Welcome, Guest. Please login or register.

Author Topic: hyperthreading  (Read 2236 times)

Description:

0 Members and 1 Guest are viewing this topic.

Offline Zac67

  • Hero Member
  • *****
  • Join Date: Nov 2004
  • Posts: 2890
    • Show only replies by Zac67
Re: hyperthreading
« Reply #29 from previous page: February 17, 2007, 09:28:00 AM »
Yes, that pretty much matches my memory, too.  :idea:
 

Offline Azryl

  • Sr. Member
  • ****
  • Join Date: Jul 2004
  • Posts: 395
    • Show only replies by Azryl
    • http://www.azryl.com
Re: hyperthreading
« Reply #30 on: February 17, 2007, 10:14:45 AM »
I will wade in here with a few rambles of my own

Guys... go look at the chip die used to manufacture the motorola 68040 and later 68060

now check the manufacturing die for the 386 and 486 chips

Motorola built beautiful chips while the x86 were nothing but horrible cludges with mixed memory models and old instructions left over from previous chips for backward compatability. They look ugly!


To then say in the same breath that the new x86 chips have more horsepower than any old 68060 chip. Again check the manufacturers die, the design of the x86 chip from pentium onwards reflects an amazing amount of design features/layout and functionality with the motorola 68060 and the modern PPC chips.

Basically, all the newer x86 chips are a risc style core running emulated old microcode instructions thru multiple pipelines wrapped inside caches to maximise speed.

How did windows achieve its multitasking... oh yeah, using a flat memory architecture which was not available before the 386 chip.. but amazingly the 68000 processor line had this from conception. So did the MOS 6502 where as Intel seemed to be off on its own as far as memory design and implimentation was concerned.

If you want to compare apple to oranges, check whats inside. It does make a world of difference.

Thats the chip hardware. Software.... well


Simply... Amiga OS is a micro kernel. It does amazingly well with few resources. But you have to write smarter, better constructed programs. Harder to do, even harder to debug.

Windows/OS2/Unix/MacOS/Linux's are macro kernels. Control all the resources and parcel out whats needed for the programs. Has memory protection, programs dont have to be as smart or as aware of the OS requirements. Easier to program and debug.

Basically.. micro kernel is one cop looking after a sandbox full of kids. Keeps moderate control but cant keep on eye on every kid all the time.

Macro kernel is a sandbox with cops lined up all around the edge watching the kids inside. Any kid get out of hand, the cops pounce and eject for bad behaviour.

Come on people, dont just sit there sprouting more and more drivel about something from 1992 or so.. dig deeper, check out what came before, what was the normal, what could a person walk into a shop and buy. Then that person take home and actually use!

Check 60's then the 70's then the 80's before you amazingly bore us all with your opinion of what now is better!

You cannot compare two totally different computer systems with kernel philospohy's from decades apart. Its just not fair

Every computer will do its job at its maximum efficiency until it physically fails.

Flame me, bore me... geez just get informed before you do either

Az



Completely useless? I can always be used as a bad example  :lol:
 

Offline Karlos

  • Sockologist
  • Global Moderator
  • Hero Member
  • *****
  • Join Date: Nov 2002
  • Posts: 16867
  • Country: gb
  • Thanked: 4 times
    • Show only replies by Karlos
Re: hyperthreading
« Reply #31 on: February 17, 2007, 10:55:59 AM »
I think you'll find it's the OP who was attempting to compare a CPU over a decade old with more recent technology and others (like myself) pointing out it was a complete fallacy to do so.

Don't get me wrong on this, I've always preferred the 680x0 architecture to the x86, I also prefer the PPC architecture to the x86, but these preferences are basically moot in todays world. Current x86 technology is absolutely nothing like it was a few years ago. You simply cannot compare 68060 performance, even at a hypothetical normalized clock rate, to something like the Core2 Duo.

It is the AmigaOS exec kernel that makes OS3.x (and 4.0) and Quark (I think it's called) in MOS and whatever underpins AROS that makes these systems so responsive under normal use.

Note that "normal" use refers to a system simply spending its time waiting for events to happen and processing them promptly when they arrive.

The very instant you give a system a compute bound task, the CPU performance becomes the dominant factor and then you will see why it is simply insane to attempt a comparison of 68060, x86 and PPC.
int p; // A
 

Offline bloodline

  • Master Sock Abuser
  • Hero Member
  • *****
  • Join Date: Mar 2002
  • Posts: 12113
    • Show only replies by bloodline
    • http://www.troubled-mind.com
Re: hyperthreading
« Reply #32 on: February 17, 2007, 11:00:34 AM »
Quote

Azryl wrote:
I will wade in here with a few rambles of my own

Guys... go look at the chip die used to manufacture the motorola 68040 and later 68060

now check the manufacturing die for the 386 and 486 chips


All I can see are a lot of very slight engravings on a shiny lump od silicon... nothing different bwteen them really.

Quote

Motorola built beautiful chips while the x86 were nothing but horrible cludges with mixed memory models and old instructions left over from previous chips for backward compatability. They look ugly!


intel, with a large installed customer base, decided that backward compatibility was important. Motorola start a fresh with the 68000... the 6800 was never a popular as the 6502... so there was less incentive to remain compatible.

Quote

To then say in the same breath that the new x86 chips have more horsepower than any old 68060 chip. Again check the manufacturers die, the design of the x86 chip from pentium onwards reflects an amazing amount of design features/layout and functionality with the motorola 68060 and the modern PPC chips.


You might have to reword this paragrah, I'm having trouble reading it.

Quote


Basically, all the newer x86 chips are a risc style core running emulated old microcode instructions thru multiple pipelines wrapped inside caches to maximise speed.


You do realise that's how the 68060 worked, right?

The "RICS style Core" is actually a simplified x86... optimised to execute commonly used instructions and without all the addressing modes common to CISC CPUs.

Microcode is rarely used on a modern x86, except on very complex often unused instructions.

Quote

How did windows achieve its multitasking... oh yeah, using a flat memory architecture which was not available before the 386 chip.. but amazingly the 68000 processor line had this from conception. So did the MOS 6502 where as Intel seemed to be off on its own as far as memory design and implimentation was concerned.


intel had backwards compatibility to think about, which seems like that was a design win given the fact almost all computers now use them. I think we can all agree that the 286 was a horrible design... but so what, intel fixed their mistakes with the 386.

The 386 was so good, intel were still manufacturing it upto the early part of this year! Motorola abandoned the 68000 long before that.

Quote

If you want to compare apple to oranges, check whats inside. It does make a world of difference.

Thats the chip hardware. Software.... well


Simply... Amiga OS is a micro kernel. It does amazingly well with few resources. But you have to write smarter, better constructed programs. Harder to do, even harder to debug.


AmigaOS isn't a micro kernel for the simple reason there is no distinction between kernel space and user space... AmigaOS is a very strange beast, designed around a number of hardware limitations and with little thought to the future. It's quite brilliant from an academic point of view and thats why I love to study it.

Quote

Windows/OS2/Unix/MacOS/Linux's are macro kernels. Control all the resources and parcel out whats needed for the programs. Has memory protection, programs dont have to be as smart or as aware of the OS requirements. Easier to program and debug.


Windows and MacOS X are hybrid kernels, with micro kernel and monolithic kernel features, to get the best of both worlds. OSX is build upon Mach 3.0 which was one of the first microkernels ever and defined pretty much everything we consider a microkernel today.

Unix and Linux are esentialy the same thing and both are Monolithic.

MacOS (Pre OSX) was a nasty kludge.

I know nothing about OS/2.

Quote

Basically.. micro kernel is one cop looking after a sandbox full of kids. Keeps moderate control but cant keep on eye on every kid all the time.

Macro kernel is a sandbox with cops lined up all around the edge watching the kids inside. Any kid get out of hand, the cops pounce and eject for bad behaviour.


No, AmigaOS is just a sandbox full of kids... who can do what ever they want regardless of each other.
The Other OS's with memory protection have the police officer, to look after them.

Quote

Come on people, dont just sit there sprouting more and more drivel about something from 1992 or so.. dig deeper, check out what came before, what was the normal, what could a person walk into a shop and buy. Then that person take home and actually use!


Yes, I am well aware of what I could buy in 1992.

Quote

Check 60's then the 70's then the 80's before you amazingly bore us all with your opinion of what now is better!

You cannot compare two totally different computer systems with kernel philospohy's from decades apart. Its just not fair


It wouldn't be fair if the illinformed poster didn't sit there and try and tell us that an 060 was faster than a Pentium4 when the 060 is running AmigaOS... it's just not true.

The Above has been edited to stay inline with what Karlos said. As he was more accurate.

Quote

Every computer will do its job at its maximum efficiency until it physically fails.

Flame me, bore me... geez just get informed before you do either

Az



How about I just inform you?

Offline bloodline

  • Master Sock Abuser
  • Hero Member
  • *****
  • Join Date: Mar 2002
  • Posts: 12113
    • Show only replies by bloodline
    • http://www.troubled-mind.com
Re: hyperthreading
« Reply #33 on: February 17, 2007, 11:11:48 AM »
Quote

Karlos wrote:

...and whatever underpins AROS...


That would be a clone of the exec.library 39.xx :-D

Offline Karlos

  • Sockologist
  • Global Moderator
  • Hero Member
  • *****
  • Join Date: Nov 2002
  • Posts: 16867
  • Country: gb
  • Thanked: 4 times
    • Show only replies by Karlos
Re: hyperthreading
« Reply #34 on: February 17, 2007, 11:31:21 AM »
Quote

bloodline wrote:
Quote

Karlos wrote:

...and whatever underpins AROS...


That would be a clone of the exec.library 39.xx :-D


And that'll be why it's nice and nippy.
int p; // A
 

Offline bloodline

  • Master Sock Abuser
  • Hero Member
  • *****
  • Join Date: Mar 2002
  • Posts: 12113
    • Show only replies by bloodline
    • http://www.troubled-mind.com
Re: hyperthreading
« Reply #35 on: February 17, 2007, 11:42:51 AM »
Quote

Karlos wrote:
Quote

bloodline wrote:
Quote

Karlos wrote:

...and whatever underpins AROS...


That would be a clone of the exec.library 39.xx :-D


And that'll be why it's nice and nippy.


Nippy, but as stable as a three legged horse playing twister in a force 7 gale... ;-)

Offline Karlos

  • Sockologist
  • Global Moderator
  • Hero Member
  • *****
  • Join Date: Nov 2002
  • Posts: 16867
  • Country: gb
  • Thanked: 4 times
    • Show only replies by Karlos
Re: hyperthreading
« Reply #36 on: February 17, 2007, 11:46:10 AM »
:lol:

Now that is a phrase I *have* to use at work next week to summarise my analysis of one of our legacy systems :-D
int p; // A
 

Offline bloodline

  • Master Sock Abuser
  • Hero Member
  • *****
  • Join Date: Mar 2002
  • Posts: 12113
    • Show only replies by bloodline
    • http://www.troubled-mind.com
Re: hyperthreading
« Reply #37 on: February 17, 2007, 11:51:31 AM »
Quote

Karlos wrote:
:lol:

Now that is a phrase I *have* to use at work next week to summarise my analysis of one of our legacy systems :-D


Hehehe,  yeah, AROS isn't actualy that bad... but I needed to use the phrase :-)

Offline bloodline

  • Master Sock Abuser
  • Hero Member
  • *****
  • Join Date: Mar 2002
  • Posts: 12113
    • Show only replies by bloodline
    • http://www.troubled-mind.com
Re: hyperthreading
« Reply #38 on: February 17, 2007, 12:35:56 PM »
Quote

Karlos wrote:

Don't get me wrong on this, I've always preferred the 680x0 architecture to the x86, I also prefer the PPC architecture to the x86, but these preferences are basically moot in todays world.


Actually, the x86 (i386) is really similar to the the 68000... True the syntax of both NASM and GAS suck when compared to Motorola syntax, but once you've set the GDT so that both the data and code segments fill the same 32bit address space... you won't really find much difference, except only having 8 gerneal purpose regs (fixed in AMD64, though I've not tried programing in long mode yet...) which is not so hard as the x86 has lots of stack manipulation instructions. The x86 has so much documentation and example code, you can pick it up in a a very short space of time.

PPC on the other hand is totally alien... I tried and tried about 8 or 9 years ago to get my head around it... but it just doesn't make sense to me.

Offline DonnyEMU

  • Hero Member
  • *****
  • Join Date: Sep 2002
  • Posts: 650
    • Show only replies by DonnyEMU
    • http://blog.donburnett.com
Re: hyperthreading
« Reply #39 on: February 17, 2007, 12:52:05 PM »
I'd also like to point out that someone earlier talked about VGA standards vs the Amiga. Remember that most VGA graphics cards in those days meant a 256 color 320x200 display and still a 16 color 640x480 mode. In other words an 8-bit graphics display. We didn't really have a "bitblit" command or a real blitter or anything like a "BIMMER" until at least 1994-1995 (Windows 95), until then things like off-screen writing wasn't even possible in windows on most graphics cards and drivers.. I know this because I used to write software that required this and we got stung Christmas 1994 by people returning software that didn't run on their windows box because they hadn't updated their graphics card driver..

Even until 1994 Amiga had the high end of things functionally until then.
======================================
Don Burnett Developer
http://blog.donburnett.com
don@donburnett.com
======================================
 

Offline bloodline

  • Master Sock Abuser
  • Hero Member
  • *****
  • Join Date: Mar 2002
  • Posts: 12113
    • Show only replies by bloodline
    • http://www.troubled-mind.com
Re: hyperthreading
« Reply #40 on: February 17, 2007, 12:57:50 PM »
Quote

DonnyEMU wrote:
I'd also like to point out that someone earlier talked about VGA standards vs the Amiga. Remember that most VGA graphics cards in those days meant a 256 color 320x200 display and still a 16 color 640x480 mode. In other words an 8-bit graphics display. We didn't really have a "bitblit" command or a real blitter or anything like a "BIMMER" until at least 1994-1995 (Windows 95), until then things like off-screen writing wasn't even possible in windows on most graphics cards and drivers.. I know this because I used to write software that required this and we got stung Christmas 1994 by people returning software that didn't run on their windows box because they hadn't updated their graphics card driver..

Even until 1994 Amiga had the high end of things functionally until then.



Commodore could have made a fotrune selling Amiga Chipset on a Multimedia board for the PC...

-Edit- I'm pretty sure a friend of mine had showed me an SVGA flight sim on his PC in 1992... it might have been later though...

Offline SamuraiCrow

  • Hero Member
  • *****
  • Join Date: Feb 2002
  • Posts: 2280
  • Country: us
  • Gender: Male
    • Show only replies by SamuraiCrow
Re: hyperthreading
« Reply #41 on: February 17, 2007, 02:18:20 PM »
There was a third party extension for DOS that had up to 800x600x256 but it had no blitter (nor copper for that matter  :lol: ).  I can't think of the acronym that it went by but it was pretty-much useless.  There were a few "Windows accelerated" graphics chips that worked with Windows 3.1, however, that had primative blitters.
 

Offline Zac67

  • Hero Member
  • *****
  • Join Date: Nov 2004
  • Posts: 2890
    • Show only replies by Zac67
Re: hyperthreading
« Reply #42 on: February 17, 2007, 02:41:55 PM »
Quote

bloodline wrote:

Quote
Basically, all the newer x86 chips are a risc style core running emulated old microcode instructions thru multiple pipelines wrapped inside caches to maximise speed.


You do realise that's how the 68060 worked, right?


This is not true. All instructions are directly run through microcode. The '060 is running a multi-scalar, non-RISC core very much like the (first) Pentium. There's no out-of-order execution either.
 

Offline CannonFodder

  • Hero Member
  • *****
  • Join Date: Sep 2003
  • Posts: 1115
    • Show only replies by CannonFodder
Re: hyperthreading
« Reply #43 on: February 17, 2007, 02:43:31 PM »
Aren't we forgetting the darling of the demoscene Mode X?
People are hostile to what they do not understand - Imam Ali ibn Abi Talib(AS)
 

Offline CannonFodder

  • Hero Member
  • *****
  • Join Date: Sep 2003
  • Posts: 1115
    • Show only replies by CannonFodder
Re: hyperthreading
« Reply #44 on: February 17, 2007, 02:54:03 PM »
Quote
-Edit- I'm pretty sure a friend of mine had showed me an SVGA flight sim on his PC in 1992... it might have been later though...


The Wing Commander games were out in 1992 IIRC.

And this was out in '93. ;-)
People are hostile to what they do not understand - Imam Ali ibn Abi Talib(AS)