Welcome, Guest. Please login or register.

Author Topic: To Vampire or not  (Read 3753 times)

Description:

0 Members and 1 Guest are viewing this topic.

Offline Iggy

  • Hero Member
  • *****
  • Join Date: Aug 2009
  • Posts: 5348
    • Show only replies by Iggy
Re: To Vampire or not
« Reply #149 from previous page: June 17, 2017, 11:42:47 PM »
Quote from: Kremlar;827279
Priority is the reason.  With tons of users clamoring for AGA support and native display over HDMI, plus the drive to release a standalone product, why would he focus on an FPU which only a handful of users need?  Especially when they are selling them as fast as they can make them, why would you expect them to deviate from their current path?

They are focusing on what they, and most users, consider most important first.


Handful of users? FPU capability would have utility for ALL users.
Although I will admit AGA has appeal, certainly more than new multimedia instructions or 16 bit sound (both features that aren't supported by ANY existing software unless you consider NG software).
"Not making any hard and fast rules means that the moderators can use their good judgment in moderation, and we think the results speak for themselves." - Amiga.org, terms of service

"You, got to stem the evil tide, and keep it on the the inside" - Rogers Waters

"God was never on your side" - Lemmy

Amiga! "Our appeal has become more selective"
 

Offline OldAmigan

Re: To Vampire or not
« Reply #150 on: June 18, 2017, 12:01:41 AM »
Hi Iggy,

So where do we find these existing routines for FPU ?
 

Offline kolla

Re: To Vampire or not
« Reply #151 on: June 18, 2017, 12:23:42 AM »
Another binary causing 8000 000B on Vampire...

http://aminet.net/package/util/libs/WizardLibrary
B5D6A1D019D5D45BCC56F4782AC220D8B3E2A6CC
---
A3000/060CSPPC+CVPPC/128MB + 256MB BigRAM/Deneb USB
A4000/CS060/Mediator4000Di/Voodoo5/128MB
A1200/Blz1260/IndyAGA/192MB
A1200/Blz1260/64MB
A1200/Blz1230III/32MB
A1200/ACA1221
A600/V600v2/Subway USB
A600/Apollo630/32MB
A600/A6095
CD32/SX32/32MB/Plipbox
CD32/TF328
A500/V500v2
A500/MTec520
CDTV
MiSTer, MiST, FleaFPGAs and original Minimig
Peg1, SAM440 and Mac minis with MorphOS
 

Offline Iggy

  • Hero Member
  • *****
  • Join Date: Aug 2009
  • Posts: 5348
    • Show only replies by Iggy
Re: To Vampire or not
« Reply #152 on: June 18, 2017, 12:52:47 AM »
Quote from: OldAmigan;827282
Hi Iggy,

So where do we find these existing routines for FPU ?

Seriously?
In many MP3 players, a few video games I can think of (like ClickBoom's port of Quake), most 3D graphic software...
If Gunnar is having a problem finding packages that can use fpu routines, he isn't trying very hard.
« Last Edit: June 18, 2017, 01:09:42 AM by Iggy »
"Not making any hard and fast rules means that the moderators can use their good judgment in moderation, and we think the results speak for themselves." - Amiga.org, terms of service

"You, got to stem the evil tide, and keep it on the the inside" - Rogers Waters

"God was never on your side" - Lemmy

Amiga! "Our appeal has become more selective"
 

Offline Crom00

  • Hero Member
  • *****
  • Join Date: Nov 2005
  • Posts: 1234
    • Show only replies by Crom00
Re: To Vampire or not
« Reply #153 on: June 18, 2017, 03:45:20 AM »
The team views AGA and chipset dev priority right now. I have seen the team work and they know their stuff. Look how much has been done by so few.
 

Offline Niding

  • Hero Member
  • *****
  • Join Date: Sep 2004
  • Posts: 566
    • Show only replies by Niding
Re: To Vampire or not
« Reply #154 on: June 18, 2017, 08:16:47 AM »
Quote from: Iggy;827278
Yes it is BOTH, and as a "drama queen", I haven't bought one yet.
Further, while you and Niding my both be unapologetic fanboys, its pretty easy to poke holes in your statements.

Gunnar didn't need other programmers or code to test an fpu, there are plenty of available packages that use floating point routines already in existence.

If he's having trouble getting it done, just admit it.
Excuses about priorities or statements about the current purchasers are merely obfuscation.

And, as Niding admitted, its obviously a useful feature, as even with its higher clock speed, a Vampire takes MUCH longer to perform fpu commands via emulation.

As an enduser with virtually no coding skills, and absolutely no developer skill/expirience I have no idea what Gunnars request entails. He did point out that it would require committing to significant time offered and asm.
Noone has taken him up on it.
Im not going to pretend I know what he needs from the coder, do you?

Either way, with noone stepping up, they seem intent on following their own priorities, which I see no need to be negative about, since I contribute absolutly zero to the work, which I think is the case for the majority of us.

Why you feel inclined to call me a fanboi just cause I accept they follow their own internal map is beyond me.
I could rant and rave about xyz not being done yet, but it wont make development process go any faster, and it just demoralizes the team. They are actually humans afterall.
 

Offline ALB42

Re: To Vampire or not
« Reply #155 on: June 18, 2017, 10:00:12 AM »
Quote from: Niding;827293
As an enduser with virtually no coding skills, and absolutely no developer skill/expirience I have no idea what Gunnars request entails. He did point out that it would require committing to significant time offered and asm.
Noone has taken him up on it.
Im not going to pretend I know what he needs from the coder, do you?

Either way, with noone stepping up, they seem intent on following their own priorities, which I see no need to be negative about, since I contribute absolutly zero to the work, which I think is the case for the majority of us.

Why you feel inclined to call me a fanboi just cause I accept they follow their own internal map is beyond me.
I could rant and rave about xyz not being done yet, but it wont make development process go any faster, and it just demoralizes the team. They are actually humans afterall.

Why noone step up to do it. Because he demands something the most people do not want:

see here this old "Deleted" apollo FPU discussion

http://apollo-core.com/knowledge.php?b=4¬e=5385&order=&x=5

see the 3rd entry.
he wants something like RIVA, so an incompatible Powerapplication only fast/only work on Apollo therefore it has to be plain written in assembler. when you browse further in this thread you will find a lot offers for routines (raytracing, 3d stuff usually) to test the implementation.
I even would do that, even in assembler if I have to be... but only to develop an incompatible new instruction set like AMMX.. no thanks.
 

guest11527

  • Guest
Re: To Vampire or not
« Reply #156 on: June 18, 2017, 10:11:38 AM »
Quote from: OldAmigan;827282
So where do we find these existing routines for FPU ?

As far as I understand, there is currently a prototype 80-bit (extended precision) FPU which supports the basic operations +,-,*,/ and sqrt, but none of the transcendental math functions of the 68881/882. Support functions for the latter are certainly available, you can still download the 68060 FPSP (floating point support package) from Motorola/Freescale/whatever they are called today.

From what I understand the 80bit FPU is, however, no longer on Gunnar's agenda. Here is more looking into a 64 bit (double precision only) FPU for which some FPGAs offer direct support (i.e. no need to sacrifice gates for it as it is already pre-routed on the chip). The advantage would be higher speed, and a high degree of parallelism (i.e. vector units).

Drawback is that it is only 64 bit. It makes probably not much of a practical difference (mathieeedoubbas is 64bit only, too), though for the 68060FPSP, you'd likely need extended precision in some places internally to support 64 bit double precision as output.

I talked with Gunnar about the possibilities (full floating point support with software as in the 68060 with assembler support, full floating point support with a microcode engine using CORDIC) and I personally would prefer the 80bit unit even it is slower due to full compatibilitiy.

For example, if the currently existing 80 bit FPU would offer the same software traps as the 68060, it would be a rather immediate step to port the 68060FPSP over, though the software layer could be a bit simpler with smarter hardware. Due to the simpicity of the 68060 stack frame, the 68060 has to decode all unsupported FPU instructions by hand, including address resolution and operand lookup - which costs more performance than necessary. The current core could for example already decode operands, avoiding the double work.

In the end, it got nowhere since it was not really clear to me where Gunnar would like to go (probably 64 bit only and ditch the existing 80 bit core - but then what about the transcendental functions?).
 

Offline kolla

Re: To Vampire or not
« Reply #157 on: June 18, 2017, 10:30:33 AM »
People don't grasp that the FPU Gunnar wants is not the FPU found in 060, 040 or 68882 - he wants to build a new one, his own super duper FPU that only works with software written specifically for it. Like AMMX. No support for existing compilers, only assembler. And he wants someone to step up and write a killer app for it, so its superiority can be demonstrated. Well, I have Riva, the AMMX killer app, and it is rather lame as it only plays mpeg1, it's like having an old peggy+ card. Using other players, like Frogger, to play other formats, and it is back to a crawl. My 060 Amigas with FPU can play a heck lot more formats at much more acceptable speed. It boils down to having every piece if software ever written for FPU rewritten for either AMMX or the new FPU, should it ever materialize, or that someone writes a emulation layer that makes use of the new FPU (and/or AMMX). In the meantime, Gunnar likes to point out that they do have a fully working and functional 68882 implementation ready, but that he doesn't like it, as it doesn't demonstrate the true raw power of apollo core.

Yeah, well, whatever. If you want to experience a truly fast and compatible Amiga setup, there is still UAE.
B5D6A1D019D5D45BCC56F4782AC220D8B3E2A6CC
---
A3000/060CSPPC+CVPPC/128MB + 256MB BigRAM/Deneb USB
A4000/CS060/Mediator4000Di/Voodoo5/128MB
A1200/Blz1260/IndyAGA/192MB
A1200/Blz1260/64MB
A1200/Blz1230III/32MB
A1200/ACA1221
A600/V600v2/Subway USB
A600/Apollo630/32MB
A600/A6095
CD32/SX32/32MB/Plipbox
CD32/TF328
A500/V500v2
A500/MTec520
CDTV
MiSTer, MiST, FleaFPGAs and original Minimig
Peg1, SAM440 and Mac minis with MorphOS
 

Offline Zooz

  • Newbie
  • *
  • Join Date: Nov 2014
  • Posts: 23
    • Show only replies by Zooz
Re: To Vampire or not
« Reply #158 on: June 18, 2017, 11:47:47 AM »
So many speculations here.

Of course that the team knows what to test for FPU. Members are not idiots, they are Amiga users. And for sharper tests they also have testcases, and knows how to write validation code. Just it takes time. When FPU will be on the work in the core, progress will be fast. At moment, the team works on something else, so we could speaks of this during 10 pages again that wont change _current_ roadmap. Let the team finish their current work instead of feeding speculations. For example, the internally 64bits or even 128bits float processing is just speculation; that this have been discussed does not mean it will. That were early discussions. That gunnar is not good in communication is a fact but that does not mean he can't change his mind, the team is more opened than that. Also, i have to tell again that FPU exists but does not fit well in the current shape and chip, but could with some strategy. You guys, Team, Members, and Users wants FPU, so there will, it can't be another way. Only thing that is true and reality is The "How" which is not yet answered.

Although I will admit AGA has appeal, certainly more than new multimedia instructions or 16 bit sound (both features that aren't supported by ANY existing software unless you consider NG software).

About the Paula improvements, there is software yet, EaglePlayer developer is working on a Pamela-Amplifier that is already working well, it is able to output modules (mods/xm/digi/...) in digital 16bits. That is not vapor-feature. It is used as a validation testcase and will be nice feature for vampire owners. Consider it brings to Vampire some Sound-card abilities, like any new soundcard on Amiga, there is no existing software, at starts, just it needs drivers or engines (like a AHI driver for example). So it seems really naysaying on this particular point.

Please, just wait that team can focus on this FPU before so much discussions.
« Last Edit: June 18, 2017, 12:24:33 PM by Zooz »
 

Offline Iggy

  • Hero Member
  • *****
  • Join Date: Aug 2009
  • Posts: 5348
    • Show only replies by Iggy
Re: To Vampire or not
« Reply #159 on: June 18, 2017, 05:46:14 PM »
Quote from: Zooz;827300
So many speculations here....
Please, just wait that team can focus on this FPU before so much discussions.


OK, I can wait.
I'm not sure how Kolla feels about it (having already made his purchase).
One thing I do hope, is that they will not forego legacy compatibility.

That would, in my opinion, reduce the value of the device.
"Not making any hard and fast rules means that the moderators can use their good judgment in moderation, and we think the results speak for themselves." - Amiga.org, terms of service

"You, got to stem the evil tide, and keep it on the the inside" - Rogers Waters

"God was never on your side" - Lemmy

Amiga! "Our appeal has become more selective"
 

Offline grond

  • Full Member
  • ***
  • Join Date: Feb 2016
  • Posts: 154
    • Show only replies by grond
Re: To Vampire or not
« Reply #160 on: June 19, 2017, 10:31:13 AM »
Quote from: kolla;827297
People don't grasp that the FPU Gunnar wants is not the FPU found in 060, 040 or 68882 - he wants to build a new one, his own super duper FPU that only works with software written specifically for it. Like AMMX. No support for existing compilers, only assembler. And he wants someone to step up and write a killer app for it, so its superiority can be demonstrated.

You understood that right but draw the wrong conclusions. If you have a superfast FPU in hardware, it is easy to map the standard FPU to it. If you do the standard FPU first and then the superfast FPU, it will mean double work.  The prioritisation of the AGA implementation over the FPU also has a lot to do with the stand-alone coming out soon.
 

guest11527

  • Guest
Re: To Vampire or not
« Reply #161 on: June 19, 2017, 11:11:49 AM »
Quote from: grond;827322
If you have a superfast FPU in hardware, it is easy to map the standard FPU to it.
Actually, not quite so. The problem is that the FPUs that come with (some) FPGAs are only 64 bit FPUs, but the Motorola FPUs are 80 bit wide. For most applications, this will not make much of a difference (but frankly, most applications do not need a FPU in first place, so the discussion is a bit academic if you ask me).

Now, while it is probably correct that programs will typically not notice the difference, there *may* be an impact in the transcendental math functions. If you expect results with error bounds in double precision, you may need intermediate results with a higher precision such as 80 bits internally, IOWs, either the CORDIC engine or the FPSP requires a bit more than 64 bits.

Now, again, one can argue why one would need such instructions in first place, and if so, why in double precision. Again, see my answer above: I personally do not see too much need in the FPU anyhow. There is mathieeedoubbas and mathieeedoubtrans, and both work fine without FPU in first place (though, the implementation CPU only implementation is rather weak).
 

Offline grond

  • Full Member
  • ***
  • Join Date: Feb 2016
  • Posts: 154
    • Show only replies by grond
Re: To Vampire or not
« Reply #162 on: June 19, 2017, 11:42:50 AM »
Quote from: Thomas Richter;827323
Actually, not quite so. The problem is that the FPUs that come with (some) FPGAs are only 64 bit FPUs, but the Motorola FPUs are 80 bit wide.

Well, your statement doesn't negate mine, it renders it more precisely as I made the statement without explicitly stating that the legacy FPU mapped onto the new FPU Hardware would then, of course, only have 64bits of precision.

Quote from: Thomas Richter;827323
For most applications, this will not make much of a difference

AFAIK UAE maps the Motorola FPU instructions to the corresponding Intel processor instructions which have 64bit precision at most and nobody has ever noticed any problem with any Amiga FPU software. 64bit of precision is still an awful lot. I think it is safe to assume that the lost 16 bits of precision will not make any difference at all.
 

Offline IanP

  • Full Member
  • ***
  • Join Date: Aug 2013
  • Posts: 132
    • Show only replies by IanP
Re: To Vampire or not
« Reply #163 on: June 19, 2017, 12:50:38 PM »
Quote from: Iggy;827284
Seriously?
In many MP3 players, a few video games I can think of (like ClickBoom's port of Quake), most 3D graphic software...
If Gunnar is having a problem finding packages that can use fpu routines, he isn't trying very hard.
Plenty of MP3 player options that don't need an FPU to choose from and with 16bit sound likely to be added to the Vampires some players may be considered obsolete if not capable of 16bit support.

I would think that there would only be a handful of games that need an FPU.

I would also think there are only an handful of masochists still using an Amiga for 3D graphics work.

The main use of an FPU is likely in the demo scene. Anecdotally much of the hardcore demo scene community are said to be "not fans" of the Vampire as it's not original Amiga/Motorola hardware (even though they may be doing most of their development on PCs). I personally, while appreciating the skills of the demo creators have no interest in installing and running the demos, I'd just watch online videos if interested in them for some reason.

Given the circumstances it's no wonder that implementing an FPU on the Vampires is not the highest priority.
 

Offline BozzerBigD

Re: To Vampire or not
« Reply #164 on: June 19, 2017, 01:22:47 PM »
@IanP

Quote
I would also think there are only an handful of masochists still using an Amiga for 3D graphics work.

I think that would be quite a fun use of a Vampire card. Some people still enjoy messing around with Lightwave on the X1000/5000 machines so why not on the Classics? Plus FPU is useful for TFX, DrawStudio 2 and now Beats of Rage!
"Art challenges technology. Technology inspires the art."

John Lasseter, Co-Founder of Pixar Animation Studios