This update changes the way we handle memory allocation. Rather
than allocating in terms of bytes, allocations are now done in
terms of objects. This is done via new set of macros that replace
the malloc_or_die() macro:
alloc_or_die(t) - allocate memory for an object of type 't'
alloc_array_or_die(t,c) - allocate memory for an array of 'c' objects of type 't'
alloc_clear_or_die(t) - same as alloc_or_die but memset's the memory to 0
alloc_array_clear_or_die(t,c) - same as alloc_array_or_die but memset's the memory to 0
All original callers of malloc_or_die have been updated to call these
new macros. If you just need an array of bytes, you can use
alloc_array_or_die(UINT8, numbytes).
Made a similar change to the auto_* allocation macros. In addition,
added 'machine' as a required parameter to the auto-allocation macros,
as the resource pools will eventually be owned by the machine object.
The new macros are:
auto_alloc(m,t) - allocate memory for an object of type 't'
auto_alloc_array(m,t,c) - allocate memory for an array of 'c' objects of type 't'
auto_alloc_clear(m,t) - allocate and memset
auto_alloc_array_clear(m,t,c) - allocate and memset
All original calls or auto_malloc have been updated to use the new
macros. In addition, auto_realloc(), auto_strdup(), auto_astring_alloc(),
and auto_bitmap_alloc() have been updated to take a machine parameter.
Changed validity check allocations to not rely on auto_alloc* anymore
because they are not done in the context of a machine.
One final change that is included is the removal of SMH_BANKn macros.
Just use SMH_BANK(n) instead, which is what the previous macros mapped
to anyhow.
function here. Remaining devices have been converted to have
device-specific functions to do the same thing with proper type
checking.
CPUs still have a set_info function but it is CPU-specific now and
no longer piggybacks on the general device function.
assumption that all device tags are unique. Specifically, the
following no longer need to provide a device type:
AM_DEVREAD/WRITE
DEVCB_DEVICE_HANDLER
devtag_get_device
devtag_reset
device_list_find_by_tag
as well as several device interfaces that referenced other devices.
Also fixed assertion due to overflow in the recent sound fix.
> On Mon, Jan 19, 2009 at 05:37:35AM -0800, R. Belmont wrote:
> > My mistake. I thought you were suggesting that we should actually
> > somehow handle malloc failures. Given that aborting is an OK way to
> > express failure, I'd suggest the return values be changed to DEFER
> > and DONT_DEFER to eliminate the conceptual imbalance of OK/DEFER.
>
> That's where comes the fact that we have 130 OK/DONT_DEFER and 1
> DEFER. It makes me think that the exceptional DEFER case should be
> handled by an exceptional function call.
>
> I know, code talks, but I'm at work right now :-)
Here we go.
OG.
interfaces when handling strings. Namely, the generic
get_info functions allocate a temporary string and the
device in question copies its string to the target,
instead of assigning a const char *. Updated all device
and sound cores to operate this way.
Added the concept of a cpu_state_table, which is
supplied by the CPU cores and which describes all the
register state accessible to the debugger and other
subsystems. The format of the table is such that most
data can be simply fetched from memory without the
further involvement of the CPU core, including the
display of common formats. Extensibility points are
available for custom display and for importing/exporting
the data to intermediate variables for more complicated
scenarios. Updated the ADSP21xx, TMS340x0, and i86 cores
to use this.
Removed the old debugger register list, which was never
used. Replaced it with using ordering from the
cpu_state_table.
Renamed REG_PC -> REG_GENPC, REG_SP -> REG_GENSP, and
REG_PREVIOUSPC -> REG_GENPCBASE. Updated a few spots
that were using these directly. Moved these definitions
into the end of the register area rather than leaving
them outside which put them in a weird range.
Updated the fcompress APIs to allow for specifying a compression level.
Removed the concept of state saving tags, which was a hack to get save states
to work with multiple CPU cores. Simplified the state saving system as a
result, performing the operation in a single pass and without allocating
a full blob of memory. Also enabled minimal compression.
Added new function cpuexec_describe_context(machine) which can be
used in logerror() and other printf-style functions to return a
description of the current CPU/PC given only the machine. Changed
several dozen sites to use this instead of directly interrogating
the activecpu.
Removed all other uses of activecpu throughout the system. Removed
activecpu from the machine structure to prevent future abuse.
Removed cpu_push_context() and cpu_pop_context(), and all call
sites.
Voodoo devices now require a CPU to be defined in the configuration
in order to know whom to steal cycles from or stall when FIFOs get
full. Updated all voodoo users to specify one.
CPD1869 devices now also require a CPU to be defined in the
configuration, in order to know which CPU's registers to fetch.
Updated all cdp1869 users to specify one.
Many other small changes to make this all work.
Sent: Saturday, December 06, 2008 4:52 PM
To: submit@mamedev.org
Cc: atariace@hotmail.com
Subject: [patch] Deprecat.h cleanup
Hi mamedev,
This patch changes some global Machine references to use machine,
device->machine, ... instead, and removes any unneeded #include
"deprecat.h" lines as well (about 10% of them in fact). It was
generated using the attached script, and then reverting some cases
where it was overzealous.
~aa
appropriate, and to keep all global variables hanging off the
machine structure. Once again, this means all state registration
call sites have been touched:
- state_save_register_global* now takes a machine parameter
- state_save_register_item* now takes a machine parameter
- added new state_save_register_device_item* which now uses
the device name and tag to generate the base name
Extended the fake sound devices to have more populated fields.
Modified sound cores to use tags from the devices and simplified
the start function.
Renumbered CPU and sound get/set info constants to align with
the device constants, and shared values where they were perfectly
aligned.
Set the type field in the fake device_configs for CPU and sound
chips to a get_info stub which calls through to the CPU and sound
specific get_info functions. This means the device_get_info()
functions work for CPU and sound cores, even in their fake state.
Changed device information getters from device_info() to
device_get_info() to match the CPU and sound macros.
Removed ATTOTIME_TO_CYCLES() and ATTOTIME_IN_CYCLES(). Replaced them
with functions in cpuexec: cpu_clocks_to_attotime() and
cpu_attotime_to_clocks(), both of which take CPU devices instead of
indexes. Updated all callers, many of which were using the functions
dubiously.
state_save_combine_module_and_tag() function in favor of passing
the tag when registering. Revisited all save state item registrations
and changed them to use the tag where appropriate.
related APIs now take a device pointer instead of an index.
All functions that take a CPU device are prefixed with cpu_*
All functions that are globally related to cpu execution
are prefixed with cpuexec_*. Below is a list of some of the
mappings:
cpu_boost_interleave -> cpuexec_boost_interleave
cpunum_suspend -> cpu_suspend
cpunum_resume -> cpu_resume
cpunum_is_suspended -> cpu_is_suspended
cpunum_get_clock -> cpu_get_clock
cpunum_set_clock -> cpu_set_clock
cpunum_get_clockscale -> cpu_get_clockscale
cpunum_set_clockscale -> cpu_set_clockscale
cpunum_get_localtime -> cpu_get_local_time
cpunum_gettotalcycles -> cpu_get_total_cycles
activecpu_eat_cycles -> cpu_eat_cycles
activecpu_adjust_icount -> cpu_adjust_icount
cpu_trigger -> cpuexec_trigger
cpu_triggertime -> cpuexec_triggertime
cpunum_set_input_line -> cpu_set_input_line
cpunum_set_irq_callback -> cpu_set_irq_callback
In addition, a number of functions retain the same name but
now require a specific CPU parameter to be passed in:
cpu_yield
cpu_spin
cpu_spinuntil_time
cpu_spinuntil_int
cpu_spinuntil_trigger
cpu_triggerint
Merged cpuint.c into cpuexec.c. One side-effect of this
change is that driver reset callbacks are called AFTER the
CPUs and devices are reset. This means that if you make
changes to the CPU state and expect the reset vectors to
recognize the changes in your reset routine, you will need
to manually reset the CPU after making the change (since it
has already been reset).
Added a number of inline helper functions to cpuintrf.h for
managing addresses
Removed cpu_gettotalcpu(). This information is rarely needed
outside of the core and can be obtained by looking at the
machine->cpu[] array.
Changed CPU interrupt acknowledge callbacks to pass a CPU
device instead of machine/cpunum pair.
Changed VBLANK and periodic timer callbacks to pass a CPU
device instead of machine/cpunum pair.
Renamed all information getters from cpu_* to cpu_get_* and
from cputype_* to cputype_get_*.
* added a set of cpu_* calls which accept a CPU device object;
these are now the preferred means of manipulating a CPU
* removed the cpunum_* calls; added an array of cpu[] to the
running_machine object; converted all existing cpunum_* calls
to cpu_* calls, pulling the CPU device object from the new
array in the running_machine
* removed the activecpu_* calls; added an activecpu member to
the running_machine object; converted all existing activecpu_*
calls to cpu_* calls, pulling the active CPU device object
from the running_machine
* changed cpuintrf_push_context() to cpu_push_context(), taking
a CPU object pointer; changed cpuintrf_pop_context() to
cpu_pop_context(); eventually these will go away
* many other similar changes moving toward a model where all CPU
references are done by the CPU object and not by index
means of setting the minimum useful scheduling quantum, and clamping
all quanta to that value.
Changed interleave/boost handling to use scheduling quanta instead
of timers.
Added machine parameter to cpu_boost_interleave.
Updated cpuexec to compute the "perfect" interleave value taking into
account the minimum number of cycles per instruction specified by the
CPU core. Updated Z80 core to indicate that the minimum cpi is 2. Fixed
incorrect minimum cpi in the 68020+ cores.
Simplified a bit of logic in cpuexec_timeslice.
fine and basic searching/playback/skipping is functional. Still a bit
glitchy.
Firefox improvements:
- removed need for deprecat.h
- memory map is complete from schematics
- gutted laserdisc hacks in favor of actual laserdisc implementation
- fixed all CPU and sound clocks
Removed old laserdsc.c implementation.
Added generic timer devices, which simply allocate a timer but don't
prime it. This is the preferred method for allocating timers, and may
eventually be the only mechanism for doing so in the future.
is either DEVICE_START_OK or DEVICE_START_MISSING_DEPENDENCY. The latter
should be returned by a device if there is another device it depends on
which hasn't been started yet. Added new flag in the device interface to
indicate whether a device has been started.
Changed laserdisc interface to explicitly specify the screen and sound
devices it should route to. Drivers no longer have to manually call
laserdisc_vsync(). Instead, the laserdisc code connects up to the routed
screen device and works based on that screen's VBLANK timing. Removed
all existing calls to laserdisc_vsync().
Changed laserdisc behavior so that it completes the previous video read
and initiates the next read at the end of VBLANK instead of the beginning.
This gives player logic time during VBLANK to alter the slider position
prior to fetching the next frame.
Added new laserdisc callback for vsync begin and changed the update
callback to be called at the end of VBLANK. Also added functions to set
the slider speed, advance the slider, and directly control the video/
audio squelch. In addition, there is a new status function to get the
slider position in general terms.
Added parameter to the VBLANK callbacks supported in emu/video.c. Updated
all callers to provide a callback value.
Fixed bug that would cause watchpoints to trigger if you had a memory
window open to the watchpoint address.
Further updates to the PR-8210 ROM simulation. Still not quite there but
the system is much better understood now. Added layout to the PR-8210
which displays the state of the front-panel LEDs.
only remaining form is the one that takes a pointer parameter.
Added macros for STATE_PRESAVE and STATE_POSTLOAD to define common
functions. Added machine parameter to these functions.
Updated all drivers and CPU/sound cores to use the new macros
and consolidate on the single function type. As a result pushed
the machine parameter through a few initialization stacks.
Removed unnecessary postload callbacks which only marked all tiles
dirty, since this is done automatically by the tilemap engine.
Added DEVINFO_INT_TOKEN_BYTES, which is queried before device
startup. It should return the size of the structure that will
be hung off of device->token. It must not return 0.
Changed the core device interface to allocate and zero a block
of memory the size of DEVINFO_INT_TOKEN_BYTES before calling
device_start. The pointer to this memory is already attached
to device->token at the time device_start is called.
The primary motivation for this is to allow the device_start
function to make internal calls and pass the device_config *
around as if the object were fully constructed. Since the token
is used to indicate whether or not a device is live, this makes
the device "live" at the time device_start is called rather
than after it completes.
Converted Centipede, as an example.
To define a scanline timer, use something like this:
MDRV_TIMER_ADD("32V", SCANLINE, generate_interrupt)
MDRV_TIMER_SCANLINE("main", 0, 16)
The first number is the first scanline the timer will fire on, the 2nd number is the increment.
So in this case, the timer will fire on 0, 16, 32, ..., 224, 240, then wrap around
because the screen is defined as 256 lines high.
The current scanline is passed to the callback in its 'param' argument
suffixed with _func. Did this throughout the core and
drivers I was familiar with.
Fixed gcc compiler error with recent render.c changes.
gcc does not like explicit (int) casts on float or
double functions. This is fracking annoying and stupid,
but there you have it.
The idea is to create extra work if a driver wants to use these and hopefully
gives an incentive to look for an alternate solution
- Added #include of deprecat.h that rely on these contructs
- Removed a bunch of unneccassary #include's from these files
- removed years from copyright notices
- removed redundant (c) from copyright notices
- updated "the MAME Team" to be "Nicola Salmoria and the MAME Team"
The attached patch adjusts most conditional logging in MAME to use the
idiom "do { if (VERBOSE) logerror x; } while (0)". This has the
benefit that the compiler checks the syntax of the logging even in the
case it will be eliminated, and in fact a number of cases here needed
adjustments to compile because of this.
