate(*this, spacenum); } break; case -1: if (spaceconfig->endianness() == ENDIANNESS_LITTLE) { if (large) memory.allocate(*this, spacenum); else memory.allocate(*this, spacenum); } else { if (large) memory.allocate(*this, spacenum); else memory.allocate(*this, spacenum); } break; } break; case 32: switch (spaceconfig->addr_shift()) { case 0: if (spaceconfig->endianness() == ENDIANNESS_LITTLE) { if (large) memory.allocate(*this, spacenum); else memory.allocate(*this, spacenum); } else { if (large) memory.allocate(*this, spacenum); else memory.allocate(*this, spacenum); } break; case -1: if (spaceconfig->endianness() == ENDIANNESS_LITTLE) { if (large) memory.allocate(*this, spacenum); else memory.allocate(*this, spacenum); } else { if (large) memory.allocate(*this, spacenum); else memory.allocate(*this, spacenum); } break; case -2: if (spaceconfig->endianness() == ENDIANNESS_LITTLE) { if (large) memory.allocate(*this, spacenum); else memory.allocate(*this, spacenum); } else { if (large) memory.allocate(*this, spacenum); else memory.allocate(*this, spacenum); } break; } break; case 64: switch (spaceconfig->addr_shift()) { case 0: if (spaceconfig->endianness() == ENDIANNESS_LITTLE) { if (large) memory.allocate(*this, spacenum); else memory.allocate(*this, spacenum); } else { if (large) memory.allocate(*this, spacenum); else memory.allocate(*this, spacenum); } break; case -1: if (spaceconfig->endianness() == ENDIANNESS_LITTLE) { if (large) memory.allocate(*this, spacenum); else memory.allocate(*this, spacenum); } else { if (large) memory.allocate(*this, spacenum); else memory.allocate(*this, spacenum); } break; case -2: if (spaceconfig->endianness() == ENDIANNESS_LITTLE) { if (large) memory.allocate(*this, spacenum); else memory.allocate(*this, spacenum); } else { if (large) memory.allocate(*this, spacenum); else memory.allocate(*this, spacenum); } break; case -3: if (spaceconfig->endianness() == ENDIANNESS_LITTLE) { if (large) memory.allocate(*this, spacenum); else memory.allocate(*this, spacenum); } else { if (large) memory.allocate(*this, spacenum); else memory.allocate(*this, spacenum); } break; } break; default: throw emu_fatalerror("Invalid width %d specified for address_space::allocate", spaceconfig->data_width()); } } } } //------------------------------------------------- // initialize - initialize the memory system //------------------------------------------------- void memory_manager::initialize() { // loop over devices and spaces within each device memory_interface_iterator iter(machine().root_device()); std::vector memories; for (device_memory_interface &memory : iter) { memories.push_back(&memory); allocate(memory); } allocate(m_machine.m_dummy_space); // construct and preprocess the address_map for each space for (auto const memory : memories) memory->prepare_maps(); // create the handlers from the resulting address maps for (auto const memory : memories) memory->populate_from_maps(); // allocate memory needed to back each address space for (auto const memory : memories) memory->allocate_memory(); // find all the allocated pointers for (auto const memory : memories) memory->locate_memory(); // disable logging of unmapped access when no one receives it if (!machine().options().log() && !machine().options().oslog() && !(machine().debug_flags & DEBUG_FLAG_ENABLED)) for (auto const memory : memories) memory->set_log_unmap(false); // register a callback to reset banks when reloading state machine().save().register_postload(save_prepost_delegate(FUNC(memory_manager::bank_reattach), this)); // dump the final memory configuration generate_memdump(machine()); // we are now initialized m_initialized = true; } //------------------------------------------------- // region_alloc - allocates memory for a region //------------------------------------------------- memory_region *memory_manager::region_alloc(const char *name, u32 length, u8 width, endianness_t endian) { osd_printf_verbose("Region '%s' created\n", name); // make sure we don't have a region of the same name; also find the end of the list if (m_regionlist.find(name) != m_regionlist.end()) fatalerror("region_alloc called with duplicate region name \"%s\"\n", name); // allocate the region m_regionlist.emplace(name, std::make_unique(machine(), name, length, width, endian)); return m_regionlist.find(name)->second.get(); } //------------------------------------------------- // region_free - releases memory for a region //------------------------------------------------- void memory_manager::region_free(const char *name) { m_regionlist.erase(name); } //------------------------------------------------- // region_containing - helper to determine if // a block of memory is part of a region //------------------------------------------------- memory_region *memory_manager::region_containing(const void *memory, offs_t bytes) const { const u8 *data = reinterpret_cast(memory); // look through the region list and return the first match for (auto ®ion : m_regionlist) if (data >= region.second->base() && (data + bytes) < region.second->end()) return region.second.get(); // didn't find one return nullptr; } //------------------------------------------------- // generate_memdump - internal memory dump //------------------------------------------------- static void generate_memdump(running_machine &machine) { if (MEM_DUMP) { FILE *file = fopen("memdump.log", "w"); if (file) { memory_interface_iterator iter(machine.root_device()); for (device_memory_interface &memory : iter) memory.dump(file); fclose(file); } } } //------------------------------------------------- // bank_reattach - reconnect banks after a load //------------------------------------------------- void memory_manager::bank_reattach() { // for each non-anonymous bank, explicitly reset its entry for (auto &bank : m_banklist) if (!bank.second->anonymous() && bank.second->entry() != BANK_ENTRY_UNSPECIFIED) bank.second->set_entry(bank.second->entry()); } //************************************************************************** // ADDRESS SPACE //************************************************************************** //------------------------------------------------- // address_space - constructor //------------------------------------------------- address_space::address_space(memory_manager &manager, device_memory_interface &memory, int spacenum, bool large) : m_config(*memory.space_config(spacenum)), m_device(memory.device()), m_addrmask(0xffffffffUL >> (32 - m_config.m_addr_width)), m_logaddrmask(0xffffffffUL >> (32 - m_config.m_logaddr_width)), m_unmap(0), m_spacenum(spacenum), m_log_unmap(true), m_name(memory.space_config(spacenum)->name()), m_addrchars((m_config.m_addr_width + 3) / 4), m_logaddrchars((m_config.m_logaddr_width + 3) / 4), m_manager(manager), m_machine(memory.device().machine()) { switch(m_config.addr_shift()) { case 3: m_direct = static_cast(new direct_read_data< 3>(*this)); break; case 0: m_direct = static_cast(new direct_read_data< 0>(*this)); break; case -1: m_direct = static_cast(new direct_read_data<-1>(*this)); break; case -2: m_direct = static_cast(new direct_read_data<-2>(*this)); break; case -3: m_direct = static_cast(new direct_read_data<-3>(*this)); break; default: fatalerror("Unsupported address shift %d\n", m_config.addr_shift()); } } //------------------------------------------------- // ~address_space - destructor //------------------------------------------------- address_space::~address_space() { switch(m_config.addr_shift()) { case 3: delete static_cast *>(m_direct); break; case 0: delete static_cast *>(m_direct); break; case -1: delete static_cast *>(m_direct); break; case -2: delete static_cast *>(m_direct); break; case -3: delete static_cast *>(m_direct); break; } } //------------------------------------------------- // adjust_addresses - adjust addresses for a // given address space in a standard fashion //------------------------------------------------- inline void address_space::adjust_addresses(offs_t &start, offs_t &end, offs_t &mask, offs_t &mirror) { // adjust start/end/mask values mask &= m_addrmask; start &= ~mirror & m_addrmask; end &= ~mirror & m_addrmask; } void address_space::check_optimize_all(const char *function, offs_t addrstart, offs_t addrend, offs_t addrmask, offs_t addrmirror, offs_t addrselect, offs_t &nstart, offs_t &nend, offs_t &nmask, offs_t &nmirror) { if (addrstart > addrend) fatalerror("%s: In range %x-%x mask %x mirror %x select %x, start address is after the end address.\n", function, addrstart, addrend, addrmask, addrmirror, addrselect); if (addrstart & ~m_addrmask) fatalerror("%s: In range %x-%x mask %x mirror %x select %x, start address is outside of the global address mask %x, did you mean %x ?\n", function, addrstart, addrend, addrmask, addrmirror, addrselect, m_addrmask, addrstart & m_addrmask); if (addrend & ~m_addrmask) fatalerror("%s: In range %x-%x mask %x mirror %x select %x, end address is outside of the global address mask %x, did you mean %x ?\n", function, addrstart, addrend, addrmask, addrmirror, addrselect, m_addrmask, addrend & m_addrmask); offs_t lowbits_mask = (m_config.data_width() >> (3 - m_config.m_addr_shift)) - 1; if (addrstart & lowbits_mask) fatalerror("%s: In range %x-%x mask %x mirror %x select %x, start address has low bits set, did you mean %x ?\n", function, addrstart, addrend, addrmask, addrmirror, addrselect, addrstart & ~lowbits_mask); if ((~addrend) & lowbits_mask) fatalerror("%s: In range %x-%x mask %x mirror %x select %x, end address has low bits unset, did you mean %x ?\n", function, addrstart, addrend, addrmask, addrmirror, addrselect, addrend | lowbits_mask); offs_t set_bits = addrstart | addrend; offs_t changing_bits = addrstart ^ addrend; // Round up to the nearest power-of-two-minus-one changing_bits |= changing_bits >> 1; changing_bits |= changing_bits >> 2; changing_bits |= changing_bits >> 4; changing_bits |= changing_bits >> 8; changing_bits |= changing_bits >> 16; if (addrmask & ~m_addrmask) fatalerror("%s: In range %x-%x mask %x mirror %x select %x, mask is outside of the global address mask %x, did you mean %x ?\n", function, addrstart, addrend, addrmask, addrmirror, addrselect, m_addrmask, addrmask & m_addrmask); if (addrmirror & ~m_addrmask) fatalerror("%s: In range %x-%x mask %x mirror %x select %x, mirror is outside of the global address mask %x, did you mean %x ?\n", function, addrstart, addrend, addrmask, addrmirror, addrselect, m_addrmask, addrmirror & m_addrmask); if (addrselect & ~m_addrmask) fatalerror("%s: In range %x-%x mask %x mirror %x select %x, select is outside of the global address mask %x, did you mean %x ?\n", function, addrstart, addrend, addrmask, addrmirror, addrselect, m_addrmask, addrselect & m_addrmask); if (addrmask & ~changing_bits) fatalerror("%s: In range %x-%x mask %x mirror %x select %x, mask is trying to unmask an unchanging address bit, did you mean %x ?\n", function, addrstart, addrend, addrmask, addrmirror, addrselect, addrmask & changing_bits); if (addrmirror & changing_bits) fatalerror("%s: In range %x-%x mask %x mirror %x select %x, mirror touches a changing address bit, did you mean %x ?\n", function, addrstart, addrend, addrmask, addrmirror, addrselect, addrmirror & ~changing_bits); if (addrselect & changing_bits) fatalerror("%s: In range %x-%x mask %x mirror %x select %x, select touches a changing address bit, did you mean %x ?\n", function, addrstart, addrend, addrmask, addrmirror, addrselect, addrselect & ~changing_bits); if (addrmirror & set_bits) fatalerror("%s: In range %x-%x mask %x mirror %x select %x, mirror touches a set address bit, did you mean %x ?\n", function, addrstart, addrend, addrmask, addrmirror, addrselect, addrmirror & ~set_bits); if (addrselect & set_bits) fatalerror("%s: In range %x-%x mask %x mirror %x select %x, select touches a set address bit, did you mean %x ?\n", function, addrstart, addrend, addrmask, addrmirror, addrselect, addrselect & ~set_bits); if (addrmirror & addrselect) fatalerror("%s: In range %x-%x mask %x mirror %x select %x, mirror touches a select bit, did you mean %x ?\n", function, addrstart, addrend, addrmask, addrmirror, addrselect, addrmirror & ~addrselect); nstart = addrstart; nend = addrend; nmask = (addrmask ? addrmask : changing_bits) | addrselect; nmirror = addrmirror | addrselect; if(nmirror && !(nstart & changing_bits) && !((~nend) & changing_bits)) { // If the range covers the a complete power-of-two zone, it is // possible to remove 1 bits from the mirror, pushing the end // address. The mask will clamp, and installing the range // will be faster. while(nmirror & (changing_bits+1)) { offs_t bit = nmirror & (changing_bits+1); nmirror &= ~bit; nend |= bit; changing_bits |= bit; } } } void address_space::check_optimize_mirror(const char *function, offs_t addrstart, offs_t addrend, offs_t addrmirror, offs_t &nstart, offs_t &nend, offs_t &nmask, offs_t &nmirror) { if (addrstart > addrend) fatalerror("%s: In range %x-%x mirror %x, start address is after the end address.\n", function, addrstart, addrend, addrmirror); if (addrstart & ~m_addrmask) fatalerror("%s: In range %x-%x mirror %x, start address is outside of the global address mask %x, did you mean %x ?\n", function, addrstart, addrend, addrmirror, m_addrmask, addrstart & m_addrmask); if (addrend & ~m_addrmask) fatalerror("%s: In range %x-%x mirror %x, end address is outside of the global address mask %x, did you mean %x ?\n", function, addrstart, addrend, addrmirror, m_addrmask, addrend & m_addrmask); offs_t lowbits_mask = (m_config.data_width() >> (3 - m_config.m_addr_shift)) - 1; if (addrstart & lowbits_mask) fatalerror("%s: In range %x-%x mirror %x, start address has low bits set, did you mean %x ?\n", function, addrstart, addrend, addrmirror, addrstart & ~lowbits_mask); if ((~addrend) & lowbits_mask) fatalerror("%s: In range %x-%x mirror %x, end address has low bits unset, did you mean %x ?\n", function, addrstart, addrend, addrmirror, addrend | lowbits_mask); offs_t set_bits = addrstart | addrend; offs_t changing_bits = addrstart ^ addrend; // Round up to the nearest power-of-two-minus-one changing_bits |= changing_bits >> 1; changing_bits |= changing_bits >> 2; changing_bits |= changing_bits >> 4; changing_bits |= changing_bits >> 8; changing_bits |= changing_bits >> 16; if (addrmirror & ~m_addrmask) fatalerror("%s: In range %x-%x mirror %x, mirror is outside of the global address mask %x, did you mean %x ?\n", function, addrstart, addrend, addrmirror, m_addrmask, addrmirror & m_addrmask); if (addrmirror & changing_bits) fatalerror("%s: In range %x-%x mirror %x, mirror touches a changing address bit, did you mean %x ?\n", function, addrstart, addrend, addrmirror, addrmirror & ~changing_bits); if (addrmirror & set_bits) fatalerror("%s: In range %x-%x mirror %x, mirror touches a set address bit, did you mean %x ?\n", function, addrstart, addrend, addrmirror, addrmirror & ~set_bits); nstart = addrstart; nend = addrend; nmask = changing_bits; nmirror = addrmirror; if(nmirror && !(nstart & changing_bits) && !((~nend) & changing_bits)) { // If the range covers the a complete power-of-two zone, it is // possible to remove 1 bits from the mirror, pushing the end // address. The mask will clamp, and installing the range // will be faster. while(nmirror & (changing_bits+1)) { offs_t bit = nmirror & (changing_bits+1); nmirror &= ~bit; nend |= bit; changing_bits |= bit; } } } void address_space::check_address(const char *function, offs_t addrstart, offs_t addrend) { if (addrstart > addrend) fatalerror("%s: In range %x-%x, start address is after the end address.\n", function, addrstart, addrend); if (addrstart & ~m_addrmask) fatalerror("%s: In range %x-%x, start address is outside of the global address mask %x, did you mean %x ?\n", function, addrstart, addrend, m_addrmask, addrstart & m_addrmask); if (addrend & ~m_addrmask) fatalerror("%s: In range %x-%x, end address is outside of the global address mask %x, did you mean %x ?\n", function, addrstart, addrend, m_addrmask, addrend & m_addrmask); offs_t lowbits_mask = (m_config.data_width() >> (3 - m_config.m_addr_shift)) - 1; if (addrstart & lowbits_mask) fatalerror("%s: In range %x-%x, start address has low bits set, did you mean %x ?\n", function, addrstart, addrend, addrstart & ~lowbits_mask); if ((~addrend) & lowbits_mask) fatalerror("%s: In range %x-%x, end address has low bits unset, did you mean %x ?\n", function, addrstart, addrend, addrend | lowbits_mask); } //------------------------------------------------- // prepare_map - allocate the address map and // walk through it to find implicit memory regions // and identify shared regions //------------------------------------------------- void address_space::prepare_map() { memory_region *devregion = (m_spacenum == 0) ? machine().root_device().memregion(m_device.tag()) : nullptr; u32 devregionsize = (devregion != nullptr) ? devregion->bytes() : 0; // allocate the address map m_map = std::make_unique(m_device, m_spacenum); // merge in the submaps m_map->uplift_submaps(machine(), m_device.owner() ? *m_device.owner() : m_device, endianness()); // extract global parameters specified by the map m_unmap = (m_map->m_unmapval == 0) ? 0 : ~0; if (m_map->m_globalmask != 0) m_addrmask = m_map->m_globalmask; // make a pass over the address map, adjusting for the device and getting memory pointers for (address_map_entry &entry : m_map->m_entrylist) { // computed adjusted addresses first adjust_addresses(entry.m_addrstart, entry.m_addrend, entry.m_addrmask, entry.m_addrmirror); // if we have a share entry, add it to our map if (entry.m_share != nullptr) { // if we can't find it, add it to our map std::string fulltag = entry.m_devbase.subtag(entry.m_share); if (manager().m_sharelist.find(fulltag.c_str()) == manager().m_sharelist.end()) { VPRINTF(("Creating share '%s' of length 0x%X\n", fulltag.c_str(), entry.m_addrend + 1 - entry.m_addrstart)); manager().m_sharelist.emplace(fulltag.c_str(), std::make_unique(m_map->m_databits, address_to_byte(entry.m_addrend + 1 - entry.m_addrstart), endianness())); } } // if this is a ROM handler without a specified region, attach it to the implicit region if (m_spacenum == 0 && entry.m_read.m_type == AMH_ROM && entry.m_region == nullptr) { // make sure it fits within the memory region before doing so, however if (entry.m_addrend < devregionsize) { entry.m_region = m_device.tag(); entry.m_rgnoffs = address_to_byte(entry.m_addrstart); } } // validate adjusted addresses against implicit regions if (entry.m_region != nullptr && entry.m_share == nullptr) { // determine full tag std::string fulltag = entry.m_devbase.subtag(entry.m_region); // find the region memory_region *region = machine().root_device().memregion(fulltag.c_str()); if (region == nullptr) fatalerror("device '%s' %s space memory map entry %X-%X references non-existant region \"%s\"\n", m_device.tag(), m_name, entry.m_addrstart, entry.m_addrend, entry.m_region); // validate the region if (entry.m_rgnoffs + m_config.addr2byte(entry.m_addrend - entry.m_addrstart + 1) > region->bytes()) fatalerror("device '%s' %s space memory map entry %X-%X extends beyond region \"%s\" size (%X)\n", m_device.tag(), m_name, entry.m_addrstart, entry.m_addrend, entry.m_region, region->bytes()); } // convert any region-relative entries to their memory pointers if (entry.m_region != nullptr) { // determine full tag std::string fulltag = entry.m_devbase.subtag(entry.m_region); // set the memory address entry.m_memory = machine().root_device().memregion(fulltag.c_str())->base() + entry.m_rgnoffs; } } // now loop over all the handlers and enforce the address mask read().mask_all_handlers(m_addrmask); write().mask_all_handlers(m_addrmask); } //------------------------------------------------- // populate_from_map - walk the map in reverse // order and install the appropriate handler for // each case //------------------------------------------------- void address_space::populate_from_map(address_map *map) { // no map specified, use the space-specific one if (map == nullptr) map = m_map.get(); // no map, nothing to do if (map == nullptr) return; // install the handlers, using the original, unadjusted memory map const address_map_entry *last_entry = nullptr; while (last_entry != map->m_entrylist.first()) { // find the entry before the last one we processed const address_map_entry *entry; for (entry = map->m_entrylist.first(); entry->next() != last_entry; entry = entry->next()) { }; last_entry = entry; // map both read and write halves populate_map_entry(*entry, read_or_write::READ); populate_map_entry(*entry, read_or_write::WRITE); populate_map_entry_setoffset(*entry); } } //------------------------------------------------- // populate_map_entry - map a single read or // write entry based on information from an // address map entry //------------------------------------------------- void address_space::populate_map_entry(const address_map_entry &entry, read_or_write readorwrite) { const map_handler_data &data = (readorwrite == read_or_write::READ) ? entry.m_read : entry.m_write; // based on the handler type, alter the bits, name, funcptr, and object switch (data.m_type) { case AMH_NONE: return; case AMH_ROM: // writes to ROM are no-ops if (readorwrite == read_or_write::WRITE) return; // fall through to the RAM case otherwise case AMH_RAM: install_ram_generic(entry.m_addrstart, entry.m_addrend, entry.m_addrmirror, readorwrite, nullptr); break; case AMH_NOP: unmap_generic(entry.m_addrstart, entry.m_addrend, entry.m_addrmirror, readorwrite, true); break; case AMH_UNMAP: unmap_generic(entry.m_addrstart, entry.m_addrend, entry.m_addrmirror, readorwrite, false); break; case AMH_DEVICE_DELEGATE: if (readorwrite == read_or_write::READ) switch (data.m_bits) { case 8: install_read_handler(entry.m_addrstart, entry.m_addrend, entry.m_addrmask, entry.m_addrmirror, entry.m_addrselect, read8_delegate(entry.m_rproto8, entry.m_devbase), data.m_mask); break; case 16: install_read_handler(entry.m_addrstart, entry.m_addrend, entry.m_addrmask, entry.m_addrmirror, entry.m_addrselect, read16_delegate(entry.m_rproto16, entry.m_devbase), data.m_mask); break; case 32: install_read_handler(entry.m_addrstart, entry.m_addrend, entry.m_addrmask, entry.m_addrmirror, entry.m_addrselect, read32_delegate(entry.m_rproto32, entry.m_devbase), data.m_mask); break; case 64: install_read_handler(entry.m_addrstart, entry.m_addrend, entry.m_addrmask, entry.m_addrmirror, entry.m_addrselect, read64_delegate(entry.m_rproto64, entry.m_devbase), data.m_mask); break; } else switch (data.m_bits) { case 8: install_write_handler(entry.m_addrstart, entry.m_addrend, entry.m_addrmask, entry.m_addrmirror, entry.m_addrselect, write8_delegate(entry.m_wproto8, entry.m_devbase), data.m_mask); break; case 16: install_write_handler(entry.m_addrstart, entry.m_addrend, entry.m_addrmask, entry.m_addrmirror, entry.m_addrselect, write16_delegate(entry.m_wproto16, entry.m_devbase), data.m_mask); break; case 32: install_write_handler(entry.m_addrstart, entry.m_addrend, entry.m_addrmask, entry.m_addrmirror, entry.m_addrselect, write32_delegate(entry.m_wproto32, entry.m_devbase), data.m_mask); break; case 64: install_write_handler(entry.m_addrstart, entry.m_addrend, entry.m_addrmask, entry.m_addrmirror, entry.m_addrselect, write64_delegate(entry.m_wproto64, entry.m_devbase), data.m_mask); break; } break; case AMH_PORT: install_readwrite_port(entry.m_addrstart, entry.m_addrend, entry.m_addrmirror, (readorwrite == read_or_write::READ) ? data.m_tag : nullptr, (readorwrite == read_or_write::WRITE) ? data.m_tag : nullptr); break; case AMH_BANK: install_bank_generic(entry.m_addrstart, entry.m_addrend, entry.m_addrmirror, (readorwrite == read_or_write::READ) ? data.m_tag : nullptr, (readorwrite == read_or_write::WRITE) ? data.m_tag : nullptr); break; case AMH_DEVICE_SUBMAP: throw emu_fatalerror("Internal mapping error: leftover mapping of '%s'.\n", data.m_tag); } } //------------------------------------------------- // populate_map_entry_setoffset - special case for setoffset //------------------------------------------------- void address_space::populate_map_entry_setoffset(const address_map_entry &entry) { install_setoffset_handler(entry.m_addrstart, entry.m_addrend, entry.m_addrmask, entry.m_addrmirror, entry.m_addrselect, setoffset_delegate(entry.m_soproto, entry.m_devbase), entry.m_setoffsethd.m_mask); } //------------------------------------------------- // allocate_memory - determine all neighboring // address ranges and allocate memory to back // them //------------------------------------------------- void address_space::allocate_memory() { auto &blocklist = manager().m_blocklist; // make a first pass over the memory map and track blocks with hardcoded pointers // we do this to make sure they are found by space_find_backing_memory first int tail = blocklist.size(); for (address_map_entry &entry : m_map->m_entrylist) if (entry.m_memory != nullptr) blocklist.push_back(std::make_unique(*this, entry.m_addrstart, entry.m_addrend, entry.m_memory)); // loop over all blocks just allocated and assign pointers from them address_map_entry *unassigned = nullptr; for (auto memblock = blocklist.begin() + tail; memblock != blocklist.end(); ++memblock) unassigned = block_assign_intersecting(memblock->get()->addrstart(), memblock->get()->addrend(), memblock->get()->data()); // if we don't have an unassigned pointer yet, try to find one if (unassigned == nullptr) unassigned = block_assign_intersecting(~0, 0, nullptr); // loop until we've assigned all memory in this space while (unassigned != nullptr) { // work in MEMORY_BLOCK_CHUNK-sized chunks offs_t curblockstart = unassigned->m_addrstart / MEMORY_BLOCK_CHUNK; offs_t curblockend = unassigned->m_addrend / MEMORY_BLOCK_CHUNK; // loop while we keep finding unassigned blocks in neighboring MEMORY_BLOCK_CHUNK chunks bool changed; do { changed = false; // scan for unmapped blocks in the adjusted map for (address_map_entry &entry : m_map->m_entrylist) if (entry.m_memory == nullptr && &entry != unassigned && needs_backing_store(entry)) { // get block start/end blocks for this block offs_t blockstart = entry.m_addrstart / MEMORY_BLOCK_CHUNK; offs_t blockend = entry.m_addrend / MEMORY_BLOCK_CHUNK; // if we intersect or are adjacent, adjust the start/end if (blockstart <= curblockend + 1 && blockend >= curblockstart - 1) { if (blockstart < curblockstart) curblockstart = blockstart, changed = true; if (blockend > curblockend) curblockend = blockend, changed = true; } } } while (changed); // we now have a block to allocate; do it offs_t curaddrstart = curblockstart * MEMORY_BLOCK_CHUNK; offs_t curaddrend = curblockend * MEMORY_BLOCK_CHUNK + (MEMORY_BLOCK_CHUNK - 1); auto block = std::make_unique(*this, curaddrstart, curaddrend); // assign memory that intersected the new block unassigned = block_assign_intersecting(curaddrstart, curaddrend, block.get()->data()); blocklist.push_back(std::move(block)); } } //------------------------------------------------- // locate_memory - find all the requested // pointers into the final allocated memory //------------------------------------------------- void address_space::locate_memory() { // once this is done, find the starting bases for the banks for (auto &bank : manager().banks()) if (bank.second->base() == nullptr && bank.second->references_space(*this, read_or_write::READWRITE)) { // set the initial bank pointer for (address_map_entry &entry : m_map->m_entrylist) if (entry.m_addrstart == bank.second->addrstart() && entry.m_memory != nullptr) { bank.second->set_base(entry.m_memory); VPRINTF(("assigned bank '%s' pointer to memory from range %08X-%08X [%p]\n", bank.second->tag(), entry.m_addrstart, entry.m_addrend, entry.m_memory)); break; } // if the entry was set ahead of time, override the automatically found pointer if (!bank.second->anonymous() && bank.second->entry() != BANK_ENTRY_UNSPECIFIED) bank.second->set_entry(bank.second->entry()); } } //------------------------------------------------- // block_assign_intersecting - find all // intersecting blocks and assign their pointers //------------------------------------------------- address_map_entry *address_space::block_assign_intersecting(offs_t addrstart, offs_t addrend, u8 *base) { address_map_entry *unassigned = nullptr; // loop over the adjusted map and assign memory to any blocks we can for (address_map_entry &entry : m_map->m_entrylist) { // if we haven't assigned this block yet, see if we have a mapped shared pointer for it if (entry.m_memory == nullptr && entry.m_share != nullptr) { std::string fulltag = entry.m_devbase.subtag(entry.m_share); auto share = manager().shares().find(fulltag.c_str()); if (share != manager().shares().end() && share->second->ptr() != nullptr) { entry.m_memory = share->second->ptr(); VPRINTF(("memory range %08X-%08X -> shared_ptr '%s' [%p]\n", entry.m_addrstart, entry.m_addrend, entry.m_share, entry.m_memory)); } else { VPRINTF(("memory range %08X-%08X -> shared_ptr '%s' but not found\n", entry.m_addrstart, entry.m_addrend, entry.m_share)); } } // otherwise, look for a match in this block if (entry.m_memory == nullptr && entry.m_addrstart >= addrstart && entry.m_addrend <= addrend) { entry.m_memory = base + m_config.addr2byte(entry.m_addrstart - addrstart); VPRINTF(("memory range %08X-%08X -> found in block from %08X-%08X [%p]\n", entry.m_addrstart, entry.m_addrend, addrstart, addrend, entry.m_memory)); } // if we're the first match on a shared pointer, assign it now if (entry.m_memory != nullptr && entry.m_share != nullptr) { std::string fulltag = entry.m_devbase.subtag(entry.m_share); auto share = manager().shares().find(fulltag.c_str()); if (share != manager().shares().end() && share->second->ptr() == nullptr) { share->second->set_ptr(entry.m_memory); VPRINTF(("setting shared_ptr '%s' = %p\n", entry.m_share, entry.m_memory)); } } // keep track of the first unassigned entry if (entry.m_memory == nullptr && unassigned == nullptr && needs_backing_store(entry)) unassigned = &entry; } return unassigned; } //------------------------------------------------- // get_handler_string - return a string // describing the handler at a particular offset //------------------------------------------------- const char *address_space::get_handler_string(read_or_write readorwrite, offs_t address) { if (readorwrite == read_or_write::READ) return read().handler_name(read().lookup(address)); else return write().handler_name(write().lookup(address)); } //------------------------------------------------- // dump_map - dump the contents of a single // address space //------------------------------------------------- void address_space::dump_map(FILE *file, read_or_write readorwrite) { const address_table &table = (readorwrite == read_or_write::READ) ? static_cast(read()) : static_cast(write()); // dump generic information fprintf(file, " Address bits = %d\n", m_config.m_addr_width); fprintf(file, " Data bits = %d\n", m_config.m_data_width); fprintf(file, " Address mask = %X\n", m_addrmask); fprintf(file, "\n"); // iterate over addresses offs_t addrstart, addrend; for (offs_t address = 0; address <= m_addrmask; address = addrend) { u16 entry = table.derive_range(address, addrstart, addrend); fprintf(file, "%08X-%08X = %02X: %s [offset=%08X]\n", addrstart, addrend, entry, table.handler_name(entry), table.handler(entry).addrstart()); if (++addrend == 0) break; } } //************************************************************************** // DYNAMIC ADDRESS SPACE MAPPING //************************************************************************** //------------------------------------------------- // unmap - unmap a section of address space //------------------------------------------------- void address_space::unmap_generic(offs_t addrstart, offs_t addrend, offs_t addrmirror, read_or_write readorwrite, bool quiet) { VPRINTF(("address_space::unmap(%s-%s mirror=%s, %s, %s)\n", core_i64_hex_format(addrstart, m_addrchars), core_i64_hex_format(addrend, m_addrchars), core_i64_hex_format(addrmirror, m_addrchars), (readorwrite == read_or_write::READ) ? "read" : (readorwrite == read_or_write::WRITE) ? "write" : (readorwrite == read_or_write::READWRITE) ? "read/write" : "??", quiet ? "quiet" : "normal")); offs_t nstart, nend, nmask, nmirror; check_optimize_mirror("unmap_generic", addrstart, addrend, addrmirror, nstart, nend, nmask, nmirror); // read space if (readorwrite == read_or_write::READ || readorwrite == read_or_write::READWRITE) read().map_range(nstart, nend, nmask, nmirror, quiet ? STATIC_NOP : STATIC_UNMAP); // write space if (readorwrite == read_or_write::WRITE || readorwrite == read_or_write::READWRITE) write().map_range(nstart, nend, nmask, nmirror, quiet ? STATIC_NOP : STATIC_UNMAP); } //------------------------------------------------- // install_device_delegate - install the memory map // of a live device into this address space //------------------------------------------------- void address_space::install_device_delegate(offs_t addrstart, offs_t addrend, device_t &device, address_map_delegate &delegate, int bits, u64 unitmask) { check_address("install_device_delegate", addrstart, addrend); address_map map(*this, addrstart, addrend, bits, unitmask, m_device, delegate); map.uplift_submaps(machine(), device, endianness()); populate_from_map(&map); } //------------------------------------------------- // install_readwrite_port - install a new I/O port // handler into this address space //------------------------------------------------- void address_space::install_readwrite_port(offs_t addrstart, offs_t addrend, offs_t addrmirror, const char *rtag, const char *wtag) { VPRINTF(("address_space::install_readwrite_port(%s-%s mirror=%s, read=\"%s\" / write=\"%s\")\n", core_i64_hex_format(addrstart, m_addrchars), core_i64_hex_format(addrend, m_addrchars), core_i64_hex_format(addrmirror, m_addrchars), (rtag != nullptr) ? rtag : "(none)", (wtag != nullptr) ? wtag : "(none)")); offs_t nstart, nend, nmask, nmirror; check_optimize_mirror("install_readwrite_port", addrstart, addrend, addrmirror, nstart, nend, nmask, nmirror); // read handler if (rtag != nullptr) { // find the port ioport_port *port = machine().root_device().ioport(device().siblingtag(rtag).c_str()); if (port == nullptr) throw emu_fatalerror("Attempted to map non-existent port '%s' for read in space %s of device '%s'\n", rtag, m_name, m_device.tag()); // map the range and set the ioport read().handler_map_range(nstart, nend, nmask, nmirror).set_ioport(*port); } if (wtag != nullptr) { // find the port ioport_port *port = machine().root_device().ioport(device().siblingtag(wtag).c_str()); if (port == nullptr) fatalerror("Attempted to map non-existent port '%s' for write in space %s of device '%s'\n", wtag, m_name, m_device.tag()); // map the range and set the ioport write().handler_map_range(nstart, nend, nmask, nmirror).set_ioport(*port); } // update the memory dump generate_memdump(machine()); } //------------------------------------------------- // install_bank_generic - install a range as // mapping to a particular bank //------------------------------------------------- void address_space::install_bank_generic(offs_t addrstart, offs_t addrend, offs_t addrmirror, const char *rtag, const char *wtag) { VPRINTF(("address_space::install_readwrite_bank(%s-%s mirror=%s, read=\"%s\" / write=\"%s\")\n", core_i64_hex_format(addrstart, m_addrchars), core_i64_hex_format(addrend, m_addrchars), core_i64_hex_format(addrmirror, m_addrchars), (rtag != nullptr) ? rtag : "(none)", (wtag != nullptr) ? wtag : "(none)")); offs_t nstart, nend, nmask, nmirror; check_optimize_mirror("install_bank_generic", addrstart, addrend, addrmirror, nstart, nend, nmask, nmirror); // map the read bank if (rtag != nullptr) { std::string fulltag = device().siblingtag(rtag); memory_bank &bank = bank_find_or_allocate(fulltag.c_str(), addrstart, addrend, addrmirror, read_or_write::READ); read().map_range(nstart, nend, nmask, nmirror, bank.index()); } // map the write bank if (wtag != nullptr) { std::string fulltag = device().siblingtag(wtag); memory_bank &bank = bank_find_or_allocate(fulltag.c_str(), addrstart, addrend, addrmirror, read_or_write::WRITE); write().map_range(nstart, nend, nmask, nmirror, bank.index()); } // update the memory dump generate_memdump(machine()); } void address_space::install_bank_generic(offs_t addrstart, offs_t addrend, offs_t addrmirror, memory_bank *rbank, memory_bank *wbank) { VPRINTF(("address_space::install_readwrite_bank(%s-%s mirror=%s, read=\"%s\" / write=\"%s\")\n", core_i64_hex_format(addrstart, m_addrchars), core_i64_hex_format(addrend, m_addrchars), core_i64_hex_format(addrmirror, m_addrchars), (rbank != nullptr) ? rbank->tag() : "(none)", (wbank != nullptr) ? wbank->tag() : "(none)")); offs_t nstart, nend, nmask, nmirror; check_optimize_mirror("install_bank_generic", addrstart, addrend, addrmirror, nstart, nend, nmask, nmirror); // map the read bank if (rbank != nullptr) { read().map_range(nstart, nend, nmask, nmirror, rbank->index()); } // map the write bank if (wbank != nullptr) { write().map_range(nstart, nend, nmask, nmirror, wbank->index()); } // update the memory dump generate_memdump(machine()); } //------------------------------------------------- // install_ram_generic - install a simple fixed // RAM region into the given address space //------------------------------------------------- void address_space::install_ram_generic(offs_t addrstart, offs_t addrend, offs_t addrmirror, read_or_write readorwrite, void *baseptr) { VPRINTF(("address_space::install_ram_generic(%s-%s mirror=%s, %s, %p)\n", core_i64_hex_format(addrstart, m_addrchars), core_i64_hex_format(addrend, m_addrchars), core_i64_hex_format(addrmirror, m_addrchars), (readorwrite == read_or_write::READ) ? "read" : (readorwrite == read_or_write::WRITE) ? "write" : (readorwrite == read_or_write::READWRITE) ? "read/write" : "??", baseptr)); offs_t nstart, nend, nmask, nmirror; check_optimize_mirror("install_ram_generic", addrstart, addrend, addrmirror, nstart, nend, nmask, nmirror); // map for read if (readorwrite == read_or_write::READ || readorwrite == read_or_write::READWRITE) { // find a bank and map it memory_bank &bank = bank_find_or_allocate(nullptr, addrstart, addrend, addrmirror, read_or_write::READ); read().map_range(nstart, nend, nmask, nmirror, bank.index()); // if we are provided a pointer, set it if (baseptr != nullptr) bank.set_base(baseptr); // if we don't have a bank pointer yet, try to find one if (bank.base() == nullptr) { void *backing = find_backing_memory(addrstart, addrend); if (backing != nullptr) bank.set_base(backing); } // if we still don't have a pointer, and we're past the initialization phase, allocate a new block if (bank.base() == nullptr && manager().m_initialized) { if (machine().phase() >= machine_phase::RESET) fatalerror("Attempted to call install_ram_generic() after initialization time without a baseptr!\n"); auto block = std::make_unique(*this, addrstart, addrend); bank.set_base(block.get()->data()); manager().m_blocklist.push_back(std::move(block)); } } // map for write if (readorwrite == read_or_write::WRITE || readorwrite == read_or_write::READWRITE) { // find a bank and map it memory_bank &bank = bank_find_or_allocate(nullptr, addrstart, addrend, addrmirror, read_or_write::WRITE); write().map_range(nstart, nend, nmask, nmirror, bank.index()); // if we are provided a pointer, set it if (baseptr != nullptr) bank.set_base(baseptr); // if we don't have a bank pointer yet, try to find one if (bank.base() == nullptr) { void *backing = find_backing_memory(addrstart, addrend); if (backing != nullptr) bank.set_base(backing); } // if we still don't have a pointer, and we're past the initialization phase, allocate a new block if (bank.base() == nullptr && manager().m_initialized) { if (machine().phase() >= machine_phase::RESET) fatalerror("Attempted to call install_ram_generic() after initialization time without a baseptr!\n"); auto block = std::make_unique(*this, address_to_byte(addrstart), address_to_byte_end(addrend)); bank.set_base(block.get()->data()); manager().m_blocklist.push_back(std::move(block)); } } } //------------------------------------------------- // install_handler - install 8-bit read/write // delegate handlers for the space //------------------------------------------------- void address_space::install_read_handler(offs_t addrstart, offs_t addrend, offs_t addrmask, offs_t addrmirror, offs_t addrselect, read8_delegate handler, u64 unitmask) { VPRINTF(("address_space::install_read_handler(%s-%s mask=%s mirror=%s, %s, %s)\n", core_i64_hex_format(addrstart, m_addrchars), core_i64_hex_format(addrend, m_addrchars), core_i64_hex_format(addrmask, m_addrchars), core_i64_hex_format(addrmirror, m_addrchars), handler.name(), core_i64_hex_format(unitmask, data_width() / 4))); offs_t nstart, nend, nmask, nmirror; check_optimize_all("install_read_handler", addrstart, addrend, addrmask, addrmirror, addrselect, nstart, nend, nmask, nmirror); read().handler_map_range(nstart, nend, nmask, nmirror, unitmask).set_delegate(handler); generate_memdump(machine()); } void address_space::install_write_handler(offs_t addrstart, offs_t addrend, offs_t addrmask, offs_t addrmirror, offs_t addrselect, write8_delegate handler, u64 unitmask) { VPRINTF(("address_space::install_write_handler(%s-%s mask=%s mirror=%s, %s, %s)\n", core_i64_hex_format(addrstart, m_addrchars), core_i64_hex_format(addrend, m_addrchars), core_i64_hex_format(addrmask, m_addrchars), core_i64_hex_format(addrmirror, m_addrchars), handler.name(), core_i64_hex_format(unitmask, data_width() / 4))); offs_t nstart, nend, nmask, nmirror; check_optimize_all("install_write_handler", addrstart, addrend, addrmask, addrmirror, addrselect, nstart, nend, nmask, nmirror); write().handler_map_range(nstart, nend, nmask, nmirror, unitmask).set_delegate(handler); generate_memdump(machine()); } void address_space::install_readwrite_handler(offs_t addrstart, offs_t addrend, offs_t addrmask, offs_t addrmirror, offs_t addrselect, read8_delegate rhandler, write8_delegate whandler, u64 unitmask) { install_read_handler(addrstart, addrend, addrmask, addrmirror, addrselect, rhandler, unitmask); install_write_handler(addrstart, addrend, addrmask, addrmirror, addrselect, whandler, unitmask); } //------------------------------------------------- // install_handler - install 16-bit read/write // delegate handlers for the space //------------------------------------------------- void address_space::install_read_handler(offs_t addrstart, offs_t addrend, offs_t addrmask, offs_t addrmirror, offs_t addrselect, read16_delegate handler, u64 unitmask) { offs_t nstart, nend, nmask, nmirror; check_optimize_all("install_read_handler", addrstart, addrend, addrmask, addrmirror, addrselect, nstart, nend, nmask, nmirror); read().handler_map_range(nstart, nend, nmask, nmirror, unitmask).set_delegate(handler); generate_memdump(machine()); } void address_space::install_write_handler(offs_t addrstart, offs_t addrend, offs_t addrmask, offs_t addrmirror, offs_t addrselect, write16_delegate handler, u64 unitmask) { offs_t nstart, nend, nmask, nmirror; check_optimize_all("install_write_handler", addrstart, addrend, addrmask, addrmirror, addrselect, nstart, nend, nmask, nmirror); write().handler_map_range(nstart, nend, nmask, nmirror, unitmask).set_delegate(handler); generate_memdump(machine()); } void address_space::install_readwrite_handler(offs_t addrstart, offs_t addrend, offs_t addrmask, offs_t addrmirror, offs_t addrselect, read16_delegate rhandler, write16_delegate whandler, u64 unitmask) { install_read_handler(addrstart, addrend, addrmask, addrmirror, addrselect, rhandler, unitmask); install_write_handler(addrstart, addrend, addrmask, addrmirror, addrselect, whandler, unitmask); } //------------------------------------------------- // install_handler - install 32-bit read/write // delegate handlers for the space //------------------------------------------------- void address_space::install_read_handler(offs_t addrstart, offs_t addrend, offs_t addrmask, offs_t addrmirror, offs_t addrselect, read32_delegate handler, u64 unitmask) { offs_t nstart, nend, nmask, nmirror; check_optimize_all("install_read_handler", addrstart, addrend, addrmask, addrmirror, addrselect, nstart, nend, nmask, nmirror); read().handler_map_range(nstart, nend, nmask, nmirror, unitmask).set_delegate(handler); generate_memdump(machine()); } void address_space::install_write_handler(offs_t addrstart, offs_t addrend, offs_t addrmask, offs_t addrmirror, offs_t addrselect, write32_delegate handler, u64 unitmask) { offs_t nstart, nend, nmask, nmirror; check_optimize_all("install_write_handler", addrstart, addrend, addrmask, addrmirror, addrselect, nstart, nend, nmask, nmirror); write().handler_map_range(nstart, nend, nmask, nmirror, unitmask).set_delegate(handler); generate_memdump(machine()); } void address_space::install_readwrite_handler(offs_t addrstart, offs_t addrend, offs_t addrmask, offs_t addrmirror, offs_t addrselect, read32_delegate rhandler, write32_delegate whandler, u64 unitmask) { install_read_handler(addrstart, addrend, addrmask, addrmirror, addrselect, rhandler, unitmask); install_write_handler(addrstart, addrend, addrmask, addrmirror, addrselect, whandler, unitmask); } //------------------------------------------------- // install_handler64 - install 64-bit read/write // delegate handlers for the space //------------------------------------------------- void address_space::install_read_handler(offs_t addrstart, offs_t addrend, offs_t addrmask, offs_t addrmirror, offs_t addrselect, read64_delegate handler, u64 unitmask) { offs_t nstart, nend, nmask, nmirror; check_optimize_all("install_read_handler", addrstart, addrend, addrmask, addrmirror, addrselect, nstart, nend, nmask, nmirror); read().handler_map_range(nstart, nend, nmask, nmirror, unitmask).set_delegate(handler); generate_memdump(machine()); } void address_space::install_write_handler(offs_t addrstart, offs_t addrend, offs_t addrmask, offs_t addrmirror, offs_t addrselect, write64_delegate handler, u64 unitmask) { offs_t nstart, nend, nmask, nmirror; check_optimize_all("install_write_handler", addrstart, addrend, addrmask, addrmirror, addrselect, nstart, nend, nmask, nmirror); write().handler_map_range(nstart, nend, nmask, nmirror, unitmask).set_delegate(handler); generate_memdump(machine()); } void address_space::install_readwrite_handler(offs_t addrstart, offs_t addrend, offs_t addrmask, offs_t addrmirror, offs_t addrselect, read64_delegate rhandler, write64_delegate whandler, u64 unitmask) { install_read_handler(addrstart, addrend, addrmask, addrmirror, addrselect, rhandler, unitmask); install_write_handler(addrstart, addrend, addrmask, addrmirror, addrselect, whandler, unitmask); } //----------------------------------------------------------------------- // install_setoffset_handler - install set_offset delegate handlers for the space //----------------------------------------------------------------------- void address_space::install_setoffset_handler(offs_t addrstart, offs_t addrend, offs_t addrmask, offs_t addrmirror, offs_t addrselect, setoffset_delegate handler, u64 unitmask) { VPRINTF(("address_space::install_setoffset_handler(%s-%s mask=%s mirror=%s, %s, %s)\n", core_i64_hex_format(addrstart, m_addrchars), core_i64_hex_format(addrend, m_addrchars), core_i64_hex_format(addrmask, m_addrchars), core_i64_hex_format(addrmirror, m_addrchars), handler.name(), core_i64_hex_format(unitmask, data_width() / 4))); offs_t nstart, nend, nmask, nmirror; check_optimize_all("install_setoffset_handler", addrstart, addrend, addrmask, addrmirror, addrselect, nstart, nend, nmask, nmirror); setoffset().handler_map_range(nstart, nend, nmask, nmirror, unitmask).set_delegate(handler); } //************************************************************************** // MEMORY MAPPING HELPERS //************************************************************************** //------------------------------------------------- // find_backing_memory - return a pointer to // the base of RAM associated with the given // device and offset //------------------------------------------------- void *address_space::find_backing_memory(offs_t addrstart, offs_t addrend) { VPRINTF(("address_space::find_backing_memory('%s',%s,%08X-%08X) -> ", m_device.tag(), m_name, addrstart, addrend)); if (m_map == nullptr) return nullptr; // look in the address map first for (address_map_entry &entry : m_map->m_entrylist) { if (entry.m_memory != nullptr && addrstart >= entry.m_addrstart && addrend <= entry.m_addrend) { VPRINTF(("found in entry %08X-%08X [%p]\n", entry.m_addrstart, entry.m_addrend, (u8 *)entry.m_memory + address_to_byte(addrstart - entry.m_addrstart))); return (u8 *)entry.m_memory + address_to_byte(addrstart - entry.m_addrstart); } } // if not found there, look in the allocated blocks for (auto &block : manager().m_blocklist) if (block->contains(*this, addrstart, addrend)) { VPRINTF(("found in allocated memory block %08X-%08X [%p]\n", block->addrstart(), block->addrend(), block->data() + address_to_byte(addrstart - block->addrstart()))); return block->data() + address_to_byte(addrstart - block->addrstart()); } VPRINTF(("did not find\n")); return nullptr; } //------------------------------------------------- // space_needs_backing_store - return whether a // given memory map entry implies the need of // allocating and registering memory //------------------------------------------------- bool address_space::needs_backing_store(const address_map_entry &entry) { // if we are sharing, and we don't have a pointer yet, create one if (entry.m_share != nullptr) { std::string fulltag = entry.m_devbase.subtag(entry.m_share); auto share = manager().shares().find(fulltag.c_str()); if (share != manager().shares().end() && share->second->ptr() == nullptr) return true; } // if we're writing to any sort of bank or RAM, then yes, we do need backing if (entry.m_write.m_type == AMH_BANK || entry.m_write.m_type == AMH_RAM) return true; // if we're reading from RAM or from ROM outside of address space 0 or its region, then yes, we do need backing memory_region *region = machine().root_device().memregion(m_device.tag()); if (entry.m_read.m_type == AMH_RAM || (entry.m_read.m_type == AMH_ROM && (m_spacenum != 0 || region == nullptr || entry.m_addrstart >= region->bytes()))) return true; // all other cases don't need backing return false; } //************************************************************************** // BANKING HELPERS //************************************************************************** //------------------------------------------------- // bank_find_or_allocate - allocate a new // bank, or find an existing one, and return the // read/write handler //------------------------------------------------- memory_bank &address_space::bank_find_or_allocate(const char *tag, offs_t addrstart, offs_t addrend, offs_t addrmirror, read_or_write readorwrite) { // adjust the addresses, handling mirrors and such offs_t addrmask = ~addrmirror; adjust_addresses(addrstart, addrend, addrmask, addrmirror); // look up the bank by name, or else by byte range memory_bank *membank = nullptr; if (tag != nullptr) { auto bank = manager().banks().find(tag); if (bank != manager().banks().end()) membank = bank->second.get(); } else { membank = bank_find_anonymous(addrstart, addrend); } // if we don't have a bank yet, find a free one if (membank == nullptr) { // handle failure int banknum = manager().m_banknext++; if (banknum > STATIC_BANKMAX) { if (tag != nullptr) throw emu_fatalerror("Unable to allocate new bank '%s'", tag); else throw emu_fatalerror("Unable to allocate bank for RAM/ROM area %X-%X\n", addrstart, addrend); } // if no tag, create a unique one auto bank = std::make_unique(*this, banknum, addrstart, addrend, tag); std::string temptag; if (tag == nullptr) { temptag = string_format("anon_%p", bank.get()); tag = temptag.c_str(); } manager().m_banklist.emplace(tag, std::move(bank)); membank = manager().m_banklist.find(tag)->second.get(); } // add a reference for this space membank->add_reference(*this, readorwrite); return *membank; } //------------------------------------------------- // bank_find_anonymous - try to find an anonymous // bank matching the given byte range //------------------------------------------------- memory_bank *address_space::bank_find_anonymous(offs_t addrstart, offs_t addrend) const { // try to find an exact match for (auto &bank : manager().banks()) if (bank.second->anonymous() && bank.second->references_space(*this, read_or_write::READWRITE) && bank.second->matches_exactly(addrstart, addrend)) return bank.second.get(); // not found return nullptr; } //------------------------------------------------- // address_space::invalidate_read_caches -- clear // the read cache (direct) for a specific entry // or all of them //------------------------------------------------- void address_space::invalidate_read_caches() { switch(m_config.addr_shift()) { case 3: static_cast *>(m_direct)->force_update(); break; case 0: static_cast *>(m_direct)->force_update(); break; case -1: static_cast *>(m_direct)->force_update(); break; case -2: static_cast *>(m_direct)->force_update(); break; case -3: static_cast *>(m_direct)->force_update(); break; } } void address_space::invalidate_read_caches(u16 entry) { switch(m_config.addr_shift()) { case 3: static_cast *>(m_direct)->force_update(entry); break; case 0: static_cast *>(m_direct)->force_update(entry); break; case -1: static_cast *>(m_direct)->force_update(entry); break; case -2: static_cast *>(m_direct)->force_update(entry); break; case -3: static_cast *>(m_direct)->force_update(entry); break; } } void address_space::invalidate_read_caches(offs_t start, offs_t end) { switch(m_config.addr_shift()) { case 3: static_cast *>(m_direct)->remove_intersecting_ranges(start, end); break; case 0: static_cast *>(m_direct)->remove_intersecting_ranges(start, end); break; case -1: static_cast *>(m_direct)->remove_intersecting_ranges(start, end); break; case -2: static_cast *>(m_direct)->remove_intersecting_ranges(start, end); break; case -3: static_cast *>(m_direct)->remove_intersecting_ranges(start, end); break; } } //************************************************************************** // TABLE MANAGEMENT //************************************************************************** //------------------------------------------------- // address_table - constructor //------------------------------------------------- address_table::address_table(address_space &space, bool large) : m_table(1 << LEVEL1_BITS), m_space(space), m_large(large), m_subtable(SUBTABLE_COUNT), m_subtable_alloc(0) { m_live_lookup = &m_table[0]; // make our static table all watchpoints if (s_watchpoint_table[0] != STATIC_WATCHPOINT) for (unsigned int i=0; i != ARRAY_LENGTH(s_watchpoint_table); i++) s_watchpoint_table[i] = STATIC_WATCHPOINT; // initialize everything to unmapped for (unsigned int i=0; i != 1 << LEVEL1_BITS; i++) m_table[i] = STATIC_UNMAP; // initialize the handlers freelist for (int i=0; i != SUBTABLE_BASE-STATIC_COUNT-1; i++) handler_next_free[i] = i+STATIC_COUNT+1; handler_next_free[SUBTABLE_BASE-STATIC_COUNT-1] = STATIC_INVALID; handler_free = STATIC_COUNT; // initialize the handlers refcounts memset(handler_refcount, 0, sizeof(handler_refcount)); } //------------------------------------------------- // ~address_table - destructor //------------------------------------------------- address_table::~address_table() { } //------------------------------------------------- // map_range - map a specific entry in the address // map //------------------------------------------------- void address_table::map_range(offs_t addrstart, offs_t addrend, offs_t addrmask, offs_t addrmirror, u16 entry) { // convert addresses to bytes m_space.adjust_addresses(addrstart, addrend, addrmask, addrmirror); // validity checks assert_always(addrstart <= addrend, "address_table::map_range called with start greater than end"); assert_always((addrstart & (m_space.alignment() - 1)) == 0, "address_table::map_range called with misaligned start address"); assert_always((addrend & (m_space.alignment() - 1)) == (m_space.alignment() - 1), "address_table::map_range called with misaligned end address"); // configure the entry to our parameters (but not for static non-banked cases) handler_entry &curentry = handler(entry); if (entry <= STATIC_BANKMAX || entry >= STATIC_COUNT) curentry.configure(addrstart, addrend, addrmask, m_space.address_to_byte_end(addrmask)); // populate it populate_range_mirrored(addrstart, addrend, addrmirror, entry); // recompute any direct access on this space if it is a read modification m_space.invalidate_read_caches(entry); // verify_reference_counts(); } u16 address_table::get_free_handler() { if (handler_free == STATIC_INVALID) throw emu_fatalerror("Out of handler entries in address table"); u16 handler = handler_free; handler_free = handler_next_free[handler - STATIC_COUNT]; return handler; } //------------------------------------------------- // setup_range - finds an appropriate handler entry // and requests to populate the address map with // it //------------------------------------------------- void address_table::setup_range(offs_t addrstart, offs_t addrend, offs_t addrmask, offs_t addrmirror, u64 mask, std::list &entries) { // Careful, you can't shift by 64 or more u64 testmask = (1ULL << (m_space.data_width()-1) << 1) - 1; if((mask & testmask) == 0 || (mask & testmask) == testmask) setup_range_solid(addrstart, addrend, addrmask, addrmirror, entries); else setup_range_masked(addrstart, addrend, addrmask, addrmirror, mask, entries); } //------------------------------------------------- // setup_range_solid - finds an appropriate handler // entry and requests to populate the address map with // it. Replace what's there. //------------------------------------------------- void address_table::setup_range_solid(offs_t addrstart, offs_t addrend, offs_t addrmask, offs_t addrmirror, std::list &entries) { // Grab a free entry u16 entry = get_free_handler(); // Add it in the "to be setup" list entries.push_back(entry); // Configure and map it map_range(addrstart, addrend, addrmask, addrmirror, entry); } //------------------------------------------------- // setup_range_solid - finds an appropriate handler // entry and requests to populate the address map with // it. Handle non-overlapping subunits. //------------------------------------------------- namespace { struct subrange { offs_t start, end; subrange(offs_t _start, offs_t _end) : start(_start), end(_end) {} }; } void address_table::setup_range_masked(offs_t addrstart, offs_t addrend, offs_t addrmask, offs_t addrmirror, u64 mask, std::list &entries) { // convert addresses to bytes m_space.adjust_addresses(addrstart, addrend, addrmask, addrmirror); // Validity checks assert_always(addrstart <= addrend, "address_table::setup_range called with start greater than end"); assert_always((addrstart & (m_space.alignment() - 1)) == 0, "address_table::setup_range called with misaligned start address"); assert_always((addrend & (m_space.alignment() - 1)) == (m_space.alignment() - 1), "address_table::setup_range called with misaligned end address"); offs_t bytemask = m_space.address_to_byte_end(addrmask); // Scan the memory to see what has to be done std::list range_override; std::map > range_partial; offs_t base_mirror = 0; do { offs_t base_address = base_mirror | addrstart; offs_t end_address = base_mirror | addrend; do { offs_t range_start, range_end; u16 entry = derive_range(base_address, range_start, range_end); u32 stop_address = range_end > end_address ? end_address : range_end; if (entry < STATIC_COUNT || handler(entry).overriden_by_mask(mask)) range_override.push_back(subrange(base_address, stop_address)); else range_partial[entry].push_back(subrange(base_address, stop_address)); base_address = stop_address + 1; } while (base_address != end_address + 1); // Efficient method to go to the next range start given a mirroring mask base_mirror = (base_mirror + 1 + ~addrmirror) & addrmirror; } while (base_mirror); // Ranges in range_override must be plain replaced by the new handler if (!range_override.empty()) { // Grab a free entry u16 entry = get_free_handler(); // configure the entry to our parameters handler_entry &curentry = handler(entry); curentry.configure(addrstart, addrend, addrmask, bytemask); // Populate it wherever needed for (std::list::const_iterator i = range_override.begin(); i != range_override.end(); ++i) populate_range(i->start, i->end, entry); // Add it in the "to be setup" list entries.push_back(entry); // recompute any direct access on this space if it is a read modification m_space.invalidate_read_caches(entry); } // Ranges in range_partial must be duplicated then partially changed if (!range_partial.empty()) { for (std::map >::const_iterator i = range_partial.begin(); i != range_partial.end(); ++i) { // Theorically, if the handler to change matches the // characteristics of ours, we can directly change it. In // practice, it's more complex than that because the // mirroring is not saved, so we're not sure there aren't // mappings on the handler outside of the zones we're // supposed to change. So we won't do the obvious // optimization at this point. // Get the original handler handler_entry *base_entry = &handler(i->first); offs_t previous_bytemask = base_entry->bytemask(); // Grab a new handler and copy it there u16 entry = get_free_handler(); handler_entry &curentry = handler(entry); curentry.copy(base_entry); // Clear the colliding entries curentry.clear_conflicting_subunits(mask); // Reconfigure the base addresses curentry.configure(addrstart, addrend, addrmask, bytemask); // Populate it wherever needed for (const auto & elem : i->second) populate_range(elem.start, elem.end, entry); curentry.expand_bytemask(previous_bytemask); // Add it in the "to be setup" list entries.push_back(entry); // recompute any direct access on this space if it is a read modification m_space.invalidate_read_caches(entry); } } // verify_reference_counts(); } //------------------------------------------------- // verify_reference_counts - check how much of a // hash we've made of things //------------------------------------------------- void address_table::verify_reference_counts() { int actual_refcounts[SUBTABLE_BASE-STATIC_COUNT]; memset(actual_refcounts, 0, sizeof(actual_refcounts)); bool subtable_seen[TOTAL_MEMORY_BANKS - SUBTABLE_BASE]; memset(subtable_seen, 0, sizeof(subtable_seen)); for (int level1 = 0; level1 != 1 << LEVEL1_BITS; level1++) { u16 l1_entry = m_table[level1]; if (l1_entry >= SUBTABLE_BASE) { assert(m_large); if (subtable_seen[l1_entry - SUBTABLE_BASE]) continue; subtable_seen[l1_entry - SUBTABLE_BASE] = true; const u16 *subtable = subtable_ptr(l1_entry); for (int level2 = 0; level2 != 1 << LEVEL2_BITS; level2++) { u16 l2_entry = subtable[level2]; assert(l2_entry < SUBTABLE_BASE); if (l2_entry >= STATIC_COUNT) actual_refcounts[l2_entry - STATIC_COUNT]++; } } else if (l1_entry >= STATIC_COUNT) actual_refcounts[l1_entry - STATIC_COUNT]++; } if (memcmp(actual_refcounts, handler_refcount, sizeof(handler_refcount))) { osd_printf_error("Refcount failure:\n"); for(int i = STATIC_COUNT; i != SUBTABLE_BASE; i++) osd_printf_error("%02x: %4x .. %4x\n", i, handler_refcount[i-STATIC_COUNT], actual_refcounts[i-STATIC_COUNT]); throw emu_fatalerror("memory.c: refcounts are fucked.\n"); } } //------------------------------------------------- // populate_range - assign a memory handler to a // range of addresses //------------------------------------------------- void address_table::populate_range(offs_t addrstart, offs_t addrend, u16 handlerindex) { offs_t l2mask = (1 << level2_bits()) - 1; offs_t l1start = addrstart >> level2_bits(); offs_t l2start = addrstart & l2mask; offs_t l1stop = addrend >> level2_bits(); offs_t l2stop = addrend & l2mask; // sanity check if (addrstart > addrend) return; // handle the starting edge if it's not on a block boundary if (l2start != 0) { u16 *subtable = subtable_open(l1start); // if the start and stop end within the same block, handle that if (l1start == l1stop) { handler_ref(handlerindex, l2stop-l2start+1); for (int i = l2start; i <= l2stop; i++) { handler_unref(subtable[i]); subtable[i] = handlerindex; } subtable_close(l1start); return; } // otherwise, fill until the end handler_ref(handlerindex, l2mask - l2start + 1); for (int i = l2start; i <= l2mask; i++) { handler_unref(subtable[i]); subtable[i] = handlerindex; } subtable_close(l1start); if (l1start != (offs_t)~0) l1start++; } // handle the trailing edge if it's not on a block boundary if (l2stop != l2mask) { u16 *subtable = subtable_open(l1stop); // fill from the beginning handler_ref(handlerindex, l2stop+1); for (int i = 0; i <= l2stop; i++) { handler_unref(subtable[i]); subtable[i] = handlerindex; } subtable_close(l1stop); // if the start and stop end within the same block, handle that if (l1start == l1stop) return; if (l1stop != 0) l1stop--; } // now fill in the middle tables handler_ref(handlerindex, l1stop - l1start + 1); for (offs_t l1index = l1start; l1index <= l1stop; l1index++) { u16 subindex = m_table[l1index]; // if we have a subtable here, release it if (subindex >= SUBTABLE_BASE) subtable_release(subindex); else handler_unref(subindex); m_table[l1index] = handlerindex; } } //------------------------------------------------- // populate_range_mirrored - assign a memory // handler to a range of addresses including // mirrors //------------------------------------------------- void address_table::populate_range_mirrored(offs_t addrstart, offs_t addrend, offs_t addrmirror, u16 handlerindex) { // determine the mirror bits offs_t lmirrorbits = 0; offs_t lmirrorbit[32]; for (int bit = 0; bit < level2_bits(); bit++) if (addrmirror & (1 << bit)) lmirrorbit[lmirrorbits++] = 1 << bit; offs_t hmirrorbits = 0; offs_t hmirrorbit[32]; for (int bit = level2_bits(); bit < 32; bit++) if (addrmirror & (1 << bit)) hmirrorbit[hmirrorbits++] = 1 << bit; // loop over mirrors in the level 2 table u16 prev_entry = STATIC_INVALID; int prev_index = 0; for (offs_t hmirrorcount = 0; hmirrorcount < (1 << hmirrorbits); hmirrorcount++) { // compute the base of this mirror offs_t hmirrorbase = 0; for (int bit = 0; bit < hmirrorbits; bit++) if (hmirrorcount & (1 << bit)) hmirrorbase |= hmirrorbit[bit]; // invalidate any intersecting cached ranges for (offs_t lmirrorcount = 0; lmirrorcount < (1 << lmirrorbits); lmirrorcount++) { // compute the base of this mirror offs_t lmirrorbase = hmirrorbase; for (int bit = 0; bit < lmirrorbits; bit++) if (lmirrorcount & (1 << bit)) lmirrorbase |= lmirrorbit[bit]; m_space.invalidate_read_caches(addrstart + lmirrorbase, addrend + lmirrorbase); } // if this is not our first time through, and the level 2 entry matches the previous // level 2 entry, just do a quick map and get out; note that this only works for entries // which don't span multiple level 1 table entries int cur_index = level1_index(addrstart + hmirrorbase); if (cur_index == level1_index(addrend + hmirrorbase)) { if (hmirrorcount != 0 && prev_entry == m_table[cur_index]) { VPRINTF(("Quick mapping subtable at %08X to match subtable at %08X\n", cur_index << level2_bits(), prev_index << level2_bits())); // release the subtable if the old value was a subtable if (m_table[cur_index] >= SUBTABLE_BASE) subtable_release(m_table[cur_index]); else handler_unref(m_table[cur_index]); // reallocate the subtable if the new value is a subtable if (m_table[prev_index] >= SUBTABLE_BASE) subtable_realloc(m_table[prev_index]); else handler_ref(m_table[prev_index], 1); // set the new value and short-circuit the mapping step m_table[cur_index] = m_table[prev_index]; continue; } prev_index = cur_index; prev_entry = m_table[cur_index]; } // loop over mirrors in the level 1 table for (offs_t lmirrorcount = 0; lmirrorcount < (1 << lmirrorbits); lmirrorcount++) { // compute the base of this mirror offs_t lmirrorbase = hmirrorbase; for (int bit = 0; bit < lmirrorbits; bit++) if (lmirrorcount & (1 << bit)) lmirrorbase |= lmirrorbit[bit]; // populate the tables populate_range(addrstart + lmirrorbase, addrend + lmirrorbase, handlerindex); } } } //------------------------------------------------- // derive_range - look up the entry for a memory // range, and then compute the extent of that // range based on the lookup tables //------------------------------------------------- u16 address_table::derive_range(offs_t address, offs_t &addrstart, offs_t &addrend) const { // look up the initial address to get the entry we care about u16 l1entry; u16 entry = l1entry = m_table[level1_index(address)]; if (l1entry >= SUBTABLE_BASE) entry = m_table[level2_index(l1entry, address)]; // use the addrmask of the entry to set minimum and maximum bounds offs_t minscan, maxscan; handler(entry).mirrored_start_end(address, minscan, maxscan); // first scan backwards to find the start address u16 curl1entry = l1entry; u16 curentry = entry; addrstart = address; while (1) { // if we need to scan the subtable, do it if (curentry != curl1entry) { u32 minindex = level2_index(curl1entry, 0); u32 index; // scan backwards from the current address, until the previous entry doesn't match for (index = level2_index(curl1entry, addrstart); index > minindex; index--, addrstart -= 1) if (m_table[index - 1] != entry) break; // if we didn't hit the beginning, then we're finished scanning if (index != minindex) break; } // move to the beginning of this L1 entry; stop at the minimum address addrstart &= ~((1 << level2_bits()) - 1); if (addrstart <= minscan) break; // look up the entry of the byte at the end of the previous L1 entry; if it doesn't match, stop curentry = curl1entry = m_table[level1_index(addrstart - 1)]; if (curl1entry >= SUBTABLE_BASE) curentry = m_table[level2_index(curl1entry, addrstart - 1)]; if (curentry != entry) break; // move into the previous entry and resume searching addrstart -= 1; } // then scan forwards to find the end address curl1entry = l1entry; curentry = entry; addrend = address; while (1) { // if we need to scan the subtable, do it if (curentry != curl1entry) { u32 maxindex = level2_index(curl1entry, ~0); u32 index; // scan forwards from the current address, until the next entry doesn't match for (index = level2_index(curl1entry, addrend); index < maxindex; index++, addrend += 1) if (m_table[index + 1] != entry) break; // if we didn't hit the end, then we're finished scanning if (index != maxindex) break; } // move to the end of this L1 entry; stop at the maximum address addrend |= (1 << level2_bits()) - 1; if (addrend >= maxscan) break; // look up the entry of the byte at the start of the next L1 entry; if it doesn't match, stop curentry = curl1entry = m_table[level1_index(addrend + 1)]; if (curl1entry >= SUBTABLE_BASE) curentry = m_table[level2_index(curl1entry, addrend + 1)]; if (curentry != entry) break; // move into the next entry and resume searching addrend += 1; } return entry; } //------------------------------------------------- // mask_all_handlers - apply a mask to all // address handlers //------------------------------------------------- void address_table::mask_all_handlers(offs_t mask) { // we don't loop over map entries because the mask applies to static handlers as well for (int entrynum = 0; entrynum < ENTRY_COUNT; entrynum++) handler(entrynum).apply_mask(mask); } //************************************************************************** // SUBTABLE MANAGEMENT //************************************************************************** //------------------------------------------------- // subtable_alloc - allocate a fresh subtable // and set its usecount to 1 //------------------------------------------------- u16 address_table::subtable_alloc() { // loop while (1) { // find a subtable with a usecount of 0 for (u16 subindex = 0; subindex < SUBTABLE_COUNT; subindex++) if (m_subtable[subindex].m_usecount == 0) { // if this is past our allocation budget, allocate some more if (subindex >= m_subtable_alloc) { m_subtable_alloc += SUBTABLE_ALLOC; u32 newsize = (1 << LEVEL1_BITS) + (m_subtable_alloc << level2_bits()); bool was_live = (m_live_lookup == &m_table[0]); int oldsize = m_table.size(); m_table.resize(newsize); memset(&m_table[oldsize], 0, (newsize-oldsize)*sizeof(m_table[0])); if (was_live) m_live_lookup = &m_table[0]; } // bump the usecount and return m_subtable[subindex].m_usecount++; return subindex + SUBTABLE_BASE; } // merge any subtables we can if (!subtable_merge()) fatalerror("Ran out of subtables!\n"); } } //------------------------------------------------- // subtable_realloc - increment the usecount on // a subtable //------------------------------------------------- void address_table::subtable_realloc(u16 subentry) { u16 subindex = subentry - SUBTABLE_BASE; // sanity check if (m_subtable[subindex].m_usecount <= 0) fatalerror("Called subtable_realloc on a table with a usecount of 0\n"); // increment the usecount m_subtable[subindex].m_usecount++; } //------------------------------------------------- // subtable_merge - merge any duplicate // subtables //------------------------------------------------- int address_table::subtable_merge() { int merged = 0; u16 subindex; VPRINTF(("Merging subtables....\n")); // okay, we failed; update all the checksums and merge tables for (subindex = 0; subindex < SUBTABLE_COUNT; subindex++) if (!m_subtable[subindex].m_checksum_valid && m_subtable[subindex].m_usecount != 0) { u32 *subtable = reinterpret_cast(subtable_ptr(subindex + SUBTABLE_BASE)); u32 checksum = 0; // update the checksum for (int l2index = 0; l2index < (1 << level2_bits())/4; l2index++) checksum += subtable[l2index]; m_subtable[subindex].m_checksum = checksum; m_subtable[subindex].m_checksum_valid = true; } // see if there's a matching checksum for (subindex = 0; subindex < SUBTABLE_COUNT; subindex++) if (m_subtable[subindex].m_usecount != 0) { u16 *subtable = subtable_ptr(subindex + SUBTABLE_BASE); u32 checksum = m_subtable[subindex].m_checksum; u16 sumindex; for (sumindex = subindex + 1; sumindex < SUBTABLE_COUNT; sumindex++) if (m_subtable[sumindex].m_usecount != 0 && m_subtable[sumindex].m_checksum == checksum && !memcmp(subtable, subtable_ptr(sumindex + SUBTABLE_BASE), 2*(1 << level2_bits()))) { int l1index; VPRINTF(("Merging subtable %d and %d....\n", subindex, sumindex)); // find all the entries in the L1 tables that pointed to the old one, and point them to the merged table for (l1index = 0; l1index <= (0xffffffffUL >> level2_bits()); l1index++) if (m_table[l1index] == sumindex + SUBTABLE_BASE) { subtable_release(sumindex + SUBTABLE_BASE); subtable_realloc(subindex + SUBTABLE_BASE); m_table[l1index] = subindex + SUBTABLE_BASE; merged++; } } } return merged; } //------------------------------------------------- // subtable_release - decrement the usecount on // a subtable and free it if we're done //------------------------------------------------- void address_table::subtable_release(u16 subentry) { u16 subindex = subentry - SUBTABLE_BASE; // sanity check if (m_subtable[subindex].m_usecount <= 0) fatalerror("Called subtable_release on a table with a usecount of 0\n"); // decrement the usecount and clear the checksum if we're at 0 // also unref the subhandlers m_subtable[subindex].m_usecount--; if (m_subtable[subindex].m_usecount == 0) { m_subtable[subindex].m_checksum = 0; u16 *subtable = subtable_ptr(subentry); for (int i = 0; i < (1 << LEVEL2_BITS); i++) handler_unref(subtable[i]); } } //------------------------------------------------- // subtable_open - gain access to a subtable for // modification //------------------------------------------------- u16 *address_table::subtable_open(offs_t l1index) { u16 subentry = m_table[l1index]; // if we don't have a subtable yet, allocate a new one if (subentry < SUBTABLE_BASE) { int size = 1 << level2_bits(); u16 newentry = subtable_alloc(); handler_ref(subentry, size-1); u16 *subptr = subtable_ptr(newentry); for (int i=0; i 1) { u16 newentry = subtable_alloc(); // allocate may cause some additional merging -- look up the subentry again // when we're done; it should still require a split subentry = m_table[l1index]; assert(subentry >= SUBTABLE_BASE); assert(m_subtable[subentry - SUBTABLE_BASE].m_usecount > 1); int size = 1 << level2_bits(); u16 *src = subtable_ptr(subentry); for(int i=0; i != size; i++) handler_ref(src[i], 1); memcpy(subtable_ptr(newentry), src, 2*size); subtable_release(subentry); m_table[l1index] = newentry; m_subtable[newentry - SUBTABLE_BASE].m_checksum = m_subtable[subentry - SUBTABLE_BASE].m_checksum; subentry = newentry; } // mark the table dirty m_subtable[subentry - SUBTABLE_BASE].m_checksum_valid = false; // return the pointer to the subtable return subtable_ptr(subentry); } //------------------------------------------------- // subtable_close - stop access to a subtable //------------------------------------------------- void address_table::subtable_close(offs_t l1index) { // defer any merging until we run out of tables } //------------------------------------------------- // handler_name - return friendly string // description of a handler //------------------------------------------------- const char *address_table::handler_name(u16 entry) const { // banks have names if (entry >= STATIC_BANK1 && entry <= STATIC_BANKMAX) for (auto &info : m_space.manager().banks()) if (info.second->index() == entry) return info.second->name(); // constant strings for static entries if (entry == STATIC_INVALID) return "invalid"; if (entry == STATIC_NOP) return "nop"; if (entry == STATIC_UNMAP) return "unmapped"; if (entry == STATIC_WATCHPOINT) return "watchpoint"; static char desc[4096]; handler(entry).description(desc); if (desc[0]) return desc; return "???"; } //------------------------------------------------- // address_table_read - constructor //------------------------------------------------- address_table_read::address_table_read(address_space &space, bool large) : address_table(space, large) { // allocate handlers for each entry, prepopulating the bankptrs for banks for (int entrynum = 0; entrynum < ARRAY_LENGTH(m_handlers); entrynum++) { u8 **bankptr = (entrynum >= STATIC_BANK1 && entrynum <= STATIC_BANKMAX) ? space.manager().bank_pointer_addr(entrynum) : nullptr; m_handlers[entrynum] = std::make_unique(space.data_width(), space.endianness(), bankptr); } // we have to allocate different object types based on the data bus width switch (space.data_width()) { // 8-bit case case 8: m_handlers[STATIC_UNMAP]->set_delegate(read8_delegate(FUNC(address_table_read::unmap_r), this)); m_handlers[STATIC_NOP]->set_delegate(read8_delegate(FUNC(address_table_read::nop_r), this)); m_handlers[STATIC_WATCHPOINT]->set_delegate(read8_delegate(FUNC(address_table_read::watchpoint_r), this)); break; // 16-bit case case 16: m_handlers[STATIC_UNMAP]->set_delegate(read16_delegate(FUNC(address_table_read::unmap_r), this)); m_handlers[STATIC_NOP]->set_delegate(read16_delegate(FUNC(address_table_read::nop_r), this)); m_handlers[STATIC_WATCHPOINT]->set_delegate(read16_delegate(FUNC(address_table_read::watchpoint_r), this)); break; // 32-bit case case 32: m_handlers[STATIC_UNMAP]->set_delegate(read32_delegate(FUNC(address_table_read::unmap_r), this)); m_handlers[STATIC_NOP]->set_delegate(read32_delegate(FUNC(address_table_read::nop_r), this)); m_handlers[STATIC_WATCHPOINT]->set_delegate(read32_delegate(FUNC(address_table_read::watchpoint_r), this)); break; // 64-bit case case 64: m_handlers[STATIC_UNMAP]->set_delegate(read64_delegate(FUNC(address_table_read::unmap_r), this)); m_handlers[STATIC_NOP]->set_delegate(read64_delegate(FUNC(address_table_read::nop_r), this)); m_handlers[STATIC_WATCHPOINT]->set_delegate(read64_delegate(FUNC(address_table_read::watchpoint_r), this)); break; } // reset the byte masks on the special handlers to open up the full address space for proper reporting m_handlers[STATIC_UNMAP]->configure(0, space.addrmask(), ~0, ~0); m_handlers[STATIC_NOP]->configure(0, space.addrmask(), ~0, ~0); m_handlers[STATIC_WATCHPOINT]->configure(0, space.addrmask(), ~0, ~0); } //------------------------------------------------- // address_table_read - destructor //------------------------------------------------- address_table_read::~address_table_read() { } //------------------------------------------------- // handler - return the generic handler entry for // this index //------------------------------------------------- handler_entry &address_table_read::handler(u32 index) const { assert(index < ARRAY_LENGTH(m_handlers)); return *m_handlers[index]; } //------------------------------------------------- // address_table_write - constructor //------------------------------------------------- address_table_write::address_table_write(address_space &space, bool large) : address_table(space, large) { // allocate handlers for each entry, prepopulating the bankptrs for banks for (int entrynum = 0; entrynum < ARRAY_LENGTH(m_handlers); entrynum++) { u8 **bankptr = (entrynum >= STATIC_BANK1 && entrynum <= STATIC_BANKMAX) ? space.manager().bank_pointer_addr(entrynum) : nullptr; m_handlers[entrynum] = std::make_unique(space.data_width(), space.endianness(), bankptr); } // we have to allocate different object types based on the data bus width switch (space.data_width()) { // 8-bit case case 8: m_handlers[STATIC_UNMAP]->set_delegate(write8_delegate(FUNC(address_table_write::unmap_w), this)); m_handlers[STATIC_NOP]->set_delegate(write8_delegate(FUNC(address_table_write::nop_w), this)); m_handlers[STATIC_WATCHPOINT]->set_delegate(write8_delegate(FUNC(address_table_write::watchpoint_w), this)); break; // 16-bit case case 16: m_handlers[STATIC_UNMAP]->set_delegate(write16_delegate(FUNC(address_table_write::unmap_w), this)); m_handlers[STATIC_NOP]->set_delegate(write16_delegate(FUNC(address_table_write::nop_w), this)); m_handlers[STATIC_WATCHPOINT]->set_delegate(write16_delegate(FUNC(address_table_write::watchpoint_w), this)); break; // 32-bit case case 32: m_handlers[STATIC_UNMAP]->set_delegate(write32_delegate(FUNC(address_table_write::unmap_w), this)); m_handlers[STATIC_NOP]->set_delegate(write32_delegate(FUNC(address_table_write::nop_w), this)); m_handlers[STATIC_WATCHPOINT]->set_delegate(write32_delegate(FUNC(address_table_write::watchpoint_w), this)); break; // 64-bit case case 64: m_handlers[STATIC_UNMAP]->set_delegate(write64_delegate(FUNC(address_table_write::unmap_w), this)); m_handlers[STATIC_NOP]->set_delegate(write64_delegate(FUNC(address_table_write::nop_w), this)); m_handlers[STATIC_WATCHPOINT]->set_delegate(write64_delegate(FUNC(address_table_write::watchpoint_w), this)); break; } // reset the byte masks on the special handlers to open up the full address space for proper reporting m_handlers[STATIC_UNMAP]->configure(0, space.addrmask(), ~0, ~0); m_handlers[STATIC_NOP]->configure(0, space.addrmask(), ~0, ~0); m_handlers[STATIC_WATCHPOINT]->configure(0, space.addrmask(), ~0, ~0); } //------------------------------------------------- // address_table_write - destructor //------------------------------------------------- address_table_write::~address_table_write() { } //------------------------------------------------- // handler - return the generic handler entry for // this index //------------------------------------------------- handler_entry &address_table_write::handler(u32 index) const { assert(index < ARRAY_LENGTH(m_handlers)); return *m_handlers[index]; } //************************************************************************** // DIRECT MEMORY RANGES //************************************************************************** //------------------------------------------------- // direct_read_data - constructor //------------------------------------------------- template direct_read_data::direct_read_data(address_space &space) : m_space(space), m_ptr(nullptr), m_addrmask(space.addrmask()), m_addrstart(1), m_addrend(0), m_entry(STATIC_UNMAP) { } //------------------------------------------------- // ~direct_read_data - destructor //------------------------------------------------- template direct_read_data::~direct_read_data() { } //------------------------------------------------- // set_direct_region - called by device cores to // update the opcode base for the given address //------------------------------------------------- template bool direct_read_data::set_direct_region(offs_t address) { // find or allocate a matching range direct_range *range = find_range(address, m_entry); // if we don't map to a bank, return false if (m_entry < STATIC_BANK1 || m_entry > STATIC_BANKMAX) { // ensure future updates to land here as well until we get back into a bank m_addrend = 0; m_addrstart = 1; return false; } u8 *base = *m_space.manager().bank_pointer_addr(m_entry); // compute the adjusted base offs_t maskedbits = address & ~m_space.addrmask(); const handler_entry_read &handler = m_space.read().handler_read(m_entry); m_addrmask = handler.addrmask(); u32 delta = handler.addrstart() & m_addrmask; if(AddrShift < 0) delta = delta << iabs(AddrShift); else if(AddrShift > 0) delta = delta >> iabs(AddrShift); m_ptr = base - delta; m_addrstart = maskedbits | range->m_addrstart; m_addrend = maskedbits | range->m_addrend; return true; } //------------------------------------------------- // find_range - find a byte address in a range //------------------------------------------------- template typename direct_read_data::direct_range *direct_read_data::find_range(offs_t address, u16 &entry) { // determine which entry address &= m_space.m_addrmask; entry = m_space.read().lookup_live_nowp(address); // scan our table for (auto &range : m_rangelist[entry]) if (address >= range.m_addrstart && address <= range.m_addrend) return ⦥ // didn't find out; create a new one direct_range range; m_space.read().derive_range(address, range.m_addrstart, range.m_addrend); m_rangelist[entry].push_front(range); return &m_rangelist[entry].front(); } //------------------------------------------------- // remove_intersecting_ranges - remove all cached // ranges that intersect the given address range //------------------------------------------------- template void direct_read_data::remove_intersecting_ranges(offs_t addrstart, offs_t addrend) { // loop over all entries for (auto & elem : m_rangelist) { // loop over all ranges in this entry's list for (auto range = elem.begin(); range!=elem.end();) { // if we intersect, remove if (addrstart <= range->m_addrend && addrend >= range->m_addrstart) range = elem.erase(range); else range ++; } } } template class direct_read_data<3>; template class direct_read_data<0>; template class direct_read_data<-1>; template class direct_read_data<-2>; template class direct_read_data<-3>; //************************************************************************** // MEMORY BLOCK //************************************************************************** //------------------------------------------------- // memory_block - constructor //------------------------------------------------- memory_block::memory_block(address_space &space, offs_t addrstart, offs_t addrend, void *memory) : m_machine(space.machine()), m_space(space), m_addrstart(addrstart), m_addrend(addrend), m_data(reinterpret_cast(memory)) { offs_t const length = space.address_to_byte(addrend + 1 - addrstart); VPRINTF(("block_allocate('%s',%s,%08X,%08X,%p)\n", space.device().tag(), space.name(), addrstart, addrend, memory)); // allocate a block if needed if (m_data == nullptr) { if (length < 4096) { m_allocated.resize(length); memset(&m_allocated[0], 0, length); m_data = &m_allocated[0]; } else { m_allocated.resize(length + 0xfff); memset(&m_allocated[0], 0, length + 0xfff); m_data = reinterpret_cast((reinterpret_cast(&m_allocated[0]) + 0xfff) & ~0xfff); } } // register for saving, but only if we're not part of a memory region if (space.manager().region_containing(m_data, length) != nullptr) VPRINTF(("skipping save of this memory block as it is covered by a memory region\n")); else { int bytes_per_element = space.data_width() / 8; std::string name = string_format("%08x-%08x", addrstart, addrend); space.machine().save().save_memory(&space.device(), "memory", space.device().tag(), space.spacenum(), name.c_str(), m_data, bytes_per_element, (u32)length / bytes_per_element); } } //------------------------------------------------- // memory_block - destructor //------------------------------------------------- memory_block::~memory_block() { } //************************************************************************** // MEMORY BANK //************************************************************************** //------------------------------------------------- // memory_bank - constructor //------------------------------------------------- memory_bank::memory_bank(address_space &space, int index, offs_t addrstart, offs_t addrend, const char *tag) : m_machine(space.machine()), m_baseptr(space.manager().bank_pointer_addr(index)), m_index(index), m_anonymous(tag == nullptr), m_addrstart(addrstart), m_addrend(addrend), m_curentry(BANK_ENTRY_UNSPECIFIED) { // generate an internal tag if we don't have one if (tag == nullptr) { m_tag = string_format("~%d~", index); m_name = string_format("Internal bank #%d", index); } else { m_tag.assign(tag); m_name = string_format("Bank '%s'", tag); } if (!m_anonymous && space.machine().save().registration_allowed()) space.machine().save().save_item(&space.device(), "memory", m_tag.c_str(), 0, NAME(m_curentry)); } //------------------------------------------------- // memory_bank - destructor //------------------------------------------------- memory_bank::~memory_bank() { } //------------------------------------------------- // references_space - walk the list of references // to find a match against the provided space // and read/write //------------------------------------------------- bool memory_bank::references_space(const address_space &space, read_or_write readorwrite) const { for (auto &ref : m_reflist) if (ref->matches(space, readorwrite)) return true; return false; } //------------------------------------------------- // add_reference - add a new reference to the // given space //------------------------------------------------- void memory_bank::add_reference(address_space &space, read_or_write readorwrite) { // if we already have a reference, skip it if (references_space(space, readorwrite)) return; m_reflist.push_back(std::make_unique(space, readorwrite)); } //------------------------------------------------- // invalidate_references - force updates on all // referencing address spaces //------------------------------------------------- void memory_bank::invalidate_references() { // invalidate all the direct references to any referenced address spaces for (auto &ref : m_reflist) ref->space().invalidate_read_caches(); } //------------------------------------------------- // set_base - set the bank base explicitly //------------------------------------------------- void memory_bank::set_base(void *base) { // nullptr is not an option if (base == nullptr) throw emu_fatalerror("memory_bank::set_base called nullptr base"); // set the base and invalidate any referencing spaces *m_baseptr = reinterpret_cast(base); invalidate_references(); } //------------------------------------------------- // set_entry - set the base to a pre-configured // entry //------------------------------------------------- void memory_bank::set_entry(int entrynum) { // validate if (m_anonymous) throw emu_fatalerror("memory_bank::set_entry called for anonymous bank"); if (entrynum < 0 || entrynum >= int(m_entry.size())) throw emu_fatalerror("memory_bank::set_entry called with out-of-range entry %d", entrynum); if (m_entry[entrynum].m_ptr == nullptr) throw emu_fatalerror("memory_bank::set_entry called for bank '%s' with invalid bank entry %d", m_tag.c_str(), entrynum); m_curentry = entrynum; *m_baseptr = m_entry[entrynum].m_ptr; // invalidate referencing spaces invalidate_references(); } //------------------------------------------------- // expand_entries - expand the allocated array // of entries //------------------------------------------------- void memory_bank::expand_entries(int entrynum) { // allocate a new array and copy from the old one; zero out the new entries int old_size = m_entry.size(); m_entry.resize(entrynum + 1); memset(&m_entry[old_size], 0, (entrynum+1-old_size)*sizeof(m_entry[0])); } //------------------------------------------------- // configure_entry - configure an entry //------------------------------------------------- void memory_bank::configure_entry(int entrynum, void *base) { // must be positive if (entrynum < 0) throw emu_fatalerror("memory_bank::configure_entry called with out-of-range entry %d", entrynum); // if we haven't allocated this many entries yet, expand our array if (entrynum >= int(m_entry.size())) expand_entries(entrynum); // set the entry m_entry[entrynum].m_ptr = reinterpret_cast(base); // if the bank base is not configured, and we're the first entry, set us up if (*m_baseptr == nullptr && entrynum == 0) *m_baseptr = m_entry[entrynum].m_ptr; } //------------------------------------------------- // configure_entries - configure multiple entries //------------------------------------------------- void memory_bank::configure_entries(int startentry, int numentries, void *base, offs_t stride) { // fill in the requested bank entries (backwards to improve allocation) for (int entrynum = startentry + numentries - 1; entrynum >= startentry; entrynum--) configure_entry(entrynum, reinterpret_cast(base) + (entrynum - startentry) * stride); } //************************************************************************** // MEMORY REGIONS //************************************************************************** //------------------------------------------------- // memory_region - constructor //------------------------------------------------- memory_region::memory_region(running_machine &machine, const char *name, u32 length, u8 width, endianness_t endian) : m_machine(machine), m_name(name), m_buffer(length), m_endianness(endian), m_bitwidth(width * 8), m_bytewidth(width) { assert(width == 1 || width == 2 || width == 4 || width == 8); } //************************************************************************** // HANDLER ENTRY //************************************************************************** //------------------------------------------------- // handler_entry - constructor //------------------------------------------------- handler_entry::handler_entry(u8 width, endianness_t endianness, u8 **rambaseptr) : m_populated(false), m_datawidth(width), m_endianness(endianness), m_addrstart(0), m_addrend(0), m_addrmask(~0), m_bytemask(~0), m_rambaseptr(rambaseptr), m_subunits(0), m_invsubmask(0) { } //------------------------------------------------- // ~handler_entry - destructor //------------------------------------------------- handler_entry::~handler_entry() { } //------------------------------------------------- // copy - copy another handler_entry, but only // if it is populated and constitutes of one or // more subunit handlers //------------------------------------------------- void handler_entry::copy(handler_entry *entry) { assert(entry->m_populated); assert(entry->m_subunits); assert(!entry->m_rambaseptr); assert(!m_populated); m_populated = true; m_datawidth = entry->m_datawidth; m_endianness = entry->m_endianness; m_addrstart = entry->m_addrstart; m_addrend = entry->m_addrend; m_addrmask = entry->m_addrmask; m_bytemask = entry->m_bytemask; m_rambaseptr = nullptr; m_subunits = entry->m_subunits; memcpy(m_subunit_infos, entry->m_subunit_infos, m_subunits*sizeof(subunit_info)); m_invsubmask = entry->m_invsubmask; } //------------------------------------------------- // reconfigure_subunits - reconfigure the subunits // to handle a new base address //------------------------------------------------- void handler_entry::reconfigure_subunits(offs_t addrstart) { s32 delta = addrstart - m_addrstart; for (int i=0; i != m_subunits; i++) m_subunit_infos[i].m_offset += delta / (m_subunit_infos[i].m_size / 8); } //------------------------------------------------- // configure_subunits - configure the subunits // and subshift array to represent the provided // mask //------------------------------------------------- void handler_entry::configure_subunits(u64 handlermask, int handlerbits, int &start_slot, int &end_slot) { u64 unitmask = ((u64)1 << handlerbits) - 1; assert(handlermask != 0); // compute the maximum possible subunits int maxunits = m_datawidth / handlerbits; assert(maxunits > 1); assert(maxunits <= ARRAY_LENGTH(m_subunit_infos)); int shift_xor_mask = m_endianness == ENDIANNESS_LITTLE ? 0 : maxunits - 1; // walk the handlermask to find out how many we have int count = 0; for (int unitnum = 0; unitnum < maxunits; unitnum++) { u32 shift = unitnum * handlerbits; u32 scanmask = handlermask >> shift; assert((scanmask & unitmask) == 0 || (scanmask & unitmask) == unitmask); if ((scanmask & unitmask) != 0) count++; } if (count == 0 || count > maxunits) throw emu_fatalerror("Invalid subunit mask %s for %d-bit space", core_i64_hex_format(handlermask, m_datawidth / 4), m_datawidth); // make sure that the multiplier is a power of 2 int multiplier = count; while (maxunits % multiplier != 0) multiplier++; // fill in the shifts int cur_offset = 0; start_slot = m_subunits; for (int unitnum = 0; unitnum < maxunits; unitnum++) { u32 shift = (unitnum^shift_xor_mask) * handlerbits; if (((handlermask >> shift) & unitmask) != 0) { m_subunit_infos[m_subunits].m_addrmask = m_bytemask / (maxunits / multiplier); m_subunit_infos[m_subunits].m_mask = unitmask; m_subunit_infos[m_subunits].m_offset = cur_offset++; m_subunit_infos[m_subunits].m_size = handlerbits; m_subunit_infos[m_subunits].m_shift = shift; m_subunit_infos[m_subunits].m_multiplier = multiplier; m_subunits++; } } end_slot = m_subunits; // compute the inverse mask m_invsubmask = 0; for (int i = 0; i < m_subunits; i++) m_invsubmask |= u64(m_subunit_infos[i].m_mask) << m_subunit_infos[i].m_shift; m_invsubmask = ~m_invsubmask; } //------------------------------------------------- // clear_conflicting_subunits - clear the subunits // conflicting with the provided mask //------------------------------------------------- void handler_entry::clear_conflicting_subunits(u64 handlermask) { // A mask of 0 is in fact an alternative way of saying ~0 if (!handlermask) { m_subunits = 0; return; } // Start by the end to avoid unnecessary memmoves for (int i=m_subunits-1; i>=0; i--) if (((handlermask >> m_subunit_infos[i].m_shift) & m_subunit_infos[i].m_mask) != 0) { if (i != m_subunits-1) memmove (m_subunit_infos+i, m_subunit_infos+i+1, (m_subunits-i-1)*sizeof(m_subunit_infos[0])); remove_subunit(i); } // compute the inverse mask m_invsubmask = 0; for (int i = 0; i < m_subunits; i++) m_invsubmask |= u64(m_subunit_infos[i].m_mask) << m_subunit_infos[i].m_shift; m_invsubmask = ~m_invsubmask; } //------------------------------------------------- // overriden_by_mask - check whether a handler with // the provided mask fully overrides everything // that's currently present //------------------------------------------------- bool handler_entry::overriden_by_mask(u64 handlermask) { // A mask of 0 is in fact an alternative way of saying ~0 if (!handlermask) return true; // If there are no subunits, it's going to override if (!m_subunits) return true; // Check whether a subunit would be left for (int i=0; i != m_subunits; i++) if (((handlermask >> m_subunit_infos[i].m_shift) & m_subunit_infos[i].m_mask) == 0) return false; return true; } //------------------------------------------------- // description - build a printable description // of the handler //------------------------------------------------- void handler_entry::description(char *buffer) const { if (m_subunits) { for (int i=0; i != m_subunits; i++) { if (i) *buffer++ = ' '; buffer += sprintf (buffer, "%d:%d:%x:%d:%x:%s", m_subunit_infos[i].m_size, m_subunit_infos[i].m_shift, m_subunit_infos[i].m_offset, m_subunit_infos[i].m_multiplier, m_subunit_infos[i].m_addrmask, subunit_name(i)); } } else strcpy (buffer, name()); } //************************************************************************** // HANDLER ENTRY READ //************************************************************************** //------------------------------------------------- // copy - copy another handler_entry, but only // if it is populated and constitutes of one or // more subunit handlers //------------------------------------------------- void handler_entry_read::copy(handler_entry *entry) { handler_entry::copy(entry); handler_entry_read *rentry = static_cast(entry); m_read = rentry->m_read; for(int i = 0; i < m_subunits; ++i) { switch(m_subunit_infos[i].m_size) { case 8: m_subread[i].r8 = rentry->m_subread[i].r8; break; case 16: m_subread[i].r16 = rentry->m_subread[i].r16; break; case 32: m_subread[i].r32 = rentry->m_subread[i].r32; break; } } } //------------------------------------------------- // name - return the handler name, from the // appropriately-sized delegate //------------------------------------------------- const char *handler_entry_read::name() const { switch (m_datawidth) { case 8: return m_read.r8.name(); case 16: return m_read.r16.name(); case 32: return m_read.r32.name(); case 64: return m_read.r64.name(); } return nullptr; } //------------------------------------------------- // subunit_name - return the handler name, from the // appropriately-sized delegate of a subunit //------------------------------------------------- const char *handler_entry_read::subunit_name(int entry) const { switch (m_subunit_infos[entry].m_size) { case 8: return m_subread[entry].r8.name(); case 16: return m_subread[entry].r16.name(); case 32: return m_subread[entry].r32.name(); case 64: return m_subread[entry].r64.name(); } return nullptr; } //------------------------------------------------- // remove_subunit - delete a subunit specific // information and shift up the following ones //------------------------------------------------- void handler_entry_read::remove_subunit(int entry) { int moving = m_subunits - entry - 1; if (moving) { memmove(m_subread+entry, m_subread+entry+1, moving*sizeof(m_subread[0])); } m_subunits--; } //------------------------------------------------- // set_delegate - set an 8-bit delegate, and // configure a stub if necessary //------------------------------------------------- void handler_entry_read::set_delegate(read8_delegate delegate, u64 mask) { // error if no object if (!delegate.has_object()) throw emu_fatalerror("Attempted to install delegate '%s' without a bound object", delegate.name()); // make sure this is a valid size assert(m_datawidth >= 8); // if mismatched bus width, configure a stub if (m_datawidth != 8) { int start_slot, end_slot; configure_subunits(mask, 8, start_slot, end_slot); for (int i=start_slot; i != end_slot; i++) { m_subread[i].r8 = delegate; } if (m_datawidth == 16) set_delegate(read16_delegate(&handler_entry_read::read_stub_16, delegate.name(), this)); else if (m_datawidth == 32) set_delegate(read32_delegate(&handler_entry_read::read_stub_32, delegate.name(), this)); else if (m_datawidth == 64) set_delegate(read64_delegate(&handler_entry_read::read_stub_64, delegate.name(), this)); } else { m_read.r8 = delegate; } } //------------------------------------------------- // set_delegate - set a 16-bit delegate, and // configure a stub if necessary //------------------------------------------------- void handler_entry_read::set_delegate(read16_delegate delegate, u64 mask) { // error if no object if (!delegate.has_object()) throw emu_fatalerror("Attempted to install delegate '%s' without a bound object", delegate.name()); // make sure this is a valid size assert(m_datawidth >= 16); // if mismatched bus width, configure a stub if (m_datawidth != 16) { int start_slot, end_slot; configure_subunits(mask, 16, start_slot, end_slot); for (int i=start_slot; i != end_slot; i++) { m_subread[i].r16 = delegate; } if (m_datawidth == 32) set_delegate(read32_delegate(&handler_entry_read::read_stub_32, delegate.name(), this)); else if (m_datawidth == 64) set_delegate(read64_delegate(&handler_entry_read::read_stub_64, delegate.name(), this)); } else { m_read.r16 = delegate; } } //------------------------------------------------- // set_delegate - set a 32-bit delegate, and // configure a stub if necessary //------------------------------------------------- void handler_entry_read::set_delegate(read32_delegate delegate, u64 mask) { // error if no object if (!delegate.has_object()) throw emu_fatalerror("Attempted to install delegate '%s' without a bound object", delegate.name()); // make sure this is a valid size assert(m_datawidth >= 32); // if mismatched bus width, configure a stub if (m_datawidth != 32) { int start_slot, end_slot; configure_subunits(mask, 32, start_slot, end_slot); for (int i=start_slot; i != end_slot; i++) { m_subread[i].r32 = delegate; } if (m_datawidth == 64) set_delegate(read64_delegate(&handler_entry_read::read_stub_64, delegate.name(), this)); } else { m_read.r32 = delegate; } } //------------------------------------------------- // set_delegate - set a 64-bit delegate //------------------------------------------------- void handler_entry_read::set_delegate(read64_delegate delegate, u64 mask) { // error if no object if (!delegate.has_object()) throw emu_fatalerror("Attempted to install delegate '%s' without a bound object", delegate.name()); // make sure this is a valid size assert(m_datawidth >= 64); m_read.r64 = delegate; } //------------------------------------------------- // set_ioport - configure an I/O port read stub // of the appropriate size //------------------------------------------------- void handler_entry_read::set_ioport(ioport_port &ioport) { m_ioport = &ioport; if (m_datawidth == 8) set_delegate(read8_delegate(&handler_entry_read::read_stub_ioport, ioport.tag(), this)); else if (m_datawidth == 16) set_delegate(read16_delegate(&handler_entry_read::read_stub_ioport, ioport.tag(), this)); else if (m_datawidth == 32) set_delegate(read32_delegate(&handler_entry_read::read_stub_ioport, ioport.tag(), this)); else if (m_datawidth == 64) set_delegate(read64_delegate(&handler_entry_read::read_stub_ioport, ioport.tag(), this)); } //------------------------------------------------- // read_stub_16 - construct a 16-bit read from // 8-bit sources //------------------------------------------------- u16 handler_entry_read::read_stub_16(address_space &space, offs_t offset, u16 mask) { u16 result = space.unmap() & m_invsubmask; for (int index = 0; index < m_subunits; index++) { const subunit_info &si = m_subunit_infos[index]; u32 submask = (mask >> si.m_shift) & si.m_mask; if (submask) { offs_t aoffset = offset * si.m_multiplier + si.m_offset; u8 val; val = m_subread[index].r8(space, aoffset & si.m_addrmask, submask); result |= val << si.m_shift; } } return result; } //------------------------------------------------- // read_stub_32 - construct a 32-bit read from // 8-bit and 16-bit sources //------------------------------------------------- u32 handler_entry_read::read_stub_32(address_space &space, offs_t offset, u32 mask) { u32 result = space.unmap() & m_invsubmask; for (int index = 0; index < m_subunits; index++) { const subunit_info &si = m_subunit_infos[index]; u32 submask = (mask >> si.m_shift) & si.m_mask; if (submask) { offs_t aoffset = offset * si.m_multiplier + si.m_offset; u16 val = 0; switch (si.m_size) { case 8: val = m_subread[index].r8(space, aoffset & si.m_addrmask, submask); break; case 16: val = m_subread[index].r16(space, aoffset & si.m_addrmask, submask); break; } result |= val << si.m_shift; } } return result; } //------------------------------------------------- // read_stub_64 - construct a 64-bit read from // 8-bit, 16-bit and 32-bit sources //------------------------------------------------- u64 handler_entry_read::read_stub_64(address_space &space, offs_t offset, u64 mask) { u64 result = space.unmap() & m_invsubmask; for (int index = 0; index < m_subunits; index++) { const subunit_info &si = m_subunit_infos[index]; u32 submask = (mask >> si.m_shift) & si.m_mask; if (submask) { offs_t aoffset = offset * si.m_multiplier + si.m_offset; u32 val = 0; switch (si.m_size) { case 8: val = m_subread[index].r8(space, aoffset & si.m_addrmask, submask); break; case 16: val = m_subread[index].r16(space, aoffset & si.m_addrmask, submask); break; case 32: val = m_subread[index].r32(space, aoffset & si.m_addrmask, submask); break; } result |= u64(val) << si.m_shift; } } return result; } //************************************************************************** // HANDLER ENTRY WRITE //************************************************************************** //------------------------------------------------- // copy - copy another handler_entry, but only // if it is populated and constitutes of one or // more subunit handlers //------------------------------------------------- void handler_entry_write::copy(handler_entry *entry) { handler_entry::copy(entry); handler_entry_write *wentry = static_cast(entry); m_write = wentry->m_write; for(int i = 0; i < m_subunits; ++i) { switch(m_subunit_infos[i].m_size) { case 8: m_subwrite[i].w8 = wentry->m_subwrite[i].w8; break; case 16: m_subwrite[i].w16 = wentry->m_subwrite[i].w16; break; case 32: m_subwrite[i].w32 = wentry->m_subwrite[i].w32; break; } } } //------------------------------------------------- // name - return the handler name, from the // appropriately-sized delegate //------------------------------------------------- const char *handler_entry_write::name() const { switch (m_datawidth) { case 8: return m_write.w8.name(); case 16: return m_write.w16.name(); case 32: return m_write.w32.name(); case 64: return m_write.w64.name(); } return nullptr; } //------------------------------------------------- // subunit_name - return the handler name, from the // appropriately-sized delegate of a subunit //------------------------------------------------- const char *handler_entry_write::subunit_name(int entry) const { switch (m_subunit_infos[entry].m_size) { case 8: return m_subwrite[entry].w8.name(); case 16: return m_subwrite[entry].w16.name(); case 32: return m_subwrite[entry].w32.name(); case 64: return m_subwrite[entry].w64.name(); } return nullptr; } //------------------------------------------------- // remove_subunit - delete a subunit specific // information and shift up the following ones //------------------------------------------------- void handler_entry_write::remove_subunit(int entry) { int moving = m_subunits - entry - 1; if (moving) { memmove(m_subwrite+entry, m_subwrite+entry+1, moving*sizeof(m_subwrite[0])); } m_subunits--; } //------------------------------------------------- // set_delegate - set an 8-bit delegate, and // configure a stub if necessary //------------------------------------------------- void handler_entry_write::set_delegate(write8_delegate delegate, u64 mask) { assert(m_datawidth >= 8); // if mismatched bus width, configure a stub if (m_datawidth != 8) { int start_slot, end_slot; configure_subunits(mask, 8, start_slot, end_slot); for (int i=start_slot; i != end_slot; i++) { m_subwrite[i].w8 = delegate; } if (m_datawidth == 16) set_delegate(write16_delegate(&handler_entry_write::write_stub_16, delegate.name(), this)); else if (m_datawidth == 32) set_delegate(write32_delegate(&handler_entry_write::write_stub_32, delegate.name(), this)); else if (m_datawidth == 64) set_delegate(write64_delegate(&handler_entry_write::write_stub_64, delegate.name(), this)); } else { m_write.w8 = delegate; } } //------------------------------------------------- // set_delegate - set a 16-bit delegate, and // configure a stub if necessary //------------------------------------------------- void handler_entry_write::set_delegate(write16_delegate delegate, u64 mask) { assert(m_datawidth >= 16); // if mismatched bus width, configure a stub if (m_datawidth != 16) { int start_slot, end_slot; configure_subunits(mask, 16, start_slot, end_slot); for (int i=start_slot; i != end_slot; i++) { m_subwrite[i].w16 = delegate; } if (m_datawidth == 32) set_delegate(write32_delegate(&handler_entry_write::write_stub_32, delegate.name(), this)); else if (m_datawidth == 64) set_delegate(write64_delegate(&handler_entry_write::write_stub_64, delegate.name(), this)); } else { m_write.w16 = delegate; } } //------------------------------------------------- // set_delegate - set a 32-bit delegate, and // configure a stub if necessary //------------------------------------------------- void handler_entry_write::set_delegate(write32_delegate delegate, u64 mask) { assert(m_datawidth >= 32); // if mismatched bus width, configure a stub if (m_datawidth != 32) { int start_slot, end_slot; configure_subunits(mask, 32, start_slot, end_slot); for (int i=start_slot; i != end_slot; i++) { m_subwrite[i].w32 = delegate; } if (m_datawidth == 64) set_delegate(write64_delegate(&handler_entry_write::write_stub_64, delegate.name(), this)); } else { m_write.w32 = delegate; } } //------------------------------------------------- // set_delegate - set a 64-bit delegate //------------------------------------------------- void handler_entry_write::set_delegate(write64_delegate delegate, u64 mask) { assert(m_datawidth >= 64); m_write.w64 = delegate; } //------------------------------------------------- // set_ioport - configure an I/O port read stub // of the appropriate size //------------------------------------------------- void handler_entry_write::set_ioport(ioport_port &ioport) { m_ioport = &ioport; if (m_datawidth == 8) set_delegate(write8_delegate(&handler_entry_write::write_stub_ioport, ioport.tag(), this)); else if (m_datawidth == 16) set_delegate(write16_delegate(&handler_entry_write::write_stub_ioport, ioport.tag(), this)); else if (m_datawidth == 32) set_delegate(write32_delegate(&handler_entry_write::write_stub_ioport, ioport.tag(), this)); else if (m_datawidth == 64) set_delegate(write64_delegate(&handler_entry_write::write_stub_ioport, ioport.tag(), this)); } //------------------------------------------------- // write_stub_16 - construct a 16-bit write from // 8-bit sources //------------------------------------------------- void handler_entry_write::write_stub_16(address_space &space, offs_t offset, u16 data, u16 mask) { for (int index = 0; index < m_subunits; index++) { const subunit_info &si = m_subunit_infos[index]; u32 submask = (mask >> si.m_shift) & si.m_mask; if (submask) { offs_t aoffset = offset * si.m_multiplier + si.m_offset; u8 adata = data >> si.m_shift; m_subwrite[index].w8(space, aoffset & si.m_addrmask, adata, submask); } } } //------------------------------------------------- // write_stub_32 - construct a 32-bit write from // 8-bit and 16-bit sources //------------------------------------------------- void handler_entry_write::write_stub_32(address_space &space, offs_t offset, u32 data, u32 mask) { for (int index = 0; index < m_subunits; index++) { const subunit_info &si = m_subunit_infos[index]; u32 submask = (mask >> si.m_shift) & si.m_mask; if (submask) { offs_t aoffset = offset * si.m_multiplier + si.m_offset; u16 adata = data >> si.m_shift; switch (si.m_size) { case 8: m_subwrite[index].w8(space, aoffset & si.m_addrmask, adata, submask); break; case 16: m_subwrite[index].w16(space, aoffset & si.m_addrmask, adata, submask); break; } } } } //------------------------------------------------- // write_stub_64 - construct a 64-bit write from // 8-bit, 16-bit and 32-bit sources //------------------------------------------------- void handler_entry_write::write_stub_64(address_space &space, offs_t offset, u64 data, u64 mask) { for (int index = 0; index < m_subunits; index++) { const subunit_info &si = m_subunit_infos[index]; u32 submask = (mask >> si.m_shift) & si.m_mask; if (submask) { offs_t aoffset = offset * si.m_multiplier + si.m_offset; u32 adata = data >> si.m_shift; switch (si.m_size) { case 8: m_subwrite[index].w8(space, aoffset & si.m_addrmask, adata, submask); break; case 16: m_subwrite[index].w16(space, aoffset & si.m_addrmask, adata, submask); break; case 32: m_subwrite[index].w32(space, aoffset & si.m_addrmask, adata, submask); break; } } } }