Mounting real CF eats up a lot of RAM

[fitzsteve]

Hi Toni,

I repeated the first test exactly using PFS3, I think it's the one bundled with Classic WB I have here (vers5.3)



Memory usage as you can see is very similar!

[Toni Wilen]

Thanks, it is time to check PFS3 sources

[fitzsteve]

Originally Posted by Toni Wilen

Thanks, it is time to check PFS3 sources

Cool! Let me know if you need a dump of the RDB before I nuke the CF

SFS seems to have similar memory usage, maybe a little less. When I set up CF's for 2mb Amiga's I always use FFS for this reason.

Of course when you have luscious amounts of Fast Ram it isn't an issue!

[Toni Wilen]

Found this:

	/* sanity checks. If HDToolbox default of 30, then 150,
	 * otherwise round in range 70 -- 600
	 */
	if (i==30) i=150;
	if (i<MIN_BUFFERS) i = MIN_BUFFERS;
	if (i>MAX_BUFFERS) i = MAX_BUFFERS;

So 30 becomes 150.. Do you get better memory usage if you change buffers to 29 or 31?

(MIN_BUFFERS is 10 in PFS3AIO, not 70)

[fitzsteve]

Usage is slightly better with 31 buffers, now 589032 in use however I had 80 buffers in my last test that had 620712 usage.

[Amiga1992]

Glad to see some progress and that it wasn't me being an idiot, rather some sort of problem! I can't believe I didn't notice in my real Amiga though. Maybe because I have 2MB of RAM and I never get into Workbench.
Then again i haven't tested in it again, gotta get to it once I fix this faulty picoPSU I received

[edit] tested on my real A600 with 1MB chip, indeed 600+KB are being eaten up!

[ajk]

This must be what I ran across too. I had put my A600 away and wasn't yet able to check the exact amount of memory taken but the results here sound about the same.

[Amiga1992]

Yup sounds about right, has to be a PFS3 problem then.
Is there anything I can do to help sort out this problem?
Should I switch back to FFS? (UGH)

[FOL]

Originally Posted by Akira

Yup sounds about right, has to be a PFS3 problem then.
Is there anything I can do to help sort out this problem?
Should I switch back to FFS? (UGH)

If your using ACA card, you can use SFS.

[fitzsteve]

Originally Posted by FOL

If your using ACA card, you can use SFS.

SFS has similar memory usage in my experience

[Amiga1992]

Originally Posted by FOL

If your using ACA card, you can use SFS.

As said in my original post, this is a STANDARD Amiga 600, no expansions at all. And I doubt I'll install any.

[Toni Wilen]

I am going to debug and hopefully it can be fixed/worked around.

I have already quickly checked PFS3 sources for all memory allocations, but didn't notice anything that could allocate that much memory. Buffer cache also seems to work as expected.

[Toni Wilen]

I tried my PFS3 (official 18.5) 20M HDF and booted without startup-sequence it had about 710k of RAM free (out of 1M), with 80 buffers so it does not always use "too much" memory.

[Amiga1992]

It might depend on the configuration, then.
Would the behaviour be the same whether it is an HDF or a real drive connected through USB?

[Toni Wilen]

Originally Posted by Akira

It might depend on the configuration, then.
Would the behaviour be the same whether it is an HDF or a real drive connected through USB?

It was standard A600 config with IDE. Perhaps it has something to do with size of partition or some other HD specific attribute. Probably need to test different partition sizes next.

HDF vs drive shouldn't make any difference.

[fitzsteve]

Something interesting.

If I set the RDB & partitions with 3.1's HDToolBox instead of 3.9's memory usage is less!

So for PFS3-AIO, 2 equal partitions, 31 buffers each:

[Amiga1992]

I don't remember what I used to set this one up, probably 3.x HDToolbox. I can try redoing it with 2.x HDToolbox.

[thomas]

I am still interested in analysing your RDB...

[Amiga1992]

OH sorry Thomas, I forgot because I ran into a bunch of other hardware trouble :/
I'll do it today and post it here.

[rhester72]

If he's got a particularly large (or multiple) filesystem drives loaded in RDB, wouldn't they necessarily have to be loaded into RAM, possibly explaining the discrepancy? They have to go *somewhere*.

Rodney