Re: Wireless - switchless kernal mod

On Mon, Apr 9, 2018 at 8:44 AM,  <afachat@gmx.de> wrote:
> On Montag, 9. April 2018 02:05:43 CEST Nejat Dilek wrote:
>> On Tue, Apr 3, 2018 at 7:44 PM, Mike Naberezny <mike@naberezny.com> wrote:
>> > I used an AVR pin change interrupt to detect an edge on a past project.
>> > The code was in assembly.  The AVR has a large number of working
>> > registers so I reserved one of them for the pin change interrupt service
>> > routine.  The very first instruction of the ISR, before preserving
>> > registers or anything else, was to read the port into that dedicated
>> > register.  This was to capture the state of the pin as close in time as
>> > possible to when the edge occurred. The ISR would then compare the value
>> > to detect the edge and either continue or bail out.
>> >
>> > Source code:
>> > https://github.com/mnaberez/vwradio/blob/b06c3a95f9b345dc4330cc3a8751b3f62
>> > 44829ba/avr_volume/firmware/m62419fp_spi.asm#L50-L61
>> >
>> > This is only an approximation of edge detection.  Several processor cycles
>> > will occur between the edge and that first instruction of the ISR
>> > capturing
>> > the port value.  The pin may change state again during that time.  In my
>> > project, the signal I was sampling was slow enough that this was
>> > acceptable.
>> >
>> > Regards,
>> > Mike
>>
>> Nice clever approach! I'm familiar with the syntax but I never handled
>> these microcontrollers with only assembly. So I'll definitely go along
>> the C route even if I hate soldering bodge wires.
>
> You can have an example of a mixed ASM/C approach here https://github.com/
> fachat/XD2031/tree/master/firmware/xs1541 in atn.S where I use the pin change
> interrupt to detect the ATN line going low. The few following assembly code
> lines set DATA / NRFD low like the 1541's/2031's XOR gates on the IEC/IEEE bus
> would do so the C64/PET does not trigger on a "device not present". All the
> rest of the code is in C.
>
> In the interrupt I only save the registers that are necessary in the interrupt
> routine, which makes it rather small and fast.
>

Well I'm only familiar with the syntax actually :) Out of laziness, if
possible I prefer the shortest route that is possible. Of course there
are side advantages like portability, readability and ease of
maintenance.
In this project speed is the last priority and I want to avoid strict
timing requirements.

Bu these are good options, the last time I used inline assembly was to
sync to the PAL C64's clock with an Atmega328 with such a construct.

#define WAIT_1_CYCLE asm("nop\n nop\n nop\n nop\n nop\n nop\n nop\n
nop\n nop\n nop\n nop\n nop\n nop\n nop\n nop\n nop\n")
#define WAIT_2_CYCLE WAIT_1_CYCLE;WAIT_1_CYCLE;
#define WAIT_3_CYCLE WAIT_2_CYCLE;WAIT_1_CYCLE;
#define WAIT_4_CYCLE WAIT_3_CYCLE;WAIT_1_CYCLE;asm("nop\n");
#define WAIT_5_CYCLE WAIT_4_CYCLE;WAIT_1_CYCLE;

Regards,

Nejat