Thanks to Aaron Giles who made me think about a different approach.
This is a rewrite from scratch for rom devices. It uses a generic
template to implement rom devices which is used together with a
description struct to define a rom device. This leads to highly
efficient code since all information is available at compile time.
This is also a step forward to support tristate outputs. All rom devices
covered by this approach have tristate or open collector outputs and
thus all code changes to support tristate outputs can now be made
consistently in one file.
Code cleanup to better separate the following stages:
- parsing
- setup
- run
In addition preliminary native tristate support was added. Not yet
production ready, please don't use it.
- working macro level parameters
- simplified code significantly
- fixed a number of hidden issue
- dead code removal
It is now possible to define netlist level parameters using:
static NETLIST_START(chip)
DEFPARAM(parameter, 123)
RES(R1, $(@.parameter))
....
NETLIST_END()
NETLIST_START(main)
chip(X1)
PARAM(X1.parameter, 1000)
...
NETLIST_END()
This will pass on 1000 to R1 as a parameter.
"@." will resolve to the current namespace. In this case it is X1, thus
"X1.parameter. This is during parsing. During evalution, i.e. device
creation, $(X1.parameter) is evaluated to 1000.
This also means, that global parameters are possible and accessible ...
DEFPARAM(P, 234)
...
RES(R1, $(P))
or going forward
RES(R1, 1000)
CAP(C1, 1.0 / $(R1.R))
This opens up the path to more possibilities ...
static NETLIST_START(DM9456_DIP)
DEFPARAM(MODEL, "74XX")
DM9456_GATE(X1)
DM9456_GATE(X2)
PARAM(X1.MODEL, $(@.MODEL))
PARAM(X2.MODEL, $(@.MODEL))
...
NETLIST_END()
NETLIST_START(main)
DM9456_DIP(X1)
PARAM(X1.MODEL, "LS74XX")
...
NETLIST_END()
The code also has been prepared to allow something along the lines of
LIBENTRY_MODEL(DM74LS00_DIP, DM7400_DIP, "LS74XX")
to define a LS version on top of an DM7400_DIP bare TTL implementation.
This however will need more efforts to existing devices to honour some
timing adjustment in the model definition. It will already honour
different output specifications between LS series and bare TTL if
connected to analog components.
This a big forward to allow reuse of code. Still a longer way to go
but the foundation there.
Also brings quite a number of simplifications and dead code removal.
- removed family_setter
- naming alignment, family becomes model.
- architecture cleanups.)
- reviewed reset logic.
- pass truthtable family as string to factory.
This is another set of changes on the path to align logic families and
models. As a side effect, the object model now makes a clear
distinction between analog models and logic models.
The number of macros in nl_base.h has decreased significantly due to
these changes.
Reduce the usage of NETLIB_CONSTRUCTOR_DERIVED to implement
dip package objects. This aligns the programmatic approach
with the approach used in the macro folder, i.e. using netlists.
Untangle model definition and interpretation as well.
This changes are a first step in aligning models and families and thus
simplify the source further.
nlwav now supports 16 and 32 bit integer format (wav16s, wav32s)
as well as 32 bit float format (wav32f).
The "wav" format is no longer supported.
Added support for high and low pass filtering the wav output.
Dynamic volume adjustment (use "-a 0")
--hpboost can suppress spikes at the beginning of the file (<10ms)
The addition of the float format simplifies debugging significantly
since it covers the whole dynamic format.
Added nlwav to the local VS 2019 build.
This change makes state saving contract based. Objects which need to
save state need to have the following template member:
template <typename ST>
void save_state(ST &st)
{
/* Example */
st.save_item(m_p, "m_p");
st.save_item(m_buf, "m_buf");
}
This member function is called when the object is passed to the
state manager save function.
The two sources act as voltage sources, though noise may also be
injected as conductivy or current noise.
SYS_NOISE_MT_U: Mersenne Twister uniform noise
SYS_NOISE_MT_N: Mersenne Twister normal noise
nld_sys_noise is templated:
using NETLIB_NAME(sys_noise_mt_u) =
NETLIB_NAME(sys_noise)<plib::mt19937_64,
plib::uniform_distribution_t>;
Thus the approach is scalable. The implementation is state save aware,
and thus reproducible results are guaranteed.
An example use case is provided as well, see examples/noise.cpp.
The model is driven by three additional model parameters: NBV, IBV, BV.
Example code how to define a typical 7.5V Zener diode:
ZDIODE(ZD, "D(BV=7.5 IBV=0.01 NBV=3)")
or
NET_MODEL("A1234 D(BV=7.5 IBV=0.01 NBV=3)")
ZDIODE(ZD, "A1234")
The device can be found in nlm_other.cpp.
Removed nld_ne566.*
Added SYS_SW, SYS_SW2 and SYS_COMP. These are single switch,
alternating switch and a analog comparator with digital outputs.
Renamed RES_SWITCH to SYS_SW.
Added example ne566.cpp in netlist/examples.
Documentation should be created on Windows using
make mingw PARAMS=doc
Admittedly this is ugly and the makefile needs to be changed to
use something like "make HOSTOS=windows doc".
Make some calls private which were public before.
Eliminate functions which would be executed only once.
Make terminal reference functions in twoterm.h return const.
Clearly identify the use cases which need a non-const ref
to terminals.
Calls to the solver used different paths. These changes unify these
calls and provide a common call for discontinuous state changes on
analog components (change_state).
"Edge cases" are compiles were the floting point type used for the core
may be different from the one used by solvers, e.g. double/float.
This does not happen in MAME, here currently this is always
double/double. Since floats are better suited for GPUs whose double
performance is limited it was time to review the code.
This commit fixes warnings about type conversions and precision loss.
In addition there is better support for INT128 and FLOAT128.
The MAT solver has seen some minor performance increases.
Added a define NL_USE_ACADEMIC_SOLVERS and disabled it in the build.
This will not compile in solvers which are either illustrative or only
perform for large sparse matrices like GMRES.
This commit simplifies the creation of static solvers. For this to
happen the following changes were made:
- nltool does no longer support the "-f netlist_file.cpp" option.
All parts of netlist which used the option now expect the file
to be specified after all other options.
Before: nltool -t 10 -f somenetlist.cpp
Now : nltool -t 10 somenetlist.cpp
- The static compile command now supports an "--output" option
to specify the file to be created and accepts multiple input
files. To be create the static solver file for MAME use the
script provided or:
./nltool --cmd static
--output=src/lib/netlist/generated/static_solvers.cpp
src/mame/audio/nl_*.cpp src/mame/machine/nl_*.cpp
- Updated documentation and examples provided by nltool --help
The CD4016 is an analog switch IC. Document an experimental approach
to implement the analog switch as a 3 terminal element which is
completely being dealt with as part as the linear system.
The intention was to improve convergence when the switch is in a
feedback loop. One example are two-opamp tridiagonal wave generators.
Unfortunately the approach did not work out and in addition was
performing far worse than the net-separating original code.
Also updated comment in nld_generic_models.h.
This commit introduces precompiled static solver code. Due to
additional optimizations the compiler can use because the detail
calculation steps for the solution are known e.g. the kidniki netlist
sees a 100% speed increase.
In all environments (windows/*nix/osx) the source for the static
solver code can be created using
bash src/lib/netlist/nl_create_mame_solvers.sh
This will create src/lib/netlist/generated/static_solvers.cpp which is
compiled into the mame binary.
The script is just a temporary workaround. The intention is that nltool
whill be able to create this file with one call.
There are other improvements in this commit speeding up the processing
of timestep and dynamic calculations.
The only link dependency left is emu_fatalerror. The format handlers
really shouldn't be throwing fatal errors at all - they should just
fail to load an image they can't handle. Maybe the interface should
be enhanced to allow better propagation of errors up from the format
handlers so they can be displayed in a UI rather than just logged.
The only other two dependencies are on logmacro.h (pure macros) and
the PAIR type from emucore.h (possibly worth moving to util).
../../../../../src/lib/formats/apridisk.cpp: In member function 'virtual bool apridisk_format::load(io_generic*, uint32_t, floppy_image*)':
../../../../../src/lib/formats/apridisk.cpp:99:7: error: 'fatalerror' was not declared in this scope; did you mean 'strerror'?
Document linearized diode model in the range exceeding maximum
dissipation. The intention is to have a faster convergence but this yet
not really is observable.
* Add doxygen comments for bit manipulation functions
* Add an overload of BIT that works like the AArch64 UBFX instruction
* Kill off some of the silly concatenating overloads for emu_file::open
* Make searchpath acually useful for devices
This is a checkpoint - I'm planning to improve ROM loading behaviour at
least a little.
Change color levels for williams2 hardware to better color display.
Output levels are computed in netlist. These levels are aligned to
LTspice modelling I did.
Also improved the modelling to do what the actual hardware for mysticm
did: Color decoding depends on W13..W11. These are actually flip screen
versions of V13..V11. Fixed treatment on Bit0 although not used by the
game code.
Added R,G,B gain and offset sliders for the williams2 hardware. MAME
only supports one setting for all channels. This is not good enough
to come close to video recordings found on youtube.
Minor changes to nltool to better measure time taken to run a
simulation.
This example shows how the netlist tools can be used to derive
RGB tables from a schematic. All steps are given in
serc/lib/netlist/examples/turkey_shoot.cpp
Using this output format nlwav will pick point samples at regular
intervals from log files. This can be used to e.g. calculate
RGB output levels with netlist.
- Use initializer_list in object_array_t.
- Added constructor which accepts a format string for object names.
- Started work on logic_input8 which provides 8 logic outputs.
(nw) I've triple-checked the code and it looks sane, but I'm still a bit nervous about changing
such a fundamental part of MAME. If integer formatting appears to have changed after this, let
me know ASAP. Also removed a workaround for older versions of GNU libstdc++ with incomplete
C++14 support.
It is necessary to return a higher score on success, higher than
returned by the default method, in order for a general 'identify' to
succeed over competitive matches.
Add a get_track_format() method that can be overridden to supply format
variations for any track and head. The code is generalised to account for such
variations. The default method returns the passed format, so this change is
neutral for existing formats.
Simplify the FLEX DSK format code. There are now simply format variation
descriptions for the second track that have the sector ID continuing in
sequence from the first track, rather than specialized code.
Extend the FLEX format to support variations in the sectors ID of the first
two sectors. The FLEX 6800 boot sectors have IDs based at zero rather than
one. Extend the FLEX format to support variations for which the first track,
on both sides, is single density on an otherwise double density disk which was
historically a common format.
Extend the OS9 disk format to support variations for which the first track, on
only the first side, is single density on an otherwise double density
disk. OS9 for the SWTPC and Gimix typically used such formats.
Extend the OS9 disk format to support variations with a base sector ID of zero
in contrast to the existing COCO OS9 format which uses a based sector ID of
one. The OS9 format identification code is extended to rely on the optional
information stored in the OS9 LSN0 header to identify COCO format disks, and
all COCO format disks appear to have this optional information in a regular
enough format.
Parameters are now passed to the netlist core as strings. During netlist
creation they are evaluated as functions. This opens the path to
parameters on subdevice level.
Examples:
PARAM(device.XY, (1+2*0.005))
RES(R1, 2.05*RES_K(1)+1)
In addition the commit contains dead code removal.
- Fixed some clang lint warnings
- Removed dead code
- Experimental parser code to allow calculations in parameter value.
This already works for compiled netlists. These changes are
currently disabled. Updated pong netlist (and CRC/SHA) to work
with this new code.
"./nltool -c convert -f UA741.mod" now runs without errors. UA741.mod is
the official TI SPICE subckt for the 741 opamp.
This commit fixes a number of issues and adds linear POLY support for
E/F SPICE models.
This commit fixes an issue with parsing netlists. To cut a long story
short: The separation of nets into independant groups of nets failed
under some circumstances for four terminal devices (controlled
voltage/current devices). Everything worked as long as an external
feedback loop existed. Without external feedback loop, the separation
would either fail or create "false" separated nets.
This fix also highlighted an issue with cmos switches (4066/4016).
There is a slight chance that other bugs may surface due to this change.
- add CD4006 and CD4070 devices
- add TL084 opamp model
- Clock now supports proxies, i.e. can be connected to
analog devices.
- Fixed netlists using CLOCK
- added some comments
- removed a forgotten header file.
- remove pthrow trampline as proposed by Vas.
- identify throwing code by adding noexcept(false)
- move "connected term" information to setup code.
- srcclean
The APPLE2_TRACK_COUNT macro was not being expanded here, at least on gcc8.
This resulted in the string "APPLE2_TRACK_COUNT" being within the specification
and that in turn broke option_resolution::lookup_in_specification leading to
an assertion fail with the debug checks enabled. So bake in the numeric
constant to at least get things working again.
Increase the time resolution from 1 nano second to 100 pico seconds.
Make sure that icount and netlist internal time are better synched by
tracking the remainder of the division.
Fixed the netlist sound device. There is a one sample overflow every 13
seconds at 48000 Hz due to integer truncation which is now ignored.
Added more doxygen documentation.
Long story: The diode model uses log-stepping as proposed in "Circuit
Simulation" (Farid N. Najm) page 183. If the previous voltage was
sufficiently negative the new voltage after a log step could not have
increased enough to alter matrix and rhs vector.
This bug surfaced in stuntcyc video signal mixing which uses a diode for
the composite signal.
Introduce an additional absolute time type netlist_time_ext to identify
whether absolute or relative time is used in the netlist code.
Extend ptime code to allow operations between ptime derived types with
different internal types.
In addition rewrote main queue serve loops. Adds a very small
performance increase.
Added netlist version information. This is used to enforce the
invalidation of save states when the major/minor netlist version
changes.
This catches edge cases for which neither the size or names of saved
items changes during releases.
* Got rid of some more simple_list in core debugger code
* Fixed a buffer overrun in wavwrite (buffer half requried size)
* Slightly reduced dependencies and overhead in wavwrite
* Made new disassembly windows in Qt debugger default to current CPU
- Removed DUMMY_INPUT. NC (not connected) pins should now use NC_PIN.
If a NC_PIN is actually connected, an error will be logged and
validation will fail.
- Enabled "extended" validation. This will catch now if power terminals
are not connected.
- Added const and noexcept where appropriate.
- Removed dead code.
- Fixed the 7414 Schmitt-Trigger device to use nld_power_pins
- Devices ttlhigh and ttlhow are no longer automatically created.
- All logic input devices (e.g. TTL_INPUT, LOGIC_INPUT) now need to have
their power terminals (VCC, GND) connected. This opens the route
for more appropriate proxy devices but comes at a cost. If the
connections are omitted your circuit will not work as expected.
Example:
LOGIC_INPUT(I_SD0, 1, "AY8910PORT")
NET_C(VCC, I_SD0.VCC)
NET_C(GND, I_SD0.GND)
- Updated all netlists.
- Removed proxy information from terminal objects. This was replaced by
a lookup hash whose life-span does not exceed netlest setup.
These changes enable the removal of a number of hacks from the
source going forward.
- fixed a code in the netlist creation which caused multiple proxies
to be created for output->terminal connections. A nice side effect of
this fix is a performance increase ~9% for kidniki and ~4% for pong.
Speaking about pong ... maximum is 490%. Dice is running at
280 FPS/60 FPS = 466%, however without any analog emulation.
- Replaced NL_NOEXCEPT with noexcept. assert is now exception-free.
- cppcheck and lint fixes.
- move memory pool to netlist_state_t removing one static allocation.
- add memory allocation stats to verbose output
- nl_assert no longer throws, first step to remove NL_EXCEPT macro.
This commit is a first step towards using formulas in parameters, i.e.
MAINCLOCK(clock, 20 * 30)
The intention is to improve readability and scalability.
Since device registration already provides all necessary information
about parameters, the code to create an include file for all
devices has been improved. Long term, this will remove the need for
device specific header files.
In addition going forward devices will accept either no connections or
all specified connections, i.e.
TTL_7400_NAND(name, chip1.2, chip2.3)
or
TTL_7400_NAND(name)
NET_C(...)
NET_C(...)
This will allow to remove all duplicate definitions which are currently
necessary, i.e. TTL_7400_NAND/TTL_7400_GATE
- comment style migration continues.
- Fixed a two bugs in the truthtable ignore inputs code
- refactored the truthtable code a bit for better readability.
- updated netlist specific gitignore.
- "nltool -c docheader" now scans sources and creates
usage focussed doxy documentation for devices. Very early
stage, but works. For an example, please see ne555 source.
- Started migrating to pure C++, i.e. "//" comments.
- Various documentation fixes.
- Added cppcheck configuration to netlist/build
- Some smaller code changes.
- more const
- explicitly raise exceptions instead of leaving this to log.fatal()
- correct a number of cppcheck findings.
- dead code removal
- clang lint corrections, e.g. include order
This fixes a case where:
* m_soundbuf_samples == processedsamples
* processedsamples > 0
* processedsamples * stream->channels() == m_soundbuf.size()
In this scenario, the std::memmove() would do nothing (moving zero
bytes), but the operator[] on the second parameter to std::memmove()
overflows the array. This can be benign in optimized builds (because
the third parameter to std::memmove() is 0), but on debugging builds
this can cause an assert.
Both compiling the core and the shaders with __float128 now work.
The support was added to be ready to deal with academic edge cases.
Performance drops to 10% of double - thus disabled by default.
__float128 is a gnu extension delivering true 128 bit floating point
support. Currently not supported by clang. In addition, the quadmath
library needs to be linked. For the time being therefore disabled.
Renamed cast member of the constants struct to magic to clearly identify
magic numbers.
Introduced nlconst struct inheriting from plib::constants<nl_fptype> to
make code better understandable.
The newly added RELTOL and VNTOL parameters implement Newton convergence
checks comparable following other SPICE implementations.
The ACCURACY solver parameter now is only used for convergence checks in
iterative solvers.
In addition, type safety was significantly improved and a lot of "magic"
numbers are identifiable now.
- Added new solver parameter FPTYPE. This determines in which floating
point domain the linear system is solved. May be one of "FLOAT",
"DOUBLE" or "LONGDOUBLE"
- Added option "--fperr" to nltool. This enables floating point
exceptions. This helps debugging the code under gdb.
The purpose of this going forward is to have more choice in
optimization. Non-dynamic systems should be just fine in the float
domain. Dynamic systems (i.e. diodes, bjts, mosfets) should in general
work with double. Certain edge cases may require long double resolution.
Added ability to compile using float instead of double. Specifically the
the solver as well as the infrastructure now can have their own floating
point type. Currently this is only an academic exercise since
numerically demanding circuits like kidniki only work with double/double
support. Using float here is pushing numerical stability over the
limits.
The long term design goal is too have the matrix type (double/float)
being a parameter.
- solver: align matrix population along the various solvers
- solver: delete dead code
- renamed nl_double to nl_fptype and use nl_fptype where previously
double has been used.
- renamed param_double_t to param_fp_t
- Removed code no longer used
- Add noexcept where appropriate
- split pparser.[c|h] into ppreprocessor and ptokenizer
- smaller optimizations, e.g. use of std::size_t
- fix lint warnings
The builtin preprocessor now behaves closer to cpp:
- supports macro parameters, i.e. define x(a) a
- supports stringification, i.e. define x(a) #a
- supports concatenation, i.e. define x(a) a ## _ext
In addition, error reporting now provides a source context including the
include history.
fundamental change to show device delegates are configured.
Device delegates are now aware of the current device during
configuration and will resolve string tags relative to it. This means
that device delegates need a device to be supplied on construction so
they can find the machine configuration object. There's a
one-dimensional array helper to make it easier to construct arrays of
device delegates with the same owner. (I didn't make an n-dimensional
one because I didn't hit a use case, but it would be a simple addition.)
There's no more bind_relative_to member - just call resolve() like you
would for a devcb. There's also no need to cast nullptr when creating a
late bind device delegate. The flip side is that for an overloaded or
non-capturing lambda you'll need to cast to the desired type.
There is one less conditional branch in the hot path for calls for
delegates bound to a function pointer of member function pointer. This
comes at the cost of one additional unconditional branch in the hot
path for calls to delegates bound to functoids (lambdas, functions that
don't take an object reference, other callable objects). This applies
to all delegates, not just device delegates.
Address spaces will now print an error message if a late bind error is
encountered while installing a handler. This will give the range and
address range, hopefully making it easier to guess which memory map is
faulty.
For the simple case of allowing a device_delegate member to be
configured, use a member like this:
template <typename... T> void set_foo(T &&...args) { m_foo_cb.set(std::forward<T>(args)...); }
For a case where different delegates need to be used depending on the
function signature, see src/emu/screen.h (the screen update function
setters).
Device delegates now take a target specification and function pointer.
The target may be:
* Target omitted, implying the current device being configured. This
can only be used during configuration. It will work as long as the
current device is not removed/replaced.
* A tag string relative to the current device being configured. This
can only be used during configuration. It will not be callable until
.resolve() is called. It will work as long as the current device is
not removed/replaced.
* A device finder (required_device/optional_device). The delegate will
late bind to the current target of the device finder. It will not
be callable until .resolve() is called. It will work properly if the
target device is replaced, as long as the device finder's base object
isn't removed/replaced.
* A reference to an object. It will be callable immediately. It will
work as long as the target object is not removed/replaced.
The target types and restrictions are pretty similar to what you already
have on object finders and devcb, so it shouldn't cause any surprises.
Note that dereferencing a device finder will changes the effect. To
illustrate this:
...
required_device<some_device> m_dev;
...
m_dev(*this, "dev")
...
// will late bind to "dev" relative to *this
// will work if "dev" hasn't been created yet or is replaced later
// won't work if *this is removed/replaced
// won't be callable until resolve() is called
cb1.set(m_dev, FUNC(some_device::w));
...
// will bind to current target of m_dev
// will not work if m_dev is not resolved
// will not work if "dev" is replaced later
// will be callable immediately
cb2.set(*m_dev, FUNC(some_device::w));
...
The order of the target and name has been reversed for functoids
(lambdas and other callable objects). This allows the NAME macro to
be used on lambdas and functoids. For example:
foo.set_something(NAME([this] (u8 data) { m_something = data; }));
I realise the diagnostic messages get ugly if you use NAME on a large
lambda. You can still give a literal name, you just have to place it
after the lambda rather than before. This is uglier, but it's
intentional. I'm trying to drive developers away from a certain style.
While it's nice that you can put half the driver code in the memory map,
it detracts from readability. It's hard to visualise the memory range
mappings if the memory map functions are punctuated by large lambdas.
There's also slightly higher overhead for calling a delegate bound to a
functoid.
If the code is prettier for trivial lambdas but uglier for non-trivial
lambdas in address maps, it will hopefully steer people away from
putting non-trivial lambdas in memory maps.
There were some devices that were converted from using plain delegates
without adding bind_relative_to calls. I fixed some of them (e.g.
LaserDisc) but I probably missed some. These will likely crash on
unresolved delegate calls.
There are some devices that reset delegates at configuration complete or
start time, preventing them from being set up during configuration (e.g.
src/devices/video/ppu2c0x.cpp and src/devices/machine/68307.cpp). This
goes against the design principles of how device delegates should be
used, but I didn't change them because I don't trust myself to find all
the places they're used.
I've definitely broken some stuff with this (I know about asterix), so
report issues and bear with me until I get it all fixed.
- Added support for line markers to the preprocessor and parser.
- Added support for include processing to the preprocessor.
- Moved sources base type to plib to be used for preprocessor includes.
This enables to include e.g. from rom memory regions.
- Renamed some defines
- solver now uses dynamic allocation on systems larger than 512x512
- fixed osx build
- moved nl_lists.h classes to plists.h
- fixed netlist makefile clint section
- readability and typos
- optimized the core queue dispatching logic. Minor performance
increase.
- fixed a number of bugs in parray. Now parray<double, 0> will be purely
dynamic allocation with the number of elements passed in the
constructor.
- Added noexpr where appropriate.
- Simplified the queue
Checked with gcc-7 (ubuntu), gcc-9, clang-10, macosx clang 10, mingw
cross compile on linux.
- moved netlists out of driver code into audio/ or machine/ as
nl_xxx.cpp files.
- identified and documented extended validation
- updated arcade, mess and nl targets
- Fix SUBMODEL
- move to strongly typed matrix sort constant
- extend maximum matrix size to 512x512
- optionally do parallel processing based on total operations
- templatize GMRES solver loops
(nw) This has been a long time coming but it's here at last. It should
be easier now that logerror, popmessage and osd_printf_* behave like
string_format and stream_format. Remember the differences from printf:
* Any object with a stream out operator works with %s
* %d, %i, %o, %x, %X, etc. work out the size by magic
* No sign extending promotion to int for short/char
* No widening/narrowing conversions for characters/strings
* Same rules on all platforms, insulated from C runtime library
* No format warnings from compiler
* Assert in debug builds if number of arguments doesn't match format
(nw) Also removed a pile of redundant c_str and string_format, and some
workarounds for not being able to portably format 64-bit integers or
long long.
* Fixed a compiltion bug under clang-cl
Oddly, this problem does not seem to manifest under clang on
gcc.godbolt.org. I suspect that this might be related to the fact that
sizeof(std::size_t) != sizeof(long) on Windows.
* Couriersud feedback
Some unknown system library seems to force the use of the global locale
on OSX. This is not the case for other *nix or Windows builds. This
commit fixes this by forcing the C locale in pfmt.
* hp9825: fixed a bug in 9825t
* hp9845: TACO driver re-written from scratch, DC100 tape separated into
a new device, various adaptations
* hp9845: "new TACO" renamed to just "TACO"
This now works with both single density and double density floppy disks, and
dynamically identifies boot sector IDs required for 6800 booting, and supports
writing back to the 'dsk' image files.
The UniFLEX disk format is not compatible with the Flex format. Significantly it
does not use a mix of single density for booting on some double density disks
which makes it simpler - hardware required a new boot ROM to run UniFLEX.
Further, the UniFLEX sector size is 512 bytes versus 256 for Flex, and the
UniFLEX 'SIR' info sector record is completely different to the info on Flex
disk, and the file system format is also not at all compatible.
Thus the UniFlex format can rely largely on the WD17xx format, with an
overload to handle the sector numbering on the second side continuing from the
first side (one feature in common with the Flex format). This gives a quick
'save' capability and shares code.
Support for 8" disks is included as this was the initial distribution format
and the only one found so far.
The CPU tries to read data in a tight loop, so there must be
some way for it to be halted until data is ready. The current
solution works good enough to boot CP/M.
Also:
- Change FDC to KR1818VG93
- Change disk format from 3.5" to 5.25"
- Add single sided disk format
- Add softlist
Current TZX specification says: "Header blocks have 8063 and data blocks have 3223 pilot pulses"
It also says: "The pilot tone consists in 8063 pulses if the first data byte (flag byte) is < 128, 3223 otherwise"
(unlike tap format that if flag byte is 0 indicates header block and data block otherwise).
CMOS 40xx and 4316 power pins fixed.
Also fixed gcc-9 error. clang++ complains about unreachable code in
nl_base.cpp line 480 even if double parantheses are used. Assigning the
define to a local variable and testing this local variable works. Weird.
The truthtable implementation of 74107 (JK-Flipflop) is included
for educational purposes to demonstrate how to implement state
holding devices as truthtables. It will completely nuke
performance for pong and therefore is disabled.
- converted NL_MAX_LINK_RESOLVE_LOOPS into a netlist parameter.
- Reduced potential bit-rot.
- nltool -v --version now displays values of all compile time defines.
There are still far too many compile time defines. However, most of them
ensure and test future scalability.
* Start fixing OS9 disk handling
* Got it working, now need to refine and test
* Almost complete. Doing more testing...
* Tested aginst a varient of sotware. Solved bug. Cleaned up tabs.
* Turned on sector interleaving in OS9_DSK and retested.
funworld.cpp cleanup: (nw)
* remove revision history - we use version control for a reason
* don't #define things before inluding the PCH, and don't #define generic names before #including anything
Protected defines in nl_config.h with ifdefs. Added a define to disable
queue statistics during compile. This is only needed during development.
Documented performance improvement efforts so I don't try this again.
Runtime performance statistics can now be enabled with nltool
option "-s". To enable those with MAME you need to run
NL_STATS=1 ./mamenl64 -v -oslog game
This fixes an over simplification. Logic devices implicitly assumed that
GND/VDD actually is connected to GND(i.e. 0V). There is no immediate
benefit from this change. It is a preparation for the future
scalability. Now all power terminals (typically 7/14, 8/16) have to be
explicitly connected to the supply rails.
Also added a validation mode to the netlist core. This is not
intended for running, but solely to better indentify pins which
are not properly connected.
mame -validate now also checks all netlist devices. It does this
by constructing a temporary netlist.
This commit also fixes some memory leaks and a bad bug which
surfaced in validation.
LM3900 model 3 only has half the number of BJTs compared to model 1
but delivers comparable results for Money Money.
Model 3 follows the datasheet.
I left code for Model 0 and 2 in for educational reasons.
improvements.
- fix MB3614 parameter
- Added VARCLOCK which derives step size from function
- optimized function handling in CS and VS
- fixed a bug in ppreprocessor
- add trunc to pfunction
- added opamp_amplification_curve to derive characteristic
amplification curve
- Untie diode code
- Fix some typos
- add TYPE=2 opamp model which omits output voltage limitation
Useful for determining causes of non-convergence
- Fix MB3614 opamp parameters to match datasheet
modelling. [Couriersud]
Added global NETLIST.DEFAULT_MOS_CAPMODEL parameter. Setting this to
zero disables using capitance modelling in mos models.
On a per mos device basis this can be achieved by adding CAPMODEL=0 to
the model definition, e.g. MOSFET(X, "NMOS(CAPMODEL=0)")
Improve MOSFET convergence by using log-stepping.
This fixes a rounding issue in the ptime code. This bug surfaced when I
tested netlist with picosecond resolution in ptime.
This will have a small impact on every driver using netlist. For
breakout, it required to adjust the color overlay.
This is a significant improvement to the MOS transistor model. It adds
modelling of the Meyer capacitance model.
This is a somewhat academic addition since the effects occur on a
nanosecond time scale and have a huge impact on performance. I plan to
make the capacitance model selectable. Both on a model level as well as
by introducing a global solver parameter.
The model delivers comparable results to LTSpice.
This effectively reverts b380514764 and
c24473ddff, restoring the state at
598cd52272.
Before pushing, please check that what you're about to push is sane.
Check your local commit log and ensure there isn't anything out-of-place
before pushing to mainline. When things like this happen, it wastes
everyone's time. I really don't need this in a week when real work™ is
busting my balls and I'm behind where I want to be with preparing for
MAME release.
- added MOSFET model. Currently capacitances are not modelled.
This is a 3-pin model (Bulk connected to Source) with provisions to
extend it to 4-pin at a later stage.
- Add a capacitor generic model which is charge conserving.
Switch netlist to use this model instead of constant capacity model.
- Start putting constants into a central place.
Please expect minor timing differences due to a different numerical
path.
The cmos inverter example illustrates the analog implementation of a
cmos inverter gate. These were used a lot back in the 70s/80s to
generate sinus waves. The model should also be able to better emulate
4066 analog switches.
The addition of a relatively simple capacitor model is planned at a
later stage.
Expect everything from the MOSFET model at the current stage. Wrong
results as well as convergence issues and crashes.
MAME crashed if user was trying to load a dsk image with more heads or tracks than disk device supports.
Now the error 'Incompatible image format' is raised.
Despite the overhead - two devices more per transistor - this addition
significantly reduces computing time on switching conditions by reducing
the needed Newton-Raphson loops dramatically.
- 80 column card with SDX in CP/M mode.
- ROM/RAM banking fixed for CP/M, and MTX500 now correctly detected.
- Support for Type 03 and Type 07 .mfloppy images.
- Added alternate MTX2 romset (German).
- Keyboard ROM now selected in Configuration.
- Quickload .RUN files.
Change the sign of go (or in other terms a12 and a21 matrix stencil
elements). This should make further optimization of matrix population
easier.
In addition hopefully improve the readability of the code by sacrifying
overloads for more verbose member names.
Memory management in plib is now alignment-aware. All allocations
respect c++11 alignas. Selected classes like parray and aligned_vector
also provide hints (__builtin_assume_aligned) to g++ and clang.
The alignment optimizations have little impact on the current use cases.
They only become effective on bigger data processing.
What has a measurable impact is memory pooling. This speeds up netlist
games like breakout and pong by about 5%.
Tested with linux, macosx and windows cross builds. All features are
disabled since I can not rule out they may temporarily break more exotic
builds.
Set USE_MEMPOOL to 1 to try this (max 5% performance increase).
For mingw, there is no alignment support. This triggers -Wattribute
errors which due to -Werror crash the build.
At least on macosx memory used by an object seems to be invalidated
before the dtor is executed. This of course is deadly for child objects
with references to the parent-in-deletion which may call back into the
parent.
One of the worst issues I had to fix. Ever. Lesson learnt: No tricks in
dtors. Never.
startup strategies. This determines the order of device triggering.
0: Full - trigger all delegates. Next all devices not touched.
1: Backwards - trigger all devices backwards (only update delegate)
2: Forward - trigger all devices forward (only update delegate)
- convert macros to c++ code.
- order of device creation should not depend on std lib.
- some state saving cleanup.
- added support for clang-tidy to makefile.
- modifications triggered by clang-tidy-9.
Useful for debugging purposes in the end - but not performance.
/*! Store input values in logic_terminal_t.
*
* Set to 1 to store values in logic_terminal_t instead of
* accessing them indirectly by pointer from logic_net_t.
* This approach is stricter and should identify bugs in
* the netlist core faster.
* By default it is disabled since it is not as fast as
* the default approach.
*
*/
#define USE_COPY_INSTEAD_OF_REFERENCE (0)
- Removed dead code.
- nltool now adds a define NLTOOL_VERSION. This can be tested in
netlists. It is used in kidniki to ensure I stop committing
debug parameters.
- Optimized the proposal for no-deactivate hints.
- Documented in breakout that hints were manually optimized.
- Minor optimizations in the order of 2% enhancement.
Still some work ahead to separate interface from execution. This is a
preparation to switch to another sparse matrix format easily which may
be better suited for parallel processing.
On the linear algebra side there are some nice additions:
- Two additional sort modes: One tries to obtain a upper left identity
matrix, the other prefers a diagonal band matrix structure. Both deliver
slightly better performance than just sorting.
- Parallel execution analysis for Gaussian elimination and LU solve.
This determines which operations may be done independently.
All of this is not really useful right now. The matrix sizes are below
100 nets. I estimate that we at least need four times more so that CPU
parallel processing overhead pays off. For GPU, add another order. But
it's nice to have code which may scale.
- src/lib/netlist/solver/mat_cr.h:143:32: error: call to 'abs' is ambiguous
- src/lib/netlist/solver/nld_ms_direct.h:62:19: error: non-type template argument evaluates to 18446744073709551488, which cannot be narrowed to type 'int' [-Wc++11-narrowing]
* Split up the different parts of ICO loading in the menus (locating files, scaling, drawing, etc.)
* Added icon support to software selection menu
* Added support for more ICO file variants, including PNG-in-ICO (new DIB parser is overkill for ICO but I can factor it out for BMP loading at some point)
* Added favourites filter for software menus - includes software that's favourited on any system, so GBC includes DMG favourties and vice versa
* Eliminated unnecessary member variables and O(n) walks in software selection menu
* Made the menus' cached texture structures a bit more efficient
This is an effort to separate netlist creation from netlist execution.
The primary target is to avoid that code which will only run during
execution is able to call setup code and thus create ugly hacks.
The schematic for the tp1983 contains an error. R19 is connected to GND.
This will never work since when Q goes low, the reset pulse will dead
lock Q to low. R19 needs to be connected to VCC (5V). This will generate
the proper, high-pass filtered pulse.
Further fixes for the 7497 as well.
