Dave,
Looking at the source code I don't see any issues.
Can you post a a little of your code ? And maybe the section of .spin (pasm) code that it generates ?
That would help a lot.
Hi
I expect its to do with the waitcnt . Maybe 80 clocks is not long enough for LMM to complete.
Its seems to print gibberish ever 50 or so seconds (nowhere near exactly that) which is about how long it takes to count 2^32 at 80Mhz.
'test of rctime on thermister 10k ohm at 25 degs c
'220 ohm resisitor from pin 24 to thermister and 0.1 uF cap in parallel-other end to 0v
DEVICE P8X32A, XTAL1, PLL16X
FREQ 80_000_000
therm PIN 24 low
pcrlf SUB 0
sertx SUB 1
print SUB 1
pstr SUB 1
strslt hub string(20)
temp var long
stmp1 var long
stmp2 var long
' ======================================================================
' this works- LMM doesn't (bet I've done something daft!!!!)
PROGRAM Start 'LMM
' ======================================================================
START:
pause 5000
pstr "starting"
pcrlf
do
high therm
pause 5
rctime therm,1,temp
print temp
pstr " "
pause 1000
loop
end
'========== SUBS ===========================
sub pstr 'send serial string
stmp1=__param1
do
rdbyte stmp1,stmp2
if stmp2=0 then exit
sertx stmp2
inc stmp1
loop
endsub
'----------------
sub print
strslt=str __param1,8
pstr strslt
endsub
'------------------------
sub sertx 'send serial byte
SEROUT 30,T9600,__param1
endsub
'---------------------------
sub pcrlf
sertx 13
sertx 10
endsub
Dave,
You are correct. 80 clocks is not quite enough time to get through the loop in LMM.
I will have to change RCTIME to measure in 2uSec increments.
As a workaround you could use in-line assembly.
as you can see the one byte string " " and "a" are problematic while the longer ones are fine.
The " " string with one space in produces &L( and the "a" produces nothing at all!
Thanks for looking into these querks- really appreciate all your hard work.
In PropBasic a single character in quotes is a byte value.
More than one character produces a pointer to the address of the string.
If you want to pass a single character to a SUB that is expecting a string you can append \000.
You would have to use:
pstr " \000" for space or pstr "a\000" for "a".
You might get away with testing the parameter to see if it is less than 128 because strings are stored at the end of the program (I think...) so as long as there is at least 128 bytes of code, it should work.
sub pstr 'send serial string
IF __param1 < 128 THEN
SERTX __param1
ELSE
stmp1=__param1
do
rdbyte stmp1,stmp2
if stmp2=0 then exit
'phex stmp2
sertx stmp2
inc stmp1
loop
ENDIF
endsub
Thanks for explaining- I have another question regarding LMM
While running in LMM mode how much of the cog is used to "run" the LMM code (which can, I assume use the whole of the hub- barring hub variables etc.) and how much is left for cog vars? I am writing some code that 'could' use a large number of instant access cog vars but 'could' instead use hub variables, but at the expense of 'looking' less obvious. If you say there is enough room for 150 cog vars left while running LMM that might do. Hope I explained that clearly!
edit/
Also, whereas I can access hub variables by using a name or index from a base variable- can this be done in the cog vars. If I define an array date_time(7) VAR LONG then each long can be indexed, but if I want to also directly acces via year, month, day, etc is there a way- I cant assign a name to each of the indexed vars.
The cog code to run LMM is pretty minimum. There should be plenty of room for 150 longs.
I don't recommend using COG arrays, it requires self-modifying-code and uses precious code space (for non-LMM code).
I never use cog arrays, so I'm not sure if this will work but you can try:
Thanks- pretty much as I thought.
Having lots of fun trying to convert some pbasic code to prop. Its all quite do-able but looks rather different. I know propbasic has some commands like rctime and PULSOUT, pause etc and others have made libraries and examples of alternatives to lookup, lookdown etc, to achieve the same ends- all is good.
Comments
I'm using the rctime function to read a thermistor (in conjunction with a 0.1uf capacitor).
It works great in cog mode but not in LMM mode.
Dave
I will look into it and let you know what I find.
Bean
Looking at the source code I don't see any issues.
Can you post a a little of your code ? And maybe the section of .spin (pasm) code that it generates ?
That would help a lot.
Bean
I expect its to do with the waitcnt . Maybe 80 clocks is not long enough for LMM to complete.
Its seems to print gibberish ever 50 or so seconds (nowhere near exactly that) which is about how long it takes to count 2^32 at 80Mhz.
the relevant spin bits....
cog version:-
mov __temp2,#5 'pause 5 mov __temp1,_1mSec add __temp1,cnt __L0002 waitcnt __temp1,_1mSec djnz __temp2,#__L0002 mov __temp1,therm 'rctime therm,1,temp andn dira,__temp1 mov temp,#0 mov __temp2,cnt adds __temp2,#80 __L0003 adds temp,#1 waitcnt __temp2,#80 and __temp1,ina NR, WZ IF_NZ jmp #__L0003 mov __param1,temp 'print temp call #printThis is LMM version:-
mov __temp2,#5 'pause 5 mov __temp1,_1mSec add __temp1,cnt __L0002 waitcnt __temp1,_1mSec djnz __temp2,#_LMM_JUMP long @@@__L0002 mov __temp1,therm 'rctime therm,1,temp andn dira,__temp1 mov temp,#0 mov __temp2,cnt adds __temp2,#80 __L0003 adds temp,#1 waitcnt __temp2,#80 and __temp1,ina NR, WZ IF_NZ sub __PC,#(($-__L0003)*4)+4 mov __param1,temp 'print tempHope that helps
Dave
You are correct. 80 clocks is not quite enough time to get through the loop in LMM.
I will have to change RCTIME to measure in 2uSec increments.
As a workaround you could use in-line assembly.
ASM mov __temp1,therm 'rctime therm,1,temp andn dira,__temp1 mov temp,#0 mov __temp2,cnt adds __temp2,#160 __RCTIMEFIX adds temp,#1 waitcnt __temp2,#160 and __temp1,ina NR, WZ IF_NZ sub __PC,#(($-__RCTIMEFIX)*4)+4 ENDASMIf you try it, let me know if it works.
Remember the value will be in 2uSec units instead of 1uSec units.
Bean
Yes that works- BUT- got another
it seems in LMM one byte strings are corrupt- I have an example
produces...
as you can see the one byte string " " and "a" are problematic while the longer ones are fine.
The " " string with one space in produces &L( and the "a" produces nothing at all!
Thanks for looking into these querks- really appreciate all your hard work.
Dave
More than one character produces a pointer to the address of the string.
If you want to pass a single character to a SUB that is expecting a string you can append \000.
You would have to use:
pstr " \000" for space or pstr "a\000" for "a".
You might get away with testing the parameter to see if it is less than 128 because strings are stored at the end of the program (I think...) so as long as there is at least 128 bytes of code, it should work.
sub pstr 'send serial string IF __param1 < 128 THEN SERTX __param1 ELSE stmp1=__param1 do rdbyte stmp1,stmp2 if stmp2=0 then exit 'phex stmp2 sertx stmp2 inc stmp1 loop ENDIF endsubBean
Thanks for explaining- I have another question regarding LMM
While running in LMM mode how much of the cog is used to "run" the LMM code (which can, I assume use the whole of the hub- barring hub variables etc.) and how much is left for cog vars? I am writing some code that 'could' use a large number of instant access cog vars but 'could' instead use hub variables, but at the expense of 'looking' less obvious. If you say there is enough room for 150 cog vars left while running LMM that might do. Hope I explained that clearly!
edit/
Also, whereas I can access hub variables by using a name or index from a base variable- can this be done in the cog vars. If I define an array date_time(7) VAR LONG then each long can be indexed, but if I want to also directly acces via year, month, day, etc is there a way- I cant assign a name to each of the indexed vars.
Dave
I don't recommend using COG arrays, it requires self-modifying-code and uses precious code space (for non-LMM code).
I never use cog arrays, so I'm not sure if this will work but you can try:
name VAR LONG MyArray(2)
to create an alias.
Bean
Thanks- pretty much as I thought.
Having lots of fun trying to convert some pbasic code to prop. Its all quite do-able but looks rather different. I know propbasic has some commands like rctime and PULSOUT, pause etc and others have made libraries and examples of alternatives to lookup, lookdown etc, to achieve the same ends- all is good.
Dave
Is there a way to restart the basic program- resetting stack etc as in the 'RUN' command of old.
Will entering 'goto start' achieve this?
Dave
In LMM you should be able to reset the call stack by doing:
Bean