@iggy

Hey! Look carefully in front of you! There's some SARCASM!

@iggy

im sure that on big single thread 8ghz cpu will be more faster than 4 core 2ghz one (if with same instructions, cache, ecc ecc )

(for sure a 4 ghz 386 will be slower than 2 ghz i7 :P)

Last edited by tlosm on 23-Jun-2017 at 03:07 PM.

_________________
I love Amiga and new hope by AmigaNG
A 500 + ; CDTV; CD32;
PowerMac G5 Quad 8GB,SSD,SSHD,7800gtx,Radeon R5 230 2GB;
MacBook Pro Retina I7 2.3ghz;
#nomorea-eoninmyhome

@tlosm

I keep waiting for those 10 GHz Pentium 4s that Intel promised us.

Sorry, I actually missed the sarcasm.
Honestly, for desktop use, I'd take a lower core count and a higher maximum frequency any day.

Unfortunately we hit a wall in regard to maximum clock speeds quite awhile ago.
Really, for most uses, four threads seems to take care of almost everything.

And I'm current using an FX-8370, which rarely keeps all eight cores occupied.
So I'm still unsure if an eight core Ryzen (which would support sixteen threads) really makes good economic sense when I can get a six core much cheaper (and still have a 50% improvement in the number of threads supported).

What we really need to focus on (as programmers) is coding in smaller, reusable modules (surprising like the considerations I needed to keep in mind with early multitasking OS').
We have, if anything, an excess capacity to run concurrent processes, but programs just aren't being constructed to work in multiple directions at once.

As an analogy, we need to go from whistling a tune to conducting an orchestra.

When the Motorola 68040 processor was introduced, it included the FPU internally. Most instructions and numeric representation modes from the 68881 were supported in hardware, but some were not, and were emulated in software.

Has anyone done any comparison of an '040 or below FPU-intensive application vs. Vampire?

If a FPU instruction is encountered, the chip will either handle it or an exception is raised an a software subroutine handles it.

This is what the math libraries are for...

The OS may report that no FPU is present, however, that doesn't mean that the instructions aren't handled. If your software uses FPU instructions and it doesn't run - shame on your software.

Last edited by Lou on 23-Jun-2017 at 06:14 PM.

@RodTerl

Aint Hyperthreading just Intels copyrighted term to enable x86 to do half decent Enforced Preemptive Multitasking anyway?
No, it's not, "Preemptive Multitasking", it basicity interrupt driven tasking where one task can run for some time, then the next get some time, have often task get to do something is controlled by task priority.

The Amiga "Preemptive Multitasking", is a software implementation for a single core CPU.

Dual core is basically two CPU's cooperating, allowing 2 program to run at same time parallel to etch other. so a dual core can theoretically do twice the amount of work,. (if the program never need to coordinate things )

Hyper threading also allows two program to run parallel. But there is a difference, it's not a full extra core instead CPU cache is shared, between the hardware threads (so this are simple cores with shared CPU cache), this makes it not as fast as dual core, but is less complicated design less heat at lower cost, so Hyper threading is basically a poor man's dual core. It can give about 15-30% increases in speed.

Modern Intel CPU's is hyper threaded and has lot of cores.

SMP is required in any of the cases, to make use of it, and for legacy support, it is debatable.

Last edited by NutsAboutAmiga on 24-Jun-2017 at 11:05 AM.
Last edited by NutsAboutAmiga on 24-Jun-2017 at 11:03 AM.

_________________
http://lifeofliveforit.blogspot.no/
Facebook::LiveForIt Software for AmigaOS

x

Last edited by NutsAboutAmiga on 23-Jun-2017 at 06:42 PM.

_________________
http://lifeofliveforit.blogspot.no/
Facebook::LiveForIt Software for AmigaOS

Quote:

This is what the math libraries are for...

Well no, the math libraries are there when you need to calculate point numbers, while slow it’s the only way you can do it unless you’re a professor in math's.

You do not get the precision, by using integer number, integer number does not allow exponential numbers v*(10^n), so dealing real tiny, number or really big numbers, is a real challenge.

When I wonted some precision coding 68000 assembler I remember work number that was 100 times value, this was so that did not lose too much information in numbers. After doing my calculation normally divide the number by 100, to get the final number, so I was basically faking two decimals behind the dot. Divisions and multiplications are considered slow, to it not good workaround for not having a FPU.

Last edited by NutsAboutAmiga on 23-Jun-2017 at 06:43 PM.

_________________
http://lifeofliveforit.blogspot.no/
Facebook::LiveForIt Software for AmigaOS

Quote:

Aros is in no way as dead as amiga ppc or morphs.

Quote:

It's about as useless as running OS4 on current PPC CPU's.

The atmosphere here is absolutely charming. It's a wonder that all the Amiga fans(?) here haven't slit their wrists or at least found a more enjoyable hobby...


@bennymee Quote:

It would be more interested in how much speed could be gained or which compiler is used to enable hypertreading on 68080, then reading 1 anti OS4 comment here and one in the Apollo forum....

You can always vote with your feet, and stop visiting Amiga sites with bad vibes. There are at least a few sites where being an arse h*le isn't acceptable behaviour.

Last edited by ChrisH on 23-Jun-2017 at 07:12 PM.

_________________
Author of the PortablE programming language.
It is pitch black. You are likely to be eaten by a grue...

@ChrisH

Quote:

The atmosphere here is absolutely charming. It's a wonder that all the Amiga fans(?) here haven't slit their wrists or at least found a more enjoyable hobby...

Isn't JUST so cheerful?
AROS isn't as "dead" as OS4 or MorphOS.

Are any of them "dead"?
There are multiple derivatives of AROS, the X86 version leading the way to NG SMP, while the 68K version might just free us from the restraints imposed by Hyperion or Cloanto AND run on new, MUCH faster hardware.
OS4 seems to be advancing faster than in than in the past, with some nice new video card drivers and OpenGL ES support.
And the MorphOS community (which unlike the other two teams, tends to keep its developments quiet, until they are ready for distribution) IS doing quite well (trust me, I just exchanged messages with a lead developer today).

Plus, the Amiga trademark might actually be freed from the clutches of Amiga Inc soon.

And so we get dysfunctional crap like this instead of positive posts.

Hey, instead of all of us cutting our wrists, why don't the negative BS posters do it and save us the grief?

Im a Vampire owner, but I do agree with ChrisH in that some people cant help themselves with regards to putting XYZ branch of the Amiga down.

It makes me wonder about the people behind the keyboards.

But users like Dan Wood/Kookytech is 1 in a 1000 I guess, that are actually positive to all hobbies and branches.

@ChrisH

Why did you skip the post niko was replying to? Was it to prove some form of moral superiority?

Quote:

@ferrels But havent most aros developers left? Isent aros mostlys abandoned ref to latest News.

Now I will go back to not giving a crap what happens in this "community" :)

Last edited by terminills on 23-Jun-2017 at 08:14 PM.

_________________
Support AROS sponsor a developer.

"AROS is prolly illegal ~ Evert Carton" intentionally quoted out of context for dramatic effect

@NutsAboutAmiga

Quote:

Hyper threading also allows two program to run parallel. But there is difference, it's not a full extra core instead CPU cache is shared, between the hardware threads (simple cores without shared CPU cache), this makes it not as fast as dual core, but is less complicated design less heat at lower cost, so Hyper threading is basically a poor man's dual core. It can give about 13% increases in speed.

Not to be harsh, but where did you take that 13% of speed increase with multithreading? Since, by definition, SMT (or ht like you prefer) uses free resources on the pipelines of superscalar architectures, the speed increase only depends on the busyness of these pipelines, which vary. You'll find times when a single, SMT enabled core will compute two threads almost at the same speed, and others when spare time on the pipeline will be scarce, BUT in any case a 13% average result is a lot FAR from reality.

@iggy

Quote:

And I'm current using an FX-8370, which rarely keeps all eight cores occupied. So I'm still unsure if an eight core Ryzen (which would support sixteen threads) really makes good economic sense when I can get a six core much cheaper (and still have a 50% improvement in the number of threads supported).

There are good chances a 4 core Ryzen would outperform the FX8730 in almost every task
http://cpu.userbenchmark.com/Compare/AMD-Ryzen-5-1400-vs-AMD-FX-8350/3922vs1489

But, obviously, I wouldn't step back from a real 8-core processor (although FX cores are real cores for integers only, while a single FPU is shared among every couple of cores) to a 4core/8thread processor to make an upgrade. Ryzen's IPC value is far better than FX one, both in integer and FP so you recoup what you loose. Point is that I wouldn't buy anything lesser than a 6core/12thread processor today, just to be future-proof enough for the next years.

@paolone

That is exactly what I'm considering, a six core.
Great pricing, support for twelve threads.
Ryzen is more efficient than the FX, but the four cores seem like they would be cutting it to close.

If Gunnar really wants to impress me, he'll start to incorporate cache into the Apollo design.
We're going to need that as the cpu scales up anyway.

IF he figure out how to retain legacy fpu capability, I'm all for the high tech floating point unit he is thinking of, as well as the new multimedia instructions.
And hopefully the design will be as superscalar as possible.

Remember, the cpu in the A1222 is only a 2 core 32 bit, with a weak fpu, no hyperthreading and a relatively low clock speed.

Given time, its not impossible that an Apollo based device could at least match that performance (if not exceed it).

@iggy

about .
i have an i7 2012 macbook . it gave me in single thread qemu 40.000 point in benchmark where 94.000 is the single thread 970MP ...
probably a new machine will result in 80.000 point in qemu and now qemu have ppc64 mmtcg what mean? you will able to have 16 cpu emulated one for every thread host on qemu... it mean you will be able to have osx leopard ppc more faster than a real 970MP in integer. it mean qemu now is dozen time more faster than a1222 in calcs

Last edited by tlosm on 24-Jun-2017 at 07:30 AM.

_________________
I love Amiga and new hope by AmigaNG
A 500 + ; CDTV; CD32;
PowerMac G5 Quad 8GB,SSD,SSHD,7800gtx,Radeon R5 230 2GB;
MacBook Pro Retina I7 2.3ghz;
#nomorea-eoninmyhome

@tlosm

Amazing! What benchmark do you use? However benchmarks sometimes can show faster results under emulation. UAE JIT under Linux on A1222 also shows strenght for someting like 68040 at 70 MHz. But real world applications are something a little bit different......

Quote:

tlosm wrote: @iggy about . i have an i7 2012 macbook . it gave me in single thread qemu 40.000 point in benchmark where 94.000 is the single thread 970MP ... probably a new machine will result in 80.000 point in qemu and now qemu have ppc64 mmtcg what mean? you will able to have 16 cpu emulated one for every thread host on qemu... it mean you will be able to have osx leopard ppc more faster than a real 970MP in integer. it mean qemu now is dozen time more faster than a1222 in calcs

@tlosm

Amazing! What benchmark do you use? However benchmarks sometimes can show faster results under emulation. UAE JIT under Linux on A1222 also shows strenght for someting like 68040 at 70 MHz. But real world applications are something a little bit different......

Quote:

tlosm wrote: @iggy about . i have an i7 2012 macbook . it gave me in single thread qemu 40.000 point in benchmark where 94.000 is the single thread 970MP ... probably a new machine will result in 80.000 point in qemu and now qemu have ppc64 mmtcg what mean? you will able to have 16 cpu emulated one for every thread host on qemu... it mean you will be able to have osx leopard ppc more faster than a real 970MP in integer. it mean qemu now is dozen time more faster than a1222 in calcs

@iggy

Quote:

Given time, its not impossible that an Apollo based device could at least match that performance (if not exceed it).

They are now at 1/20 of 460EX 1.1 GHz performance (not counting FPU of course)...

@dooz

you can use speedometer on os 9 *i use it for testing"
and geekbench if you need to test osx for testing

i understand that the performances are true because use the software too. eg softwindows, vpc are running ok under qemu or you can try games

im using now qemu 2.9.0.

Last edited by tlosm on 24-Jun-2017 at 10:14 AM.
Last edited by tlosm on 24-Jun-2017 at 10:12 AM.

_________________
I love Amiga and new hope by AmigaNG
A 500 + ; CDTV; CD32;
PowerMac G5 Quad 8GB,SSD,SSHD,7800gtx,Radeon R5 230 2GB;
MacBook Pro Retina I7 2.3ghz;
#nomorea-eoninmyhome

@paolone

https://en.wikipedia.org/wiki/Hyper-threading

"According to Intel, the first hyper-threading implementation used only 5% more die area than the comparable non-hyperthreaded processor, but the performance was 15–30% better."

Hemm most read 15 and 30 and taken one number from one number and other from other number, you know I'm dyslectic.

_________________
http://lifeofliveforit.blogspot.no/
Facebook::LiveForIt Software for AmigaOS