Firmware developent status

[AndreyB]

I've merged all the downloadable files on the build server into one single http://rusefi.com/build_server/rusefi_bundle.zip

A tiny perk of this file is that in contains a subfolder with current date & time. I hope I did not break too many links

[AndreyB]

Time has come to unify indexation everywhere where live humans can see it - counting of cylinders/coils/channels should always begin from #1.

For now I've changed some lines in showconfig and I bet it would be a mess for some time.

[AndreyB]

A new option if your coil does not like longer dwell while cranking:

Code: Select all

	engineConfiguration->useConstantDwellDuringCranking = true;
	engineConfiguration->ignitionDwellForCrankingMs = 5;

[AndreyB]

It's impossible to see, but that's a 6 cylinder BMW which is rusEfi #10

[video][/video]

[meXanicus]

Cool!

[kb1gtt]

Yes very cool. I see the car got so excited it blew it's hood off Kind of like it's saying, see what I have inside

[AndreyB]

There is a critical issue with the firmware - I recommend against starting engines, any pin can stick open. More on this @ https://sourceforge.net/p/rusefi/tickets/107/
This project needs more software developers

I've also added a feature into dev console - it would check at the server if a critical issue is pending and show a warning on startup.

[AndreyB]

Here are my S2R for that #107 issue:

1) set_engine_type 25
2) enable self_stimulation
3) rpm 2000
4) writeconfig
5) reboot the chip
6) wait for between 3 and 30 minutes - RED led

[AndreyB]

I love C, but

Say (stkalign_t *)(__get_SP() - sizeof(struct intctx)) = 0x2001FEF4

Say stkalign_t * otp->p_stklimit = 0x2001FC00

What would be
int remaingStack = (stkalign_t *)(__get_SP() - sizeof(struct intctx)) - otp->p_stklimit;

Do you think you know the answer?

[kb1gtt]

That's easy it's 42. http://en.wikipedia.org/wiki/42_(number) For those that haven't read the 5 book trilogy, known as the hitch hikes guide to the galaxy, I'd recommend it.

[AndreyB]

The https://sourceforge.net/p/rusefi/tickets/107/ critical issue is now RESOLVED. I have two boards running for 9 hours both.

The root cause was irq nesting - in debug I've experienced up to 4 IRQs nesting one within another, most of them had no reason to have higher priority then others. As a fix I've assigned same priority to all IRQs. If that would have negative effects on performance I would revisit this, but anyway no reason to have six different levels of priority. The evil with nesting is that the planets have to align right for you to test this condition and there is nothing you can do to get it sooner then later.

[AndreyB]

Anyone to participate in dev console development? Log viewer needs some TLC - for instance here's a fresh log
https://svn.code.sf.net/p/rusefi/code/trunk/misc/logs/e34_60_2/MAIN_rfi_report_2014-11-12_21_37.csv

Need to implement gauges in synch with digital sniffer scrolling, digital sniffer scrolling key stoke etc.

[AndreyB]

@ made some usability improvements to the dev console. Thank you Sébastien!

[3400tZ]

Hey Andrey, well thanks to you for this great project! I will try to get some more patches to you in the next few days hopefully. I have a couple more improvements I want to do

[AndreyB]

A lot of process was made this weekend, two nasty bugs fixed.

I really need a week without new major defects, that would be so nice...

[AndreyB]

I am again banging my head over some performance Still sporadic trigger synchronization issues with 60/2 @ 5000 rpm. Again not exactly sure what is the issue about.

[AndreyB]

I am again banging my head over some performance Still sporadic trigger synchronization issues with 60/2 @ 5000 rpm. Again not exactly sure what is the issue about.

We only have 100uS between rise and fall of a 60-2 trigger @ 5000 rpm, that's 18000 (18K) CPU ticks.

The way the firmware is designed right now, it wants to do a lot of calculations for the whole next engine cycle within the 1st tooth handling. triggerinfo now shows ignitionMathTime value which is the duration of this calculation in systicks. Right now this value is about 5K cpu ticks and that's a lot, 5K is too close to 18K.

[AndreyB]

On the one hand, I've improved performance. On the other hand, I am not sure if that was the problem - the problem is still there

[AndreyB]

Hell Yes!

I had to scalp a lot of excess in the code - less method calls more macro, less pointers more global variables. Also a huge part of the problem was a less-then-correct gap setting for 60/2 trigger (was 2.5+-25%. should be 3+-25%). https://sourceforge.net/p/rusefi/tickets/115/ is resolved

2014-11-26 17_36: EngineState: isError No/total errors=79 ord_err=3/total revolutions=176872/self=No
2014-11-26 17_36: EngineState: sn=Yes ignitionMathTime=1380 schTime=1409 triggerMaxDuration=21524
2014-11-26 17_36: EngineState: maxLockTime=8417 / maxTriggerReentraint=0
2014-11-26 17_36: EngineState: maxEventQueueTime=2465

79 errors over 176872 cycles is good enough for now. triggerMaxDuration could and should be lowered further but good enough for now.

[puff]

congrats!

[AndreyB]

Ops, there was a bug in 'fuelpumpbench' & 'fanbench' commands. This explains some of the weird stuff recently...

Now fixed.

[AndreyB]

New command: stopengine

[AndreyB]

[video][/video]

Note the delay between LCD turning on and the status LED (top-right corner of the frame) turning on. (the slow camera cannot shoot the blinking but it's there)

Also note the spark channel #1 being ON on startup (bottom-left corner of the frame)

https://sourceforge.net/p/rusefi/tickets/120/

I kinda hope I know where the bug is, wish I had noticed it before

[AndreyB]

Idle solenoid is now off while the engine is not spinning - #117

[AndreyB]

[video][/video]

[AndreyB]

set_idle_pwm PWM command now accepts a numeric percent value, for example

set_idle_pwm 75.125

would keep the valve 75.125% open

[AndreyB]

[video][/video]

[AndreyB]

Happy new year to me:

Code: Select all

                    // todo: improve this hack
                    int i = response.indexOf("2014: ");
                    if (i != -1)
                        response = response.substring(i + 6);

"2014:" as in "year 2014"... Not any more, time to fix this hack in dev console log viewer

[AndreyB]

Usability: there is now a tiny rusefi_console.exe launcher inside rusefi_bundle.zip - that's for people who do not have Java installed.

[AndreyB]

Usability improvement: dev console now remembers last open tab & last entered command. The addition of configuration saving opens potential for more usability improvements.

New TODO usability tickets:
https://sourceforge.net/p/rusefi/tickets/134/
https://sourceforge.net/p/rusefi/tickets/135/
https://sourceforge.net/p/rusefi/tickets/136/