| |
Fun things to do over the weekend |
| |
Phlamethrower (14:05 16/8/2008)
Phlamethrower (12:58 19/8/2008)
andrew (21:42 19/8/2008)
Phlamethrower (22:18 19/8/2008)
andrew (23:07 19/8/2008)
Phlamethrower (00:15 20/8/2008)
andrew (22:42 20/8/2008)
adrianl (11:02 21/8/2008)
Phlamethrower (12:12 21/8/2008)
adrianl (20:56 21/8/2008)
Phlamethrower (22:32 21/8/2008)
Wrath (07:27 24/8/2008)
|
| |
|
Jeffrey Lee |
Message #108024, posted by Phlamethrower at 14:05, 16/8/2008 |
 Hot Hot Hot Hot Hot Hot Hot Hot Hot Hot Hot Hot Hot stuff
Posts: 15100
|
1. Party like it's 1993
2. Discover that Red Squirrel isn't quite good enough to pull off the timing tricks for the palette swaps |
| |
 [ Log in to reply ] |
| |
|
Jeffrey Lee |
Message #108071, posted by Phlamethrower at 12:58, 19/8/2008, in reply to message #108024 |
Hot Hot Hot Hot Hot Hot Hot Hot Hot Hot Hot Hot Hot stuff
Posts: 15100
|
Of course, now I'm sat here wondering how good it would be possible to make a game like that look if you were to write it now on an Iyonix. |
| |
[ Log in to reply ] |
| |
|
Andrew |
Message #108085, posted by andrew at 21:42, 19/8/2008, in reply to message #108071 |
 Handbag Boi
Posts: 3439
|
Bare minimum SF3000:TNG resolution, effects and colours. |
| |
[ Log in to reply ] |
| |
|
Jeffrey Lee |
Message #108087, posted by Phlamethrower at 22:18, 19/8/2008, in reply to message #108085 |
Hot Hot Hot Hot Hot Hot Hot Hot Hot Hot Hot Hot Hot stuff
Posts: 15100
|
Hopefully something a bit better than that - I'm thinking more along the lines of a one-up on what Iron Dignity would have been.
If you stick to a game engine that's basically the same as Lotus II/E-Type/etc (i.e. all it does is render each scanline at a different X offset to create the illusion of the road bending) then you'll have a lot of horsepower left over to do graphical effects on the road surface. You could probably do bump-mapping, highlights from the sun, reflections of cars in puddles, etc. without running into any real problems.
Or you could just stick with an engine identical to the old one and run it at 1920x1200 instead (which could be quite a fun bit of code to write too - but perhaps a bit trickier since it would probably have to emulate their palette swapping system so partial screen redraws can be used to get around the slow PCI write speed) |
| |
[ Log in to reply ] |
| |
|
Andrew |
Message #108090, posted by andrew at 23:07, 19/8/2008, in reply to message #108087 |
Handbag Boi
Posts: 3439
|
SF:TNG cripples the Iyonix with any optional graphics enhancement turned on. I didn't realise the Iyonix had particularly slow video speed though. I thought it was with the game being a port or not quite finished. Iron Dignity set the bar for graphics you're right - it ran on a RiscPC! |
| |
[ Log in to reply ] |
| |
|
Jeffrey Lee |
Message #108092, posted by Phlamethrower at 00:15, 20/8/2008, in reply to message #108090 |
Hot Hot Hot Hot Hot Hot Hot Hot Hot Hot Hot Hot Hot stuff
Posts: 15100
|
My numbers are in no way 100% scientific, but using a tight loop of assembler:
* A RiscPC can read and write RAM at around 36MB/sec
* A RiscPC can write to VRAM at around 20MB/sec. Read speed seems to be only 12.5MB/sec.
* An Iyonix seems to be able to read from RAM at around 160MB/sec and write at around 114MB/sec.
* An Iyonix can write to PCI RAM at around 50MB/sec, which is quite bad as it is when compared to the RiscPC's RAM:VRAM ratio. But read speed (something which I've only just tested today) seems even worse - an absolutely abysmal 5.6MB/sec. I've checked this a few times (including using *save to save a fair-size chunk of screen memory to a RAM disc) and it seems correct.
So I guess if the SF3000:TNG enhancements involve extra reads from screen memory then it's to be expected for it to severly impact performance.
If the IOP's DMA channels could be used without causing audio corruption then things would be a lot better. |
| |
[ Log in to reply ] |
| |
|
Andrew |
Message #108107, posted by andrew at 22:42, 20/8/2008, in reply to message #108092 |
Handbag Boi
Posts: 3439
|
This needs raising on the Iyonix list and the newsgroups, IMO. |
| |
[ Log in to reply ] |
| |
|
Adrian Lees |
Message #108108, posted by adrianl at 11:02, 21/8/2008, in reply to message #108092 |
Member
Posts: 1637
|
All of those numbers can be increased using PLD when reading from SDRAM and DMA when writing to screen, and - amusingly - it's better to save to HD rather than RAM if you want to measure screen read speed because the RAMFS code is appallingly slow (IIRC about 17MB/s loading a large file, I have a local tweak somewhere that makes it use the AAU for large transfers and it's much, much faster).
If TNG is reading from the video memory it should be updated, it's going to be pretty poor on any modern PC architecture (being the overcomplicated beasts that they are!).
As for the audio corruption, I've not looked into that further... it could just be something really simple like a need to change the threshold settings in the audio DMA master. |
| |
[ Log in to reply ] |
| |
|
Jeffrey Lee |
Message #108109, posted by Phlamethrower at 12:12, 21/8/2008, in reply to message #108108 |
Hot Hot Hot Hot Hot Hot Hot Hot Hot Hot Hot Hot Hot stuff
Posts: 15100
|
As for the audio corruption, I've not looked into that further... it could just be something really simple like a need to change the threshold settings in the audio DMA master. Well tell me how to do that (or point me towards the right docs) and I'll look at it  |
| |
[ Log in to reply ] |
| |
|
Adrian Lees |
Message #108111, posted by adrianl at 20:56, 21/8/2008, in reply to message #108109 |
Member
Posts: 1637
|
Well, now, you'd be needing documentation on the ALi M1535+ South Bridge, and though I have it I don't think it's freely redistributable. It was once readily accessible on a FTP site but I can't find any mention on the web atm, so happy hunting. |
| |
[ Log in to reply ] |
| |
|
Jeffrey Lee |
Message #108113, posted by Phlamethrower at 22:32, 21/8/2008, in reply to message #108111 |
Hot Hot Hot Hot Hot Hot Hot Hot Hot Hot Hot Hot Hot stuff
Posts: 15100
|
Well, now, you'd be needing documentation on the ALi M1535+ South Bridge, and though I have it I don't think it's freely redistributable. Confidentiality of other people's documentation doesn't seem to worry nvidia - http://www.nvidia.com/docs/CP/29993/m1535dm_web.pdf
Unfortunately that doc doesn't contain any useful information 
I've had a look at some linux sources, and the IR driver sets the DMA threshold, but the sound driver doesn't. It also doesn't define the address of the threshold register. So either the audio DMA threshold can't be changed, or I'm looking in the wrong place, or they just haven't put any code in to change it since it seems to work OK on linux. |
| |
[ Log in to reply ] |
| |
|
Nathan |
Message #108138, posted by Wrath at 07:27, 24/8/2008, in reply to message #108090 |
Member
Posts: 155
|
SF:TNG cripples the Iyonix with any optional graphics enhancement turned on. I didn't realise the Iyonix had particularly slow video speed though. I thought it was with the game being a port or not quite finished. Iron Dignity set the bar for graphics you're right - it ran on a RiscPC! SF:TNG was/is finished. It was converted to be played on the Omega and was going to have specifics in there to only work with the Omega. This is why the intro has the FlaYmz group stuff on it as FlaYmz were tied to MicroDigital.
I must say that Lee Noar did a FANTASTIC job on the conversion. It's amazing it got done at all, the majority of the 3DO documentation is patented and not released so Lee had to do a fair amount of guess work. I do know that the video side of things was complicated. |
| |
[ Log in to reply ] |
| |
|
