devf-ram

I’m doing some performance test with devf-ram. I’ve created a 128Meg ram
disk.
The command flashctl takes 26 seconds to “mount” the disk. But what I don’t
understand is during that 26 seconds no CPU cyle are consumed. Machine is
sitting idle.

That doesn’t sound normal to me.

  • Mario

Hi Mario,

I am tring to find out more info on this. However from what I found is that
this driver treats ram like it is flash, so it has to walk through the whole
thing like it was flash (which is not very fast or efficient).

I will see about getting some more details on this issue for you.

Erick.


Mario Charest <mcharest@clipzinformatic.com> wrote:


I’m doing some performance test with devf-ram. I’ve created a 128Meg ram
disk.
The command flashctl takes 26 seconds to “mount” the disk. But what I don’t
understand is during that 26 seconds no CPU cyle are consumed. Machine is
sitting idle.

That doesn’t sound normal to me.

  • Mario

“Hardware Support Account” <hw@qnx.com> wrote in message
news:9pfl76$gk6$1@nntp.qnx.com

Hi Mario,

I am tring to find out more info on this. However from what I found is
that
this driver treats ram like it is flash, so it has to walk through the
whole
thing like it was flash (which is not very fast or efficient).

That’s my point. If the CPU is busy walking through the whole thing it
should
be very busy doing so (Efsys.ram should use 100% of the CPU)

I will see about getting some more details on this issue for you.

Erick.


Mario Charest <> mcharest@clipzinformatic.com> > wrote:


I’m doing some performance test with devf-ram. I’ve created a 128Meg
ram
disk.
The command flashctl takes 26 seconds to “mount” the disk. But what I
don’t
understand is during that 26 seconds no CPU cyle are consumed. Machine
is
sitting idle.

That doesn’t sound normal to me.

  • Mario

The way to make it faster would be to make the “sectors” smaller. For
example, ATMEL flash has 256 byte sectors, which make it quite fast compared
to other manufacturers’ parts.

“Hardware Support Account” <hw@qnx.com> wrote in message
news:9pfl76$gk6$1@nntp.qnx.com

Hi Mario,

I am tring to find out more info on this. However from what I found is
that
this driver treats ram like it is flash, so it has to walk through the
whole
thing like it was flash (which is not very fast or efficient).

I will see about getting some more details on this issue for you.

Erick.


Mario Charest <> mcharest@clipzinformatic.com> > wrote:


I’m doing some performance test with devf-ram. I’ve created a 128Meg
ram
disk.
The command flashctl takes 26 seconds to “mount” the disk. But what I
don’t
understand is during that 26 seconds no CPU cyle are consumed. Machine
is
sitting idle.

That doesn’t sound normal to me.

  • Mario

“Issam Haddad” <ihaddad@asurent.com> wrote in message
news:9pft0j$nkr$1@inn.qnx.com

The way to make it faster would be to make the “sectors” smaller. For
example, ATMEL flash has 256 byte sectors, which make it quite fast
compared
to other manufacturers’ parts.

Unless that was change, the SDK for flash isnt’ design to work well
with such small blocks. In my case sectors size is unrelated, we are
talking about ram which doesn’t suffer delays for erase/write operations

“Hardware Support Account” <> hw@qnx.com> > wrote in message
news:9pfl76$gk6$> 1@nntp.qnx.com> …
Hi Mario,

I am tring to find out more info on this. However from what I found is
that
this driver treats ram like it is flash, so it has to walk through the
whole
thing like it was flash (which is not very fast or efficient).

I will see about getting some more details on this issue for you.

Erick.


Mario Charest <> mcharest@clipzinformatic.com> > wrote:


I’m doing some performance test with devf-ram. I’ve created a 128Meg
ram
disk.
The command flashctl takes 26 seconds to “mount” the disk. But what I
don’t
understand is during that 26 seconds no CPU cyle are consumed.
Machine
is
sitting idle.

That doesn’t sound normal to me.

  • Mario

    \

Mario,

Are you using 6.1.0? If so, why don’t you use devb-ram? It is based
off of the io-blk/cam architecture, so it is a lot faster than devf-ram.

Previously, Mario Charest wrote in qdn.public.qnxrtp.os:

“Issam Haddad” <> ihaddad@asurent.com> > wrote in message
news:9pft0j$nkr$> 1@inn.qnx.com> …
The way to make it faster would be to make the “sectors” smaller. For
example, ATMEL flash has 256 byte sectors, which make it quite fast
compared
to other manufacturers’ parts.


Unless that was change, the SDK for flash isnt’ design to work well
with such small blocks. In my case sectors size is unrelated, we are
talking about ram which doesn’t suffer delays for erase/write operations

“Hardware Support Account” <> hw@qnx.com> > wrote in message
news:9pfl76$gk6$> 1@nntp.qnx.com> …
Hi Mario,

I am tring to find out more info on this. However from what I found is
that
this driver treats ram like it is flash, so it has to walk through the
whole
thing like it was flash (which is not very fast or efficient).

I will see about getting some more details on this issue for you.

Erick.


Mario Charest <> mcharest@clipzinformatic.com> > wrote:


I’m doing some performance test with devf-ram. I’ve created a 128Meg
ram
disk.
The command flashctl takes 26 seconds to “mount” the disk. But what I
don’t
understand is during that 26 seconds no CPU cyle are consumed.
Machine
is
sitting idle.

That doesn’t sound normal to me.

  • Mario





    \


-Jay.

It would have been nice to have this utility referenced (somewhere) in the
docs… I’m glad I try to keep up to date with this newsgroup.

What are the options for this (undocumented) utility?

“Jay Greig” <greig@qnx.com> wrote in message
news:Voyager.011009101239.11526C@node341.ott.qnx.com

Mario,

Are you using 6.1.0? If so, why don’t you use devb-ram? It is based
off of the io-blk/cam architecture, so it is a lot faster than devf-ram.

Previously, Mario Charest wrote in qdn.public.qnxrtp.os:

“Issam Haddad” <> ihaddad@asurent.com> > wrote in message
news:9pft0j$nkr$> 1@inn.qnx.com> …
The way to make it faster would be to make the “sectors” smaller. For
example, ATMEL flash has 256 byte sectors, which make it quite fast
compared
to other manufacturers’ parts.


Unless that was change, the SDK for flash isnt’ design to work well
with such small blocks. In my case sectors size is unrelated, we are
talking about ram which doesn’t suffer delays for erase/write operations

“Hardware Support Account” <> hw@qnx.com> > wrote in message
news:9pfl76$gk6$> 1@nntp.qnx.com> …
Hi Mario,

I am tring to find out more info on this. However from what I found
is
that
this driver treats ram like it is flash, so it has to walk through
the
whole
thing like it was flash (which is not very fast or efficient).

I will see about getting some more details on this issue for you.

Erick.


Mario Charest <> mcharest@clipzinformatic.com> > wrote:


I’m doing some performance test with devf-ram. I’ve created a
128Meg
ram
disk.
The command flashctl takes 26 seconds to “mount” the disk. But
what I
don’t
understand is during that 26 seconds no CPU cyle are consumed.
Machine
is
sitting idle.

That doesn’t sound normal to me.

  • Mario










    -Jay.

“Jay Greig” <greig@qnx.com> wrote in message
news:Voyager.011009101239.11526C@node341.ott.qnx.com

Mario,

Are you using 6.1.0? If so, why don’t you use devb-ram?

Now I will :wink:

Are the doc people watching this?

It is based
off of the io-blk/cam architecture, so it is a lot faster than devf-ram.

Previously, Mario Charest wrote in qdn.public.qnxrtp.os:

“Issam Haddad” <> ihaddad@asurent.com> > wrote in message
news:9pft0j$nkr$> 1@inn.qnx.com> …
The way to make it faster would be to make the “sectors” smaller. For
example, ATMEL flash has 256 byte sectors, which make it quite fast
compared
to other manufacturers’ parts.


Unless that was change, the SDK for flash isnt’ design to work well
with such small blocks. In my case sectors size is unrelated, we are
talking about ram which doesn’t suffer delays for erase/write operations

“Hardware Support Account” <> hw@qnx.com> > wrote in message
news:9pfl76$gk6$> 1@nntp.qnx.com> …
Hi Mario,

I am tring to find out more info on this. However from what I found
is
that
this driver treats ram like it is flash, so it has to walk through
the
whole
thing like it was flash (which is not very fast or efficient).

I will see about getting some more details on this issue for you.

Erick.


Mario Charest <> mcharest@clipzinformatic.com> > wrote:


I’m doing some performance test with devf-ram. I’ve created a
128Meg
ram
disk.
The command flashctl takes 26 seconds to “mount” the disk. But
what I
don’t
understand is during that 26 seconds no CPU cyle are consumed.
Machine
is
sitting idle.

That doesn’t sound normal to me.

  • Mario










    -Jay.

Previously, Issam Haddad wrote in qdn.public.qnxrtp.os:

It would have been nice to have this utility referenced (somewhere) in the
docs… I’m glad I try to keep up to date with this newsgroup.

What are the options for this (undocumented) utility?

use devb-ram

devb-ram Ram Driver.

Syntax:
devb-ram [cam cam_options][ram ram_options…][blk io-blk_options…] &

Options: All options are separated by commas.

cam options:
quiet Be quiet: don’t display any information on startup.
verbose Be verbose: display full information about units (devices)
on startup.

ram options
The ram options control the drivers interface to RAM.

capacity=capacity Specify the capacity of the RAM drive in 512 byte
blocks (default 4096).
Examples:
devb-ram &

“Jay Greig” <> greig@qnx.com> > wrote in message
news:> Voyager.011009101239.11526C@node341.ott.qnx.com> …
Mario,

Are you using 6.1.0? If so, why don’t you use devb-ram? It is based
off of the io-blk/cam architecture, so it is a lot faster than devf-ram.

Previously, Mario Charest wrote in qdn.public.qnxrtp.os:

“Issam Haddad” <> ihaddad@asurent.com> > wrote in message
news:9pft0j$nkr$> 1@inn.qnx.com> …
The way to make it faster would be to make the “sectors” smaller. For
example, ATMEL flash has 256 byte sectors, which make it quite fast
compared
to other manufacturers’ parts.


Unless that was change, the SDK for flash isnt’ design to work well
with such small blocks. In my case sectors size is unrelated, we are
talking about ram which doesn’t suffer delays for erase/write operations

“Hardware Support Account” <> hw@qnx.com> > wrote in message
news:9pfl76$gk6$> 1@nntp.qnx.com> …
Hi Mario,

I am tring to find out more info on this. However from what I found
is
that
this driver treats ram like it is flash, so it has to walk through
the
whole
thing like it was flash (which is not very fast or efficient).

I will see about getting some more details on this issue for you.

Erick.


Mario Charest <> mcharest@clipzinformatic.com> > wrote:


I’m doing some performance test with devf-ram. I’ve created a
128Meg
ram
disk.
The command flashctl takes 26 seconds to “mount” the disk. But
what I
don’t
understand is during that 26 seconds no CPU cyle are consumed.
Machine
is
sitting idle.

That doesn’t sound normal to me.

  • Mario










    -Jay.

    \


-Jay.

Mario Charest <mcharest@clipzinformatic.com> wrote:

: “Jay Greig” <greig@qnx.com> wrote in message
: news:Voyager.011009101239.11526C@node341.ott.qnx.com
:> Mario,
:>
:> Are you using 6.1.0? If so, why don’t you use devb-ram?

: Now I will :wink:

: Are the doc people watching this?

Yes. I’ll see that it gets added to our (rather long) to-do list. Thanks.


Steve Reid stever@qnx.com
TechPubs (Technical Publications)
QNX Software Systems

Previously, Mario Charest wrote in qdn.public.qnxrtp.os:

“Jay Greig” <> greig@qnx.com> > wrote in message
news:> Voyager.011009101239.11526C@node341.ott.qnx.com> …
Mario,

Are you using 6.1.0? If so, why don’t you use devb-ram?

Now I will > :wink:

Are the doc people watching this?

There was a problem report filed for this a while ago (by me) - the
doc has been made, just waiting for next release/update.

It is based
off of the io-blk/cam architecture, so it is a lot faster than devf-ram.

Previously, Mario Charest wrote in qdn.public.qnxrtp.os:

“Issam Haddad” <> ihaddad@asurent.com> > wrote in message
news:9pft0j$nkr$> 1@inn.qnx.com> …
The way to make it faster would be to make the “sectors” smaller. For
example, ATMEL flash has 256 byte sectors, which make it quite fast
compared
to other manufacturers’ parts.


Unless that was change, the SDK for flash isnt’ design to work well
with such small blocks. In my case sectors size is unrelated, we are
talking about ram which doesn’t suffer delays for erase/write operations

“Hardware Support Account” <> hw@qnx.com> > wrote in message
news:9pfl76$gk6$> 1@nntp.qnx.com> …
Hi Mario,

I am tring to find out more info on this. However from what I found
is
that
this driver treats ram like it is flash, so it has to walk through
the
whole
thing like it was flash (which is not very fast or efficient).

I will see about getting some more details on this issue for you.

Erick.


Mario Charest <> mcharest@clipzinformatic.com> > wrote:


I’m doing some performance test with devf-ram. I’ve created a
128Meg
ram
disk.
The command flashctl takes 26 seconds to “mount” the disk. But
what I
don’t
understand is during that 26 seconds no CPU cyle are consumed.
Machine
is
sitting idle.

That doesn’t sound normal to me.

  • Mario










    -Jay.

    \


-Jay.

Thanks Jay. I’d figured this out on the x86 host (with “use”), and made it
work on the ppc target.

I know you guys have a lot of work on your hands – we do appreciate it.

“Jay Greig” <greig@qnx.com> wrote in message
news:Voyager.011009142024.11526J@node341.ott.qnx.com

Previously, Mario Charest wrote in qdn.public.qnxrtp.os:

“Jay Greig” <> greig@qnx.com> > wrote in message
news:> Voyager.011009101239.11526C@node341.ott.qnx.com> …
Mario,

Are you using 6.1.0? If so, why don’t you use devb-ram?

Now I will > :wink:

Are the doc people watching this?

There was a problem report filed for this a while ago (by me) - the
doc has been made, just waiting for next release/update.


It is based
off of the io-blk/cam architecture, so it is a lot faster than
devf-ram.

Previously, Mario Charest wrote in qdn.public.qnxrtp.os:

“Issam Haddad” <> ihaddad@asurent.com> > wrote in message
news:9pft0j$nkr$> 1@inn.qnx.com> …
The way to make it faster would be to make the “sectors” smaller.
For
example, ATMEL flash has 256 byte sectors, which make it quite
fast
compared
to other manufacturers’ parts.


Unless that was change, the SDK for flash isnt’ design to work well
with such small blocks. In my case sectors size is unrelated, we
are
talking about ram which doesn’t suffer delays for erase/write
operations

“Hardware Support Account” <> hw@qnx.com> > wrote in message
news:9pfl76$gk6$> 1@nntp.qnx.com> …
Hi Mario,

I am tring to find out more info on this. However from what I
found
is
that
this driver treats ram like it is flash, so it has to walk
through
the
whole
thing like it was flash (which is not very fast or efficient).

I will see about getting some more details on this issue for
you.

Erick.


Mario Charest <> mcharest@clipzinformatic.com> > wrote:


I’m doing some performance test with devf-ram. I’ve created a
128Meg
ram
disk.
The command flashctl takes 26 seconds to “mount” the disk.
But
what I
don’t
understand is during that 26 seconds no CPU cyle are consumed.
Machine
is
sitting idle.

That doesn’t sound normal to me.

  • Mario










    -Jay.






    -Jay.