Fix interupt -> interrupt typo, including in some function names.
Signed-off-by: Sami Liedes <sliedes@cc.hut.fi>
Do not let code clear the UDRE flag in USART.
It's a read only location and clearing it may break software that
assumes it cannot be cleared.
Signed-off-by: Sami Liedes <sliedes@cc.hut.fi>
Merge remote-tracking branch 'sliedes/to-upstream' into dev-home
Tweak struct avr_t to always include a pointer to trace data structure.
Previously the size of struct avr_t was dependent on whether the
preprocessor macro CONFIG_SIMAVR_TRACE was defined. This was problematic
since it meant that applications that link to simavr need to define it
the same way as it is defined in simavr Makefiles.
Change it to a single pointer to a struct, which is allocated iff
tracing is compiled in and a NULL pointer otherwise.
Signed-off-by: Sami Liedes <sliedes@cc.hut.fi>
Automate test cases.
This patch implements a framework for test cases. Each test case is
compiled into an individual executable. All test cases can be run by
invoking the command `make run_tests', which runs the shell script
run_tests.
Signed-off-by: Sami Liedes <sliedes@cc.hut.fi>
Include sim_gdb.h from sim_gdb.c. Also fix prototype in .h.
sim_gdb.c did not include sim_gdb.h. Also there was a slight
difference in the prototype in the header and the implementation in
the .c, causing a warning.
Signed-off-by: Sami Liedes <sliedes@cc.hut.fi>
uart: Touchup on the xon/xoff IRQs
Set/reset XON & XOFF irqs following the status
of the fifo. It means they can be added to a VCD
file and help debug a "part".
Signed-off-by: Michel Pollet <buserror@gmail.com>
uart: Added a comment
As to why somehow we don'r write a value in the interupt
clearing logic
Signed-off-by: Michel Pollet <buserror@gmail.com>
twi: Placeholder example board
Uses AVR application note implementation of master/slave mode
Placeholder, non functional
Signed-off-by: Michel Pollet <buserror@gmail.com>
twi: New master implementation
Yanked the old overengineered version, replaced it with a KISS
version. Untested, work in progress. Needs a new testing harness
Signed-off-by: Michel Pollet <buserror@gmail.com>
regbits: Added a "raw" variant to get/set
Allows comparing to "real life" constants, used in Twi module
Signed-off-by: Michel Pollet <buserror@gmail.com>
part: Made irq_names static
Have to to have several of them...
Signed-off-by: Michel Pollet <buserror@gmail.com>
Makefile: Fix for x86_64
silly architecture names...
Signed-off-by: Michel Pollet <buserror@gmail.com>
Makefile updates
Changed the makefiles to derivate the target architecture from
gcc itself, and compile all the objects into obj-$arch. This
allows several compilers to be used to build simavr, at the same
time without interferences.
This allow arm cross compilation to live in the same tree as
the x86 one...
Signed-off-by: Michel Pollet <buserror@gmail.com>
Updated gitignore
Added obj-* and *.vcd
Signed-off-by: Michel Pollet <buserror@gmail.com>
Remove unused variable from avr_service_interrupts().
The done flag is nowadays redundant as it's never read after it's set.
Signed-off-by: Sami Liedes <sliedes@cc.hut.fi>
Fix LD, ST, LPM, STS, MUL to take the correct number of cycles.
They are now the ATMega timings. Need to figure out how to take
ATTiny into account. Before this change they weren't correct for
either.
Signed-off-by: Sami Liedes <sliedes@cc.hut.fi>
fix typos in comments.
Signed-off-by: Sami Liedes <sliedes@cc.hut.fi>
Silence gcc warnings by using PRIu64 in CRASH() printf().
Actually we define a PRI_avr_cycle_count preprocessor macro in
sim_avr.h in the style of PRI* in inttypes.h.
Signed-off-by: Sami Liedes <sliedes@cc.hut.fi>
Fix compilation with TRACE on.
The MOV instruction's TRACE() used the undefined variable vd.
Signed-off-by: Sami Liedes <sliedes@cc.hut.fi>
Fix compilation on non-32-bit platforms.
The addition of -Werror to the compilation flags broke building on
64-bit hosts because casts between pointers and integers of different
sizes cause warnings on gcc -Wall. Fix this by always casting the
pointers to intptr_t first.
Signed-off-by: Sami Liedes <sliedes@cc.hut.fi>
irq: Add names to most io module external irqs
Nothing to be shown for this work so far, the cool stuff hopefully
will come later...
Signed-off-by: Michel Pollet <buserror@gmail.com>
doc: Updated callgraph .pdf
Just a minor update
Signed-off-by: Michel Pollet <buserror@gmail.com>
examples: Updated to new part prototypes
Followup on previous checkin, the board examples now use
"names" when possible.
Signed-off-by: Michel Pollet <buserror@gmail.com>
parts: Updated to name their IRQs
First bits of the system that provide names for the IRQs
Signed-off-by: Michel Pollet <buserror@gmail.com>
simavr: Updare simavr core to new IRQ prototypes
Mostly prototypes changes and some gratuitous reformatting
Signed-off-by: Michel Pollet <buserror@gmail.com>
irq: Introduce a "irq pool"
The goal is to be able to name IRQs, so a couple fields are
added.
It is not functional for now, as the task is to first convert
all the code to the new prototype.
Also introduce a 2011 way of formatting prototypes
Signed-off-by: Michel Pollet <buserror@gmail.com>
simavr: Added -Werror and fixed warnings
Added -Werror to compilation to make sure the next wave of changes
can't easily break things
Signed-off-by: Michel Pollet <buserror@gmail.com>
core: Add a muxer for IO writes
It turns out that some core (tiny85 at the very least) are competing
for IO writes on some register.
This patch introduces a muxer callback that can call multiple clients
without impacting performance for the rest of the cores.
The system could be extended for IO read if it proves necessary.
Signed-off-by: Michel Pollet <buserror@gmail.com>
core: Introduce run() and sleep() callbacks
It seems that the default 'run_avr' function was penalized by
some tests made that were only useful when gdb was active.
This patch introduces two version of the critical run_avr, one for
"raw" running, and one for gdb.
It also introduces a sleep() callback for similar reasons. Ir can
also be used to run unit testing at 100% speed instead of having
random sleep code during execution.
Signed-off-by: Michel Pollet <buserror@gmail.com>
Fix SBIS etc trace output.
SBIS used to print the io register wrong. The other fixes add spaces
in the correct places in "Will branch" and "Will not branch" messages.
Signed-off-by: Sami Liedes <sliedes@cc.hut.fi>
avr_loadcode(): If too much code, bail out instead of crashing.
Check that the code to be loaded actually fits in the flash.
Signed-off-by: Sami Liedes <sliedes@cc.hut.fi>
avr_register_io_[read,write](): fail if overriding earlier registration.
The avr_register_io_*() functions do not maintain a chain of
callbacks. This can be confusing if someone like me tries to use them
to listen to some I/O events (for which you should use the avr_irq_*
stuff).
Make this more explicit by checking in the register functions if we
already have a callback and failing with an error message if so.
Signed-off-by: Sami Liedes <sliedes@cc.hut.fi>
board_hd44780: New sample code
Emulated a complete HD44780 LCd screen, and uses a standard
AVR driver to write on it.
Signed-off-by: Michel Pollet <buserror@gmail.com>
vcd: Increase trace buffer size
Ran into the 128 limit at some point..
Signed-off-by: Michel Pollet <buserror@gmail.com>
Makefiles: Some fixups
+ add -fPIC to -shared for the shared library
+ .axf rule try to comoile all dependencies
Signed-off-by: Michel Pollet <buserror@gmail.com>
Merge git://gitorious.org/~luki/simavr/lukis-simavr into dev-home
Makefile: Add building of shared library on linux
Disabled for non linux platform, as it need more work
than just a gcc -shared
Author: Martin Glueck <martin@mangari.org>
Edit: Michel Pollet <buserror@gmail.com>
core: Added ATtiny x4
Untested. Also added to sim_avr.c
Author: Martin Glueck <martin@mangari.org>
Edit: Michel Pollet <buserror@gmail.com>
core: Remove the run() callback
run() callback was unused but was still using cycles
Signed-off-by: Michel Pollet <buserror@gmail.com>
Makefiles: Explicitely link opengl
needed with modern linker
Signed-off-by: Michel Pollet <buserror@gmail.com>
timer: Fix a subtle off-by-one bug in TCNT reading.
There's a subtle bug in TCNT (timer counter) reading which only causes
the counter to ever run from 0 to TOP-1, while it should stay in the
TOP value for a full timer cycle before resetting to 0 (but the
interrupt needs to come when TOP is first reached).
Interestingly, this causes Arduino's micros() function to occasionally
return incorrect values. micros() function relies on TCNT0 not
transitioning from <= 254 to 0 in only a few cycles with /64
prescaler.
Signed-off-by: Sami Liedes <sliedes@cc.hut.fi>
timers: don't reconfigure for minor changes
Prevents the timer to be reset when changing non important
bits, like the edge detector etc
Signed-off-by: Michel Pollet <buserror@gmail.com>
timers: Print a warning on unsuported timer modes
and also don't [rint one for "mormal" mode
Signed-off-by: Michel Pollet <buserror@gmail.com>
Store the next cycle count where the earliest cycle timer needs to be run.
This speeds up cycle timer processing significantly and almost doubles
the speed of the overall simulation.
Signed-off-by: Sami Liedes <sliedes@cc.hut.fi>
Rework avr_service_interrupts() to use ffs() too for speed.
Use ffs() for finding timers from cycle timer map.
This optimization speeds up avr_cycle_timer_process() considerably. It
used to be a major bottleneck in my test case (52% of CPU time spent
in that routine), now it's about half that.
Return zeros when GDB reads just past end of stack.
GDB likes to read the topmost value of stack when using the
"stepi" (step instruction) command. Unfortunately this does not work
when the stack is empty, causing an error message and GDB confusion.
Hack the GDB stub code to just return zeros if GDB tries to read two
bytes just past the end of stack.
NOTE: This may need to be modified to support also longer reads for
bigger AVRs where code pointers are longer than two bytes (e.g.
ATMega2560). As simavr doesn't support atm2560 yet, I'm implementing
it this way now.
Enable interrupt handling for Fast PWM timers too
Arduino seems to use that.
UART: Fixed UDRC/E behaviour
The interrupt is suposed to fire as soon as it is
enabled is the UDR is empty.
Signed-off-by: Michel Pollet <buserror@gmail.com>
cycle timers: Added avr_cycle_timer_status()
Allows verification wether a timer is running, and
until when it is.
Signed-off-by: Michel Pollet <buserror@gmail.com>
simduino: added init/deinit functions for avr_t initialization
fixes a crash (free of mmap'ed memory)
nicer cleanup by munmap + close the file
Signed-off-by: Christian Balles <code@ballessay.de>
core: added extra init/deinit functions to avr_t
allows programs like the simduino example to modify the init/deinit function
to fix a crash
Signed-off-by: Christian Balles <code@ballessay.de>
misc: Typos
Fixed a few comments
Signed-off-by: Michel Pollet <buserror@gmail.com>
core: Added a bit of documentation
Fixed some of the copy/paste errors in comments
Signed-off-by: Michel Pollet <buserror@gmail.com>
gdb: Adds the 'reset' command
Apparently needed for compatibility with AVRStudio debugger
Signed-off-by: Michel Pollet <buserror@gmail.com>
console: Implement debug console register
Used in pair with the console register declaration, just
prints what's sent to that debug register.
Signed-off-by: Michel Pollet <buserror@gmail.com>
mcu_section: Added New CONSOLE Tag
CONSOLE tag allow declaring a register for dumping debug output
oh the terminal without using an UART
Also cleaned up and documented the header file further
Signed-off-by: Michel Pollet <buserror@gmail.com>
tests: Update them so they work...
They relied on bugs that were fixed in simavr since..
Signed-off-by: Michel Pollet <buserror@gmail.com>
IOs: Use new IRQ alloc function
All converted to the brand new way...
Signed-off-by: Michel Pollet <buserror@gmail.com>
IRQs: A bit of sanity when freeing
Check the parameters...
Signed-off-by: Michel Pollet <buserror@gmail.com>
simavr: Deallocate resources
Deallocate memory and IO modules at teardown
Signed-off-by: Michel Pollet <buserror@gmail.com>
eeprom: Added a dealloc function
frees the eeprom memory at teardown time
Signed-off-by: Michel Pollet <buserror@gmail.com>
ios: Added a way to set and teardown modules IRQs
Added a function to set the io module IRQ, and added a way
to deallocate them at teardown time.
Signed-off-by: Michel Pollet <buserror@gmail.com>
misc: Made (most) of the headers c++ friendly
No functional changes
Signed-off-by: Michel Pollet <buserror@gmail.com>
misc: Fixed clang static analizer warnings
One genuine bug too. The rest was cosmetic.
Signed-off-by: Michel Pollet <buserror@gmail.com>
ioport: Added an IRQ for the port pin direction changes
New IRQ for client code (ie "board code") will be called
when the AVR firmware changes the DDR register.
Signed-off-by: Michel Pollet <buserror@gmail.com>
ioport: Add an ioctl to get the port state
Example:
for (int i = 'A'; i <= 'F'; i++) { avr_ioport_state_t state;
if (avr_ioctl(AVR_IOCTL_IOPORT_GETSTATE(i), &state) == 0)
printf("PORT%c %02x DDR %02x PIN %02x\n",
state.name, state.port, state.ddr, state.pin);
}
Signed-off-by: Michel Pollet <buserror@gmail.com>
misc: Typos etc
Makefile now uses $(CC) so it works with clang static analyzer.
Signed-off-by: Michel Pollet <buserror@gmail.com>
core: CALL/RET instructions fix
Swap LSB and MSB of return address on the stack.
Thanks to "Peter" on the simavr mailing list.
Signed-off-by: Michel Pollet <buserror@gmail.com>
core: Add PORTA support for Attiny2313
The Attiny2313 has a PORTA, the pins for it are normally used for
reset and the crystal, but it is possible to use them if you use the
internal R/C oscilator.
Simavr didn't support this port on a Attiny2313 yet, so I added it myself.
Signed-off-by: Wander Winkelhorst <w.winkelhorst@gmail.com>
Signed-off-by: Michel Pollet <buserror@gmail.com>
timer: Fix for older AVRs & tinies
They lack WGM bits
Signed-off-by: Michel Pollet <buserror@gmail.com>
core: Fix SBRS/C for 32 bits instructions
Untested, but matches the datasheet
Signed-off-by: Michel Pollet <buserror@gmail.com>
core: Fix MULS register
MULS uses 16+ registers.
Signed-off-by: Michel Pollet <buserror@gmail.com>
ADC: Fix ADLAR bit handling
Was shifting when zero, not one.
Signed-off-by: Michel Pollet <buserror@gmail.com>
timer: Also reconfigure on writing the WGM bits
Some broken code configure the clock source before
setting the waveform bits.
Signed-off-by: Michel Pollet <buserror@gmail.com>
ioports: Cleanup and implement PINx toggle
Turns out writing 1 to PINx register toggles the
bit in PORTx. Did'nt know it worked like that...
Signed-off-by: Michel Pollet <buserror@gmail.com>
uart: Now use avr_clear_interrupt_if
New function simplifies clearing pending bits
Signed-off-by: Michel Pollet <buserror@gmail.com>
timer: Allow clearing of the interupts
Allow exclicit clearing of interupts by writing 1
to the pending bit.
Signed-off-by: Michel Pollet <buserror@gmail.com>
interrupts: Mark interupts as pending even if not enabled...
.. Then trigger them as soon as they are enabled again
There is a bit of performance hit on this, since the bitmask
array used before can now be full of pending bits that won't
ever be cleared..
Signed-off-by: Michel Pollet <buserror@gmail.com>
ICP: Added Input capture pin support to timers
Added (untested as of now) Input Pin Capture to the 16 bits
timers. The "edge" flag is also handled. The code is untested
for now, it will need an "example" board that will be
checked in later.
Signed-off-by: Michel Pollet <buserror@gmail.com>
cores: Added tinyX5 and mega128 ADC bits
Not tested
Signed-off-by: Michel Pollet <buserror@gmail.com>
misc: Warning fix
Makes gcc happy
Signed-off-by: Alessio Igor Bogani <abogani@texware.it>
Signed-off-by: Michel Pollet <buserror@gmail.com>
misc: More warning fixes
#warning "This file has been moved to <util/delay.h>."
Signed-off-by: Alessio Igor Bogani <abogani@texware.it>
Signed-off-by: Michel Pollet <buserror@gmail.com>
misc: Fix warning replacing %d with %zu for size_t
Warning fixes.
Signed-off-by: Alessio Igor Bogani <abogani@texware.it>
Signed-off-by: Michel Pollet <buserror@gmail.com>
doc: Updated callgraph .pdf
Hadn't been updated for some time...
Signed-off-by: Michel Pollet <buserror@gmail.com>
core: Fixes SBCI
Thanks to <evgeny.chukreev@gmail.com>
Signed-off-by: Michel Pollet <buserror@gmail.com>
test: Add a test/example for the ADC subsystem
Loads ADC values, also reads the V1.1 reference voltage of an Atmega644
Signed-off-by: Michel Pollet <buserror@gmail.com>
cores: Add (some) of the ADC mux data
Add ADC Mux data to the x8 and x4 cores. The others still needs
to be done. Also filled the reference voltages.
Signed-off-by: Michel Pollet <buserror@gmail.com>
ADC: Fully functional core
simavr ADC allows external code to feed real voltages to the
simulator, and the simulator uses it's 'real' reference voltage
to do the right thing and return the 'proper' 10 bits ADC value
to the AVR firmware.
To send values to the ADC, register your code to wait for the
ADC_IRQ_OUT_TRIGGER irq, and at that point send any of the
ADC_IRQ_ADC* with Millivolts as value.
External trigger is not done yet.
Signed-off-by: Michel Pollet <buserror@gmail.com>
core+elf: Add fields for the MCU voltages
You can now specify AVR_MCU_VOLTAGES(vcc, avcc, aref) in millivolts
in your firmware to set them into the simavr core.
Prerequisite for ADC VREF support.
Signed-off-by: Michel Pollet <buserror@gmail.com>
run_avr: Add a way to trace just interruots vectors
Calling run_avr with -ti <value> will trace the scheduling
of that pacticular vector.
You can have up to 8 traces.
Signed-off-by: Michel Pollet <buserror@gmail.com>
timer: Fixed an off by one
That was breaking the timers badly in corner cases
Signed-off-by: Michel Pollet <buserror@gmail.com>
cores: Updated comparators
Also filled up the ones for tinyx5, megax4, tiny2313
Signed-off-by: Michel Pollet <buserror@gmail.com>
timer: Changed timer config feom a bit array to a normal regbit
Since all the timer seems to have these 2 bits declared
in the same way anyway...
Signed-off-by: Michel Pollet <buserror@gmail.com>
ELF: Fixes a bad malloc
How it didn't crash before, no idea...
Signed-off-by: Michel Pollet <buserror@gmail.com>
mega128: Fixes timer comparators
Bits are in the right order now
Signed-off-by: Michel Pollet <buserror@gmail.com>
twi: Stripped down TWI module to fix compile
New system will take it's place soon
Signed-off-by: Michel Pollet <buserror@gmail.com>
example: Added OC2A to timer test
Uses OC2A toggle with a VCD trace to demo the new compators.
Signed-off-by: Michel Pollet <buserror@gmail.com>
cores: Refactored timer comparators
Based on patch by <tomi.leppikangas@gmail.com> but fixed the other
cores to fix a few cut/paste issues.
Signed-off-by: Michel Pollet <buserror@gmail.com>