@Rogue & HyperionMP:

In that case, I stand (actually, I sit, watching "A New Hope" with a short intermediate "Troops" (the excellent SW parody), but anyway) corrected.

Thank you VERY much for explaining it, now maybe this could shut the trolls up? If DMA has been tested with OS4 I'm happy, and I suspect a LOT of other people are content as well

_________________
This weeks pet peeve:
Using "voltage" instead of "potential", which leads to inventing new words like "amperage" instead of "current" (I, measured in A) or possible "charge" (amperehours, Ah or Coulomb, C). Sometimes I don't even know what people mean.

@Rogue

Quote:

There is a DMA driver, however, there is absolutely no reason to introduce yet another unknown. In order to work, the driver does not need to support DMA. We're taking a one-feature-at-a-time approach where the PIO modes are supposed to work before new functionality is introduced.. FWIW, the DMA driver seems to work just fine, but like I said it's not required for a prelininary pre-release version.

But once it is (release version), you might get into the problem, that if issues turn up,
they will delay the release version (unless it was decided that an OS4-release without
DMA driver is okay). You then might take a couple of weeks to fix/workaround the issues
-> OS4 also delayed once more at least this couple of few weeks.

If OTOH you had *started* doing this dma drivers or even just dma test programs half
a year or a year ago (warned by this long time it is taking to get them right in Linux)
yould would have in total much more time to get it right/tested without much delaying
OS4 release itself, and neither necessarily OS4 prerelease.

"started" as in interleaving work on dma, while having time for it, because maybe being forced to suspend your actual work on part XYZ, until some other coder is done with stuff
XZY2.

@Georg

Quote:

If OTOH you had *started* doing this dma drivers or even just dma test programs half a year or a year ago (warned by this long time it is taking to get them right in Linux) yould would have in total much more time to get it right/tested without much delaying OS4 release itself, and neither necessarily OS4 prerelease.

"If", yes. If we had the people to spare, then we would have started a lot of things half a year or a year ago. However, we haven't.

Besides, there had been ongoing work on the Linux kernel, and given the little resources we have there was definitely no need to duplicate work.

Quote:

"started" as in interleaving work on dma, while having time for it, because maybe being forced to suspend your actual work on part XYZ, until some other coder is done with stuff XZY2.

Thanks for the lecture in software management. The keyword is "having time". If the DMA drivers are going to delay OS 4, then so be it. It is very likely they won't because as I said, there is already a testbed implementation that seems to work (whereas testing will only prove the presense of bugs, not their absense).

_________________
Seriously, if you want to contact me do not bother sending me a PM here. Write me a mail

@DrBombcrater
Quote:

Remember, this is a problem that's supposed to only manifest itself under very heavy loads. These days it's quite common for the north bridge to have to deal up to 60MB/sec of traffic on each IDE channel, a saturated PCI bus, and lots of AGP traffic, while the CPU is at 100% usage. All at the same time.

I would guess that the only reason anyone should be so overwhelmingly concerned about such things is if they are designing an application that pushes the design limits of the hardware.
Does your testing of the current OS4 pre-pre-release suggest that PIO doesn't offer sufficient performance to support your application?

Please, may we have status update(s)? Pretty please, Hyperion?

hehe.. next big reality tv show: Hyperion OS4 development team!

Hidden cameras set up all over Hyperion's offices give over the shoulder's views of what the Hyperion developers are up to...

This would be a 24/7 cam system that's pay-per-view, so all the amigans could pay to watch OS4 being developed.

_________________
Men who have girlies in their avatars are Girliemen!

@harsh

Quote:

Does your testing of the current OS4 pre-pre-release suggest that PIO doesn't offer sufficient performance to support your application?

I've never even seen OS4 and certainly not tested it, but to suggest that PIO may be okay for anything but the most trivial usage is just plain nuts. PIO mode has been left in modern IDE drives and controllers purely as a way of loading in just enough of the OS to get DMA working.

Any appliction that uses the hard disk a lot (which includes things as mundane as e-mail clients) will suffer dramatic loss of performance in PIO, and anything real-time probably won't work at all. PIO also has a huge negative impact on multi-tasking.

Would you be happy if MP3s started skipping because you copied some big files to the hard drive, that your CD-RW drive turns out coasters if you do anything while it's burning or run it at more than a crawl, or that playing DVDs gives a jerky mess?

DMA mode is necessity. If it doesn't work the A1 will be nothing more than a joke and OS4.0 will sink faster than the Titanic. Regardless of who's fault it is, if this problem really exists it needs to be fixed - now!.

@Darth_X

Quote:

Hidden cameras set up all over Hyperion's offices give over the shoulder's views of what the Hyperion developers are up to...

Hehe. I'd pay good money for that. I can just picture Rogue throwing stuff around his office and screaming "That Bombcrater freak is on my case, again! Agghh!"

_________________
Who do you serve, and who do you trust? - Galen

@DrBombcrater

Quote:

DMA mode is necessity. If it doesn't work the A1 will be nothing more than a joke and OS4.0 will sink faster than the Titanic. Regardless of who's fault it is, if this problem really exists it needs to be fixed - now!.

DMA is a necessity I agree, however as has been pointed out, this isn't a necessity for the pre-release beta. It is exactly that, a Beta. This means that things may not be working 100%, the system may crash and features may be missing.

Let the guys finish one problem before they potentially create another

Kelvin

_________________

I don't mean to interrupt the discussion regarding DMA on the A1, but
I would like some sort of news on the status of the classic PPC version
of OS4. It is neat hearing about all the the A1/OS4 news, but as a
cyberstorm user, I am feeling a little left out. Just like the beginning
of this thread asked, a current status report for us classic PPC users
would be very helpful. Has there been any public beta demostrations
for the classic version?

C Snyder

_________________
"Space Mountain might be the oldest ride in the park, but it still has one of the longest lines!" - Ric Flair

@gemini

Preseumably classic Amiga setups like for example an A1200 BPPC running OS4 will not support DMA regardless of whether OS4 is capable of it or not, because of the hardware issue? Or have I misunderstood how this would work.

Can I ask also, how does OS4 fare with hacks such as the PowerFlyer which clips over the Gayle Chip? Is the BPPC version of OS4 taking these kinds of things into consideration or do they make absolutely no difference to the running of OS4 anyway?

I'd be interested to learn more about the complications that BPPC/CSPPC versions of OS4 are giving the Hyperion staff!

Brian

@BrianHoskins

The honest answer is I don't know, hopefully Hyperion can tell us!

_________________

@DrBombcrater

Quote:

I've never even seen OS4 and certainly not tested it, but to suggest that PIO may be okay for anything but the most trivial usage is just plain nuts. PIO mode has been left in modern IDE drives and controllers purely as a way of loading in just enough of the OS to get DMA working.

The Prerelease will not have DMA, but as I already suggested, there is already a testbed implementation. But the last thing is that a bug in the driver will be used to cry "DMA bug" again, and therefore it will undergo through testing before reelase.

Quote:

If it doesn't work the A1 will be nothing more than a joke

It already works. The sound card drivers as well as the ethernet drivers already use it, ok?

Quote:

OS4.0 will sink faster than the Titanic

OS 4 is not tied to a specific harddisk. FYI, look at the reports of people having DMA problems and lockups on the Pegasos (even the Pegasos II), and tell me that this is a joke too.

Quote:

Regardless of who's fault it is, if this problem really exists it needs to be fixed - now!.

Is anyone of you actually listening to what I am saying? Probably not.

Quote:

Hehe. I'd pay good money for that. I can just picture Rogue throwing stuff around his office and screaming "That Bombcrater freak is on my case, again! Agghh!"

Hardly. It takes more to get me upset, much more to get me screaming, and a lot more to make me throw stuff around.

_________________
Seriously, if you want to contact me do not bother sending me a PM here. Write me a mail

@Darth_X

Quote:

Hidden cameras set up all over Hyperion's offices give over the shoulder's views of what the Hyperion developers are up to...

This could be the biggest show in your life, if you had one

_________________
Seriously, if you want to contact me do not bother sending me a PM here. Write me a mail

_________________
"Life," said Marvin dolefully, "loathe it or ignore it, you can't like it."

@Rogue

It might be a good idea to, once you have made the IDE DMA driver, send it off to the earlybirds to try on their systems - I don't know how many you have on your internal beta team, but I'd be surprised if they were more than a few dozen, and there are a few hundred earlybird owners. That many people playing with it will show up the common usage bugs much sooner, and from a PR point of view you can say "we've had people testing this for weeks, and fixed all the problems they found".

OBTW, the radeon PLL thing - was it an endianness issue when copying the BIOS tables to the video driver? I've got a bet with a friend, and I'd like to know who won

@DrBombcrater

I have a friend whom is of a mind that carbureted automobiles can't reasonably function because they don't have a mighty engine control computer that modern fuelies do. What you're stating seems borne out of this same kind of prejudice.

Buffered I/O can be a magical thing.

The mighty Commodore 64 did pretty well with serially connected drives even in a market that already had access to DMA-based devices. Theory is on your side with respect to what is ideal, but the marketplace and theories about what it "needs" don't often play well together.

Great journeys begin with a single step but one must take that step. Even IBM released their first PCs with cassette tape storage. You'll not go far trying to score a hole-shot in today's market. Look what it did for NeXT.

The hardware is there and the software has demonstrated itself to a large extent. This should be enough for a pre-release. Remember that this is not a release candidate that we're talking about. Microsoft has been showing "Longhorn" for months now, but they don't plan on releasing it for at least three years. They've even done pre-releases of Service Pack 2 for Windows XP even though they lack imperative security measures that are promised in the release (I'm guessing they don't want the security defeated before the patch is released).

I've been tinkering with Amigas since Kickstarts were numbered in the upper .8x range. Back then, the argument was the same: The Atari 520ST was based on CP/M68K (a proven operating system based on the venerable CP/M for the Intel 8080 from which DOS was unabashedly ripped off), DMA hard drive access, VGA display, Yamaha sound chip, DOS format floppies and VT52 emulation built into the boot ROM. Clearly it would quickly and decisively win the battle against the Amiga 1000... Does anyone remember the Apple IIgs???

I submit that your theory about complete DMA support in the pre-release being critical to the long-term success of OS4 and the next generation of Amiga computers is similarly destined.

@harsh

100% correct. There's something else that AmigaOS has that is pretty unique - MaxTransfer. By transferring data in bursts, you get no loss of interactivity even when using polled or programmed IO devices. Anyone who used an A1200 or A4000 (with their programmed IO IDE controllers) will attest that amigas don't freeze up when using PIO modes, even under heavy load. There's no denying that DMA is an improvement due to lower CPU use, but when you're running AmigaOS that is basically the only improvement.

@Darth_X

Quote:

hehe.. next big reality tv show: Hyperion OS4 development team!

Ala "American Chopper", I can see it now "Belgian Programmers". Just hope that it is better than it sounds.

_________________

@CodeSmith

Quote:

It might be a good idea to, once you have made the IDE DMA driver, send it off to the earlybirds to try on their systems

We'll see than

Quote:

OBTW, the radeon PLL thing - was it an endianness issue when copying the BIOS tables to the video driver? I've got a bet with a friend, and I'd like to know who won

Sorry to disappoint you, but that would have been too easy The PLL values where actually right from the beginning, it was something about the order in which they where written. Don't ask me for details, I don't know them. But the good thing is they're working now.

_________________
Seriously, if you want to contact me do not bother sending me a PM here. Write me a mail

@Argo

Quote:

Ala "American Chopper", I can see it now "Belgian Programmers". Just hope that it is better than it sounds.

(Assumes Hercule Poirot voice) I'm a german, not a belgian.

_________________
Seriously, if you want to contact me do not bother sending me a PM here. Write me a mail

@Rogue

Quote:

But the good thing is they're working now.

I think this calls for a dancing nana.



Wonderful.

Chris

_________________
"Miracles we do at once, the impossible takes a little longer" - AJS on Hyperion
Avatar is Tabitha by Eric W Schwartz