Menu
▾
▴
Re: [Contiki-developers] Fwd:Fwd:Fwd:RegardingRS232supportforAtmega2560...
|
From: David K. <da...@em...> - 2014-04-18 20:41:07
|
If printf("Hello\n") does not work then stdout is not directed to the right
port. rs232_print and printf should be calling the exact same same
character output routine.
If printf works then printf_P(PSTR("Hello\n")) should also work, the PSTR
macro just forces an address in program flash memory, which printf_P uses to
fetch the string.
However printf_P("Hello") or printf(PSTR("Hello")) won't work, the _P
routines expect an address in program flash. There are a lot of those,
memcpy_P, strcmp_P, etc.
The PRINTF macro is just a convenient way to avoid typing PSTR all the time.
-----Original Message-----
From: P Gupta
Sent: Friday, April 18, 2014 4:19 PM
To: Contiki developer mailing list
Subject: Re:
[Contiki-developers]Fwd:Fwd:Fwd:RegardingRS232supportforAtmega2560...
Dear David Sir,
Thanks for those gems of inputs, by those I am very easily now able to get
the memory usage in decreasing order and am at work in optimizing those
which are memory monsters :)
Also, I am still not able to get rid of rs232_print() for getting prints;
PRINTF and printf_P(STR, ...) are really not working for me, and this is
getting really clumsy for me, cos whereever I want to see print (as in uip,
etc.) PRINTF is used and those aren't giving out print even if I define
DEBUG=1; what am I missing? Also, Arduino by default configures port 0 for
debug, so have now dedicated RS232 port 1 for XBee.
With regards,
P Gupta
On Sat, Apr 19, 2014 at 1:02 AM, David Kopf <da...@em...> wrote:
The UART needs to be within 5% of nominal, see
http://electronics.stackexchange.com/questions/27763/using-the-atmega328-with-the-internal-oscillator
The internal oscillator depends greatly on temperature and to a lesser
extent on Vcc and could vary by 10% over the allowed operating range. The
factory value for OSCCAL is typically specified as 3% at 25°C and 5 volts.
If there is an accurate crystal reference it does not take long to calibrate
to within 1% under the current operating conditions.
Timing accuracy is useful for some applications, for example turning the
radio on at precise times to detect a transmission. A 1% error is not so
great, 10 milliseconds per second, but parts per million can be had by
comparing the phase of the cpu clock to an external watch crystal and
wobbling OSCCAL around the average value, see
https://github.com/contiki-os/contiki/wiki/RDC-Phase-optimization
I have no experience with external RAM, don't know how easy it is to use but
surely adds some complication. If at all possible I'd get it going with
reduced memory by decreasing table and buffer sizes. To see what is using
RAM look at the map file or run avr-objdump on the elf file.
For example a webserver build that overflows RAM:
$ avr-size -C --mcu=atmega128rfa1 webserver6.avr-atmega128rfa1
AVR Memory Usage
----------------
Device: atmega128rfa1
Program: 66834 bytes (51.0% Full)
(.text + .data + .bootloader)
Data: 19840 bytes (121.1% Full)
(.data + .bss + .noinit)
EEPROM: 41 bytes (1.0% Full)
(.eeprom)
The css_style sheet is taking 2571 bytes of RAM because I forget to compile
it using the flash attribute:
dak@Ginger ~/contiki/examples/webserver-ipv6
$ avr-objdump -t --section=.data webserver6.avr-atmega128rfa1 |
sort -r --key=3
webserver6.avr-atmega128rfa1: file format elf32-avr
00800200 l d .data 00000000 .data
00801032 l O .data 00000a0b data_style_css
00800c56 l O .data 000003dc data_index_html
0080062f l O .data 0000030e data_files_shtml
0080095b l O .data 000002fb data_header_html
...
Some RAM could be freed by reducing the length of the packet queue, or
neighbor or route table sizes. rxframe is 389 bytes showing it allows three
buffered 802.15.4 packets, going to single buffering would save 256 bytes:
$ avr-objdump -t --section=.bss webserver6.avr-atmega128rfa1 |
sort -r --key=3
webserver6.avr-atmega128rfa1: file format elf32-avr
00801dda l d .bss 00000000 .bss
0080295a l O .bss 00000a2c packets_memb_memb_mem
00802281 l O .bss 00000690 buframmem_memb_mem
008042a2 g O .bss 00000500 uip_aligned_buf
00803391 l O .bss 00000500 sicslowpan_aligned_buf
00804b0a g O .bss 000003c0 uip_ds6_nbr_cache
00803955 l O .bss 000003ac routememb_memb_mem
00801de0 l O .bss 00000208 conns_memb_mem
00803d11 l O .bss 00000200 parent_memb_memb_mem
00804010 g O .bss 00000186 rxframe
00804975 g O .bss 0000011b uip_ds6_if
00804869 g O .bss 000000fa uip_udp_conns
00802143 l O .bss 000000b2 packetbuf_aligned
...
-----Original Message-----
From: P Gupta
Sent: Friday, April 18, 2014 10:38 AM
To: Contiki developer mailing list
Subject: Re:
[Contiki-developers]Fwd:Fwd:Fwd:RegardingRS232supportforAtmega2560...
Dear David Sir,
Thanks for the inputs, I am testing the things. I haven't really done
anything with the OSCCAL for arduino mega (like in avr-raven
calibrate_rc_osc_32k() is called in initialize). Why is it actually required
to be done, give me some insight.
One more important thing I would like to discuss. It is regarding adding
more memory to my Arduino Mega 2560. Off late, I have realized, that 8KB RAM
is a bit less when using memb memory allocations. Need few KBs more I guess.
There's a way to it, and that is by using 23K256 or 23LC1024 SRAM chip by
Microchip which is SPI-compatible bus interface and adds 256Kbit/1 Mbit of
memory. Now coming to Contiki, if I mount this chip on Arduino on SPI bus,
for all the static memory related things, the internal memory is used by the
OS and it's processes, while all the dynamic memory management done by memb
(allocation, free) gets done from the external RAM on SPI ?
With regards,
P Gupta
On Thu, Apr 17, 2014 at 11:51 PM, David Kopf <da...@em...> wrote:
The RPL mote should send a couple of DIS packets at startup and the jackdaw
in sniffer mode will report those as long as the radio channel is correct.
That lets you debug the tx routine without even implementing the rx. In
network mode the jackdaw will ignore packets with mismatching PANid,
destination address, or checksum but raw packet mode 'r' will report the
rest to wireshark.
Junk characters are usually a baud rate mismatch. Verify the CPU frequency
is what you think it is (e.g. turn on an LED, loop F_CPU times, turn off
LED). If that frequency is right then the UART should output with the baud
rate you set but it won't be exact unless using a crystal having a multiple
of the baud rate. Try 9600 and increase from there. 57600 using the
internal oscillator may require a change in OSCCAL.
The UART output is shifted to RS232 levels, yes?
-----Original Message-----
From: P Gupta
Sent: Wednesday, April 16, 2014 6:02 PM
To: Contiki developer mailing list
Subject: Re:
[Contiki-developers]Fwd:Fwd:Fwd:RegardingRS232supportforAtmega2560...
Dear David Sir,
I was also thinking on the same grounds, thanks for confirming. The program
uses mmem_* functions, hence, removing those would be difficult (and I have
already reduced the mmem by MMEM_CONF_SIZE=512). Keeping it aside, I tried
out simple-udp-rpl unicast sender with mmem commented from arduino-main and
the board didn't hang (thus it's indeed stack overflow problem). I will see
the way in which I can reduce the memory usage of the actual program.
Meanwhile, the simple-udp-rpl unicast sender is now running on the board and
as expected nothing over air (the rzusb I am having doesn't show any packets
on wireshark. Possibly cos of xbee driver (which is still untested) ?. How
to go about debugging that ? The prints coming on minicom for unicast-sender
are also unreadable, junk characters.
With regards,
P Gupta
------------------------------------------------------------------------------
Learn Graph Databases - Download FREE O'Reilly Book
"Graph Databases" is the definitive new guide to graph databases and their
applications. Written by three acclaimed leaders in the field,
this first edition is now available. Download your free book today!
http://p.sf.net/sfu/NeoTech
_______________________________________________
Contiki-developers mailing list
Con...@li...
https://lists.sourceforge.net/lists/listinfo/contiki-developers
------------------------------------------------------------------------------
Learn Graph Databases - Download FREE O'Reilly Book
"Graph Databases" is the definitive new guide to graph databases and their
applications. Written by three acclaimed leaders in the field,
this first edition is now available. Download your free book today!
http://p.sf.net/sfu/NeoTech
_______________________________________________
Contiki-developers mailing list
Con...@li...
https://lists.sourceforge.net/lists/listinfo/contiki-developers
------------------------------------------------------------------------------
Learn Graph Databases - Download FREE O'Reilly Book
"Graph Databases" is the definitive new guide to graph databases and their
applications. Written by three acclaimed leaders in the field,
this first edition is now available. Download your free book today!
http://p.sf.net/sfu/NeoTech
_______________________________________________
Contiki-developers mailing list
Con...@li...
https://lists.sourceforge.net/lists/listinfo/contiki-developers
------------------------------------------------------------------------------
Learn Graph Databases - Download FREE O'Reilly Book
"Graph Databases" is the definitive new guide to graph databases and their
applications. Written by three acclaimed leaders in the field,
this first edition is now available. Download your free book today!
http://p.sf.net/sfu/NeoTech
_______________________________________________
Contiki-developers mailing list
Con...@li...
https://lists.sourceforge.net/lists/listinfo/contiki-developers
|