On Sat, 13 Sep 2003 21:39:47 +0200, "Jérôme Delacotte"
<challengejs (AT) free (DOT) fr> wrote:

In which environment ? PROGRESS 6 on DOS or PROGRESS 9 on a 8
processor HP-UX ?

Was your variable declared with no-undo?



if no-undo was not used then the loop becomes part of a transaction.


--
Posted via http://dbforums.com

"Jérôme Delacotte" <challengejs (AT) free (DOT) fr> writes:

Try "no-undo" qualifier on the declaration of "i".

It's optimised to nothing. Infinitely faster.

/"\ Bernd Felsche - Innovative Reckoning, Perth, Western Australia
\ / ASCII ribbon campaign | I'm a .signature virus!
X against HTML mail | Copy me into your ~/.signature
/ \ and postings | to help me spread!

Jérôme,

Unlike P4GL, the C compiler is most likely recognizing that the do block is
empty and that it can be replaced with "i = 100001", the final value of i.

Regardless of validity of above, the P4GL is definately at lease one or more
orders of magnitude SLOWER than native executables.

However, very little time is spent doing native execution while waiting for
disk drive heads to move while they are reading and/or writing your data.

Joe

"Jérôme Delacotte" <challengejs (AT) free (DOT) fr> wrote

stefan.lang (AT) t-online (DOT) de wrote:
Hi.

IMHO

DO i = 1 to 100000:
END.

is a single transaction, so it does not matter if You
use or not the NO-UNDO

REPEAT i = 1 to 100000:
END.
is a 1000000 transactions:

/* TEST 1: REPEAT & NO-UNDO*/

repeat:
if etime(yes) = 0 then.
def var i as int no-undo.
repeat i = 1 to 100000: end.
disp etime(yes).
end.

1.960
1.551
1.381
1.443
1.503
1.429
1.388
1.715
.....

/* test 2: REPEAT */

repeat:
if etime(yes) = 0 then.
def var i as int.
repeat i = 1 to 100000: end.
disp etime(yes).
end.


1.483
1.554
1.625
1.375
1.375
1.539
1.391
.....

/* test 3: DO */
repeat:
if etime(yes) = 0 then.
def var i as int.
do i = 1 to 100000: end.
disp etime(yes).
end.

1.236
1.239
1.349
1.242
1.357
1.248
1.241
.....

/* test 4: DO & NO-UNDO */
repeat:
if etime(yes) = 0 then.
def var i as int NO-UNDO.
do i = 1 to 100000: end.
disp etime(yes).
end.

1.246
1.256
1.237
1.234
1.234
1.238
1.236
1.234
.....

IMHO:
1. NO-UNDO does not matter when DO is used.
It does matter (20-30% faster) when REPEAT is used.
2. DO is a single transaction,
3. REPEAT is a multiple transactions.
4. Progress 4GL is not a language for math operation, but
for databases. These 1.5 seconds does not matter when
4GL is used for normal database processing.

TESTED ON: RS/6000 B50 running AIX 4.3.1.0 PRGS 8.3E02
with regards

what are you talking about ? progress is interpreted not compiled !
it's 4gl, what did you expect ?

progress compiles code into action cells that drives an interpreter
unlike c that compiles code into actual machine code

yeah ! i didn't get an a+ for a long time

just curious, what are you writting that you have to fight over

every millisecond ? is it something to do with graphics ? any how good luck !