FreeBSD and processors.

sossego

Retired from the forums
You know what? I'm thinking, "Match the code to the instruction set and you will have an efficient machine."
  1. AMD64/i386 Complex instruction. These are good for conversion and compression.
  2. Power-PC/POWER Load, store, reduced, and looped instructions such as those for sound effects, gaming, animation sequences.
  3. SPARC64/UltraSPARC Large load delivery and handling.
 
Not everyone can afford a SPARC/POWER server.

In my old job we had lots of POWER6 (and POWER5) machines. To be honest, we had only IBM stuff. In my new job, we have only amd64/i386 architecture. Although a SPARC would be beneficial for us (database-wise), management said it's too expensive (but they also said MySQL is not up to the challenge - who understands them).
 
One could use EBay to purchase older machines such as a SunBlade 1000/1500/2000/- or other types- and then install FreeBSD to it and the required database software. This would cut down on costs.

I either bought mine from Craigslist: iMac G4 & PowerMac G4 QuickSilver one purchase & a G3 B&W another purchase & two Dell Dimensions, bought it from ebay: two SunBlade 1000 machines, found it in the trash: a few laptops, or it was given to me: AMD64 Sempron machine & a few laptops.
 
Match the operating system design to the purpose of the machine. What do you want to do? Think about that. Look at the capabilities of the OS then test the limits. I think we forget about this at times and need to realize that there is no such thing as a general purpose system.
 
We are beginning to change our focus from a centralized isolationist view to one that is open and accepting yet respectful. No longer are the questions about the quantity of information and processes a system can do but about what the system can do and be the best in a given situation. Confusing? Not really, continue reading.

The OS has a design for a specific purpose. Learn that purpose and apply it. E.g.
  • OpenBSD is made for security, works great on SPARC and UltraSPARC/SPARC64 machines.
  • Debian works well on ARM devices, is good for simple storage.
  • Windows is good for presentations, works good on AMD64 and i386 with two or more cores and/or processors.
  • MacOSX is good for media production, works on AMD64 and Power(PC)64.
  • QNX is good for embedded machines.
The same standards can be applied to virtualization, networking, programming, and other aspects of the computer field.

On Virtualization.
  • Each system is set up to virtualize or simulate certain environments.
  • FreeBSD jails, chroots, usermode Linux, and Solaris containers separate processes yet share the kernel.
  • Xen and KVM are similar.
On networking.
  • Certain protocols are better for certain types of information.
We must move towards making systems compatible by emphasizing the strengths of each within every field. People are afraid because they remain uninformed and misinformed. It is your choice to teach others.
 
There will never be complete unanimous consensus on anything when it comes to the human race.

Other opinions are welcomed.
 
When the members of a community are more open and accepting, ideas shared become the basis for newer ideas. If one person, one school of thought, or one political ideology is allowed to overtake an idea, the following result is a stagnation and regression of ideas. You may not agree with the use of system A or the use of certain applications on system B but, allow others to develop on these platforms and be willing to give positive feedback.

This thread is not a strict standard on the proper way to do things but one single perspective when it comes to the computing community.
 
I hope it is OK to comment on some points you made in this thread, since you are the sole poster up to now.

sossego said:
There will never be complete unanimous consensus on anything when it comes to the human race.
I have to disagree! First, because I think there ARE fundamental values each&everyone should agree on. Second, this is the standard proof (Goedel) that there can be unsolveable things even in simple systems. Because if each one agreed on something, if there is consensus, then the hypothesis would be proven wrong by proving it right.

sossego said:
When the members of a community are more open and accepting, ideas shared become the basis for newer ideas. If one person, one school of thought, or one political ideology is allowed to overtake an idea, the following result is a stagnation and regression of ideas. You may not agree with the use of system A or the use of certain applications on system B but, allow others to develop on these platforms and be willing to give positive feedback.

This thread is not a strict standard on the proper way to do things but one single perspective when it comes to the computing community.

There is an analogy to this in social interactions and computer systems. When I look around in this country, with special attention to the timeline and progress, I can see patterns. There are the old trading routes which were established in the stone age and are still here today. There was trade in flint, amber, salt, ... and the routes which were taken were best for that days. Later came the 'Hanse', providing faster and further trade. Along the trade routes not only came goods, but also people. And with them came new ideas, new views and sometimes new plagues. People living along these routes, or even more so where they crossed, had to learn to deal with others, accept new ideas, ... Everyone who participated in that made progress, profit was made, things learned. Tolerance formed.

Those who were sitting in their mountain valley did not, and that mindset which was formed by that shows to this day.

Now for the parallels into the realms of computers. In the beginning we had an abundance of different makers, operating systems, architectures. Cross pollination was common, a lot of different outfits made profit. Computing was fun.

Then there set in a process which also is known from the 'Human World'. While small trade, small business could make a lot of people pretty well to do, it was not enough for some. But to become more wealthy you need other people to make your money. Thus you need to destroy their seperate base of living. Form guilds to regulate crafts, and in the end the same process ends in patents for if. Today we have only one architecture on the desktop and mostly one operating system. In the 'human world' this is also a recipe for desaster. A certain amount of organisation is necessary, true, because else the $VIKINGS come around to collect. More, and it would/will collapse on it's own.

In computer terms, you can substitute patent lawyers/trolls and the bureaucracy of big companies. That's what big corporations fear, that they can be outmaneuvered by smaller and much faster ones. They do not have to fear the raiders. But: how long would it take to turn something like IBM around? Or change desktop computing to the MIPS architecture? To ensure the dominance, measures are taken. The same as those in the 'human world'. Laws, guilds, patents, religion, raiding, war, diplomacy, bribery. To fight dominance, to be free, also the same things need to be done. Self-sustainability, open minds, awareness, planning, tolerance, loose and close cooperation.
 
You forgot about @da1.

One does not just learn and continue, you must be able to learn continuously.
 
Last edited by a moderator:
Yes, sorry!

I did not see that posting in scrolling trough, seems I need new glasses.

But as I tried to point out, the same principles are at work on the digital world and the 'normal' world.
Monocultures may be good for troughput, if nothing goes wrong. But they are vulnerable to deseases, diversity is your friend there. Also, one size fits nobody. X86 is optimized for one kind of workload while other architectures are tailored to other topics. X86 has good code density, which helps if memory is tight and memory bandwith can be the bottleneck. But all in all, it is a horrible architecture.

RISC designs like MIPS/PPC/SPARC, which I really like for their architecture, prefer higher memory bandwidth due to slightly larger code. MIPS is IMHO the most 'pure' RISC of these, the original had one clock per instruction for all instructions. Power has instructions which take longer, even variable amounts of clocks. It is not nice to write a scheduler for them, but it can boost performance a lot.

In the end I would prefer these over x86/x64 and also ARM for their internal beauty, but hardware support/availability is somewhat limited. Let's see where the chinese take the MIPS64 cores to, something that runs 64 cores of them is rumored to be around. I would not mind using one for my daily work.

Is it madness to go against the flow? No. There is no better way to reach the source. And laughing about the next Wintel-trojan is fun, isn't it?
 
The computer is becoming wearable. I've wondered, "Why isn't there a glove-mouse?"
 
sossego said:
The computer is becoming wearable. I've wondered, "Why isn't there a glove-mouse?"

Oddly enough, many feel that their keyboard is almost that. I remember in the DOS days, the mouse was a pain. Later, I dreaded having to resort to the keyboard for anything other than typing. Now, again the mouse has become a pain. Using only laptops now, the pointing stick and touch pads make coping with point clicking much easier. It still amazes me the difference, moving away from the mouse makes. Long live key-bindings. The keyboard is your glove.

I would like to see the open source world, develop its own hardware. That is, what the original post leads me to think about. Hardware of a certain kind, will perform better than others at some tasks. I would like to see more open-source hardware. Very module and versatile. Yes, lacking because it isn't geared in any specific direction, other that keeping the developer in mind. So much of what we have at our disposal, is geared towards the needs or demands of commerce. Cheaper parts for mass production. Or, more power to support inefficient code.

However, it is rather amusing to see outdated machines serving their users better than many new purchases. This just because the outdated machine is running an OS that the user knows how to use well, while also providing access to high quantities of open-source applications. And, the new purchase has every convenience of non-thought interfacing. Often, with the hindrance of financial consideration whenever a new applications is required . But, not everyone has the time to learn beyond their immediate computer need.
 
Perhaps not very related but the POWER architecture is planning to be opened by IBM.
The OpenPOWER consortium has been joined by some quite large companies including NVIDIA and Google.

I think this could be good for a change from x86 or ARM (which always seems locked down). Saying that, wasn't SPARC meant to be quite open? I would love a modern Thinkpad with a SPARC or POWER processor! ;)
 
kpedersen said:
I think this could be good for a change from x86 or ARM (which always seems locked down). Saying that, wasn't SPARC meant to be quite open? I would love a modern Thinkpad with a SPARC or POWER processor! ;)

No, thanks! It'd be a nightmare for software developers.
 
Surely the C / C++ compiler protects us from the underlying architecture (*snigger*).

In all honesty, it would indeed put more stress on developers but it would at least force us to write more portable code.
 
expl said:
No, thanks! It'd be a nightmare for software developers.

The compilers are available and most code will work on most architectures. Those developers who were dependent upon one system and one- plus the 64 bit extension- architecture will just have to change.
th
 
Haswell CPUs are about 200 bucks (just built a machine a couple of weeks ago - Core i5-4430). Ivy/Sandy bridge are likely cheaper.

Finding anything close in terms of bang for buck, hardware support, etc. is going to be very difficult. Even including old Sparc or Power hardware from Ebay (which obviously has no warranty, spare parts harder to source, etc.).

Yes, in an ideal world the other architectures are cleaner internally, more elegantly designed, etc. In reality, Intel and AMD are just WAY faster in terms of MIPS/$ and more power efficient.

Yes, it's a horrible hackish architecture, but it gets the job done. And if you're a user or high level language programmer, the internals are pretty irrelevant, outside of idealism or religious zealotry. :)

I think the "You shall deal with it" graphic applies here instead :D
 
kpedersen said:
I would love a modern Thinkpad with a SPARC or POWER processor! ;)
Me too, that would be great. There are laptops with MIPS cores, but I'm waiting for the 4-core to become available.

throAU said:
Yes, it's a horrible hackish architecture, but it gets the job done. And if you're a user or high level language programmer, the internals are pretty irrelevant, outside of idealism or religious zealotry. :)

I think the "You shall deal with it" graphic applies here instead :D

I remember the x86 architecture being describes as "difficult to explain and impossible to love". That's true. I don't need to love the architecture to use it, but it would make things better IMHO. I would like to see some MIPS, SPARC or PowerPC cores being manufactured in the same process as the new intel chips. Then you could compare the power consumption and speed 1:1.
 
Crivens said:
I would like to see some MIPS, SPARC or PowerPC cores being manufactured in the same process as the new intel chips. Then you could compare the power consumption and speed 1:1.

As would I, the geek in me hates x86/x64. It's just so... ugly. Though I hear much of the brain damage was fixed in native x64 mode (I haven't had anything to do with x64 assembler though to comment further).

BUT... and this is a big but: It will never happen. Intel are way ahead in terms of fabrication and Intel want to pump out x86/x64. Rather - most people want x86/x64 compatible. Intel themselves have tried to kill x86 on numerous occasions, and failed - the 960, 860, ia64, etc.

In the real world, being able to run x86 software at native speed is a huge win anyway. Later Intel stuff is all RISC internally anyway, and in terms of percentage of CPU resources allocated for the instruction decoder (to translate from CISC front end to RISC internally) vs. the rest of the core, it is quite small, and will become smaller as caches and other CPU features grow. The other benefit is CISC has higher code density and therefore smaller process size and better caching.

So the current situation isn't all bad.

If I'm honest, it's why I actually pulled the trigger on a Mac quite late. The safety net of being able to run generic x86 code at native speed was a major plus vs. the old PPC Macs - much as the Power architecture is cleaner, etc. As always, the processor that runs the code you need to run fastest will win. Currently for the vast majority of code out there, that is x86/x64.
 
sossego said:
The computer is becoming wearable. I've wondered, "Why isn't there a glove-mouse?"

And we turn back the wheel of time for almost 20 years, and we find this.
 
throAU said:
Yes, in an ideal world the other architectures are cleaner internally, more elegantly designed, etc. In reality, Intel and AMD are just WAY faster in terms of MIPS/$ and more power efficient.

Yes, it's a horrible hackish architecture, but it gets the job done. And if you're a user or high level language programmer, the internals are pretty irrelevant, outside of idealism or religious zealotry. :)

Excellent sum up.
 
Back
Top