Eingrenzung eines Merlin-Errors

... oder wo passierts ...

jomixl
Benutzer
Avatar
Gender:
Posts: 225
Registered: 02 / 2008
Subject:

Eingrenzung eines Merlin-Errors

 ·  Posted: 13.04.2018 - 09:02  ·  #1
Hallo zusammen,
ich musste ein altes Mega16 Programm erweitern und bin jetzt fast an der Flash-Grenze.
Darum hab ich mal den Optimizer probiert. Der schafft zwar ca. 30% Platz aber dafür läuft das Programm auch nicht mehr. Nun will ich versuchen das Problem einzugrenzen.
Das Programm besteht aus mehreren Units und ich hab keinen Plan wie ich jetzt vorgehen soll.
Hab bisher auch keinen Schalter gefunden mit dem ich ein Unit vom Optimizer sperren kann oder hab ich da was übersehen?

Danke für eure Vorschläge
Gruß Joachim
Merlin
Administrator
Avatar
Gender:
Age: 24
Posts: 1409
Registered: 03 / 2005
Subject:

Re: Eingrenzung eines Merlin-Errors

 ·  Posted: 13.04.2018 - 11:11  ·  #2
Hi Jachim.

The optimiser transcends units, so omitting units makes no sense. I have a whole range of options to narrow things down. The attached are in English, I am afraid, but describe how the problem can be narrowed. If you know roughly where the error lies (e.g. in a particular function) you can send the asm and dsm files to me straight away and I will look to see if I can see anything obvious.

There is also a new version of optimiser about to be released this weekend which resolves a few small issues. Again if you send me the asm and dsm files via pm I will run your program against the new version and see if your program is affected.

Regards

Merlin.
Attachments
Disabling parts of the optimiser

Each part of the optimiser can be separately disabled using flags. This is to help isolate issues that might be difficult to find otherwise. It should not be used for 'production code'.

Each flag is of the form

{$OPTI xxx}

To isolate the part of the optimiser causing the problem, put the following lines at the top of your main code.

{$OPTI NO_REDUNDANT_FRAME_USAGE_OPT}
{$OPTI NO_OPT_OFFSET_LOADS}
{$OPTI NO_CHECK_REDUNDANT_LOAD}
{$OPTI NO_CHECK_REDUNDANT_FRAME_LOAD}
{$OPTI NO_CHECK_REDUNDANT_SAVE}
{$OPTI NO_CHECK_MOV}
{$OPTI NO_CHECK_RETURN_REGS}
{$OPTI NO_LOOP_OPT}
{$OPTI NO_OPT_LOOP_LOADS}
{$OPTI NO_BRANCH_OPT}
{$OPTI NO_CSE_OPT}
{$OPTI No_Remove_Code_Islands}
{$OPTI No_Opt_Final_MOV}
{$OPTI NO_COM_OPT}
{$OPTI NO_OPT_TST}
{$OPTI NO_BITTEST_OPT}
{$OPTI NO_OPT_CPI}
{$OPTI NO_CALL_FRAME_OPT}
{$OPTI NO_CALL_FRAME_OPT_LOCAL}
{$OPTI NO_OPT_LOCAL_VARS}
{$OPTI NO_OPT_LOCAL_VARS2}
{$OPTI NO_OPT_LOCAL_VARS3}
{$OPTI NO_ALLOW_INLINE}
{$OPTI No_Phase_4}
{$OPTI No_Phase_3}
{$OPTI No_Phase_2_OR_3}
{$OPTI No_Phase_1_2_OR_3}

This effectively disables every part of the optimiser, so a compile and optimise now should have the same effect as just compiling.

To reintroduce an optimisation comment out the line associated with that optimisation, for example

//{$OPTI No_Phase_1_2_OR_3}

or

{ $OPTI No_Phase_1_2_OR_3}
(note the space between { and $)

Reintroduce the optimisations one by one until the problem re-appears, and advise me of the results. I recommend reintroducing from the bottom upwards.

Merlin.
Filename: Disabling Optimiser Parts.txt
Filesize: 1.57 KB
Title:
Download counter: 44
Special Debug Mode
==================

Where a problem proves very difficult to isolate, a special debug mode is available.

It will typically take about 10 tests to isolate the problem down to a small nuber of lines.

Add the line

{$OPTI SPECIAL_DEBUG}

to the source. Then compile and optimise as usual. Then test the result. Approximately half of the program will not be optimised.

Now compile and optimise again. This time you will see a new screen pop up asking whether the test failed or passed.

If it failed, the other half of the program will be tested. Typically it will pass, but of course the problem may exist in both halves.

Repeat the process, making a note of the result each time until the optimiser tells you that testing is complete. Each iteration will narrow the fault down to a smaller set of lines.

Then zip the sub-directory MERLIN_DEBUG and send to Merlin.
Filename: special debug mode.txt
Filesize: 907 Byte
Title:
Download counter: 45
jomixl
Benutzer
Avatar
Gender:
Posts: 225
Registered: 02 / 2008
Subject:

Re: Eingrenzung eines Merlin-Errors

 ·  Posted: 13.04.2018 - 11:46  ·  #3
hi merlin,
thanks for your answer! I'll try to findout where the error occures.

joachim
allan
Benutzer
Avatar
Gender: n/a
Age: 64
Posts: 11
Registered: 10 / 2016
Subject:

Re: Eingrenzung eines Merlin-Errors

 ·  Posted: 14.04.2018 - 15:55  ·  #4
Hi Merlin!

Looks like the latest optimizer version 3.5.7.3 corrupts Dataflash SPI driver or something in FlashIO.pas
Something happens with DataFlash address, data is read from wrong location.

regards,
Allan
Merlin
Administrator
Avatar
Gender:
Age: 24
Posts: 1409
Registered: 03 / 2005
Subject:

Re: Eingrenzung eines Merlin-Errors

 ·  Posted: 14.04.2018 - 16:28  ·  #5
Hello Allen.

Thanks for letting me know. Could you send me an example asm and dsm files via pm.

Do you know if earlier versions exhibited this problem?

Thanks

Merlin.
Selected quotes for multi-quoting:   0

Registered users in this topic

Currently no registered users in this section

The statistic shows who was online during the last 5 minutes. Updated every 90 seconds.
MySQL Queries: 15 · Cache Hits: 15   96   111 · Page-Gen-Time: 0.021332s · Memory Usage: 2 MB · GZIP: on · Viewport: SMXL-HiDPI