There are multiple issues with the current device callbacks:
* They always dispatch through a pointer-to-member
* Chained callbacks are a linked list so the branch unit can't predict the early
* There's a runtime decision made on the left/right shift direction
* There are runtime NULL checks on various objects
* Binding a lambda isn't practical
* Arbitrary transformations are not supported
* When chaining callbacks it isn't clear what the MCFG_DEVCB_ modifiers apply to
* It isn't possible to just append to a callback in derived configuration
* The macros need a magic, hidden local called devcb
* Moving code that uses the magic locals around is error-prone
* Writing the MCFG_ macros to make a device usable is a pain
* You can't discover applicable MCFG_ macros with intellisense
* Macros are not scoped
* Using an inappropriate macro isn't detected at compile time
* Lots of other things
This changeset overcomes the biggest obstacle to remving MCFG_ macros
altogether. Essentially, to allow a devcb to be configured, call
.bind() and expose the result (a bind target for the callback). Bind
target methods starting with "set" repace the current callbacks; methods
starting with "append" append to them. You can't reconfigure a callback
after resolving it. There's no need to use a macro matching the
handler signatures - use FUNC for everything. Current device is implied
if no tag/finder is supplied (no need for explicit this).
Lambdas are supported, and the memory space and offset are optional.
These kinds of things work:
* .read_cb().set([this] () { return something; });
* .read_cb().set([this] (offs_t offset) { return ~offset; });
* .write_cb().set([this] (offs_t offset, u8 data) { m_array[offset] = data; });
* .write_cb().set([this] (int state) { some_var = state; });
Arbitrary transforms are allowed, and they can modify offset/mask for example:
* .read_cb().set(FUNC(my_state::handler)).transform([] (u8 data) { return bitswap<4>(data, 1, 3, 0, 2); });
* .read_cb().set(m_dev, FUNC(some_device::member)).transform([] (offs_t &offset, u8 data) { offset ^= 3; return data; });
It's possible to stack arbitrary transforms, at the cost of compile
time (the whole transform stack gets inlined at compile time). Shifts
count as an arbitrary transform, but mask/exor does not.
Order of mask/shift/exor now matters. Modifications are applied in the
specified order. These are NOT EQUIVALENT:
* .read_cb().set(FUNC(my_state::handler)).mask(0x06).lshift(2);
* .read_cb().set(FUNC(my_state::handler)).lshift(2).mask(0x06);
The bit helper no longer reverses its behaviour for read callbacks, and
I/O ports are no longer aware of the field mask. Binding a read
callback to no-op is not supported - specify a constant. The GND and
VCC aliases have been removed intentionally - they're TTL-centric, and
were already being abused.
Other quirks have been preserved, including write logger only logging
when the data is non-zero (quite unhelpful in many of the cases where
it's used). Legacy syntax is still supported for simple cases, but will
be phased out. New devices should not have MCFG_ macros.
I don't think I've missed any fundamental issues, but if I've broken
something, let me know.
* Added ps2timer device to encapsulate Playstation 2 timers.
* Temporarily hacked R5900 core to always have scratchpad RAM mapped at 0x70000000.
* Added reference counting to divtlb so that it does not unmap pages that are still shared with other entries.
* Added a considerable amount of logging to ps2sony.cpp.
-mips3: Added basic Emotion Engine support. [Ryan Holtz]
* Added S bit to TLB mapping.
* Added support for VSUB, VIADD, VSQI, VISWR, VOR, LQ, SQ, MFSA, MTSA, MFHI1, MFLO1, MULT1, DIV1, DIVU1, PEXTLW, PADDUW, PMFHI,
PMFLO, PCPYLD, PCPYUD, SQC2, LQC2 opcodes. [Ryan Holtz]
This makes these commands useful and consistent with dump in a virtual memory environment. One minor issue is what to do with save for an unmapped memory address: the approach taken here is to write the space.unmap() value, which seems the least harmful. On load, unmapped addresses are not written to, meaning that save/load with a constant address map work as expected.
confusing people for far too long. Yes, this is a change in behaviour.
Add a valdiation check for ROMs with BIOS flag set that are
unselectable, fix the things it uncovers.
(nw) Fix other random stuff.
devfind: revert previous change - if finders aren't set in stone after device_resolve_objects it's going to lead to all kinds of hard-to-diagnose bugs (I'll add more checks for this some time after release)
(nw) also clean up rotting tabulation
- PULSE_LINE is no longer a value. Existing uses have been changed to pulse_input_line with attotime::zero as the second argument.
- Formerly only INPUT_LINE_NMI and INPUT_LINE_RESET were allowed with PULSE_LINE. INPUT_LINE_NMI no longer receives special handling; instead, CPU devices must specify which of their input lines are edge-triggered and thus may be used with zero-width pulses by overriding the execute_input_edge_triggered predicate. INPUT_LINE_RESET is still special-cased, however.
- execute_default_irq_vector now allows a different default vector to be specified for each input line. This added flexibility may or may not prove useful.
This reverts commit c83e2a853d.
Revert "fix compile. (nw)"
This reverts commit a259ba3e36.
GCC is being bad and allowing invalid C++ that other compilers reject.
different devices at a different execution point), don't try to generate
a negative number of samples [O. Galibert]
Not sure why that never happened before. Oh well.
* Get rid of implicit prefix for GFX decode names and prefix them all
* Get rid of special macro for adding GFXDECODE in favour of constructor
* Make empty GFX decode a static member of interface
* Allow palette to be specified to GFXDECODE as a device finder
* Removed diserial.h from emu.h as it's used relatively infrequently
Also fix darkseal and vaportra propely. The palette device
automatically attaches itself to a share with matching tag. The correct
solution here is to rename one or the other out of the way, since it was
never attached to a share before.
* add topcat template
HP topcat was an ASIC used on HP900/300 graphics cards.
Signed-off-by: Sven Schnelle <svens@stackframe.org>
* hook up topcat asic to HP98544
Signed-off-by: Sven Schnelle <svens@stackframe.org>
* topcat: add basic configuration macros for fb planes, height and width
Signed-off-by: Sven Schnelle <svens@stackframe.org>
* hp98544: move logic to topcat video driver
Preparation to support multi plane graphic cards
like the HP98543/98545/98547.
Signed-off-by: Sven Schnelle <svens@stackframe.org>
* xtal: add 35.904MHz XTAL
* topspeed.cpp : Minor cleanups, Add arrays for reduce duplicates
* topspeed.cpp : Fix compile
* romload.h : Add macro for when 64 bit ROM data bus case
Current syntax: MCFG_DEVICE_REPLACE(tag_or_finder, TYPE, ...)
Next-generation syntax: TYPE(config.replace(), tag_or_finder, ...)
(nw) Kill off some more low-value macros that aren't needed any more,
and get rid of the token-pasting voodoo and casts in the discrete sound
macros.
Fix MT06964
Fix µPD7759 class hierarchy, and reset callback before resolving it
(fixed assert in Sega C2)
Remove some more low-value device add indirection macros, default some
more clocks
Make cards inherit clock from slot by default
Start replacing special device macros with additional constructors,
starting with ISA, INTELLEC 4 and RS-232 buses.
Allow an object finder to take on the target of another object finder.
(For a combination of the previous two things in action, see either the
INTELLEC 4 driver, or the Apple 2 PC Exporter card. Also check out
looping over a device finder array to instantiate devices in some
places. Lots of things no longer need to pass tags around.)
Start supplying default clocks for things that have a standard clock or
have all clocks internal.
Eliminate the separate DEV versions of the DEVCB_ macros. Previously,
the plain versions were a shortcut for DEVICE_SELF as the target. You
can now supply a string tag (relative to current device being
configured), an object finder (takes on the base and relative tag), or
a reference to a device/interface (only do this if you know the device
won't be replaced out from under it, but that's a safe assumption for
your subdevices). In almost all cases, you can get the effect you want
by supplying *this as the target.
Eliminate sound and CPU versions of macros. They serve no useful
purpose, provide no extra checks, make error messages longer, add
indirection, and mislead newbies into thinking there's a difference.
Remove a lot of now-unnecessary ":" prefixes binding things relative to
machine root.
Clean up some miscellaneous rot.
Examples of new functionality in use in (some more subtle than others):
* src/mame/drivers/intellec4.cpp
* src/mame/drivers/tranz330.cpp
* src/mame/drivers/osboren1.cpp
* src/mame/drivers/zorba.cpp
* src/mame/devices/smioc.cpp
* src/devices/bus/a2bus/pc_xporter.cpp
* src/devices/bus/isa/isa.h
* src/devices/bus/isa/isa.h
* src/devices/bus/intellec4/intellec4.h
* Allows defaulted clocks (see subtle example with vboy)
* Allows additional constructors (see RS232 port in tranz330)
* Allows use of device finder in place of tag in MCFG_DEVICE_ADD
* Requires out-of-line destructor for devices using incomplete types
* Requires XTAL or explicit u32 for clocks for devices with private types
Devices must still define the standard constructor. When writing
additional constructors, be aware that the constructor runs before
device_add_mconfig in the context of the existing device, not the new
device. See osborne1, zorba, tranz330, and vboy for examples of this in
use. Compilation is a bit slower, but this is temporary while
refactoring is in progress.
Eliminated the need for MCFG_SOUND_ROUTE_EX.
Removed macros from slot option configuration - they just obfuscated
code and slowed it down with needless dynamic casts, but didn't actually
simplify it.
finder. This works outside machine configuration context so the
workarounds in ATA HLE and MSX slots are no longer necessary. It also
allows reduction in tag repetition in machine configuration (see
converted osborne1.cpp, zorba.cpp or the more extreme tranz330.cpp).
Allow reimagined device instantiation to take a device finder based on
current device being configured to reduce repetition (see tranz330.cpp).
Also, look Ma - no magic prologue!
Slot card additions run in the context of the slot itself, which isn't
entirely intuitive. Slot configuration needs a bunch of other cleanup
anyway.
a destructor for a literal type due to the implicit nothrow. It's just
not worth the trouble it's causing.
In file included from ../../../../../src/emu/emu.h:83:
../../../../../src/emu/mconfig.h:70:5: error: '~token' has a non-throwing exception specification but can still throw [-Werror,-Wexceptions]
assert(m_device == m_host.m_current_device);
^
In file included from ../../../../../src/emu/emu.h:29:
../../../../../src/emu/emucore.h:230:48: note: expanded from macro 'assert'
#define assert(x) do { if (!(x)) throw emu_fatalerror("assert: %s:%d: %s", __FILE__, __LINE__, #x); } while (0)
^
In file included from ../../../../../src/emu/emu.h:83:
../../../../../src/emu/mconfig.h:66:3: note: destructor has a implicit non-throwing exception specification
~token()
^
1 error generated.
the finder's owner. This meand you no longer need to care about the
your relationship to the object being configured and a lot of ^ and :
can disappear. There's a bit reduction in string pasting in macros from
this.
Yes, I have to make this apply to devcb etc. as well, but that's a job
for another day.
There's probably at least one thing broken by this where optional
objects are involved. Most things can be solved by just getting rid of
the now-problematic ^ and : prefixes.
- Write callbacks can now be configured as `OUTPUT("item_name")`. This behaves equivalently to a zero-dimensional `output_finder`, while eliminating the need to instantiate and resolve this in driver classes separately from the callback itself.
- The width of a callback's default mask now properly depends on its type (as was half-implemented before), instead of always being reset to 0xffffffffffffffff when actually configured. This allows MCFG_DEVCB_INVERT to work with line write callbacks as one might logically expect.
Concrete device types now have a call operator that instantiates a
device.
This change means you *must* use DECLARE_DEVICE_TYPE to declare the
public interface of your device, even if it's device_t. If you want
to use private implementation classes, use DEFINE_DEVICE_TYPE_PRIVATE
and instantiate the object finders.
This change is intended to expedite debugging of software written for the TMPZ84C015 or similar Z80-based controllers which use 8-bit I/O addressing for the on-chip peripherals but may use either 8-bit or 16-bit addressing externally.
* Move around the debugger hooks to get a small but measurable performance increase
* Remove emucore from external tools
* Improve performance of DSP16 interpreter a little by generating six variants of execution loop
* M1COMM: update simulation based on real firmware (nw)
- read partial frames correctly now
- added VSYNC packets (framesync currently disabled as this can cause
MAME to freeze and we have no way to tell if the socket is still open)
* M2COMM: update simulation (nw)
- read partial frames correctly now
- added VSYNC packets (framesync currently disabled as this can cause
MAME to freeze and we have no way to tell if the socket is still open)
* M1COMM, M2COMM: add config option to sync frames over network (nw)
* M2COMM: another update to the simulation.
- added relay mode (used by stcc)
- added "connection loss"
* M1COMM: update to simulation (nw)
- better sync
- detect lost connection
* M2COMM: use osd_file rather than emu_file for better control (nw)
* M2COMM: handle connection loss in a a more elegant way (nw)
* M1COMM: use osd_file rather than emu_file for better control (nw)
* S32COMM: updated simulation (nw)
- handle connection loss
- use osd_file rather than emu_file for better control
ASCII.
This has not been done unilaterally - I have the support of @galibert,
@Tafoid, and @rb6502 to do something about the current free-for-all.
The trouble with the ROM label field in MAME is that it serves multiple
competing purposes: it's supposed to identify the device in the original
system, and also act as a filename when searching for media image files
to load. It also has to appear in listings of needed/missing files
(e.g. in cases where the image _isn't_ found).
To identify the original device, the ROM label field in MAME often
contains text derived from some combination of one or more of the text
on a label if present, the silkscreen on an IC package, the location on
the circuit board, and the device designation. There's no standard for
the order in which these appear and how they're separated. Some people
add arbitrary filename extensions and other annotations.
There are practical limitations on what can appear in the string, given
it's used as a filename:
* Path/name length limits.
* Restrictions on characters that can appear in a filename.
* Practicality of using the filename in a command-line environment.
* Ambiguity when describing a filename.
Filesystems themselves typically restrict characters in filenames:
* Windows defines MAX_PATH as 260 characters - longer paths are
difficult to use with Win32 APIs and don't work properly in Windows
Explorer
* Most filesystems don't allow ^@ or the path separator in names.
* Windows doesn't allow C0 control characters or <>:"/\|?* characters in
filenames.
* Filesystems may have collation, e.g. FAT16 is case-folding, NTFS and
HFS+ are case-preserving but case-insensitive, while EXT and XFS are
case-sensitive.
* Filesystems may perform Unicode normalisation, e.g. NTFS forces NFC,
HFS+ forces NFD, while ZFS stores filenames as supplied at creation,
but may be configured to apply normalisation when testing equality.
Shells use various ASCII characters for special purposes:
* C0 control characters for line editing and control (e.g. ^C to cancel
a line, ^V for control charecter escape, ^R for history search).
* The "'\ chracaters for quoting/escaping.
* The ><| characters for redirection.
* The *?[] characters for pattern matching.
* The ${}~ characters for variable substitution/sequence expansion.
* The ! or ^ characters for history substitution.
* The ()` characters for controlling subshells.
* The %& characters for job control.
* The ; character as a command separator.
* The # character for comments.
There's also the issue of whether users across a range of locales will
be able to type/display characters. We still don't have good support
for Unicode console output on Windows (std::wcout doesn't seem to work
properly), many users don't install C/J/K fonts, and many users aren't
comfortable entering text in unfamiliar languages. This means we're
limited to printable ASCII for practical purposes.
The practical limitations mean the subset of "safe" characters is
limited to ASCII digits, either uppercase or lowercase English Latin
(but not both due to collation behaving differently across systems), and
the +,-.=_ punctuation chracters. We've decided on lowercase, digits,
and safe punctuation. In addition to this, spaces are allowed, as they
can be quoted/escaped easily enough if no other special characters are
used.
There have been some arguments that allowing uppercase is "more
accurate", but in practical terms it doesn't add much value. A string
in a C++ program can't represent layout, relative size of text, colour
and shape of the label, text font, graphics, and many other details. It
also does nothing to address labels with text outside the English Latin
alphabet (e.g. labels with Chinese ideographs). Besides missing
information, the lack of hard and fast rules means you need to intuit
what a label string in MAME is trying to represent. There is simply no
substitute for photographs. There wasn't even any consistency in case
within individual machine sets. For example, several games in
vigilant.cpp had inconsistent case for "ic" vs "IC" in designation
suffixes, and ibm6850.cpp had inconsistent case for filename extensions
withing a set. There were sets that used uppercase for text from the
label but not from the part number/PCB location, and vice versa. It was
a huge mess.
There's some merit to the idea of allowing a wider variety of characters
in the label strings in the source, and mapping to a more restricted set
when searching for files. However it creates more issues than it
solves. It would require a change to the XML output to provide both the
label and filename, and a corresponding change to external ROM
management tools. It would be impractical to do for software lists,
because it would require ROM management tools to implement the exact
same mapping algorithm as MAME.
But that aside, actually doing useful mapping would be impractical.
What would you do with C/J/K ideographs, like the chip labelled
東方不敗 (Dongfang Bubai)? There's no intuitive way to do the mapping
wtihout incluing something like Unihan data, which would add a lot of
bloat. Even the, without a language hint the Romanisation would be less
than ideal in many cases (using Chinese reading for Japanese text and
vice versa). There's still the messy issue of filesystem collation to
deal with.
We do allow full Unicode in comments in the source. If you want to
provide a more detailed description of a ROM label, that's the place for
it. You've got more characters available, and the possibility of using
mulitple lines. There are too many other competing requirements on the
label field in the ROM definitions.
* Turn deprecated declataion warnings on by default and make them non-fatal
* Make output_finder iterable in algorithms and range-based for loops
* Replace a lot of set_something with output_finder
The new cswidth address map constructor method overrides the masking normally performed on narrow-width accesses. This entailed a lot of reconfiguration to make the shifting and masking of subunits independent operations. There is unlikely to have any significant performance impact on drivers that don't frequently reconfigure their memory handlers.
* Fixed building using system utf8proc
* Fixed building using system portaudio
* Allow using system-wide asio headers (1.11.0 or higher required).
* Allow using system-wide glm headers
* Allow using system-wide rapidjson headers
* Missed a couple escape sequences. (nw)
* A little more escaping, acronym fixes, fix oddity in symlist (nw)
* Update debugger internal help to match docs (nw)
* Lowercasing for CPU in command parameters, fix casing on ASCII. (nw)
- Make the global flipping functions of driver_device protected so as not to be accessible from within subdevices
- Eliminate the flip_screen_set_no_update kludge
This allows for the much more natural "import another map and patch
it" structure, or "cover a whole region then punch holes in it". Our
previous first-entry-wins rule was always a surprise to newcomers, and
oldcomers too.
* various reorganization of radica and vtech stuff
* missed this (nw)
* correct file (nw)
* newlines and stuff (nw)
* less c_str (nw)
* worse (IMHO) filenames (nw)
* format got messed up (nw)
* some bits for golden tee (nw)
* get us renderng something in rad_gtg (nw)
* some basic inputs (nw)
* further improvements to the Golden Tee Home Edition (radica eu3a14)
added Radica Sensible Soccer [Sean Riddle]
* tilebase handling (nw)
* golden tee home video improvements (nw)
Beware, the device context does not follow in MCFG_FRAGMENT_ADD
anymore due to the prototype change. So creating a device then
configuring through a fragment doesn't work as-is. The simplest
solution is just to add a MCFG_DEVICE_MODIFY at the start of the
fragment with the correct tag.
Revert "Removal of voltage_regulator_device (nw)"
This reverts commit 1af133752a.
Revert "New way to provide DAC reference inputs (nw)"
This reverts commit 1c6a7ab40c.
- Introduce MCFG_SOUND_REFERENCE_INPUT to provide fixed inputs through the resampler, eliminating the need for the "voltage regulator" device
- Replace memset use in sound.cpp with std::fill
This was my third implementation of this concept. The previous two involved attaching sound streams to the dummy device (which required giving it device_sound_interface and other modifications).
* Added fallback_artwork and override_artwork as MAME options to allow default artwork to be loaded.
* Removed debug testing code.
* - Allow loading of built-in layouts even if override_artwork is specified.
- Allow loading of fallback_artwork if only default view have been found.
- Fixed order of built-in layouts with regards to fallback_artwork as agreed upon the forums.
* Changed |= true to = true, and changed override artwork so it only checks for default.lay if the <machine name>.lay is not found.
- c65: Remove provisional XTAL definition as requested
- vt240: Correct some clocks and add MCFG_SCREEN_RAW_PARAMS
- tv950: Configure keyboard MCU (not hooked up yet)
- wy55: Release year guessed from ad
* Basic anchor links for FAQ page
* Add verbose logging for CFG files
We already have verbose logging for INI files that get parsed, so having CFGs get similar treatment is useful.
please people, remember to keep source UTF-8 and if you're committing on behalf of others, clean up indents to meet MAME conventions
anyone can run srcclean over a submission and see what will get hit
- reset scheduler savestate to what it was for years before rewind
-- changing saved variables should be done after thorough testing. right now, adding some vars breaks some machines, adding other vars breaks others
- switch to megabyte-wise capacity
-- savestate size greatly differs between machines, relying on state count is unstable
- switch to internal indexing
-- no longer depends on inaccurate machine time
- rewind accelerator key in debugger (Ctrl+F11)
- report capacity hit (once), with some useful info
- make error reports saner
- mention rewind and rewind_capacity in the docs
This reverts commit 6084751936.
This is pointless pollution of the global namespace when you can call
bitswap<4>(...) or bitswap<12>(...) directly. The existing BITSWAP8,
BITSWAP16, etc. were just kept for compatibility with code written
before we had variadic templates available.
- Allow endianness and data/address width to be altered during configuration
- Raise memory_space_config from private to protected so it can be overridden
- Make entire interface optional (as needed by one device to be added soon)
- Use interface_post_load instead of explicitly registered delegate
- Only call rom_bank_updated when bank actually changes
- Remove prototypes for nonexistent functions
When you load a state, icount (*icountptr) would remain whatever it was before loading, messing with the remaining cycles and with the amount of code executed per run() call. This introduced non-determinism and badly influenced usage of savestates while debugging. machine().time() would also return wrong values after that, since it adds remaining cycles.
This starts the work requested in #2398.
How RAM states work.
Implemented using util::vectorstream. Instead of dumping m_save.m_entry_list to file, it writes them as binary to vectorstream. Compression is not used, as it would slow down the process. The header is written as usual, also in binary. When a state is loaded, the savestate data gets binary-read from vectorstream.
How rewind works.
Rewind is optional, it can be turned off through MAME GUI while not running. Rewind capacity is available there too. Rewind step hotkey is available from the standard hotkey menu. In the debugger you have the "rewind" command ("rw" shortcut) that works the same as the hotkey.
Every time you advance a frame (pause step), rewinder captures a RAM savestate of the frame you were at. It does the same when you do step into/over/out in the debugger. Every time it captures a new state (and when you unpause), it marks as invalid all its states that go after the current machine time, because input might change, so they are not relevant anymore. It keeps their buffers allocated though, for future use. When rewinder runs out of allowed amount of savestates it can have, it invalidates the first state in the list and tosses its unique_ptr to the end of the list, then it uses its buffer to capture a new state. When you hit the rewind step key, or use "rewind" command in the debugger, it loads a state that is immediately before the current machine time. Invalid states between valid ones are not allowed to appear, as that breaks rewinder integrity and causes problems. Rewinder keeps its own set of ram states as a vector of unique_ptr's. All rewinder operations and errors get reported using machine().popmessage().
* direct_read_data is now a template which takes the address bus shift
as a parameter.
* address_space::direct<shift>() is now a template method that takes
the shift as a parameter and returns a pointer instead of a
reference
* the address to give to {read|write}_* on address_space or
direct_read_data is now the address one wants to access
Longer explanation:
Up until now, the {read|write}_* methods required the caller to give
the byte offset instead of the actual address. That's the same on
byte-addressing CPUs, e.g. the ones everyone knows, but it's different
on the word/long/quad addressing ones (tms, sharc, etc...) or the
bit-addressing one (tms340x0). Changing that required templatizing
the direct access interface on the bus addressing granularity,
historically called address bus shift. Also, since everybody was
taking the address of the reference returned by direct(), and
structurally didn't have much choice in the matter, it got changed to
return a pointer directly.
Longest historical explanation:
In a cpu core, the hottest memory access, by far, is the opcode
fetching. It's also an access with very good locality (doesn't move
much, tends to stay in the same rom/ram zone even when jumping around,
tends not to hit handlers), which makes efficient caching worthwhile
(as in, 30-50% faster core iirc on something like the 6502, but that
was 20 years ago and a number of things changed since then). In fact,
opcode fetching was, in the distant past, just an array lookup indexed
by pc on an offset pointer, which was updated on branches. It didn't
stay that way because more elaborate access is often needed (handlers,
banking with instructions crossing a bank...) but it still ends up with
a frontend of "if the address is still in the current range read from
pointer+address otherwise do the slowpath", e.g. two usually correctly
predicted branches plus the read most of the time.
Then the >8 bits cpus arrived. That was ok, it just required to do
the add to a u8 *, then convert to a u16/u32 * and do the read. At
the asm level, it was all identical except for the final read, and
read_byte/word/long being separate there was no test (and associated
overhead) added in the path.
Then the word-addressing CPUs arrived with, iirc, the tms cpus used in
atari games. They require, to read from the pointer, to shift the
address, either explicitely, or implicitely through indexing a u16 *.
There were three possibilities:
1- create a new read_* method for each size and granularity. That
amounts to a lot of copy/paste in the end, and functions with
identical prototypes so the compiler can't detect you're using the
wrong one.
2- put a variable shift in the read path. That was too expensive
especially since the most critical cpus are byte-addressing (68000 at
the time was the key). Having bit-adressing cpus which means the
shift can either be right or left depending on the variable makes
things even worse.
3- require the caller to do the shift himself when needed.
The last solution was chosen, and starting that day the address was a
byte offset and not the real address. Which is, actually, quite
surprising when writing a new cpu core or, worse, when using the
read/write methods from the driver code.
But since then, C++ happened. And, in particular, templates with
non-type parameters. Suddendly, solution 1 can be done without the
copy/paste and with different types allowing to detect (at runtime,
but systematically and at startup) if you got it wrong, while still
generating optimal code. So it was time to switch to that solution
and makes the address parameter sane again. Especially since it makes
mucking in the rest of the memory subsystem code a lot more
understandable.
- Define alternate XTAL for MIKBUG version (nw)
- Add RAM configuration and MC14411 device (nw)
- Reduce terminal baud rate to the supported maximum of 1200 (nw)
- Add notes for future reference (nw)
Disassemblers are now independant classes. Not only the code is
cleaner, but unidasm has access to all the cpu cores again. The
interface to the disassembly method has changed from byte buffers to
objects that give a result to read methods. This also adds support
for lfsr and/or paged PCs.
