@pavlor

not every comparation is a comparation

natami only existed in form of a few prototype boards, Tina not more than a idea

in this sense selling a couple of hundred computers is a big success

@OlafS25

Quote:

in this sense selling a couple of hundred computers is a big succes

That was exactly my point.

@cdimauro

Quote:

cdimauro wrote: @Dandy You'll find a lot of fun transmitting the video signal from the loco to the monitor, since the XCore can transfer only 15MB/s to the CPU/memory.

Not sure if the video signal from the loce actually has to be transmitted to the monitor via xCORE. There are possibly better ways of transmitting the signal - xCORE will possibly just be used to start/stop transmission...

Regarding the transfer rate of 15 mB/s between cpu and xCORE:
A-Eon stated that "In the AmigaOne X1000 we have one quarter of these connected to the CPU local bus for direct communication with the system".
A web search for "local of the PA6T-1682M - transfer rates" brought me to the PWRficientTM Evaluation Kit. There I found some information regarding bus speeds and a schematic:

Quote:

PWRficientTM Evaluation Kit wrote: ... PROCESSOR: ... The PA6T-1682M includes a SoC providing 22 SERDES lanes allowing for up to 55 gigabits per second transfer rates between the SoC and external devices. ... MEMORY: ... high bandwidth Flash memory ... Autostore NOVRAM provides conventional SRAM fast access timing ... ARCHITECTURE: high-performance CPU and memory design ... The AOP4 processor’s SoC architecture includes a high-speed bus, capable of sustained block copy transfers exceeding 7.5 gigabytes per second, and a SERDES interface with 22 lanes, providing a peak bandwidth of over 55 gigabits per second. ... External I/O includes a VME interface, supporting VME64, 2eVME and 2eSST protocol, and sustained transfer rates exceeding 100 MB per second. ...

According to the schematcs "Autostore NOVRAM" (which "provides conventional SRAM fast access timing") is connected to the "Local Bus".
Static RAM (SRAM) has access times as low as 10 nanoseconds.
This makes it hard to believe that memory access is really as slow as 15MB/s.

The other transfer rates figure I found is related to "External I/O", connected to a so called "SERDES" lane and comes up with transfer rates exceeding 100 mB/s.
"SERDES interface with 22 lanes, providing a peak bandwidth of over 55 gigabits per second"

Quote:

cdimauro wrote: Anyway, do you think that your project cannot be realized using some other solutions (like the one which I reported, for example)?

Most likely there isn't just one way of doing it. We have a saying here: "Many roads lead to Rome!"...

Quote:

cdimauro wrote: Quote: Dandy wrote: You're talking about the memory of the A1X1k - and not about the xCORE's own RAM, aren't you? Exactly. ...

Well, I had the xCORE's own RAM in mind here...

Quote:

cdimauro wrote: Something which can be realized with a better synergy of the XCore with the main system. The problem here is that the communication channel between the two is too much limited (see above: 15MB/s of bandwidth).

I doubt this figure being correct, as I described above.

Quote:

cdimauro wrote: So, it's much more convenient taking some external GPIO board, and connecting it to a PC with an USB2 connector, for example.

As I stated above: there isn't just one way of doing it.

Quote:

cdimauro wrote: I wish you the best for solving it.

Thanks a lot for the good wishes, mate!

Quote:

cdimauro wrote: But nowadays you can have more spending much less.

Yes - but "Amiga" is my hobby.
If I would follow the "have more spending much less"-path I would not be here and the Amiga community would loose another member.
I hope this is not what you desire...

Quote:

cdimauro wrote: Consider, also, that SDKs for such boards usually support only Windows and/or Linux. You can have hard time trying to use something different. I'm not saying that it's not possible, but only that the experience can be worse.

Yes, I considered that. But it seems as if Lyle Haze has put up an XMOS compiler that can be accessed from any network-connected AmigaOS machine - just that it is not publically available, because it's running on an Atom powered netbook:
Lyle Haze posting regarding XMOS compiler

Furthermore I found a very interesting article by Lyle Haze about programming XENA:
Writing apps for Xena

_________________
Ciao

Dandy
__________________________________________
If someone enjoys marching to military music, then I already despise him.
He got his brain accidently - the bone marrow in his back would have been sufficient for him!
(Albert Einstein)

- (edit) never mind..

Last edited by wawa on 10-Aug-2015 at 02:01 PM.

@Dandy

>This makes it hard to believe that memory access is really as slow as 15MB/s.

The limitting side is the xcore model used on nemo.

IIRC, From some schematic + xcore reference manual I once understood that theorethically 30MB/s is maximum for the connection implemented on nemo for xcore communication. Then, for real life, 15MB of bi-directional communication would seem sane.

more:

the built in RAM of the xcore is split for the cores and HW threads of the chip. So one process/task/thread on xcore has very small memory fragment to use, the memory of other HW threads is available only via serial link between threads (IIRC). xcore can be configured & program to have more RAM, but it's not a powerfull solution. xena is really a control/simple bridge for stuff.

(there are other xcore chips to use, if one wants to connect to faster DATA path like PCIex1, for example.
someone should be able to build xorro card that has pciex1 etc. capable xcore chip that networks automatically with xena. (I'm not going to try to think any practical use for such thing... except some PCIex1 analyzator... ))

Last edited by KimmoK on 10-Aug-2015 at 01:09 PM.
Last edited by KimmoK on 10-Aug-2015 at 01:00 PM.

_________________
- KimmoK
// For freedom, for honor, for AMIGA
//
// Thing that I should find more time for: CC64 - 64bit Community Computer?

@wawa

"meant for desktop use"

In 1985: GUI, sound, colors, multitasking, mouse... none of those were meant for desktop use. That must be what killed the Amiga.

_________________
- KimmoK
// For freedom, for honor, for AMIGA
//
// Thing that I should find more time for: CC64 - 64bit Community Computer?

@Rose
>>I toy with the idea of a computer with a "GEEK PORT".
>And this is how you make basic functionality costing $10 to other to cost $100 to amigans.

Where can I buy easy to use & reliable "geek port" for $10?

Recently replaced $600 NI HW with $70 HW, but over USB it's a kludge & conflicts with other USB devices!

Last edited by KimmoK on 10-Aug-2015 at 01:21 PM.

_________________
- KimmoK
// For freedom, for honor, for AMIGA
//
// Thing that I should find more time for: CC64 - 64bit Community Computer?

@lylehaze

Thank you for the post & info refresh.

_________________
- KimmoK
// For freedom, for honor, for AMIGA
//
// Thing that I should find more time for: CC64 - 64bit Community Computer?

@cdimauro

Quote:

cdimauro wrote: @Dandy ... You can take them from third-parties, but a user wants them.

???
What do you mean with that?
That I'm not a user (as I took 3rd party stuff)?

Quote:

cdimauro wrote: Icaros (an AROS "distro"), for example, integrates a lot of stuff on top of AROS because users want and need them all (even more).

Not sure how exactly Icaros works...

Quote:

cdimauro wrote: However implementing things like 64-bit and SMP for AOS means losing its backward compatibility with existing applications (both PowerPC and 68K).

Hmm - I got the impression that the Power Architecture compatible P.A. Semi Platform Processor is able to execute both - 32 Bit AND 64 Bit code:

Quote:

Power Architecture compatible P.A. Semi Platform Processor wrote: "This new P.A. Semi Platform Processor is Power Architecture compatible, and runs new or existing 32 or 64-bit PowerPC code."

Quote:

cdimauro wrote: That's why you'll not see a 64-bit version of it, and only a multicore (not SMP) support will be added.

Must have missed that implementation of 64 bitnes got removed from the OS4.2 roadmap. Do you have a link to this (bad) news?
AMP is also O.K. ...

Quote:

cdimauro wrote: ... However in this case common sizes are measured in terms of hundred of GB, and not TB (too much expensive).

Hmmm - actual offer of a PC/electronics reseller here in Cologne:
3000 gB Western Digital Red WD30EFRX.....................122,90 €
1000 gB Crucial CCT1000BX100SSD1 BX100 SSD.......357,90 €

So 'real' (spinning) harddrives are a lot cheaper than SSDs - in this case the SSD is roughly 8.74 times more expensive per tB than a mechanical Harddisk...

_________________
Ciao

Dandy
__________________________________________
If someone enjoys marching to military music, then I already despise him.
He got his brain accidently - the bone marrow in his back would have been sufficient for him!
(Albert Einstein)

@KimmoK

Quote:

In 1985: GUI, sound, colors, multitasking, mouse... none of those were meant for desktop use. That must be what killed the Amiga.

what?

@wawa

Just tried to point how irrelevan this is:
>>>xmos is definitely not meant for desktop use. also ppc isnt meant for desktop use

xmos or ppc are not in the way of AOS success on desktop, other matters are.


(If AOS4 would already be mature, capable and SW rich enough for mainstream professional use or mainstream gaming use, THEN eg. ppc might be bottlenec.)

_________________
- KimmoK
// For freedom, for honor, for AMIGA
//
// Thing that I should find more time for: CC64 - 64bit Community Computer?

@KimmoK

Quote:

xena is really a control/simple bridge for stuff.

YES! and that is all it was ever intended to be. If you want high speed, or huge amounts, of data transfer you put that on the xorro card and control it with xena.

The strong point of Xena is the very fast rise and fall times of its pins, and it is edge triggered. A rising signal on a pin can trigger Xena to do something else (programmable) before a steady state is reached, and of course no need to wait for a falling edge (or the other way around, again programmable).

So, a detected input event can trigger a data transfer to Nemo much faster than the pin clocks could ever operate. It is NOT a magic chip (like op-amps) that can automatically speed up data buses.

It is what it is. That does not mean that people desiring it do something more should be treated badly or the inclusion of it on the board for their desired want was wrong, it only suggests that they require more information before delving into a project.

@Dandy

Start a thread "What Xena can and cannot do".

If for no other reason just to see how soon The Fart shows up and turns it to stink.

_________________
Tinkering with computers.

@Signal

Quote:

Start a thread "What Xena can and cannot do". If for no other reason just to see how soon The Fart shows up and turns it to stink.

I suppose this is a real possibility, so I should suggest an existing on-topic thread for this purpose:

XMOS - what is it all about?

Although I realize it's a mortal sin to post in on-topic threads, so I have little hope...

Another thread with on-topic content

#6

Last edited by number6 on 10-Aug-2015 at 03:57 PM.

_________________
This posting, in its entirety, represents solely the perspective of the author.
*Secrecy has served us so well*

@KimmoK

Quote:

xmos or ppc are not in the way of AOS success on desktop, other matters are.

they may not be main issues, but they add up on the rest, to the point where you ask yourself, when finally decissions will start to be made with actual advantages for the platform and widest understood consumer satisfaction in mind, rather than.. well what are actually the reasons for having components on board, never to be seriously used by anyone??

take into account example of acube and its sam460cr (cost reduced?) where a it seems that even a marginal price drop while compromising few board elements may be considered neccessary.

Quote:

(If AOS4 would already be mature, capable and SW rich enough for mainstream professional use or mainstream gaming use, THEN eg. ppc might be bottlenec.)

ppc is the only platform this os and its software runs. its not some minor handicap that can be addresed later whe´n other problems are solved, since this is a major limiting factor. the os is obviously not portable and not endian agnostic. as soon as you have addresed other issues, and made your os "capable" you will have to start the whole effort from the scratch, if every work done till this point was done with almost non existant hardware in mind.

@number6

Quote:

number6 wrote: @Signal Although I realize it's a mortal sin to post in on-topic threads, so I have little hope... #6

Amen.

_________________
Tinkering with computers.

@wawa

I thought the entire exercise of rewriting AmigaOS in C was to make it more portable.

It would be interesting to see Hyperion try to build OS4 for a different big endian architecture, just for the kicks.

_________________
B5D6A1D019D5D45BCC56F4782AC220D8B3E2A6CC

@KingKong

Quote:

KingKong wrote: One aspect could be 48 bit than 64 bit, ...

This 'not 64 bit' idea is totally shit, all available memory modules are designed for 64 bit wide memory access. Designing something with less bit width would need custom build modules or speed loss during miss aligned word access.

And 'saving resources with less bits' wont help, most 64Bit CPUs doesn't need a full 64Bit address bus on their chip. Like the original 68k, that could use 32Bit addresses on software but had only pins for a 24bit address space, you don't need to have full 64 bit address pins on actual CPUs. The Intel® Pentium® Processor G6950 for example has a physical address range of 'only' 16GB, that's less then you could access with 64Bit

And all your fancy ideas need someone showing up and throw millions of dollars into a development process, for an outdated OS you want to see on a non existing CPUs.

Sorry but no one will do this, investors will not ask about hardware or OSes, they will ask about new ideas. An Internet company built around some scripts could become more valuable in short time, as the Amiga ever was.

People like you are still living in the 80ies, today the big business is not about CPUs or OSes, it's about Social Communities, Media Streaming, Cloud computing.

---

Updated after KimmoK reminded me to other 68k CPU that have an other pin-out like the original Chip had.

Last edited by OneTimer1 on 11-Aug-2015 at 11:58 PM.
Last edited by OneTimer1 on 10-Aug-2015 at 10:43 PM.

@pavlor

Quote:

pavlor wrote: Do you know other PowerPC + XMOS designed for desktop motherboards?

No they stopped making PowerPC dekstop motherboards some years ago. Even IBM and Motorola stopped selling desktop system with this architecture.

@OneTimer1

thx. finally someone new to the party that isnt actually a loony..

@OneTimer1

>Like the 68k, that could use 32Bit addresses on software but had only pins for a 24bit address space,

That is not true for 68k in general 68030, 68040 and 68060 all had 32bit address path.

(and some 68k accelerators even had 64bit DATA path to RAM)


>big business is not about CPUs or OSes, it's about Social Communities, Media Streaming, Cloud computing.

Too bad that users do not care about business?


Amiga OS users/fans see how the OS can make the user experience nicer for them.

_________________
- KimmoK
// For freedom, for honor, for AMIGA
//
// Thing that I should find more time for: CC64 - 64bit Community Computer?