 Notice. New forum software under development. It's going to miss a few functions and look a bit ugly for a while, but I'm working on it full time now as the old forum was too unstable. Couple days, all good. If you notice any issues, please contact me. |
Forum Index : Microcontroller and PC projects : PicoMite MMbasic V7.00.00 betas:
  Page 3 of 6   |
|||||
| Author | Message | ||||
| matherp Guru  Joined: 11/12/2012 Location: United KingdomPosts: 12034 |
Version V7.00.00b4 is now available at https://github.com/UKTailwind/PicoMite/releases/tag/PicoMite-V7.0.00b4 This includes more optimisation for the rp2350 and increased user heap for the rp2350 as well as fixing the bugs identified above Volhout: try la_24_2 on this at 378MHz and have a look at the cycle time.  |
||||
| hhtg1968 Senior Member  Joined: 25/05/2023 Location: GermanyPosts: 247 |
I do not close the housing of my pico 3 any longer... one beta after the other. but it is so cool to test... |
||||
karlelch Guru Joined: 30/10/2014 Location: GermanyPosts: 361 |
Thanks for the new version, Peter! I tested Version V7.00.00b4 with my current robot control software (running on a standard Pico2W). It does a lot of math, communicates with digital servos via a COM port, and sends servo data via UDP to a host PC. At first glance, I did not find any issues. The increase in speed is small but detectable, but then I tried to optimize the MMBasic code already. Options WebMite MMBasic RP2350A Edition V7.0.00b4 OPTION FLASH SIZE 4194304 OPTION LIBRARY_FLASH_SIZE 34000 OPTION COLOURCODE ON OPTION HEARTBEAT OFF OPTION CPUSPEED (KHz) 315000 OPTION DISPLAY 64, 80 OPTION WIFI Elchland3, ****************, PICO3A71AD2F089 OPTION PLATFORM walkr|server Option Compile Off Run Loop 0.364 ± 0.659 ms (n=26811, -0.166..9.877) Exec 8.926 ms (n=45, 8.853..9.015) Move 9.803 ms (n=47, 9.301..19.398) Option Compile On Loop 0.304 ± 0.288 ms (n=61485, -0.190..0.809) Exec 8.824 ms (n=93, 8.746..8.910) Move 8.310 ms (n=94, 7.903..17.666) > ?MM.Info$(compile) NONE 0 REUSED 4 STMTS 2448 RAN 1442659 MISS 156107 CODE 727922 SIZE 66228 OF 636 [PERF] elapsed=13298965 us statements=575991 findvar=1288427 (locals=903188 [2 [PERF] top commands by dispatch count: 210515 Let 165838 If 39045 Static 38715 /* 37870 Local 19493 Loop 4713 Inc 3883 Cat 1071 Poke 838 Math 558 Print 207 Pause 83 Const 76 For 57 Dim 47 Read 25 WEB 16 Do 8 Restore 7 Select Case [PERF] top SUBs by exclusive (self) time: self_us incl_us calls self_us/call name [PERF] top SUBs by call count: 19290 R._timing 19290 R.updatePosture 13361 R.spinGaitGen 357 ST.sendPacket 150 R._bodyIK 150 R._legIK 48 _update_timing 36 ST.stopServo 25 R._computeMove 25 R._executeMove 25 R.inputUpdate 25 R.generateGaitSeq 25 R.Angles2Pos 25 R._moveAllServos 15 ST.waitForStop 9 R.inputBody 9 R.inputPosition 9 R.inputWalk 4 R.resetInput 4 R.getServoState Option Compile Shadow does not report any issues. Option Compile Shadow Loop 0.352 ± 0.287 ms (n=20497, -0.147..0.857) Exec 8.795 ms (n=25, 8.718..8.898) Move 10.506 ms (n=27, 9.856..19.813) If someone is interested what the software does: I am building a new hexapod robot now with serial servos. To debug the software and improve the movements before the hardware of the robot is finished, Claude helped me to write some ROS2 software that runs on a PC, receives the servo positions from the MMBasic firmware via UDP and applies them to a model of the robot. https://youtu.be/L64OecLvrtc Cheers Thomas Edited 2026-10-03 20:29 by karlelch |
||||
| matherp Guru Joined: 11/12/2012 Location: United KingdomPosts: 12034 |
I'm now extending the compiler's capabilities. If you post your code I can analyse it and see what statements aren't being compiled and potentially include them in the next version. |
||||
| karlelch Guru Joined: 30/10/2014 Location: GermanyPosts: 361 |
Thanks - the code is in my repository https://github.com/teuler/walkr-robot/blob/master/code-mmbasic/ `hpx_srv.bas` requires `lib_input.bas` and `lib_st3215.bas` on `A:`. No additional hardware is needed |
||||
| PhenixRising Guru Joined: 07/11/2023 Location: United KingdomPosts: 2097 |
Been out of the MMBasic loop for several weeks due to big-panic work issue. I return to find this amazing insanity  Haven't found my feet yet but I tested on my RP2350 DIL module and got an immediate ~84% percent speed increase. For the sake of comparison, what was happening 20,000 times/sec is now 37,000 times/sec with OPTION COMPILE ON. But here's what shocked me; I used "crunch" in MMEdit IDE and this jumped to 57,000 times/sec  This has me confused because I would've though that the compiler would have done similar(???) Edit: Correction 53,000 times/sec. Edited 2026-10-03 22:44 by PhenixRising |
||||
| matherp Guru Joined: 11/12/2012 Location: United KingdomPosts: 12034 |
karlech You use types in your program (good) but you also use normal variables with dots in their names (bad). The compiler doesn't support types and doesn't have the ability to scan the program and work out what is an element of a type and what is a simple variable so it doesn't compile any line which include a dot in a name. As an experiment you could bulk edit the program and see what the performance gain would be. PhenixRising The impact of comments and blank lines is a limitation in b4 that will be addressed in b5 |
||||
| PhenixRising Guru Joined: 07/11/2023 Location: United KingdomPosts: 2097 |
Pete, I am giddy beyond belief. This development is simply astonishing. Many thanks X 1,000,000   |
||||
Bleep Guru Joined: 09/01/2022 Location: United KingdomPosts: 863 |
Hi Peter, Testing various programs, but I've found Chess fails in a very strange way, at a point where it has never had a problem and I can't see anything inherently wrong with it, I've tried giving all the If statements in this Function an EndIf incase that was the problem, but still get the error. Just run the chess program and then type bench at the prompt. I get the same error simply playing a game, it sometimes isn't straight away. Used to work on all previous versions, also works in Fuzix. Because of it's size and complexity, might be a good candidate for your further speed improvements? :-) Regards Kevin. chessbas.zip Edited 2026-10-04 04:24 by Bleep |
||||
| matherp Guru Joined: 11/12/2012 Location: United KingdomPosts: 12034 |
It is almost certainly running out of stack space. I'll have a look at it. Confirmed - tested on various versions. Is your code recursing more than the original? Do you have the original you can post? Edited 2026-10-04 05:10 by matherp |
||||
| gadgetjack Senior Member Joined: 15/07/2016 Location: United StatesPosts: 240 |
After flashing new firmware , I get "input not supported" on my monitor. This worked good until version 7. Is the start up screen at a higher resolution than it used to be? Confused.... |
||||
| gadgetjack Senior Member Joined: 15/07/2016 Location: United StatesPosts: 240 |
update - No version after 6.02 works on hdmi. I had not update for a while. I tried going backwards till it worked again. I tried 3 different monitors in case i had a bad one. Same on all 3. |
||||
| gadgetjack Senior Member Joined: 15/07/2016 Location: United StatesPosts: 240 |
Ok , figured it out. I tried my other board and it ran fine. It was the Pimoroni module that would not run it correctly. I guess it was too fast for it. One with the core module runs great. I will stick to the 6.02 version for the other board. |
||||
| Bleep Guru Joined: 09/01/2022 Location: United KingdomPosts: 863 |
Hi Peter, As far as I remember, no actual changes to the way the code worked were made, so as far as I remember the recursion was the same, originally on the 2040 I reduced the MAX_PLY, and GEN_STACK but now they have been put back to original. When this problem occurred I did try reducing them both again, but there was no difference. I'm away at the moment, so don't have access to my home sytem & I'm not sure I have the "original" here, I'll have a search later today. Regards Kevin OK found this, I think it might be original or close. TSCPChess.zip Edited 2026-10-04 15:49 by Bleep |
||||
| karlelch Guru Joined: 30/10/2014 Location: GermanyPosts: 361 |
I changed the variable and subroutine names that contain a `.` to `_` - expect the Types, of course. There is an improvement in the computation of the move (from ~8 to ~5 ms per round). Executing the move stays the same, which is not unexpected because here the commands are sent to the servos. Option Compile Off Loop 0.330 ± 0.289 ms (n=51368, -0.172..0.867) Exec 9.326 ms (n=95, 9.268..9.935) Move 8.538 ms (n=97, 8.145..18.644) Option Compile On Loop 0.299 ± 0.551 ms (n=49711, -0.230..9.192) Exec 8.593 ms (n=61, 8.518..9.017) Move 5.227 ms (n=63, 5.014..14.190) [PERF] elapsed=21883715 us statements=1004857 findvar=2161060 (locals=16050652 [PERF] top commands by dispatch count: 396749 Let 149461 /* 140020 If 100347 Local 100037 Static 9983 Inc 7053 Cat 2066 Math 1059 Poke 429 Print 159 Pause 156 Loop 83 Const 63 WEB 57 Dim 47 Read 18 Do 8 Select Case 8 Restore 6 Option [PERF] top SUBs by exclusive (self) time: self_us incl_us calls self_us/call name [PERF] top SUBs by call count: 49712 R__timing 45595 R_spinGaitGen 378 R__bodyIK 378 R__legIK 353 ST_sendPacket 124 _update_timing 63 R__computeMove 63 R__executeMove 63 R_inputUpdate 63 R_generateGaitSeq 63 R_Angles2Pos 63 R__moveAllServos 36 ST_stopServo 18 ST_waitForStop 10 R_inputBody 10 R_inputPosition 10 R_inputWalk 4 R_resetInput 4 R_getServoState 4 R_logServoState ?MM.Info(Compile) NONE 1 REUSED 1 STMTS 2453 RAN 2265470 MISS 197226 CODE 1306415 SIZE 77268 OF 66 |
||||
| matherp Guru Joined: 11/12/2012 Location: United KingdomPosts: 12034 |
Bleep I've run diagnostics on the chess program and basically it was always broken and always overran the allocated stack in every build. It just so happens in some builds the stack extended into areas of memory that were benign. I'll have a look at the chess program and see what can be done but on a build with extended stack it ram around 3x faster. karlelch Thanks for running the experiment. The performance improvement is good and your printout gives me more areas to look at. Please report again after I post b5 later today or tomorrow. |
||||
| karlelch Guru Joined: 30/10/2014 Location: GermanyPosts: 361 |
Glad I could help. I am travelling for a week; will be able to test new versions not before the next weekend. But then will happily do so. |
||||
homa Guru Joined: 05/11/2021 Location: GermanyPosts: 715 |
That’s also what I’d like Peter to do if the PC3 is redesigned: move the slide switch from the centre of the motherboard to the front, next to the mini-USB socket I’ve been a bit out of the loop due to work and illness, but the adventure can begin again! Matthias |
||||
twofingers Guru Joined: 02/06/2014 Location: GermanyPosts: 1782 |
And I would like a jumper header in parallel with the boot button so that an external button can be connected there if needed.  Regards Michel |
||||
| matherp Guru Joined: 11/12/2012 Location: United KingdomPosts: 12034 |
Version V7.00.00b5 is now available at https://github.com/UKTailwind/PicoMite/releases/tag/PicoMite-V7.0.00b5 More compiler goodness and some bugs fixed - see the release notes. Note Prince of Persia was broken in b3 and b4 but should be working again now |
||||
| Page 3 of 6 |
|||||
 Print this page |
 |
The Back Shed's forum code is written, and hosted, in Australia. | © JAQ Software 2026 |
