scsi v43.46 beta need testing - Page 3

pandy71 · 01 February 2011, 22:10

I mean - A500, no internal Gayle IDE thus made similar (identical from software point of view) hardware as in A600/A1200 but with OWN ROM ie with Autoconfig use.
I know that replacing module in KICK is possible.

@gulliver - thx, immediately checking Your links.

PS
what is most CPU time consuming in IDE(scsi) driver - something that can be accelerated in hardware?

altcomputing · 01 February 2011, 22:43

I see no point in where it wouldn't be possible to LoadModule the scsi.device to RAM.

Quote:

Originally Posted by pandy71

I mean - A500, no internal Gayle IDE thus made similar (identical from software point of view) hardware as in A600/A1200 but with OWN ROM ie with Autoconfig use.
I know that replacing module in KICK is possible.

@gulliver - thx, immediately checking Your links.

PS
what is most CPU time consuming in IDE(scsi) driver - something that can be accelerated in hardware?

pandy71 · 01 February 2011, 22:59

there is no need to load scsi.device to RAM (can be in KICK or in board ROM)

gulliver · 01 February 2011, 23:20

@pandy71

Ide(scsi) is always a cpu hungry interface for an Amiga unless you get some third party ide interface that somehow aleviates that situation.
Your choice on a ide(scsi.device) version depends on your hardware (cpu, ide interface, Amiga model, storage media) and if you want to reach beyond the 4GB storage space limit.
With those details we may figure out what ide(scsi.deviced) best fits your goals.

Tumo007 · 02 February 2011, 00:44

Will this update work on an A4000 ?

Cosmos · 02 February 2011, 06:54

Ok, I'll rework the A4000 version...

pandy71 · 02 February 2011, 14:36

@gulliver so my question is - maybe something in hardware (something simple like byte swapping etc) can be implemented in hardware to reduce CPU use.

Now i have few A500 without IDE controller, simple controller like http://www.students.tut.fi/~leinone3/ide/ide68k.html can be made but without KICK 2.05 requirement ie self sufficient to use Autoconfig and KICK 1.3, maybe with few improvements on hardware to offload CPU from something trivial to implement in hardware but time consuming from CPU point of view, alternatively something like OCIDEC-3 http://opencores.org/project,ata implemented in Amiga, maybe with additional hardware like C2P etc.

blasterreal · 02 February 2011, 22:06

Apollo 1240 40MHZ 32mb
Smartfilesystem 1.279
Buffer 300
-----------------------------

Scsi.device 44.2 > 2.683.980 b/s
Scsi.device 43.46b2 > 3.276.800 b/s

Good work

Foul · 02 February 2011, 22:21

waouuu ! nice boost !

Cosmos · 03 February 2011, 10:40

scsi.device v43.46 (A600 & A1200) beta 3 (14 568 bytes)

- R_BeginIO optimised, a bit faster
- Read/write function a bit faster now
- Interrupt server optimised
- Some subroutines called one time inlined again
- 1372 bytes saved

NovaCoder · 18 February 2011, 02:34

Good work Cosmos, can't wait to try this out with my new ACA + SATA SSD

gibs · 18 February 2011, 10:30

This seriously kick ass !

altcomputing · 20 February 2011, 22:51

Quote:

Originally Posted by Cosmos

scsi.device v43.46 (A600 & A1200) beta 3 (14 568 bytes)

- R_BeginIO optimised, a bit faster
- Read/write function a bit faster now
- Interrupt server optimised
- Some subroutines called one time inlined again
- 1372 bytes saved

b1 is working fine with me.

b2 and b3 is making my DH1 and later partitions refuse to mount with "DHx:Unreadable" ghosted icons instead of them, even if some of them don't cross the 4GB boundary. I've had the same problem with scsi.device 44.x, tested on a few drives, only my old Kingston 4GB CF did work with it fine, other drives refused to mount any partition other than system (first) partition.

This may have something to do with the PFS workaround introduced in b2 or inlined code I believe, but I'm not sure. I use SFS.

I remember with 44.2 i tried both SFS and FFS and they didn't work. Didn't check that with your 43.46 b2 and b3, didn't also try PFS.

mfilos · 20 February 2011, 23:12

43.46b3 works just great with PFS3 on mine with no issues.
Also tried it with FFS and SFS and also works like a charm

Cosmos · 21 February 2011, 07:17

@altcomputing

Which version of HDToolBox is using ? The last 45.6 ??

altcomputing · 21 February 2011, 09:27

Quote:

Originally Posted by Cosmos

@altcomputing

Which version of HDToolBox is using ? The last 45.6 ??

Yes, version 45.6

NathanTolbert · 21 February 2011, 23:37

Okay tested it on my A600 which doesn't have a buffered IDE interface and as such I noticed one specific thing

ACA630 (maprom on) default dh1 - 1,931,841 bytes (measured by sysinfo)
dh2 - 2,131,412 bytes
ACA630 (maprom off) 43.46b3 dh1 - 1,451,371 bytes
dh2 - 1,842,962 bytes

so for the slower ACA630 (the 25MHz version) it seems definitely a good idea to map that file into a ROM image and use the maprom to run it from memory.
I'm going to try that Wednesday after my final.

Cosmos · 28 February 2011, 11:45

scsi.device v43.46 (A600 & A1200) beta 4 (14 292 bytes)

- Some subroutines called one time inlined again
- 1648 bytes saved

FOL · 28 February 2011, 13:26

Quote:

Originally Posted by Cosmos

scsi.device v43.46 (A600 & A1200) beta 4 (14 292 bytes)

- Some subroutines called one time inlined again
- 1648 bytes saved

Little confused, Im using 44.2, does your 43.46 include all the changes that 44.2 does?

altcomputing · 28 February 2011, 13:26

Any changes related to my problem posted above? Will test in the evening.

Quote:

Originally Posted by Cosmos

scsi.device v43.46 (A600 & A1200) beta 4 (14 292 bytes)

- Some subroutines called one time inlined again
- 1648 bytes saved

02 February 2011, 22:06	#48
blasterreal Registered User Join Date: Jan 2010 Location: turkey Posts: 30	Apollo 1240 40MHZ 32mb Smartfilesystem 1.279 Buffer 300 ----------------------------- Scsi.device 44.2 > 2.683.980 b/s Scsi.device 43.46b2 > 3.276.800 b/s Good work

03 February 2011, 10:40	#50
Cosmos Banned Join Date: Jan 2007 Location: France Posts: 655	scsi.device v43.46 (A600 & A1200) beta 3 (14 568 bytes) - R_BeginIO optimised, a bit faster - Read/write function a bit faster now - Interrupt server optimised - Some subroutines called one time inlined again - 1372 bytes saved Last edited by Cosmos; 28 February 2011 at 11:46.

20 February 2011, 23:12	#54
mfilos Paranoid Amigoid Join Date: Mar 2008 Location: Athens/Greece Age: 45 Posts: 1,978	43.46b3 works just great with PFS3 on mine with no issues. Also tried it with FFS and SFS and also works like a charm Last edited by mfilos; 21 February 2011 at 08:40.

28 February 2011, 11:45	#58
Cosmos Banned Join Date: Jan 2007 Location: France Posts: 655	scsi.device v43.46 (A600 & A1200) beta 4 (14 292 bytes) - Some subroutines called one time inlined again - 1648 bytes saved Last edited by Cosmos; 17 June 2011 at 06:31.

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
SetPatch v43.6 non-beta?	Photon	request.Apps	34	02 November 2013 06:49
scsi v43.47 beta need testing	Cosmos	Coders. General	3	22 October 2011 22:18
Planet Retro Beta Testing	jimqode	News	0	30 July 2006 20:22

01 February 2011, 22:10	#41
pandy71 Registered User Join Date: Jun 2010 Location: PL? Posts: 2,810	I mean - A500, no internal Gayle IDE thus made similar (identical from software point of view) hardware as in A600/A1200 but with OWN ROM ie with Autoconfig use. I know that replacing module in KICK is possible. @gulliver - thx, immediately checking Your links. PS what is most CPU time consuming in IDE(scsi) driver - something that can be accelerated in hardware?

01 February 2011, 22:59	#43
pandy71 Registered User Join Date: Jun 2010 Location: PL? Posts: 2,810	there is no need to load scsi.device to RAM (can be in KICK or in board ROM)

01 February 2011, 23:20	#44
gulliver BoingBagged Join Date: Aug 2007 Location: The South of nowhere Age: 46 Posts: 2,358	@pandy71 Ide(scsi) is always a cpu hungry interface for an Amiga unless you get some third party ide interface that somehow aleviates that situation. Your choice on a ide(scsi.device) version depends on your hardware (cpu, ide interface, Amiga model, storage media) and if you want to reach beyond the 4GB storage space limit. With those details we may figure out what ide(scsi.deviced) best fits your goals.

02 February 2011, 00:44	#45
Tumo007 Registered User Join Date: May 2007 Location: Montreal, CANADA Posts: 56	Will this update work on an A4000 ?

02 February 2011, 06:54	#46
Cosmos Banned Join Date: Jan 2007 Location: France Posts: 655	Ok, I'll rework the A4000 version...

02 February 2011, 14:36	#47
pandy71 Registered User Join Date: Jun 2010 Location: PL? Posts: 2,810	@gulliver so my question is - maybe something in hardware (something simple like byte swapping etc) can be implemented in hardware to reduce CPU use. Now i have few A500 without IDE controller, simple controller like http://www.students.tut.fi/~leinone3/ide/ide68k.html can be made but without KICK 2.05 requirement ie self sufficient to use Autoconfig and KICK 1.3, maybe with few improvements on hardware to offload CPU from something trivial to implement in hardware but time consuming from CPU point of view, alternatively something like OCIDEC-3 http://opencores.org/project,ata implemented in Amiga, maybe with additional hardware like C2P etc.

02 February 2011, 22:21	#49
Foul Registered User Join Date: Jun 2009 Location: Perigueux/France Age: 49 Posts: 1,516	waouuu ! nice boost !

18 February 2011, 02:34	#51
NovaCoder Registered User Join Date: Sep 2007 Location: Melbourne/Australia Posts: 4,405	Good work Cosmos, can't wait to try this out with my new ACA + SATA SSD

18 February 2011, 10:30	#52
gibs Banned Join Date: Jan 2010 Location: france Posts: 932	This seriously kick ass !

21 February 2011, 07:17	#55
Cosmos Banned Join Date: Jan 2007 Location: France Posts: 655	@altcomputing Which version of HDToolBox is using ? The last 45.6 ??

21 February 2011, 23:37	#57
NathanTolbert Users Awaiting Email Confirmation Join Date: Jun 2009 Location: Olathe, Kansas Posts: 42	Okay tested it on my A600 which doesn't have a buffered IDE interface and as such I noticed one specific thing ACA630 (maprom on) default dh1 - 1,931,841 bytes (measured by sysinfo) dh2 - 2,131,412 bytes ACA630 (maprom off) 43.46b3 dh1 - 1,451,371 bytes dh2 - 1,842,962 bytes so for the slower ACA630 (the 25MHz version) it seems definitely a good idea to map that file into a ROM image and use the maprom to run it from memory. I'm going to try that Wednesday after my final.

Currently Active Users Viewing This Thread: 1 (0 members and 1 guests)