Santanu Chatterjee | 1 Oct 08:13 2005
Picon

Where is the scheduler's timer implemented?

Hi all,

Could you please tell me which file contains the implementation of the
timer that is used by the scheduler in eCos? I don't think it
is the RTC that is being used by the scheduler. The scheduler's
timer must be of much higher resolution I think.

What I actually wish to understand is how exactly the scheduler is
invoked at the scheduling instants.

Thanks in advance for your time.

Regards,
Santanu Chatterjee

--

-- 
Before posting, please read the FAQ: http://ecos.sourceware.org/fom/ecos
and search the list archive: http://ecos.sourceware.org/ml/ecos-discuss

Roy E Richardson | 2 Oct 06:18 2005
Picon
Picon

Re: Where is the scheduler's timer implemented?

The common source is in '\packages\kernel\current\src\common\clock.cxx'

By default the "timed" processes are "alarms", granularity of ticks, 10 Ms
There are also references to RTC timing granularities, but I have no 
experience
with their usage - see CYGVAR_KERNEL_COUNTERS_CLOCK, etc.

PS. For alarms, it is my personal opinion that the config option,
    CYGIMP_KERNEL_COUNTERS_SORT_LIST,
    should be enabled for best optimization of alarm processing,
    although that opinion is contrary to the provided default.

----- Original Message ----- 
From: "Santanu Chatterjee" <thisissantanu <at> gmail.com>
To: <ecos-discuss <at> ecos.sourceware.org>
Sent: Friday, September 30, 2005 11:13 PM
Subject: [ECOS] Where is the scheduler's timer implemented?

Hi all,

Could you please tell me which file contains the implementation of the
timer that is used by the scheduler in eCos? I don't think it
is the RTC that is being used by the scheduler. The scheduler's
timer must be of much higher resolution I think.

What I actually wish to understand is how exactly the scheduler is
invoked at the scheduling instants.

Thanks in advance for your time.

(Continue reading)

rpai | 3 Oct 10:06 2005
Picon

eCos clock tick Help..

hi,

   I was trying to minimise the clock tick to as low as possible . In eCos
    1 tick = 10 ms , I was trying to reduce it to low to 1ms . The
problem is I tried to get a value of how much loops a tick is and while
it is 10 ms and also tried it using 5 ms and 1 ms .

   I changed the tick by changing  CYGNUM_HAL_RTC_NUMERATOR value . but
the no of loops per tick is quite uneasy to accept . is there any other
place where change is required .

  The no of loops values are like

        for 10 ms   -> approx 4986325
             5 ms   -> approx 3816406
             1 ms   -> approx 3399930

  Can any one tell me where all i have to change to get correct values .

  the code used for this is

void init_loops_per_tick(void)
{
    int old_ticks, new_ticks;
    int i;
    cyg_uint32 j;

    if (loops_per_tick != 0) return;

    loops_per_tick = 0;
(Continue reading)

Steven_cheng | 3 Oct 10:15 2005

Re: JFFS2 questions with eCos

Hi Adnrew,

    Thank you for your reply!!
But I still have some questions about mkfs.jffs2.

Q1 :  Can you tell me what is the different betweem -e
and --eraseblock=size,
-p and --pad=size ??

Q2 : "ereaseblock=size" and "--pad=size" in the statements, the unit is
bytes or Kbytes ??

Q3 : "-e size" and "-p size" in the statements, the unit is bytes or Kbytes
??

I can't understand the meaning of the help as following:

  -p, --pad[=SIZE]        Pad output to SIZE bytes with 0xFF. If SIZE is
                          not specified, the output is padded to the end of
                          the final erase block.

  -e, --eraseblock=SIZE   Use erase block size SIZE (default: 64KiB)

Thank you very much~~

Steven Cheng

----- Original Message ----- 
From: "Andrew Lunn" <andrew <at> lunn.ch>
To: "Steven_cheng" <05071 <at> alphanetworks.com>
(Continue reading)

Andrew Lunn | 3 Oct 10:34 2005
Picon

Re: JFFS2 questions with eCos

On Mon, Oct 03, 2005 at 04:15:19PM +0800, Steven_cheng wrote:
> Hi Adnrew,
> 
>     Thank you for your reply!!
> But I still have some questions about mkfs.jffs2.

There is a man page you know....

> 
> Q1 :  Can you tell me what is the different betweem -e
> and --eraseblock=size,

None. One is a short opt and one is a long opt. Standard GNU practice.

> -p and --pad=size ??

None, although there was/is a bug which i think means one does not
work. Check the actuall size of the image created.

> Q2 : "ereaseblock=size" and "--pad=size" in the statements, the unit is
> bytes or Kbytes ??

Why not read the man page and save us all waisted time...

       Options that take SIZE arguments can be  specified  as  either  decimal
       (e.g., 65536), octal (0200000), or hexidecimal (0x1000).

 
> I can't understand the meaning of the help as following:
> 
(Continue reading)

Steven_cheng | 3 Oct 11:35 2005

Re: JFFS2 questions with eCos

Hi Andrew,

    I'm sorry to disturb you, because I take some tests
with the mkfs.jffs2.

When I use the following commands, the size of the image are
different, it is tested in i386-linux and cygwin.

(1) ./mkfs.jffs2 -b -o myjffs2_1.img -r
jffs2_dir/ --pad=4194304 --eraseblock=131072
      [image size = 4,194,304 bytes ]

(2) ./mkfs.jffs2 -b -o myjffs2_2.img -r jffs2_dir/ --pad=4194304 -e 131072
      [image size = 4,194,304 bytes  ]

(3) ./mkfs.jffs2 -b -o myjffs2_3.img -r jffs2_dir/ --pad=4194304 -e=131072
      [ error !!
       i386linux: Increasing erase size to 8KiB minimum
       cygwin: mkfs.jffs2: Unrecognisable erase size]

(4) ./mkfs.jffs2 -b -o myjffs2_4.img -r jffs2_dir/ -p 4194304 -e 131072
      [image size = 393,216 bytes]

(5) ./mkfs.jffs2 -b -o myjffs2_5.img -r jffs2_dir/ -p=4194304 -e 131072
      [image size = 320,688 bytes]

So I am not sure which is the correct usage !!

Thank yor for your reply ~~

(Continue reading)

Andrew Lunn | 3 Oct 11:58 2005
Picon

Re: JFFS2 questions with eCos

On Mon, Oct 03, 2005 at 05:35:01PM +0800, Steven_cheng wrote:
> Hi Andrew,
> 
>     I'm sorry to disturb you, because I take some tests
> with the mkfs.jffs2.
> 
> When I use the following commands, the size of the image are
> different, it is tested in i386-linux and cygwin.
> 
> (1) ./mkfs.jffs2 -b -o myjffs2_1.img -r
> jffs2_dir/ --pad=4194304 --eraseblock=131072
>       [image size = 4,194,304 bytes ]
> 
> (2) ./mkfs.jffs2 -b -o myjffs2_2.img -r jffs2_dir/ --pad=4194304 -e 131072
>       [image size = 4,194,304 bytes  ]
> 
> (3) ./mkfs.jffs2 -b -o myjffs2_3.img -r jffs2_dir/ --pad=4194304 -e=131072
>       [ error !!
>        i386linux: Increasing erase size to 8KiB minimum
>        cygwin: mkfs.jffs2: Unrecognisable erase size]
> 
> (4) ./mkfs.jffs2 -b -o myjffs2_4.img -r jffs2_dir/ -p 4194304 -e 131072
>       [image size = 393,216 bytes]
> 
> (5) ./mkfs.jffs2 -b -o myjffs2_5.img -r jffs2_dir/ -p=4194304 -e 131072
>       [image size = 320,688 bytes]
> 
> So I am not sure which is the correct usage !!
> 

(Continue reading)

huleslid | 3 Oct 16:43 2005
Picon

MMU support

Hi ,

     ECos has MMU support . that is it supports flat
address space concept . I have an application that
should have partition that will run its own process
and should not overlap .

     And I think strict usage of MMU is supported by
power pc target .

   I want to know about the more about MMU and how
could i make partitions stricly followed by eCos . and
also any material regarding will also be helpful to
me.

 Huleslid

		
__________________________________________________________ 
Yahoo! India Matrimony: Find your partner now. Go to http://yahoo.shaadi.com

--

-- 
Before posting, please read the FAQ: http://ecos.sourceware.org/fom/ecos
and search the list archive: http://ecos.sourceware.org/ml/ecos-discuss

Gary Thomas | 3 Oct 17:14 2005

Re: MMU support

On Mon, 2005-10-03 at 15:43 +0100, huleslid wrote:
> Hi ,
> 
>      ECos has MMU support . that is it supports flat
> address space concept . I have an application that
> should have partition that will run its own process
> and should not overlap .
> 
>      And I think strict usage of MMU is supported by
> power pc target .
> 
>    I want to know about the more about MMU and how
> could i make partitions stricly followed by eCos . and
> also any material regarding will also be helpful to
> me.

IMO, if you _need_ this, then you should be using Linux, not eCos.
The notion of processes and separate (or even protected) address
spaces are contrary to the basic [very simple] design of eCos.

That said, you could look at the work contributed by 3G Labs a
few years ago.  It was very ARM specific though, and to make it
work on PowerPC would be an effort.  See http://ecos.sourceware.org/contrib.html
for some details.

--

-- 
------------------------------------------------------------
Gary Thomas                 |  Consulting for the
MLB Associates              |    Embedded world
------------------------------------------------------------
(Continue reading)

huleslid | 3 Oct 18:47 2005
Picon

Re: MMU support


 hi,

    The application i want to develop is a real time
one and is prefered to be ecos while comparing other
RTOS. and it wants partioned memory that should be
strictly followed .

   how much ecos support it as lynux os have support
for it.

  I went through the link but now 3glabs is not there
any idea about it.

   I want to know what all is to be made to make
memory partion strict .

  Huleslid

		
__________________________________________________________ 
Yahoo! India Matrimony: Find your partner now. Go to http://yahoo.shaadi.com

--

-- 
Before posting, please read the FAQ: http://ecos.sourceware.org/fom/ecos
and search the list archive: http://ecos.sourceware.org/ml/ecos-discuss


Gmane