[comp.sys.amiga] Hurricane 68020 & O/S speedup

peter@dalcsug.UUCP (Peter Philip) (01/13/88)

I'm posting this for a friend who dosn't have access to the net :

---------------
  Is anyone familiar with Finally Techonologies' "Hurricane" 68020/68881 board
as mentioned in the Jan. 88 issue of AmigaWorld?  Specifically, is it 
available, and how does it compare with the CSA 68020/68881 board? (A1000)
One important question is how many wait states on their 2 Meg board?  Could
it be static RAM?

  On a related note:

  An interesting technique used to speed up accesses to the ROM
BIOS in the current crop of IBM compatible 80386 machines is to copy
the BIOS to 32 bit memory and reroute the system calls.  Even for a
relatively primitive OS like PC-DOS the speedup is quite significant.
For a machine like the Amiga, which interacts with almost all hardware and 
software (IE: multitasking overhead) through system routines, the speed gained
by moving KS/ROM routines to 32 bit/fast memory vs 16 bit could be very large.
  Just how feasable is this?  I know you lose 192K or maybe 256K, but it might
be worth it, at least as an option.
  I guess you would copy the memory directly, and then change the ROMtags.
  The only problem (other than implementation) that I can think of is that some
of the code in KS/ROM must be in a location that is accessable to the
coprocessors (e.g., like the graphics data, etc. restricted to chip memory);  
but I don't think this is true.  After all, the whole OS uses relative
addressing, right?
  Thanks for any responses.

  R. Andrew MacRae
  Dept. of Geology (Student)
  Dalhousie U., Halifax, N.S.
  Canada
---------------

Respond to the net and I will forward all replies to Andrew.
-Peter Philip 

ewhac@well.UUCP (Leo 'Bols Ewhac' Schwab) (01/14/88)

In article <271@dalcsug.UUCP> peter@dalcsug.UUCP (R. Andrew MacRae) writes:
>  Is anyone familiar with Finally Techonologies' "Hurricane" 68020/68881 board
>as mentioned in the Jan. 88 issue of AmigaWorld?  Specifically, is it 
>available, and how does it compare with the CSA 68020/68881 board? (A1000)

	I've got one (which, if you read a previous posting of mine, is the
aforementioned slightly-nefariously-acquired 68020/68881 board).  It works
wonderfully.  I don't know how it compares with CSA's unit performance-wise,
but I suspect that they are roughly equivalent.

	The Hurricane board also has the added advantage of plugging
directly inside your A1000 (replacing the 68000), as opposed to CSA's which
is a seperate card for their Turbo Highrise.

>One important question is how many wait states on their 2 Meg board?  Could
>it be static RAM?
>
	Judging from prices for Finally's 32-bit RAM as compared to CSA's
prices, I suspect that either Finally's RAM is dynamic RAM (possibly with
wait states), or that CSA is gouging.  I think it a bit of both.  I have no
facts, though (I keep meaning to call them).

>For a machine like the Amiga, which interacts with almost all hardware and 
>software (IE: multitasking overhead) through system routines, the speed gained
>by moving KS/ROM routines to 32 bit/fast memory vs 16 bit could be very large.
>  Just how feasable is this?  I know you lose 192K or maybe 256K, but it might
>be worth it, at least as an option.
>  I guess you would copy the memory directly, and then change the ROMtags.

	I just sent a letter to Dale asking why ExecBase isn't moved into
Autoconfig RAM after boot up.  He said most of the system is operational by
the time Exec gets around to Autoconfiging stuff, and thus ExecBase is too
difficult to move by that time.  (Actually, I wouldn't mind spending a week
of my time banging on Exec to get it to move ExecBase out of CHIP RAM; I'm
not doing anything else at the moment.)

	The point of this is that, if ExecBase is hard to move, the whole of
KickStart is probably next to impossible.  Nevertheless, I agree it would be
a real win to have KickStart in 32-bit RAM, if it was available.

>After all, the whole OS uses relative addressing, right?

	Somehow, I don't think so....

	Am I being overly pessimistic?

_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_
Leo L. Schwab -- The Guy in The Cape	ihnp4!ptsfa -\
 \_ -_		Recumbent Bikes:	      dual ---> !{well,unicom}!ewhac
O----^o	      The Only Way To Fly.	      hplabs / (pronounced "AE-wack")
"Work FOR?  I don't work FOR anybody!  I'm just having fun."  -- The Doctor

bryce@hoser.berkeley.edu (Bryce Nesbitt) (01/15/88)

In article <4989@well.UUCP> ewhac@well.UUCP (Leo 'Bols Ewhac' Schwab) writes:
>In article <271@dalcsug.UUCP> peter@dalcsug.UUCP (R. Andrew MacRae) writes:
>
>	I just sent a letter to Dale asking why ExecBase isn't moved into
>Autoconfig RAM after boot up.  He said most of the system is operational by
>the time Exec gets around to Autoconfiging stuff, and thus ExecBase is too
>difficult to move by that time.

HA!  Hate to contradict Mr. Luck, but I've hacked that area of the rom
into submission.  The bugs in that section kept taking out my recoverable
ram disk.  This is now fixed; and I don't read from any write-only chip
registers.

Both execbase and THE SUPERVISOR STACK end up in chip on a non-$C00000
machine.  Regardless, I'm confident that they could be moved to normal
autoconfig memory with manageable hassle.

Since I don't have "normal autoconfig" memory, I have no 
compelling reason to do this.  I am available for rent, however. :-)


>	The point of this is that, if ExecBase is hard to move, the whole of
>KickStart is probably next to impossible. 

Nah, you do it in hardware.  Just set your 32bit memory to respond to
the proper address range.


	Wait!  Sniff the air?  Am I just blowing smoke?  Hope not!

|\ /|  . Ack! (NAK, SOH, EOT)
{o O} . bryce@hoser.berkeley.EDU -or- ucbvax!hoser!bryce (or try "cogsci")
 (")
  U	"Your theory is crazy... but not crazy enought to be true." -Niels Bohr

daveh@cbmvax.UUCP (Dave Haynie) (01/15/88)

in article <271@dalcsug.UUCP>, peter@dalcsug.UUCP (Peter Philip) says:
> Keywords: 68020 AmigaDOS ROM

>   An interesting technique used to speed up accesses to the ROM
> BIOS in the current crop of IBM compatible 80386 machines is to copy
> the BIOS to 32 bit memory and reroute the system calls.  Even for a
> relatively primitive OS like PC-DOS the speedup is quite significant.
> For a machine like the Amiga, which interacts with almost all hardware and 
> software (IE: multitasking overhead) through system routines, the speed gained
> by moving KS/ROM routines to 32 bit/fast memory vs 16 bit could be very large.
>   Just how feasable is this?  I know you lose 192K or maybe 256K, but it might
> be worth it, at least as an option.

Buy a Commodore-Amiga A2620 board if you can wait a little longer.  At least
for A2000s.  This board has an MMU on it.  Among other things an MMU can do
for you, it can translate addresses.  So you allocate an appropriate 256K
block of memory.  Copy ROM into it.  Set up the MMU to translate that chunk
of memory to the normal ROM addresses.  You now have the OS in RAM, and no
problems with relocating the ROM or such things.  <poof>

> After all, the whole OS uses relative addressing, right?

Wrong.  Object files contain relocation information used by the loader to
relocate.  Once located, they aren't necessarily using relative addressing.
The problem with relative addressing on the 68000 is it limits you to
16 bits relative.

>   R. Andrew MacRae
> Respond to the net and I will forward all replies to Andrew.
> -Peter Philip 
-- 
Dave Haynie  "The B2000 Guy"     Commodore-Amiga  "The Crew That Never Rests"
   {ihnp4|uunet|rutgers}!cbmvax!daveh      PLINK: D-DAVE H     BIX: hazy
		"I can't relax, 'cause I'm a Boinger!"

sean@ms.uky.edu (Sean Casey) (01/16/88)

In article <3146@cbmvax.UUCP> daveh@cbmvax.UUCP (Dave Haynie) writes:
>Buy a Commodore-Amiga A2620 board if you can wait a little longer.  At least
>for A2000s.  This board has an MMU on it.  Among other things an MMU can do
>for you, it can translate addresses.  So you allocate an appropriate 256K
>block of memory.  Copy ROM into it.  Set up the MMU to translate that chunk
>of memory to the normal ROM addresses.  You now have the OS in RAM, and no
>problems with relocating the ROM or such things.  <poof>

Hey, CATS, is CA going to make a 68020 + MMU board for the A1000?  If
a third party can do it, CA can do it.

Sean
-- 
--  Sean Casey               sean@ms.uky.edu,  sean@ukma.bitneT
--  (the Empire guy)         {rutgers,uunet,cbosgd}!ukma!sean
--  University of Kentucky in Lexington Kentucky, USA
--  "If something can go will, it wrong."

blgardne@esunix.UUCP (Blaine Gardner) (01/17/88)

in article <3146@cbmvax.UUCP>, daveh@cbmvax.UUCP (Dave Haynie) says:
> in article <271@dalcsug.UUCP>, peter@dalcsug.UUCP (Peter Philip) says:

[Stuff about moving Kickstart to 32 bit RAM]

>> Just how feasable is this?  I know you lose 192K or maybe 256K, but it might
>> be worth it, at least as an option.
 
> Buy a Commodore-Amiga A2620 board if you can wait a little longer.  At least

I'd love to! How long, how much?

Should I hold my breath, or just start putting pennies in the piggy
bank? I know that price/delivery is not your department Dave, but an
estimate +- $100, +- 1 month would be helpful.

> for A2000s.  This board has an MMU on it.  Among other things an MMU can do
> for you, it can translate addresses.  So you allocate an appropriate 256K
> block of memory.  Copy ROM into it.  Set up the MMU to translate that chunk
> of memory to the normal ROM addresses.  You now have the OS in RAM, and no
> problems with relocating the ROM or such things.  <poof>

Sounds nice, but will Commodore support this, or will it be hackers
only?  One question though, will the wait state from the MMU cancel the
speed gain from the 32 bit RAM?
-- 
Blaine Gardner @ Evans & Sutherland    540 Arapeen Drive, SLC, Utah 84108
UUCP Addresses:  {ihnp4,ucbvax,allegra,decvax}!decwrl!esunix!blgardne
        	 ihnp4!utah-cs!esunix!blgardne        usna!esunix!blgardne
"Nobody will ever need more than 64K."    "Nobody needs multitasking on a PC."

spencer@eris (Randal m. Spencer [RmS]) (01/20/88)

Recently on *comp.sys.amiga* daveh@cbmvax.UUCP (Dave Haynie) wrote:
...in article <271@dalcsug.UUCP>, peter@dalcsug.UUCP (Peter Philip) says:
...> Keywords: 68020 AmigaDOS ROM
...
...>   An interesting technique used to speed up accesses to the ROM
...> BIOS in the current crop of IBM compatible 80386 machines is to copy
...> the BIOS to 32 bit memory and reroute the system calls.
...
...Buy a Commodore-Amiga A2620 board if you can wait a little longer.

...> -Peter Philip 
...Dave Haynie  "The B2000 Guy"     Commodore-Amiga  "The Crew That Never Rests"

Actually, I was looking at the new 68030 board that CSA was showing
at AmiExpo in LA this weekend, and there was this rumor that went
around the show that when 1.3 (1.21, whatever) comes out, they are
going to put it in 32 bit-wide rom with their board.  Not that I work
for them or anything....
-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
Randy Spencer      P.O. Box 4542   Berkeley  CA  94704        (415)222-7595 
spencer@mica.berkeley.edu        I N F I N I T Y         BBS: (415)222-9416
..ucbvax!mica!spencer            s o f t w a r e                  AAA-WH1M
-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-

daveh@cbmvax.UUCP (Dave Haynie) (01/21/88)

in article <6615@agate.BERKELEY.EDU>, spencer@eris (Randal m. Spencer [RmS]) says:
> 
> Recently on *comp.sys.amiga* daveh@cbmvax.UUCP (Dave Haynie) wrote:
> ...Buy a Commodore-Amiga A2620 board if you can wait a little longer.

> Actually, I was looking at the new 68030 board that CSA was showing
> at AmiExpo in LA this weekend, and there was this rumor that went
> around the show that when 1.3 (1.21, whatever) comes out, they are
> going to put it in 32 bit-wide rom with their board.  Not that I work
> for them or anything....

That sounds a bit silly.  You COULD build in fast 32 bit ROMs if you like,
but why not use a small chunk of your 32 bit RAM.  The 68030 has it's own
built-in MMU, which can do the same things that the C-A board's MMU can
do.  Like remapping memory.  The 32 bit ROM would make it simpler, if you
want to pay for extra circuitry, board real-estate, and the cost of very
fast ROMs.  The RAM idea is a freebie.

In either case, moving the OS into FAST 32 bit memory of some kind is a very
good idea.

> Randy Spencer      P.O. Box 4542   Berkeley  CA  94704        (415)222-7595 
-- 
Dave Haynie  "The B2000 Guy"     Commodore-Amiga  "The Crew That Never Rests"
   {ihnp4|uunet|rutgers}!cbmvax!daveh      PLINK: D-DAVE H     BIX: hazy
		"I can't relax, 'cause I'm a Boinger!"