Hello,
I have a problem at the level of the cache management of the hard disk.
I have a test which charges with the set of data overlapping time with
others.
The time of loading is 2.5 times more important than under windows.
I therefore tried to create an image boot while varying the size of the
cache of my hard disk, and that i put 0Mo or 128 Mo, i have the same time of
loading (Modified line in .build: [pri = 10o] PATH=/proc/boot
diskboot -b1 -D1 -o “devb-eide,blk cache=0m”… or [pri = 10o]
PATH=/proc/boot diskboot -b1 -D1 -o “devb-eide,blk cache=128m”, and after
under Neutrino i can see my modification with sin fonction).
Is there somebody able to explain me why?
Additionally, if my data is on DOS partition, the system takes the same time
exactly. However in theory shouldn’t it have an excess cost of decoding
there??
The maximum cache size allowed by io-blk.so is 10M. If you specify a larger
cache size, it is ignored, and the 10M size is used.
I am looking into the DOS part of the question.
Cheers,
-Brian
+===============================+
Brian K. Hlady
OEM Support Rep
QNX Software Systems, Ltd. bhlady@qnx.com
+===============================+
“CHANAUD” <lchanaud@bcisa.com> wrote in message
news:a1bs2k$vp$1@inn.qnx.com…
Hello,
I have a problem at the level of the cache management of the hard disk.
I have a test which charges with the set of data overlapping time with
others.
The time of loading is 2.5 times more important than under windows.
I therefore tried to create an image boot while varying the size of the
cache of my hard disk, and that i put 0Mo or 128 Mo, i have the same time
of
loading (Modified line in .build: [pri = 10o] PATH=/proc/boot
diskboot -b1 -D1 -o “devb-eide,blk cache=0m”… or [pri = 10o]
PATH=/proc/boot diskboot -b1 -D1 -o “devb-eide,blk cache=128m”, and after
under Neutrino i can see my modification with sin fonction).
Is there somebody able to explain me why?
Additionally, if my data is on DOS partition, the system takes the same
time
exactly. However in theory shouldn’t it have an excess cost of decoding
there??
The maximum cache size allowed by io-blk.so is 10M. If you specify a
larger cache size, it is ignored, and the 10M size is used.
Sorry, but no. It will not default (no cache= given) to using more than
10MB (even if this is <15%), but you may set it (using cache=) to any size.
It is more likely that the “blk” option is being incorrectly merged by
“diskboot” (the 6.1.1 version of io-blk.so will now handle this) … does
“pidin mem” really show different cache values? As to execution speed,
cache is not really (all things considered) going to affect this; I think
there was a text caching issue in Proc that is also addressed in 6.1.1.
Brian K. Hlady <> bhlady@qnx.com> > wrote:
The maximum cache size allowed by io-blk.so is 10M. If you specify a
larger cache size, it is ignored, and the 10M size is used.
Sorry, but no. It will not default (no cache= given) to using more than
10MB (even if this is <15%), but you may set it (using cache=) to any
size.
It is more likely that the “blk” option is being incorrectly merged by
“diskboot” (the 6.1.1 version of io-blk.so will now handle this) … does
“pidin mem” really show different cache values? As to execution speed,
cache is not really (all things considered) going to affect this; I think
there was a text caching issue in Proc that is also addressed in 6.1.1.
Where can we get QNX 6.1.1? In order to give it a try…
And check if that solves our problem.
Thanks for the answer.
Didier.
“John Garvey” <> jgarvey@qnx.com> > a écrit dans le message de news:
a1d0q5$1aq$> 1@nntp.qnx.com> …
Brian K. Hlady <> bhlady@qnx.com> > wrote:
The maximum cache size allowed by io-blk.so is 10M. If you specify a
larger cache size, it is ignored, and the 10M size is used.
Sorry, but no. It will not default (no cache= given) to using more than
10MB (even if this is <15%), but you may set it (using cache=) to any
size.
It is more likely that the “blk” option is being incorrectly merged by
“diskboot” (the 6.1.1 version of io-blk.so will now handle this) …
does
“pidin mem” really show different cache values? As to execution speed,
cache is not really (all things considered) going to affect this; I
think
there was a text caching issue in Proc that is also addressed in 6.1.1.
Currently you could grab default arguments of devb-eide (pidin arg | grep devb-eide) and launch the
disk driver before diskboot with cache size what you’d like (also put here grabed args). The next
line in .build file should start diskboot and exclude the disk driver (diskboot -b1 -D1
-xdevb-eide).
HTH,
Best regards,
Eduard.
With a PC, I always felt limited by the software available.
On Unix, I am limited only by my knowledge.
–Peter J. Schoenster <> pschon@baste.magibox.net
“Didier Harrang” <> didier.harrang@bcisa.com> > wrote in message
news:a1ebbq$n78$> 1@inn.qnx.com> …
So the big question…
Where can we get QNX 6.1.1? In order to give it a try…
And check if that solves our problem.