Recent files: https://freaks-unidos.net/cgi-bin/svnwiki/emqbit/

how-to-query-a-modem

2008-12-02T15:23:28Z

This program is a small example useful to query a modem. It might be a starting point for serial programming.

#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <errno.h>
#include <fcntl.h>
#include <sys/types.h>
#include <sys/times.h>
#include <sys/ioctl.h>
#include <sys/unistd.h>
#include <getopt.h>
#include <termios.h>

/*  This code started as a streamlined version of
 *  a C++ file that is part of a project which is
 *  is Copyright (c) 2005 Dave Hylands.
 *  http://www.davehylands.com/Software/BootHost/
 *
 *  We might need to restore more options later
 *  to allow 96008N1 and so on.
 */

#define SERIAL_BAD_BAUDRATE -1000
#define SERIAL_BAD_PORT -1001
#define SERIAL_GEN_ERROR -1002

static struct {
    speed_t speed;
    unsigned baudRate;
} BaudTable[] = {
    {
    B50, 50}, {
    B75, 75}, {
    B110, 110}, {
    B134, 134}, {
    B150, 150}, {
    B200, 200}, {
    B300, 300}, {
    B600, 600}, {
    B1200, 1200}, {
    B1800, 1800}, {
    B2400, 2400}, {
    B4800, 4800}, {
    B9600, 9600}, {
    B19200, 19200}, {
    B38400, 38400}, {
    B57600, 57600}, {
    B115200, 115200}, {
    B230400, 230400}
};

#define ARRAY_LEN(x)    ( sizeof( x ) / sizeof( x[ 0 ]))

int serial_open(char *portStr, int try_baud_rate, int verbose)
{
    speed_t baudRate = B0;
    int gPortFd = -1;
    int i;

    struct termios attr;

    for (i = 0; i < ARRAY_LEN(BaudTable); i++) {
       if (BaudTable[i].baudRate == try_baud_rate) {
            baudRate = BaudTable[i].speed;
            break;
        }
    }

    if (baudRate == B0) {
        if (verbose)
          fprintf(stderr, "Unrecognized baud rate: '%d'\n", try_baud_rate);
        return SERIAL_BAD_BAUDRATE;
    }

    if ((gPortFd = open(portStr, O_RDWR | O_EXCL)) < 0) {
        if (verbose)
          fprintf(stderr, "Unable to open serial port '%s': %s\n", portStr,
                strerror(errno));
        return SERIAL_BAD_PORT;
    }

    if (tcgetattr(gPortFd, &attr) < 0) {
        if (verbose)
          fprintf(stderr, "Call to tcgetattr failed: %s\n", strerror(errno));
        return SERIAL_GEN_ERROR;
    }

    attr.c_iflag = 0;
    attr.c_oflag = 0;
    attr.c_cflag = CLOCAL | CREAD | CS8;
    attr.c_lflag = 0;
    attr.c_cc[VTIME] = 0;       /* timeout in tenths of a second */
    attr.c_cc[VMIN] = 1;        /* Only wait for a single char */

    cfsetispeed(&attr, baudRate);
    cfsetospeed(&attr, baudRate);

    if (tcsetattr(gPortFd, TCSAFLUSH, &attr) < 0) {
        if (verbose)
          fprintf(stderr, "Call to tcsetattr failed: %s\n", strerror(errno));
        return SERIAL_GEN_ERROR;
    }

    return gPortFd;
}

/*
 * Tne next two function were written by Nelson Castillo.
 * You should not blame Dave for what he wrote here :-)
 */

int
serial_read (int fd, char *buf, int len, int ms_timeout)
{
  fd_set rfds;
  struct timeval tv;
  int retval;
  int nread = 0;

  FD_ZERO(&rfds);
  FD_SET(fd, &rfds);

  tv.tv_sec = ms_timeout / 1000;
  tv.tv_usec = (ms_timeout % 1000) * 1000;

select_next:

  /* We rely on the value of tv after the select call.
   * This is a Linux-only thing.
   * TODO: Fix. */

  errno = 0;
  retval = select(fd + 1, &rfds, NULL, NULL, &tv);

  if (retval == -1)
  {
    if (errno == EINTR)
      goto select_next;

    perror("select");
    exit(1);
  }
  else if (retval)
  {
    if (len - nread > 0)
    {
      int r;
      r = read(fd, buf + nread, len - nread);
      if (r > 0)
      {
        nread += r;

        if (len - nread >  0)
          goto select_next;
      }
      else
      {
        fprintf(stderr, __FILE__" * read returned %d\n\n", r);
        perror("read");
      }
    }
  }

  return nread;
}

int
main(int argc, char **argv)
{
 int r;
 int len;
 char buf[256];

 r = serial_open("/dev/modem", 57600, 1);
 if (r < 0)
   return 1;

 if (4 != write(r, "ATZ\r", 4)) {
   perror("write");
   return 1;
 }

 len = serial_read (r, buf, 255, 500 /* milliseconds */);
 printf("We got:");
 fwrite(buf, 1, len, stdout);

 return 0;
}

ecbot

2007-11-03T00:20:36Z

Edit Link Top

ECBOT

A GNU/Linux powered robot compatibe with the ECB_AT91 family of boards.

The ECBOT is being developed by Carlos Camargo at the National University of Colombia.

Features:

180 MHz ARM9 processor (Atmel AT91RM9200)
2 MBytes of serial flash
32MBytes of SDRAM
1 SD/MMC slot
USB 2.0 host
FPGA for image capture (image processing capable)
640x480 CMOS image sensor
6 IR proximity sensors
6 RGB leds
2 DC motor driver based in AVR microcontroller
1 AVR microcontroller for analog inputs
JTAG support

ECBOT Block diagram

at91-gpio

2007-08-22T13:42:30Z

You can use the GPIOs from user-space or from kernel space. To use them from user-space you need to use memory mapped IO.

We have some examples in our repository.

Let us know if you improve them.

tspc

2007-08-11T15:03:37Z

Edit Link Top

Quick overview

tspc is a client that will allow you to use an IPv6 Broker. By default, it uses a Freenet6 broker.

With it, you can have IPv6 in your computer in the fly. This is very useful if you're behind a NAT and you want to be able to SSH to the natted computer.

Edit Link Top

More information

tspc is a lightweight client that utilizes the Tunnel Setup Protocol (TSP) and is available on Windows 2000, NT, XP, Linux, FreeBSD, OpenBSD, NetBSD, Solaris, OS X, VxWorks and QNX. Its key features include:

IPv6 in IPv4 tunneling
IPv6 permanent (or temporary) addresses
NAT traversal

All you need is a valid IPv4 connection with access to the Internet. It does not matter if you're behind a NAT and you're using a Intranet address (such as 10.X.X.X, 172.16.X.X or 192.168.X.X), as long as you can reach the Internet with a gateway.

Edit Link Top

Use

Edit Link Top

Install

aptitude install tspc

This will install the client in your computer. When it starts, it will give you your own IPv6.

To check if you got the IPv6 address, run ifconfig, and you should see a tunnel, that allows you to connect to the IPv6 network.

tun       Link encap:UNSPEC  HWaddr 00-00-00-00-00-00-00-00-00-00-00-00-00-00-00-00  
   inet6 addr: 2001:5c0:8fff:fffe::4a21/128 Scope:Global
   UP POINTOPOINT RUNNING NOARP MULTICAST  MTU:1280  Metric:1
   RX packets:9 errors:0 dropped:0 overruns:0 frame:0
   TX packets:3 errors:0 dropped:0 overruns:0 carrier:0
   collisions:0 txqueuelen:500 
   RX bytes:656 (656.0 b)  TX bytes:240 (240.0 b)

Now you can be reached from the Internet using your IP, that in this example is 2001:5c0:8fff:fffe::4a21.

We will now provide some examples of what you can do in the IPv6 world. Remember that we will use an example IP and that you have to look up your own with ifconfig.

Also note that if you're used to be behind a firewall, you will not be protected in IPv6 unless you add new rules to your firewall. All the ports will be reachable.

Edit Link Top

ping

Ping is useful to determine if you can already reach some IPv6 host.

$ ping6 www.kame.net
PING www.kame.net(orange.kame.net) 56 data bytes
64 bytes from orange.kame.net: icmp_seq=2 ttl=50 time=414 ms
64 bytes from orange.kame.net: icmp_seq=3 ttl=50 time=378 ms

Edit Link Top

browse

With IPv6, you could see the improved version of the dancing kame at www.kame.net.

Edit Link Top

SSH

Someone can SSH into your box, using:

 ssh 2001:5c0:8fff:fffe::4a21

Edit Link Top

HTTP

Someone can see your wep page using your IP. Like this:

http://[2001:5c0:8fff:fffe::4a21]/

Remember you have to run a web server, such as Apache.

Edit Link Top

Restart the service

If you lose your IP, you can restart the tunnel.

/etc/init.d/tspc restart

Edit Link Top

Authenticated tunnels

By default you get an anonymous section. In order to keep it, you have to create an account in the IPv6 broker and then edit the file /etc/tsp/tspc.conf.

Edit Link Top

Conclusion

With little effort you can try IPv6. Let us know how it goes for you.

OpenEmbedded from scratch

2007-08-08T04:44:05Z

Note: Keeping an OE document updated is hard. Please refer to the OE website.

http://www.openembedded.org

Hardware watchdog for the AT91RM9200

2007-07-30T23:29:41Z

Andres just added a new document that explains how to enable the hardware watchdog of the at91rm9200. Check it out.

This can be useful in a critical environment.

Loading... Vote up Vote down Discussion

at91rm9200-watchdog

2007-07-30T23:16:12Z

In this page we show how to enable the watchdog timer of the at91rm200. It will allow you to do an automated system reset if a program fails to ping the watchdog.

Edit Link Top

Setup the kernel

Enable required options in your kernel .confg:

 CONFIG_WATCHDOG=y
 CONFIG_WATCHDOG_NOWAYOUT=y
 CONFIG_AT91RM9200_WATCHDOG=y

Boot the new kernel, and check for watchdog support when it boots.

 > dmesg |grep Watch
 AT91 Watchdog Timer enabled (5 seconds, nowayout)

Edit Link Top

Verify that the device file exists

This file has a major number 10, minor number 130.

 crw-rw---- 1 root root 10, 130 Dec 31 19:00 /dev/watchdog

If it doesn't exist, create it with mknod.

 > mknod  /dev/watchdog c 10 130

Edit Link Top

Build an example

Here we will show the example watchdog-test.c that lives in the Linux kernel documentation linux-2.6.xx/Documentation/watchdog/src.

/*
 * Watchdog Driver Test Program
 */

#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <unistd.h>
#include <fcntl.h>
#include <sys/ioctl.h>
#include <linux/types.h>
#include <linux/watchdog.h>

int fd;

/*
 * This function simply sends an IOCTL to the driver, which in turn ticks
 * the PC Watchdog card to reset its internal timer so it doesn't trigger
 * a computer reset.
 */

void keep_alive(void)
{
    int dummy;

    ioctl(fd, WDIOC_KEEPALIVE, &dummy);
}

/*
 * The main program.  Run the program with "-d" to disable the card,
 * or "-e" to enable the card.
 */
int main(int argc, char *argv[])
{
    fd = open("/dev/watchdog", O_WRONLY);

    if (fd == -1) {
	fprintf(stderr, "Watchdog device not enabled.\n");
	fflush(stderr);
	exit(-1);
    }

    if (argc > 1) {
	if (!strncasecmp(argv[1], "-d", 2)) {
	    ioctl(fd, WDIOC_SETOPTIONS, WDIOS_DISABLECARD);
	    fprintf(stderr, "Watchdog card disabled.\n");
	    fflush(stderr);
	    exit(0);
	} else if (!strncasecmp(argv[1], "-e", 2)) {
	    ioctl(fd, WDIOC_SETOPTIONS, WDIOS_ENABLECARD);
	    fprintf(stderr, "Watchdog card enabled.\n");
	    fflush(stderr);
	    exit(0);
	} else {
	    fprintf(stderr, "-d to disable, -e to enable.\n");
	    fprintf(stderr, "run by itself to tick the card.\n");
	    fflush(stderr);
	    exit(0);
	}
    } else {
	fprintf(stderr, "Watchdog Ticking Away!\n");
	fflush(stderr);
    }

    while (1) {
	keep_alive();
	sleep(1);
    }
}

Now compile it:

arm-softfloat-linux-gnu-gcc  watchdog-test.c -o watchdog-test

Edit Link Top

Test the watchdog

 > ./watchdog-test -e
 Watchdog card enabled.

 > ./watchdog-test
 Watchdog Ticking Away!

Interrupt the process (with Ctrl+C) and wait five seconds. The at91rm9200 watchdog should reset the board.

Edit Link Top

References

Linux kernel documentation linux-2.6.xx/Documentation/watchdog/src.

First post

2007-06-28T05:26:27Z

This is the fist post of the emQbit's weblog. We will use this weblog to keep track of what we do.

We guess this weblog will be developer-oriented, but only time will tell.

The software behind this page is Svnwiki. Nelson wrote some reasons to use Svnwiki for weblogs one year ago. Since then the software has improved a lot.

Fasten your seatbelts :)

Loading... Vote up Vote down Discussion

emqbot

2007-06-28T02:36:26Z

Edit Link Top

emQbot

Embedded robotics research
Reconfigurable computing

Edit Link Top

Details

darrell-loader

2007-06-12T21:15:00Z

Read flashing-the-ecb-at91 to know how to upload this loader to the board.

Edit Link Top

AT91RM9200 Loader

Warning: This loader is working with gcc 3.4.5. We tried gcc 4.1 and it didn't work. We have to debug this issue. If you have problems, email us and we'll send you a binary.

We maintain a loader for the AT91RM9200 that could be useful for other projects. It was started by Darrell Harmon. If you find it useful we could host it in sourceforge or in code.google.com. It requires just 10KB.

Darrell's loader - Thanks to the u-boot project
Version 1.0
RAM:32MB

1: Upload Darrell's loader to Dataflash
2: Upload u-boot to Dataflash
3: Upload Kernel to Dataflash
4: Start u-boot
5: Upload Filesystem image
6: Memory test

emQbit's weblog

2007-06-03T20:49:00Z

Hardware watchdog for the AT91RM9200

Mon Jul 30 23:29:41 2007

Andres just added a new document that explains how to enable the hardware watchdog of the at91rm9200. Check it out.

This can be useful in a critical environment.

Loading... Vote up Vote down Discussion

First post

Sun Jun 3 20:52:01 2007

This is the fist post of the emQbit's weblog. We will use this weblog to keep track of what we do.

We guess this weblog will be developer-oriented, but only time will tell.

The software behind this page is Svnwiki. Nelson wrote some reasons to use Svnwiki for weblogs one year ago. Since then the software has improved a lot.

Fasten your seatbelts :)

Loading... Vote up Vote down Discussion

Loading... Vote up Vote down

log/menu

2007-06-03T20:46:48Z

You are creating the file xsvnwiki-atom/xsvnwiki-dir. You can go to this weblog's main page

Edit Link Top

About

This is the emQbit's weblog. We're located in Bogota, Colombia.

Edit Link Top

Contact

[Archive]

Loading... Vote up Vote down Discussion

linux-kernel

2007-06-03T20:00:29Z

We will put some kernel related links here. In the NEAR future, wi will start putting some parches here. But the best thing to do is to send them upstream as soon as we can, I guess.

Edit Link Top

Where are the patches?

Russell King puts his patches here:

ftp://ftp.arm.linux.org.uk/pub/armlinux/kernel/git-cur/ Patches: http://www.arm.linux.org.uk/developer/patches/

And I guess they're agains the linux GIT tree. You can get it with:

git clone git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux-2.6.git

We migth use our own git soon.

ccache-distcc

2007-05-30T18:24:15Z

The following instructions are for Debian and they might work for Ubuntu as well.

Edit Link Top

Motivation

If you don't use ccache you'll waste time.
If you don't use distcc you'll waste time.
If you don't use ccache and distcc together, you'll waste time.

Edit Link Top

Some theory

You might want to use ccache and distcc all the time, even if you compile only from a single machine. This way, you can add nodes to the cluster anytime and the cache will be valid when you compile with a single processor some other time.

This is especially useful if you have a laptop.

Edit Link Top

Installing

In the front-end machine:

 aptitude install ccache distcc gcc g++

In the nodes:

 aptitude install distgcc gcc g++

Edit Link Top

Configuring ccache

The first thing to set is the maximum cache size. The more, the better. For me a few gigabytes are more than enough. If need to cache a few kernel compilations, it will be enough.

$ ccache --max-size 5G

You can check statistics anytime you like with ccache -s.

$ ccache -s
cache directory                     /home/galactus/.ccache
cache hit                              0
cache miss                             0
files in cache                     43454
cache size                         265.9 Mbytes
max cache size                       3.0 Gbytes

Edit Link Top

Configuring distcc

In the compilation nodes, make sure that your computer is allowed to use distcc. To configure it use:

dpkg-reconfigure distcc

Edit Link Top

The environment

In the front-end machine, you might want to use something like the following environment:

# When You're alone
# export DISTCC_HOSTS=localhost

# When you use the cluster
# Faster machines first.
export DISTCC_HOSTS="192.168.1.2 192.168.1.3 192.168.1.4 localhost"
export CCACHE_PREFIX="distcc"
#If you don't want to cache the compilation results
#export  CCACHE_DISABLE.

Edit Link Top

An example

This is an example of what you would need to do to use ccache with the Linux kernel.

--- Makefile.old        2007-05-30 10:07:00.000000000 -0500
+++ Makefile    2007-05-30 10:07:49.000000000 -0500
@@ -195,8 +195,8 @@
          else if [ -x /bin/bash ]; then echo /bin/bash; \
          else echo sh; fi ; fi)
 
-HOSTCC       = gcc
-HOSTCXX      = g++
+HOSTCC       = ccache gcc
+HOSTCXX      = ccache g++
 HOSTCFLAGS   = -Wall -Wstrict-prototypes -O2 -fomit-frame-pointer
 HOSTCXXFLAGS = -O2
 
@@ -281,7 +281,7 @@
 
 AS             = $(CROSS_COMPILE)as
 LD             = $(CROSS_COMPILE)ld
-CC             = $(CROSS_COMPILE)gcc
+CC             = ccache $(CROSS_COMPILE)gcc
 CPP            = $(CC) -E
 AR             = $(CROSS_COMPILE)ar
 NM             = $(CROSS_COMPILE)nm

In this example, we used cross-compiler.

In order to use all the nodes in the cluster, you need to call make with the "-j" flag. For instance:

crossmake -j5 vmlinux

Flashing the ECB_AT91

2007-05-12T00:54:46Z

To flash the board, you need to use the serial port.

The software you will need:

Warning: The darrell loader is working with gcc 3.4.5. We tried gcc 4.1 and it didn't work. We have to debug this issue. If you have problems, email us and we'll send you a binary.

When you compile the Darrell loader, you get a binary named loader.bin, the one you will upload.

The tricky part is the Darrel's loader. It's located in the ROM in the AT91RM9200 CPU. To have it loaded you need to perform the following steps:

Load the Darrel's loader into RAM
- Power off the board
- Short the J2 jumper
- Start a XMODEM transfer. You might want to use minicom for this task.
- Power on the board before the transfer times out.
- Wait until the u-boot image loads

When it gets loaded, it prompts the menu endlessly, you can stop that by pressing a key (ENTER is a good choice).

Remove the J2 jumper without restarting the board.

This allows the loader to store the Loader in the AT91's ROM.

Using the menu of the Loader that is running, select the first option (1)

1: Upload loader to Dataflash with vector 6 modification.
2: Upload u-boot to Dataflash.
3: Upload Linux to Dataflash
4: Start U-boot
5: Upload Filesystem image
6: Memory test

This will ask for a XMODEM transfer, the same one that was done in a previous step. In minicom, you can use Ctrl + a + s to repeat the transfer.

Once the transfer finishes, the Darrel's loader will be stored in the AT91RM9200 ROM (TODO: Is it OK to call this memory a ROM?). So it will be there when you reboot, and you'll get a nifty menu that will allow you to:

There is another method that doesn't require a Jumper. It's not easy to describe. Perhaps we should automate it using expect, as we did a long time ago for the gumstix.

Our first reflow oven (Obi-Wan)

2007-03-26T04:11:46Z

Our first reflow oven (Obi-Wan)

For the control part, we're using an Atmega16 AVR and a PC. The AVR is just an slave controlled using a serial port. We wanted to use it for the Analog-Digital conversion. In the final step, we will generate the controller.

This is our victim:

Edit Link Top

Calibration the analog-digital converter

We tested the AD-C, using ADC0 (pin 40).

The AVR manual says that the conversion formula is: ADC = (I*1024)/2.56. Our results are close, but not exact.

Volts(I)	10-bit A-D conversion (ADC)	Theory
4.43	1023 (overflow)
2.54	986	1016
1.90	739	760
1.27	491	508
0.63	243	252

The linear function is: I = M*ADC + B.
- M = 0.0025706940874036 (slope)
- B = 0.00532974427994848

Let's test it (Scheme language) :

>(define (v x) (+ B (* M x)))
> (v 243)
0.630000000000002
> (v 491)
1.26752355316285
> (v 739)
1.90504710632571
> (v 739)
1.90504710632571
> (v 986)
2.54

Ok. It works.

Now let's relate these values with the actual temperature. We calibrated the thermocouple and the linear regression (line) fits the measured data (dashed line).

The linear function is: T = M1*V + B1.
- M1 = 99.4219653179191 (slope)
- B1 = -0.66473988439313

> (define (t v) (+ (* v 99.4219653179191) -0.66473988439313))
> (t 2.31)
229.0
> (t 0.58)
57.0

It works. So, we measure the temperature using this formula: T = M1 * (M * ADC + B) + B1. That is:

T = 99.4219653179191 * (0.0025706940874036 * ADC + 0.00532974427994848) + -0.66473988439313

So, the temperature this room is now (We measured 82 with the ADC).

>(define (real adc) (+ -0.66473988439313 (* 99.4219653179191 (+ 0.00532974427994848 (* adc 0.0025706940874036)))))
> (real 82)
20.8229973554286

Nelson's watch is reporting 20.05 �C, so this is quite close. We are done with the calibration.

Edit Link Top

Controling the oven

Let's see how the oven performs heating the air inside it.

We turned it off at time 6.74, and opened the door at time 14.38.

Seems a little slow.

Edit Link Top

programador-paralelo-avr

2007-03-24T06:50:03Z

Edit Link Top

Cable paralelo

Naranja = 1 Verde = 2 Azul = 11 Blanco/Azul = 16 (reset) Caf� = 20, 21 (GND)

Extra cables:

Blanco verde = 3 Blanco cafe = 4 Blanco naranja = 5

Install:

avr-libc gcc-avr uisp

ecb-at91-v3

2007-03-19T00:52:55Z

Edit Link Top

SBC ECB_AT91 V3

Edit Link Top

Features of the Single Board Computer

200MHz ARM926EJ-S-based processor (Atmel AT91SAM9263)
- DSP Instruction Extensions
- Jazelle Technology for Java Acceleration
LCD Controller
- Supports Passive or Active Displays
- Up to 24 bits per Pixel in TFT Mode, Up to 16 bits per Pixel in STN Color Mode
- Up to 16M Colors in TFT Mode, Resolution Up to 2048x2048, Supports Virtual Scren Buffers
2D Graphics Accelerator
- Line Draw, Block Transfer, Polygon Fill, Clipping, Commands Queuing
UCB1400 Audio and Touch Screen Controller
- Integrated AC'97 Rev. 2.1 interface
- 20-bit stereo audio codec supporting input/output gain control
- 4-wire resistive touch screen interface circuit supporting position, pressure and plate resistance measurements
128 MBytes, SanDisk mDOC H3
8 MBytes of serial flash (boot room)
Up to 128MBytes of SDRAM (Supports 32M/64M)
2 MicroSD slots
2 USB 2.0 host
1 10/100 Ethernet interface
2 serial (RS232) interfaces
JTAG support

linux-cl-url-temporal

2007-03-16T04:37:47Z

Nelson Castillo es Ingeniero de Sistemas, Colombiano, con 8 a�os de experiencia en administraci�n de servidores con GNU/Linux. Actualmente se encuentra trabajando tiempo completo en emQbit, una empresa que se dedica al desarrollo de sistemas embebidos. Hace 10 a�os, con dos amigos, inici� el proyecto Dfd, un int�rprete de diagramas de flujo que es usado en ense�anza de algoritmos b�sicos.

�C�mo y por qu� llegaste a utilizar Debian?

Primero fue por curiosidad, y luego por que me pareci� una distribuci�n de mejor calidad.

Fue en el a�o 2000 � en el 2001, no recuerdo bien la fecha. Por esa �poca era estudiante y estaba trabajando como administrador de los servidores de Correo y Proxy. El primer servidor de correo de la Universidad del Magdalena lo instalamos con Debian Potato. El servidor funcionaba sin problemas y no requer�a mucha atenci�n.

El servidor Proxy estaba instalado con RedHat, que fue la primera distribuci�n que us�, por algo m�s de un a�o. Un d�a la navegaci�n en la Intranet se ca�a con frecuencia, y yo no pod�a encontrar la causa. Por accidente, en un momento logramos notar que cuando alguien hac�a una transferencia usando el protocolo FTP hacia un servidor que estaba fuera de la Intranet, el kernel se ocupaba al 100%. Me sorprendi� que un error tan grave estuviera presente en el sistema, y por lo que logr� averiguar era un error con la versi�n del kernel instalada. Al d�a siguiente ya el servidor no era RedHat, lo cambiamos por un Debian. Desde ese d�a Debian es mi distribuci�n preferida para servidores.

RedHat me parec�a una distribuci�n buena para los est�ndares de la �poca, pero Debian me pareci� de mejor calidad. Sobre todo, por el manejo de los paquetes y las dependencias entre paquetes. Un problema de este estilo es menos probable en Debian, ya que el kernel es uno de los paquetes m�s probados, y tambi�n por la din�mica de desarrollo, que hace que cuando uno descargue una versi�n estable, realmente sea estable.

Las versiones estables de Debian se liberan cuando est�n listas cuando el n�mero de errores cr�ticos se acerque a cero. Esto se puede ver siguiendo la l�nea verde en esta gr�fica. En el 2005 se ve que esta l�nea estaba cerca de 0, cuando Debian Sarge estuvo listo.

- �Ha cumplido Debian con tus expectativas?

S�.

El sistema de paquetes que desarrollaron me gusta mucho. En mi m�quina uso Debian inestable. De vez en cuando alguna funcionalidad se da�a, pero afortunadamente hasta ahora he podido mantener la misma instalaci�n, que ya tiene 2 a�os. Para muchos usos supongo que lo recomendable es usar una versi�n estable de Debian, pero es importante anotar que para muchos usuarios experimentados de Debian esto no es una restricci�n y usan Debian testing, o Debian unstable.

La parte que m�s desanima a los usuarios nuevos es lo que se demoran en salir las nuevas versiones estables, y este si es un problema de Debian.

Supongo que lo ideal es que hiciera carrera para desarrollador de Debian para ayudar con algunos paquetes, pero actualmente me quedar�a dif�cil sacar el tiempo. Tengo una tarea pendiente desde hace m�s de un a�o : crear paquetes de Debian para Svnwiki.

�Qu� otro S.O. utilizas normalmente?

Usualmente trabajo con GNU/Linux. Me gustar�a probar FreeBSD y OpenBSD.

- Sabemos que trabajas con procesadores ARM en proyectos embebidos, �por qu� utilizan Debian?

Nos gusta usar Debian por que tiene muchos paquetes listos para ser usados, y por la calidad de los mismos que es buena. Tambi�n usamos OpenEmbedded y Buildroot. Creo que estas distribuciones no compiten estre si. Son muy diferentes.

- �Existe otra distribuci�n que entregue tantos paquetes para esa arquitectura?

No estoy seguro, aunque creo que no hay otra. Una ventaja que tiene Debian es que crear paquetes para ARM muchas veces no implica un trabajo adicional para los desarrolladores.

- Sabemos que crearon una placa basada en ARM de tama�o peque�o, y adem�s, con hardware libre eso es algo extra�o ac� en Sud am�rica, �ha sido un trabajo muy complejo?

Supongo que te refieres a la Free ECB_AT91, ha sido un proyecto muy interesante para nosotros. En sur-Am�rica ya se hab�a hecho algo, como el proyecto de Flavio Ribeiro.

Usar Linux nos ha ayudado bastante a entrar en el mundo de los Embebidos. Si ya uno est� familiarizado con el n�cleo, y con herramientas de desarrollo libres, el paso a Linux Embebido no es traum�tico.

En esta pregunta me qued� corto as� que le pregunt� a Carlos Camargo, qui�n trabaj� el hardware de este desarrollo. Esta es su respuesta:

En algunos pa�ses de sur-Am�rica no estamos acostumbrados a desarrollar nuestras propias soluciones, y siempre adquirimos tecnolog�a, por esta raz�n, la apropiaci�n de tecnolog�a es un poco traum�tica, adem�s porque no existen puntos de referencias locales que ayuden en el proceso. Hablando espec�ficamente de nuestra board, el �nico problema que existi� fue en la construcci�n misma de la board. La primera vez que armamos nuestro prototipo, no funcion�, cambiamos muchos valores del dise�o, medimos en muchos puntos de la board, y al final mandamos a construir unas nuevas boards, pero cuando llegaron estas sucedi� exactamente lo mismo, �no funcionaba!. Comenzamos a hacer una serie de pruebas para determinar hasta que punto el dise�o funcionaba y despu�s de mucho probar nos dimos cuenta que el problema era de soldadura, es decir, la t�cnica de soldar que utilizamos no era la adecuada. Si en Colombia existiera experiencia previa en este tipo de tecnolog�a muy seguramente nos hubi�ramos ahorrado ese tiempo.
Una vez el prototipo alcanz� el grado de funcionalidad m�nimo, fue relativamente f�cil implementar todas las herramientas necesarias para trabajar con Linux Embebido, las listas de discusi�n son una herramienta poderosa, ya que muy seguramente alguien ha tenido el mismo problema que uno antes.

Ahora yo a�adir�a otra cosa. Una parte que nos ha costado bastantes horas de trabajo es la documentaci�n que hemos puesto en l�nea. As� que cualquier ayuda que sirva para mejorar los documentos es bienvenida. Tenemos una lista para este proyecto en Espa�ol, y otra en Ingl�s.

- �A qu� mercado est� dirigida esa placa?

Al mundo industrial y acad�mico, este Single-board computer es una herramienta de trabajo muy poderosa que puede y est� siendo utilizada por varias universidades de Colombia, Venezuela, Per�, B�lgica y Espa�a y por varias empresas en Colombia y una en Italia.

- �Alg�n proyecto que est� pronto a ver la luz?

Ya est� listo el ECB_AT91 V2, y nos falta poner m�s documentaci�n en l�nea para el uso del FPGA, del PsOC y de los otros juguetes que esta board trae.

Tenemos dos desarrollos pendientes, que creo ser�n los pr�ximos. Uno es un nuevo Single-board computer, con m�s caracter�sticas. Y la c�mara de red, que va a ser una c�mara con bastante poder de c�mputo. Esta c�mara tendr� un FPGA y un PowerPC.

- �Algo que quieras decirle a la comunidad Debian en Chile? (y tambi�n a los que no lo utilizan)

Huy, que responsabilidad :)

Primero, por supuesto una invitaci�n a probar Debian GNU/Linux, y segundo, que no se puede olvidar que Ubuntu ha logrado ser bueno gracias al trabajo de la comunidad de desarrollo de Debian.

Hay una comunidad de usuarios y desarrolladores de Debian en Colombia, que es muy amigable. Supongo que las puertas est�n abiertas para colaboraci�n en el futuro. En cuanto a Linux Embebido, los invitamos a participar en el grupo Linux en-Caja, en el que hemos logrado crear documentaci�n �til.

Y para terminar, muchas gracias por tenernos en cuenta para publicar estas opiniones en el espacio que est�n creando, que espero que les sirva para coordinar muchas actividades en el futuro pr�ximo.

eabi-oabi-benchmark

2007-03-15T19:59:49Z

Edit Link Top

Analysis: Why EABI matters

Published in LinuxDevices. (Mar. 14, 2007)

Edit Link Top

Foreword

It's usual to hear of the new ARM EABI nowadays. There are many motivations to start using it but there is one we specially like: it's much faster for floating point operations.

Many ARM cores lack of a Floating Point Unit and any software acceleration is more than welcome.

It might be hard to switch to EABI. For instance, for the Debian distribution, EABI is actually considered a new port.

Edit Link Top

Without EABI

The ARM EABI (embedded application binary interface) improves the floating point performance. So, it's not striking if you read how your processor is wasting a lot of cycles now. From the Debian ARM-EABI wiki:

The current Debian port creates hardfloat FPA instructions. FPA comes from "Floating Point Accelerator". Since the FPA floating point unit was implemented only in very few ARM cores, these days FPA instructions are emulated in kernel via Illegal instruction faults. This is of course very inefficient: about 10 times slower that -msoftfloat for a FIR test program. The FPA unit also has the peculiarity of having mixed-endian doubles, which is usually the biggest grief for ARM porters, along with structure packing issues.

So, what does it mean? It means that the compilers usually generate inscructions for a piece of harware, namely a Floating Point Unit that is not actually there!

So, when you make a floating point operation, such at 3.58*x, the CPU runs into an illegal instruction, and it raises an exception.

The kernel catches this specific exception and performs the intended float point operation, and then resumes executing the program. And this is slow because it implies a context switch.

Edit Link Top

The benchmark

We made a simple benchmark using our Open Hardware Free ECB_AT91 ARM(ARMv4t) development board and a simple benchmark we have used before : The dot product of 2 given vectors, the euclidean distance of the vectors, and the FFT algorithm (complex valued, Cooley and Tukey radix-2), The source code we used is available (GPL).

It's usual to use the number of Floating Point Operations performed by a given program for benchmarking purposes. This can be misleading because some operations (division) take more time than others (addition). That's why we will run the same program in both setups, with similar compiler flags

First we will try the Old ABI using the Debian distribution (Debian Sid). We will use an image that we bootstrapped.

For the EABI test we used OpenEmbedded (by using the �ngstr�m Distribution).

Edit Link Top

Results

MFLOPS (Millions of floating point instructions per second)

Speed-Up EABI over OABI

In each context switch both the data and instruction cache are flushed and this hurts the Old ABI's performance. You will notice it in the graphs because the performance with the old ABI does not depend on the size (N) of the input data whereas in EABI the impact of the cache in the performance is seen clearly. The dot-product performance only goes down when N > 4096 (When we use more than 16KB in memory). The processor we're using (AT91RM9200) has a 16-Kbyte data cache.

Edit Link Top

Closing Remarks

In order to benefit from the new EABI you need a distribution that supports it. Thanks to the work of the OpenEmbedded crew we got a working toolchain based on GCC. We didn't test EABI with Debian, but we expect the results to be quite similar.

Edit Link Top

About the Authors

Andr�s Calder�n and Nelson Castillo are co-founders of emQbit. Both of them are long time GNU/Linux users.

Andr�s has experience in High Performance computing and in Embedded Systems Design, DSP and FPGA programming.

Nelson has experience in Unix Network Programming, Linux driver programming and Embedded System programming.

The Free ECB_AT91 V1 was used in the tests. The design (open hardware) of this board is available online.

Edit Link Top

Recent files: https://freaks-unidos.net/cgi-bin/svnwiki/emqbit/

how-to-query-a-modem

ecbot

ECBOT

ECBOT Block diagram

at91-gpio

tspc

Quick overview

More information

Use

Install

ping

browse

SSH

HTTP

Restart the service

Authenticated tunnels

Conclusion

OpenEmbedded from scratch

Hardware watchdog for the AT91RM9200

at91rm9200-watchdog

Setup the kernel

Verify that the device file exists

Build an example

Test the watchdog

References

First post

emqbot

emQbot

Details

darrell-loader

AT91RM9200 Loader

emQbit's weblog

Hardware watchdog for the AT91RM9200

First post

log/menu

About

Contact

linux-kernel

Where are the patches?

ccache-distcc

Motivation

Some theory

Installing

Configuring ccache

Configuring distcc

The environment

An example

Flashing the ECB_AT91

Our first reflow oven (Obi-Wan)

Calibration the analog-digital converter

Controling the oven

Links

programador-paralelo-avr

Cable paralelo

ecb-at91-v3

SBC ECB_AT91 V3

Features of the Single Board Computer

linux-cl-url-temporal

eabi-oabi-benchmark

Analysis: Why EABI matters

Foreword

Without EABI

The benchmark

Results

Closing Remarks

About the Authors

References