PFS3AIO v3.1 Final

[Heiroglyph]

I hope this hasn't been answered elsewhere, I didn't see it.

The 3.1.4 hdtoolbox has a Direct Scsi option for partitions.

How does that interact with the PFS DS option?

Should we enable both? Does it matter?

I haven't tried it yet, so I don't know if this is filesystem specific.

I've always used the DS version of PFS3 in the past.

[Toni Wilen]

Originally Posted by Heiroglyph

I hope this hasn't been answered elsewhere, I didn't see it.

The 3.1.4 hdtoolbox has a Direct Scsi option for partitions.

How does that interact with the PFS DS option?

Should we enable both? Does it matter?

I haven't tried it yet, so I don't know if this is filesystem specific.

I've always used the DS version of PFS3 in the past.

It is almost same as use of PDS\3 dostype except Direct SCSI is forced. No DS compatibility test, no fallback to NSD or TD64. (This may change in the future if some annoying situations are found..)

New DS option is part filesystem and part hdtoolbox feature.

[Heiroglyph]

Ok, thanks, that helps a lot.

[Storm]

Hello all,

Is it beneficial to back up all data and reformat the drives so it has the latest version on the drive. ie. PFSDoctor shows version 19.2 rather than the older version of 18.5 that I have?

[trixster]

This was covered in post #326

[apex]

I read the manual and I know that PFS/SFS are working with a 512 blocksize.
But a CF card works internally with a 4096 blocksize. What is if I change it for
my PFS and reformat the CF partition? Any issues that can come up? Will the repair tool from you still work? Just need a final answer and will never ask.

[Toni Wilen]

Larger block sizes should work normally, internally PFS3 block size is always at least 1024 bytes (and 2048 bytes if partition >104G and so on).

[daxb]

Did you changed this behaviour or was it already valid for the old PFS3 version? The old PFS3.guide says that it can only use 512 or 1024 bytes block size.

[Toni Wilen]

I thought it was possible but it looks like 1024 byte block size is max (if less than 104G partition). Larger partitions allow higher block size, up to 4096 byte if 411G or larger.

For some reason there is no check for "too large" block size. It most likely causes crash because internal block size variable becomes zero..

It probably is possible to support up to 4096 block size using new large partition mode even if partition is smaller. Of course this will make partition incompatible with older pfs3 and pfsdoctor versions.

[daxb]

Ok, but does it make sense to use larger block size then 512 bytes with PFS3? Will PFS3 become faster (read/write)?

If you use a flash device that uses 4096 bytes per block and write many small files (<4096 bytes) it would make sense I guess but with larger files there isn't any benefit, right? Except some waste of drive space.

[Toni Wilen]

Originally Posted by daxb

Ok, but does it make sense to use larger block size then 512 bytes with PFS3? Will PFS3 become faster (read/write)?

It probably will become faster, at least when accessing large files. Tiny file access may get slower (whole block needs to be read/written even if file size is 1 byte) but it depends on device read/write speed.

Buffers also require more RAM. Number of buffers * block size = required RAM.

If you use a flash device that uses 4096 bytes per block and write many small files (<4096 bytes) it would make sense I guess but with larger files there isn't any benefit, right? Expect some waste of drive space.

There is always some performance loss if logical block size is smaller than flash sector size, writing two 512 byte blocks may require flash to do read-erase-write sequence twice (or more) for same sector.

On the other hand most Amiga IDE devices are so slow that modern flash erase and write sequence usually is much faster than Amiga IDE so only waste is number of flash write cycles but it is practically irrelevant with Amigas.

[daxb]

Originally Posted by Toni Wilen

Buffers also require more RAM. Number of buffers * block size = required RAM.

I thought the value for one buffer for PFS3 is fixed to 1024 bytes (independent from block size). At least the old PFS3.guide says it. Is that false or only true for PFS3AIO?

There is always some performance loss if logical block size is smaller than flash sector size, writing two 512 byte blocks may require flash to do read-erase-write sequence twice (or more) for same sector.

But doesn't exists flash drives that are intelligent enough to handle this in an optimized way? But as you said, for usual Amiga use case it isn't much relevant. I fear that systems exists where you can measure a (small) performance difference and on other systems there is no difference. Needs some tests.

[apex]

Using OS3.1.4 I changed the blocksize to 4096 byte.
I made a quick format and got the requester PFS 19.2 will be used.

I have 2 problems, but they are may 3.1.4 based?
Workbench says blocksize is still 512 byte, see picture.
And filename lengt under WB with PFS seems to be limited to 30?
Yes I used setfnsize and I changed it in WBPrefs to 100.
Funny, DOpus 4.18 is working with longer names.

[arti]

I can't format disk 70GB with 3.1 from aminet.
I have 0x54445303 instead of 0x50445303 ?

[talybont]

Originally Posted by arti

I can't format disk 70GB with 3.1 from aminet.
I have 0x54445303 instead of 0x50445303 ?

You need to select PFS where it says "File System:" I think.

[arti]

There is no such option..

[arti]

With old version 53 , I can choose PDS/3.

[Storm]

It’s Custom File System with DOS Type of 0x50445303. I think that is the direct SCSI type and will show up as PDS\03

[Hedeon]

Originally Posted by arti

There is no such option..

When you load the FS to the RDB you can set identifier. It looks like you set DFS/03. Which is international FFS (I think) which is the default of HDToolBox.

[arti]

It finally appeared but when booting I get Guru 80000008.
I have tried 1K, 2K , and 4K block sizes.