PFS3AIO Questions

[Toni Wilen]

Originally Posted by Scyphe

Okay. I just tried the muPFS3-020DS (18.5) and noticed no difference. I guess your version is actually containing newer code since you included all the SVN updates. Just trying out various combinations to find the fastest version for my IDE-CF (SanDisk Ultra 4GB 30MB/s) which on my stock A1200 gives 1020-1080KB/s ie. around 1MB/s.

I sold my M-Tec T1230/42MHz accelerator with 4MB Fastram and is waiting for my Blizzard MK-IV 50MHz to show up. Speed was 2MB/s with the old accelerator so it seems disk transfer speed is connected with having Fastram and perhaps a fast 030 (with MMU and FPU).

I'll go back to PFS3_aio-handler. Btw, is it auto-sensing what kind of CPU/system it's using + Identifier (50445303 = PDS3 to get DirectSCSI on A1200 IDE) or is it using a generic 68k driver for all systems whether 68000 or 68060? I gather from the very name of the file (All In One) that it's auto-sensing during the installation of the filesystem in combination with the identifier... ?

Filesystem makes no difference on raw transfer speed (unless it has bad algorithms), limit is always Amiga's very slow IDE port (or accelerator having slow path to IDE port address range).

Note that raw transfer benchmarks also bypass filesystems and access the drive directly, you would get exact same results from unformatted partitions too

PFS3AIO is compatible with all 680x0 CPUs. 68020+ only version may increase speed by 1-2% or so which would not worth the trouble and sometimes it seems GCC makes better code if it is compiled for 68000... (and it still would not have any effect on raw transfer speeds)

AIO = all IO commands supported and best option autodetected (standard 32-bit IO, TD64, NSD and DirectSCSI) and it is all AOS version compatible, including KS 1.x. Check the readme.txt for more details.

[Scyphe]

Okay. At least my old accelerator (M-Tec T1230) doubled the raw speed. The much better quality (imho the best accelerator/ram-board for A1200 ever) Blizzard MK-IV should bring back the 2MB+/s raw transfer speed.

"68020+ only version may increase speed by 1-2% or so which would not worth the trouble and sometimes it seems GCC makes better code if it is compiled for 68000... (and it still would not have any effect on raw transfer speeds)"

How about compiling for different CPUs and release them as a package for the user to decide which one? Perhaps support for an MMU or an FPU for those with the hardware can be used to improve the effectiveness? Perhaps I should be looking at getting FastATA MK-IV instead if I want more speed.

[Bamiga2002]

Does it make difference to use 0x50445303 / 0x50465303 identifier or does the autosensing feature take care of this?

[Toni Wilen]

Originally Posted by Bamiga2002

Does it make difference to use 0x50445303 / 0x50465303 identifier or does the autosensing feature take care of this?

It changes priority of autodetection. If DirectSCSI dostype, DirectSCSI is tested first (and if test passes, other tests are skipped and DS is used). If any other dostype: DirectSCSI is tested after TD64 and NSD tests, DS is only used if both TD64 and NSD are unsupported.

If all 3 tests fail and partition crosses 4G border (or is completely outside of 4G), partition is not mounted. If partition is completely inside first 4G, standard CMD io commands are used and partition is mounted.

[tomse]

I've had PFS3 to work when it was first released as open source.

using latest version of pfs3aio (2014-04-18)

I'm trying to make 2 partitions in the beginning of the drive each around 336megs in size.

The 250GB is attached to the SCSI port of the CSPPC
I've got a 1GB CF attached to the IDE port of the A4000 (OS31 + recovery tools)

Using HDInstTools, Adding filesystem: the PFS3 filesystem comes out as DOS/3 v18.5
if I change this manually to PDS/3 or PFS/3
I can't change the dostype of a partition to custom, and PDS/3 / PFS/3 won't show in the in the menus that have DOS/0, DOS/1 etc.
-- I can't manually enter PDS/3 or PFS/3 in custom, it just won't accept my input, and the field is empty after I hit return.

If I have no other filesystem in the RDB, but PFS3, and PFS3 is DOS/3, and the partition is DOS/3 I can't format the partition with pfsformat.. the tool seems to work, as in it's actually formatting, but the result is an NDOS drive.

Though using HDToolBox (os31) and I edit the identifiers in the filesystem, it returns version 18.0 and I can add that PDS/3 filesystem to the partition. (it actually shows up in the list)
(I took one of the 2 identifiers mentioned above 50445303, this one was easier to remember when walking to my Amiga, so that is the background of the choice used)

my guess is that somehow the automatic something something doesn't do it's automatic something something.

@toni...
I'm willing to do alot of testing here if required.

[Toni Wilen]

Those are hdinsttools issues. I have never used hdinsttools, sorry.

There is no header(s) or anything like that in filesystem handler executable that partitioning program could use to guess type or attributes of filesystem (unfortunately). It is just a plain executable. They can only know between "possibly FFS" and "most likely not FFS"

[tomse]

No worries then ;-)

Id like to give thanks for regularly updating it as well