COMPGROUPS.NET | Search | Post Question | Groups | Stream | About | Register

### Extracting Digits from a Number

• Email
• Follow

```Hello,

I know of a way to extract digits from a number using the %10 and
divide by 10.  But I am wondering if there is an algorithm out there
that does not use a divide by 10 feature.

The reason I ask is that I am programming on a RISC system where
division is quite expensive, and I want to be able to extract the
digits from an integer (the integer will be from 1 to 6 digits long,
and I know how many digits are in the number).

Any ideas?

```
 0
Reply aarongeier (7) 9/9/2005 2:28:26 PM

See related articles to this posting

```In article <1126276106.645791.168640@g47g2000cwa.googlegroups.com>,
Kosio <aarongeier@gmail.com> wrote:
>I know of a way to extract digits from a number using the %10 and
>divide by 10.  But I am wondering if there is an algorithm out there
>that does not use a divide by 10 feature.

>The reason I ask is that I am programming on a RISC system where
>division is quite expensive, and I want to be able to extract the
>digits from an integer (the integer will be from 1 to 6 digits long,
>and I know how many digits are in the number).

>Any ideas?

branch tables, look-up tables, repeated subtraction, "manual"
division by 10 in binary (shifts, subtractions, multiplications)

An old posting
references
"Optimizing Integer Division by a Constant Divisor" by Robert Grappel
which was apparently in Dr. Dobbs... probably 15 years ago or so.
--
"This was a Golden Age, a time of high adventure, rich living and
hard dying... but nobody thought so."      -- Alfred Bester, TSMD
```
 0
Reply roberson2 (8606) 9/9/2005 2:47:19 PM

```Kosio wrote:
> Hello,
>
> I know of a way to extract digits from a number using the %10 and
> divide by 10.  But I am wondering if there is an algorithm out there
> that does not use a divide by 10 feature.
>
> The reason I ask is that I am programming on a RISC system where
> division is quite expensive, and I want to be able to extract the
> digits from an integer (the integer will be from 1 to 6 digits long,
> and I know how many digits are in the number).
>
> Any ideas?

As this is an algorithm question, not a "C" one, you might get
more opinions in comp.programming.  You could get digits from a
table with 1000000 entries.  Or do /1000 and %1000 and get them from a
table with 1000 entries.  Since you care about efficiency, I assume
using sprintf is right out.  I'm sure there are other ways...

-David

```
 0
Reply lndresnick (326) 9/9/2005 2:54:37 PM

```In article <1126276106.645791.168640@g47g2000cwa.googlegroups.com>,
Kosio <aarongeier@gmail.com> wrote:
>
> I know of a way to extract digits from a number using the %10 and
> divide by 10.  But I am wondering if there is an algorithm out there
> that does not use a divide by 10 feature.
>
> The reason I ask is that I am programming on a RISC system where
> division is quite expensive, and I want to be able to extract the
> digits from an integer (the integer will be from 1 to 6 digits long,
> and I know how many digits are in the number).
>
> Any ideas?

Here are some thoughts, in no particular order:

1. There may be some compiler options to help you out.  A lot
of RISC compilers (incl. gcc) default to the mostly widely
compatible instruction set for a given processor.  If you are
willing to de-support older/cheaper versions of a CPU then
you may be able to take advantage of division in hardware.

8. An alternative is to ship two versions of a function, one
compiled for hardware division and one w/o.  You can either
select at install time or run time whatever works best.

2. You may be able to exploit some redundancy in the input
data.  Special-case -100 thru +100 or so in an array, and/or
cache the last several numbers outside that range.  You
don't have to speed it up for all numbers, just most of the
ones actually encountered.

5. Don't store the numbers as numbers.  Put them in bytes
or nybbles and when you need to do arithmetic on them,
the conversion to numbers will only require multiply and add,
not division, so it should be faster.

4. Depending on the abilities of the CPU and the dedication
of whoever implented stdlib.h on your platform, it is
theoretically possible that div() is faster than doing
separate % and / operations.
```
 0
Reply anon7843 (142) 9/9/2005 6:00:07 PM

```"Kosio" <aarongeier@gmail.com> writes:
> I know of a way to extract digits from a number using the %10 and
> divide by 10.  But I am wondering if there is an algorithm out there
> that does not use a divide by 10 feature.
>
> The reason I ask is that I am programming on a RISC system where
> division is quite expensive, and I want to be able to extract the
> digits from an integer (the integer will be from 1 to 6 digits long,
> and I know how many digits are in the number).

Before you consider using an algorithm other than the obvious one, do
some actual performance measurements.  If extracting digits is really
a bottleneck, then by all means do as much optimization as you need.
(I don't have any good ideas about how to do that.)  But if it's not a
bottleneck, don't bother making it faster.

Why do you need the decimal digits?  The most obvious reason is for
human-readable output; in that case, the time to perform the output is
likely to exceed the time to extract the digits.  If it's not for
human-readable output, consider whether you can use octal or

--
Keith Thompson (The_Other_Keith) kst-u@mib.org  <http://www.ghoti.net/~kst>
San Diego Supercomputer Center             <*>  <http://users.sdsc.edu/~kst>
We must do something.  This is something.  Therefore, we must do this.
```
 0
Reply kst-u (21963) 9/9/2005 8:22:15 PM

```> Before you consider using an algorithm other than the obvious one, do
> some actual performance measurements.  If extracting digits is really
> a bottleneck, then by all means do as much optimization as you need.
> (I don't have any good ideas about how to do that.)  But if it's not a
> bottleneck, don't bother making it faster.
>
> Why do you need the decimal digits?  The most obvious reason is for
> human-readable output; in that case, the time to perform the output is
> likely to exceed the time to extract the digits.  If it's not for
> human-readable output, consider whether you can use octal or
>
> --
> Keith Thompson (The_Other_Keith) kst-u@mib.org  <http://www.ghoti.net/~kst>
> San Diego Supercomputer Center             <*>  <http://users.sdsc.edu/~kst>
> We must do something.  This is something.  Therefore, we must do this.

My reasoning is that I want to send a number to a 7 segment, 8-digit
LED display, but I have to know each individual number to set the
display.  These numbers will be updated fairly often.  I am using an
SPI protocol to send data between the micro and the LED Display...
Basically I send the numbers digit by digit to the display, so I need
to know individual digits of the numbers passed to the micro.

I'll have to look more into how the division algorith works on this
compiler, but with limited ROM space on the micro (and I'm tight as it
is already), two lines of division code take up roughly 3% of my
program space.

Aaron

```
 0
Reply aarongeier (7) 9/9/2005 11:01:32 PM

```In article <1126276106.645791.168640@g47g2000cwa.googlegroups.com> "Kosio" <aarongeier@gmail.com> writes:
> I know of a way to extract digits from a number using the %10 and
> divide by 10.  But I am wondering if there is an algorithm out there
> that does not use a divide by 10 feature.
>
> The reason I ask is that I am programming on a RISC system where
> division is quite expensive, and I want to be able to extract the
> digits from an integer (the integer will be from 1 to 6 digits long,
> and I know how many digits are in the number).

If you have 64 bit integers the following will do it:

char repr[7];

void convert(unsigned long long z) {
unsigned long long a = z * 687195;
unsigned long long c;
int shift = 36;
int i;
for(i = 0; i < 6; i++) {
c = a >> shift;
repr[i] = (int)c + '0';
a = a - (c << shift);
a += (a << 2);
shift--;
}
repr[6] = 0;
}
--
dik t. winter, cwi, kruislaan 413, 1098 sj  amsterdam, nederland, +31205924131
home: bovenover 215, 1025 jn  amsterdam, nederland; http://www.cwi.nl/~dik/
```
 0
Reply Dik.Winter (1625) 9/10/2005 12:44:27 AM

```In article <1126306892.317811.213390@g44g2000cwa.googlegroups.com>,
Kosio <aarongeier@gmail.com> wrote:
>My reasoning is that I want to send a number to a 7 segment, 8-digit
>LED display, but I have to know each individual number to set the
>display.  These numbers will be updated fairly often.

In the old days, there were dedicated chips for this purpose...

>I am using an
>SPI protocol to send data between the micro and the LED Display...
>Basically I send the numbers digit by digit to the display, so I need
>to know individual digits of the numbers passed to the micro.

Does that imply that you are sending the digits in sequence?
That's a bit different than the task of being able to extract an
arbitrary digit from the middle of a number.

Does the protocol allow the least significant digit to be sent
first?

>I'll have to look more into how the division algorith works on this
>compiler, but with limited ROM space on the micro (and I'm tight as it
>is already), two lines of division code take up roughly 3% of my
>program space.

You only have about 64 instruction locations available for your
entire program??
--
I was very young in those days, but I was also rather dim.
-- Christopher Priest
```
 0
Reply roberson2 (8606) 9/10/2005 5:17:30 AM

```"Kosio" <aarongeier@gmail.com> wrote in message
> Hello,
>
> I know of a way to extract digits from a number using the %10 and
> divide by 10.  But I am wondering if there is an algorithm out there
> that does not use a divide by 10 feature.
>
> The reason I ask is that I am programming on a RISC system where
> division is quite expensive, and I want to be able to extract the
> digits from an integer (the integer will be from 1 to 6 digits long,
> and I know how many digits are in the number).
>
> Any ideas?
>

I usually use assembler when programming microcontrollers, but here is the
basic algorithm that should do what you want in C.

void Bin2Ascii (int binval)
{
int      plv[]={10,100,1000,10000,100000};
int      place=5;
int      retval;
if(binval > 999999) return;
while(place--)
{
retval=0;
while(binval>plv[place])
{
binval-=(plv[place]);
retval++;
}
printf("%d\n",retval); // Print Hundred thousands,ten
thousands,thousands,hundreds
}
printf("%d\n",binval);     // print ones
}

This should be faster and and consume less space than a general division
function.

Good luck,
Mike H.

```
 0

```Dik T. Winter wrote:

> <aarongeier@gmail.com> writes:
>  > I know of a way to extract digits from a number using the %10 and
>  > divide by 10.  But I am wondering if there is an algorithm out there
>  > that does not use a divide by 10 feature.
>  >
>  > The reason I ask is that I am programming on a RISC system where
>  > division is quite expensive, and I want to be able to extract the
>  > digits from an integer (the integer will be from 1 to 6 digits long,
>  > and I know how many digits are in the number).
>
> If you have 64 bit integers the following will do it:
>
> char repr[7];
>
> void convert(unsigned long long z) {
>     unsigned long long a = z * 687195;
>     unsigned long long c;
>     int shift = 36;
>     int i;
>     for(i = 0; i < 6; i++) {
>         c = a >> shift;
>         repr[i] = (int)c + '0';
>         a = a - (c << shift);
>         a += (a << 2);
>         shift--;
>     }
>     repr[6] = 0;
> }

Where does the black magic behind this function come from?
</OT>
--
Eric Laberge
```
 0
Reply demon_lord (11) 9/10/2005 7:42:55 PM

```In article <3HGUe.41853\$FI2.1313722@wagner.videotron.net> demon_lord@myrealbox.com writes:
> Dik T. Winter wrote:
....
> >  > I know of a way to extract digits from a number using the %10 and
> >  > divide by 10.  But I am wondering if there is an algorithm out there
> >  > that does not use a divide by 10 feature.
....
> > If you have 64 bit integers the following will do it:
> >
> > char repr[7];
> >
> > void convert(unsigned long long z) {
> >     unsigned long long a = z * 687195;
> >     unsigned long long c;
> >     int shift = 36;
> >     int i;
> >     for(i = 0; i < 6; i++) {
> >         c = a >> shift;
> >         repr[i] = (int)c + '0';
> >         a = a - (c << shift);
> >         a += (a << 2);
> >         shift--;
> >     }
> >     repr[6] = 0;
> > }
>
> Where does the black magic behind this function come from?

No black magic.  I found this kind of algorithm a long time ago for
three digit numbers.  The multiplier is 41 and the starting shift is 12.
The idea is that division by a power of 2 is cheaper than a generic
division.  So to convert an "n" digit number using multiplication and
shifts only you look for a power of 2 such that the following holds
(note that ^ means exponentiation here):
Say the power of 2 is p = 2^k
Calculate r = p/(10^(n-1)) + 1.
When r * (10^n - 1)/p == 9 you have a valid set of numbers.
You take the smallest k such that this holds (and start with the smallest
k such that p > 10^n - 1).  Your multiplier is r, the starting shift is k.

Let us try it for 3 digit numbers.  The smallest power of 2 exceeding 999
is 1024 == 2^10.  r == 11.  11 * 999 / 1024 == 10.  Wrong.  Goto 2048 ==
2^11.  r == 21.  21 * 999 / 2048 == 10.  Wrong.  Goto 4096 == 2^12.
r == 41.  41 * 999 / 4096 == 9.  Check.

And the final reason this kind of algorith works is because (from the
three conditions above) you have that for any n-digit number z:
z < r * 10^(n-1) * z / p < z + 1,
but that is just mathematics.
--
dik t. winter, cwi, kruislaan 413, 1098 sj  amsterdam, nederland, +31205924131
home: bovenover 215, 1025 jn  amsterdam, nederland; http://www.cwi.nl/~dik/
```
 0
Reply Dik.Winter (1625) 9/11/2005 1:32:55 AM

```# I know of a way to extract digits from a number using the %10 and
# divide by 10.  But I am wondering if there is an algorithm out there
# that does not use a divide by 10 feature.
#
# The reason I ask is that I am programming on a RISC system where
# division is quite expensive, and I want to be able to extract the
# digits from an integer (the integer will be from 1 to 6 digits long,
# and I know how many digits are in the number).

static void decimalassign(char *accumulator,char *value,int N) {
int n = strlen(value); if (n>N) n = N;
memset(accumulator,'0',N-n);
strcpy(accumulator+(N-n),value);
}
int i,c;
for (i=N-1,c=0; i>=0; i--) {
c = d>9; if (c) d -= 10;
accumulator[i] = d+'0';
}
}
char power2[N+1];
decimalassign(power,"1",N);
}
}

static void spacefill(char *accumulator,int N) {
int i; for (i=0; i<N-2; i++) {
if (accumulator[i]=='0') accumulator[i] = ' ';
else break;
}
}
static void zerotrim(char *accumulator,int N) {
int i; for (i=0; i<N-2; i++) {
if (accumulator[i]!='0') return accumulator+i;
}
return accumulator+N-1;
}

--
SM Ryan http://www.rawbw.com/~wyrmwif/
OOOOOOOOOO! NAVY SEALS!
```
 0
Reply wyrmwif (945) 9/11/2005 6:27:29 AM

```Kosio wrote:
>>Before you consider using an algorithm other than the obvious one, do
>>some actual performance measurements.  If extracting digits is really
>>a bottleneck, then by all means do as much optimization as you need.
>>(I don't have any good ideas about how to do that.)  But if it's not a
>>bottleneck, don't bother making it faster.
>
> My reasoning is that I want to send a number to a 7 segment, 8-digit
> LED display, but I have to know each individual number to set the
> display.  These numbers will be updated fairly often.

Let's do a timing estimate and see if there is a problem.  Unless you
are doing something like labelling frames in a high-speed motion camera,
you probably are generating data for human reading.  An update speed
faster than about 20 numbers/second simply generates a blur.  I think
you would need to be slower than about 5 numbers/second to have
intermediate numbers readable.  Let's say that you update an 8 digit
display 10 times/second (rather fast).  That would mean generating 80
digits /second.  If a 32-bit integer division routine took 400
instructions on an 8-bit processor, the conversion for one second of
that is 32 ms., which is about 3% of the processor load, which would be

```
 0

```"Thad Smith" <ThadSmith@acm.org> wrote in message
> Kosio wrote:
>>>Before you consider using an algorithm other than the obvious one, do
>>>some actual performance measurements.  If extracting digits is really
>>>a bottleneck, then by all means do as much optimization as you need.
>>>(I don't have any good ideas about how to do that.)  But if it's not a
>>>bottleneck, don't bother making it faster.
>>
>> My reasoning is that I want to send a number to a 7 segment, 8-digit
>> LED display, but I have to know each individual number to set the
>> display.  These numbers will be updated fairly often.
>
> Let's do a timing estimate and see if there is a problem.  Unless you are
> doing something like labelling frames in a high-speed motion camera, you
> probably are generating data for human reading.  An update speed faster
> than about 20 numbers/second simply generates a blur.  I think you would
> need to be slower than about 5 numbers/second to have intermediate numbers
> readable.  Let's say that you update an 8 digit display 10 times/second
> (rather fast).  That would mean generating 80 digits /second.  If a 32-bit
> integer division routine took 400 instructions on an 8-bit processor, the
> conversion for one second of display (10 updates) would take about 32000
> instructions.  At 1 mips, that is 32 ms., which is about 3% of the
>
I don't think the speed of actually updating the display is usually the
issue. In my experience the problem is that you have other very speed
intensive things to be doing so you have to cut time from any place
possible. Of course it would seem logical to use a processor better suited
to the job, but specs usually dictate what you have to work with. Most of
the time this results in using smaller/cheaper parts running at lower speeds
to consume less power...etc...
I think were getting a little off topic now ;-)

Mike H.

```
 0

13 Replies
32 Views

Similar Articles

12/6/2013 5:16:47 AM
[PageSpeed]

Similar Artilces:

From D
for """ """. 4) With D I have created an xsplit() generator, and from my tests it's quite faster than the split(), expecially if the string/lines you want to split are few hundred chars long or more (it's not faster if you want to split very little strings). So I think Python can enjoy such string method too (you can probably simulate an xsplit with a regular expression, but the same is true for some other string methods too). Bye, bearophile On Tue, 24 Jul 2007 03:19:53 -0700, bearophileHUGS wrote: > There are various things I like about the D... this is just a very small token in the whole line. (Yes, there may be exceptions, there may be syntax highlighting.) > 4) With D I have created an xsplit() generator, and from my tests it's > quite faster than the split(), expecially if the string/lines you want > to split are few hundred chars long or more (it's not faster if you want > to split very little strings). So I think Python can enjoy such string > method too (you can probably simulate an xsplit with a regular > expression, but the same is true for some other string methods too). Yea, that's a good

stuff from it
can i have my money back all i did was come across your site on top of that buy stuff from it itsoooook@yahoo.com wrote: > can i have my money back Yes. -- Michael Hoffman

from ravi
i have dowloaded the minix img from minix3 site but unable to edit it or add any file wats the reason any help pls. i m runnin it on vm ware workstation

ftp site works from CLI but not from Safari?
> download with it, but not upload. This is weird. Today it worked fine from Safari. Go figure. I don't know if it's my system, their server, or something else. The only thing I know is that they deal mostly with Windows users.

SIGSEGV from from /usr/lib/libc.2
Hi, I am getting following error... Program received signal SIGSEGV, Segmentation fault si_code: 0 - SEGV_UNKNOWN - Unknown Error. 0x79fa57c0 in __ldfcvt_r+0x2ec0 () from /usr/lib/libc.2 This comes after executing following lines of code. if ((private_fp = fopen(wuEnvFile, "r")) == NULL) printf ("%s File opened = %s\n",wuEnvFile); Could any body tell me why it comes and how i can remove this error. Thanks Navaneet Navaneet wrote: > Hi, > > I am getting following error... > > Program received signal SIGSEGV, Segmentation fault > si_code: 0 - SEGV_UNKNOWN - Unknown Error. > 0x79fa57c0 in __ldfcvt_r+0x2ec0 () from /usr/lib/libc.2 > > This comes after executing following lines of code. > > if ((private_fp = fopen(wuEnvFile, "r")) == NULL) > printf ("%s File opened = %s\n",wuEnvFile); You need to specify two strings in your printf statement to correspond to your two %s's. Robert > > Could any body tell me why it comes and how i can remove this error. > > Thanks > Navaneet > "Navaneet" <tonavaneet@gmail.com> wrote in message news

invoices from 2002 & not from 2004
I stumped [this should be easy] Can someone assist me with the query for "invoices from 2002 & not from 2004" thanks On 15 Jun 2004 10:08:15 -0700, StevenH wrote: > I stumped [this should be easy] > Can someone assist me with the query for "invoices from 2002 & not from 2004" > thanks When you write "Invoices from 2002" do you want just that one year, or from 2002 through the present? Add a new column to the query: InvoiceYear:Year([DateField]) As criteria for this column, write: 2002 To get records from 2002 and newer: >=2002 You can change the criteria to = [Enter Year] or... >= [Enter Year] and it will prompt for the year. This gives you more flexibility to get different years without changing the query. -- Fred Please only reply to this newsgroup. I do not reply to personal email. fredg <fgutkind@example.invalid> wrote in news:kycxhbla19yz.1dfdk3x2sdrs8.dlg@40tude.net: > On 15 Jun 2004 10:08:15 -0700, StevenH wrote: > >> I stumped [this should be easy] >> Can someone assist me with the query for "invoices from 2002 & >> not from 2004" thanks > > When you write

Retrieving data from a hashmap from within a JSP
Hello: I can not understand what I have done wrong. I need to test if a key/value pair exists in a hashmap from within a JSP. I though this would be simple. Struts is also involved here, but I do not believe that Struts is causing any problems. The key/values are placed in the hashmap as follows: hMap = new HashMap(20); while ( rs.next() ) { fRoles = new FormRoles(); fRoles.setFormName(rs.getString("FormName")); fRoles.setCreate( ( rs.getInt("C")==1)? true:false ); fRoles.setRead( ( rs.getInt("R")==1)? true:false ); fRoles.setUpdate( ( rs.getInt("U")==1)? true:false ); fRoles.setDelete( ( rs.getInt("D")==1)? true:false ); hMap.put(rs.getString("FormName"), fRoles); } //while user.setFormRoles( hMap); In the above code I place the FormRoles objects into the hashmap with a String as the key. I will be able to retrieve the objects with a String. Ie. if the form name is "manager", get("manager") will return the corresponding FormRole object. Correct?? Here is my JSP: <%@ taglib prefix="c"

Reading from pipe or from stdin, how?
Hello. I have a program where you specify the name of the file you want to read from command line. But, I want to modify it so that if no file is specified, it reads from stdin. so here's a code snippet: my \$file = ""; # Parse command line to get file option. open( FILE, "<", \$file) or die "Could not open \$file\n"; my \$line; while( (\$file && \$line = <FILE> ) || (!\$file && \$line = <STDIN>)) { do_big_loop_here; } This doesn't seem to work: Unix> echo "blah" | ./pxmlgrep Value of <HANDLE> construct can be "0"; test with defined() at ./ pxmlgrep line 233. Can't modify logical and (&&) in scalar assignment at ./pxmlgrep line 234, near "<FILE> ) " Execution of ./pxmlgrep aborted due to compilation errors. Any suggestions? thanks. In article <2067c26a-a7ae-44b3-abea-37701147e0e9@r34g2000vbi.googlegroups.com>, DaLoverhino <DaLoveRhino@hotmail.com> wrote: > Hello. I have a program where you specify the name of the file you > want to read from command line. But, I want to modify it so that if > no file is specified

CLI shows from H323 but not from SIP
Hi, I have an AS5350 (ios 12.2(15)T10) voip/isdn gateway. It is set up to receive calls from SIP phones, and from other similar gateways, and send them over ISDN to our PBX. The important runtime config parts look like this: isdn switch-type primary-net5 controller E1 3/1 framing NO-CRC4 pri-group timeslots 1-31 description SIP Gateway UA-1; e1-pri controller 3/1 interface Serial3/1:15 no ip address isdn switch-type primary-net5 isdn overlap-receiving isdn protocol-emulate network isdn incoming-voice modem no cdp enable ! voice-port 3/1:D input gain 12 output attenuation 5 echo-cancel coverage 32 cptone BE interface FastEthernet0/0 ip address xxx.xxx.52.147 255.255.255.240 no ip route-cache no ip mroute-cache duplex auto speed 100 ! sip-ua dial-peer voice 1301 pots destination-pattern 111T direct-inward-dial port 3/1:D prefix 0229 clid network-number 0990111 ! So if a call comes in from SIP or another gateway with the prefix 111 it will be sent to port 3/1 with prefix 0229 and CLI 0990111. This works just fine, except that for the calls originating from SIP the CLI 0990111 is NOT sent.On calls from another Cisco the CLI IS sent. The logs show the SIP

Jumpstart from CD
Is it possible to run Custom Jumpstart for Solaris 9 (4/04) right from the CD without using the Jumpstart Server(I'm afraid it will kill our already slow network)? If it is possible, where can I put the sysidcfg file and the profiles file since the CD is not writable? Can I use the floppy? How? Thanks On Thu, 13 May 2004 01:33:36 -0400, Sarah Tanembaum wrote: > If it is possible, where can I put the sysidcfg file and the profiles file > since the CD is not writable? Can I use the floppy? How? Thanks > Read the fine manuals on docs.sun.com. http://docs.sun.com/db/doc/816-7171/6md6pohpn?a=view Helmut -- Half the people you know are below average. "Sarah Tanembaum" <sarah.tanembaum@yahoo.com> wrote in message news:<2ggfljF25t1fU1@uni-berlin.de>... > Is it possible to run Custom Jumpstart for Solaris 9 (4/04) right from the > CD without using the Jumpstart Server(I'm afraid it will kill our already > slow network)? > > If it is possible, where can I put the sysidcfg file and the profiles file > since the CD is not writable? Can I use the floppy? How? Thanks As you probably know, Solaris install image (i.e. NFS