- Reaction score
- 1,697
Problem is he probably forgot alot of it since it was so long ago. Check out this article
where he talks about programming the NUON just a FYI this was not on the latest hardware Aires 3 at this time.
-----------------------------------------------------
Gamasutra: When I first met you, this was E3 maybe 1999 or 2000, I saw you at the Nuon booth, and remarked that Tempest 3000 [a Nuon launch title] seemed a lot like Tempest 2000 to me, but I appreciate that a lot more work went into it. I know it makes you mad when people say that.
JM: Well, Tempest 3000 is an evolution of Tempest 2000, really. In a way it was hamstrung by the fact that it was on quite an underpowered system. It was a real struggle to get that out of the Nuon CPU. The Nuon was "interesting" to program on, let's put it that way.
GS: Did anyone at [Nuon maker] VM Labs ever think that they were going to defeat the PlayStation?
JM: What they wanted to do was create a sideways market to the PlayStation. They wanted to get in there early and have a DVD player with additional capabilities.
The trouble with VM Labs is that they ended up being about a year late to market. That really put the kibosh on it. Not enough people took it up. If they had got it into OEMs earlier, I suspect they might have done better.
It really was quite a feat of programming to get the best out of it. If you just coded it with the standard libraries, you didn't really get a lot out of it. To get stuff out of Tempest 3000, I really had to work. That was probably the most difficult job of programming I've ever done.
GS: Really?
JM: Yeah, it was doubly parallel. For one thing, it was a VLIW architecture, which meant that for each instruction tick, you had to compose the instruction itself with little sub instructions sent to the ALU, to the memory unity, and each little part of the chip. You had to order the instructions yourself.
Plus, it wasn't just one CPU. It was an array of CPUs. If you wanted to use them all, you had to make your code so that it was never bigger than 4k, because the CPUs only had 4k of on-chip RAM. So you had to have paging schemes that paged all this in and out. Plus, you had to have a load management scheme which parsed out all the bits thrown at all the CPUs. I actually managed to make Tempest scalable.
Theoretically, if you put it on a system with 16 CPUs, it would have used them all to run more smoothly. But, in order to do that, you really had to understand how this chip worked. Of course, I worked at VM Labs, and I was there throughout its development, so I had a map of that chip in my head at all times so I could do it. But, for most people coming to develop a game, unless you wanted to get beyond just using the simple C+ libraries, most people probably wouldn't have bothered, because it was really hard work to do.
GS: And in fact most people didn't.
JM: Most people didn't, and that's why most games were a bit crap!
GS: How do you feel about the time you spent at VM Labs?
JM: I enjoyed my time at VM Labs. It was interesting because I got to see the birth of a piece of silicon and work very closely with the hardware engineers. I worked with some great guys, I had some good times. I'm proud of Tempest 3000 because on that hardware, it's pretty damn good.
Related:
Llamas In Space: Catching Up with Llamasoft's Jeff Minter
Gamasutra speaks to Llamasoft founder Jeff Minter about the upcoming Space Giraffe for Xbox 360, his work ethic, his involvement with the ill-fated Nuon 'console', and his pet sheep, in this in-depth interview.
www.gamedeveloper.com
where he talks about programming the NUON just a FYI this was not on the latest hardware Aires 3 at this time.
-----------------------------------------------------
Gamasutra: When I first met you, this was E3 maybe 1999 or 2000, I saw you at the Nuon booth, and remarked that Tempest 3000 [a Nuon launch title] seemed a lot like Tempest 2000 to me, but I appreciate that a lot more work went into it. I know it makes you mad when people say that.
JM: Well, Tempest 3000 is an evolution of Tempest 2000, really. In a way it was hamstrung by the fact that it was on quite an underpowered system. It was a real struggle to get that out of the Nuon CPU. The Nuon was "interesting" to program on, let's put it that way.
GS: Did anyone at [Nuon maker] VM Labs ever think that they were going to defeat the PlayStation?
JM: What they wanted to do was create a sideways market to the PlayStation. They wanted to get in there early and have a DVD player with additional capabilities.
The trouble with VM Labs is that they ended up being about a year late to market. That really put the kibosh on it. Not enough people took it up. If they had got it into OEMs earlier, I suspect they might have done better.
It really was quite a feat of programming to get the best out of it. If you just coded it with the standard libraries, you didn't really get a lot out of it. To get stuff out of Tempest 3000, I really had to work. That was probably the most difficult job of programming I've ever done.
GS: Really?
JM: Yeah, it was doubly parallel. For one thing, it was a VLIW architecture, which meant that for each instruction tick, you had to compose the instruction itself with little sub instructions sent to the ALU, to the memory unity, and each little part of the chip. You had to order the instructions yourself.
Plus, it wasn't just one CPU. It was an array of CPUs. If you wanted to use them all, you had to make your code so that it was never bigger than 4k, because the CPUs only had 4k of on-chip RAM. So you had to have paging schemes that paged all this in and out. Plus, you had to have a load management scheme which parsed out all the bits thrown at all the CPUs. I actually managed to make Tempest scalable.
Theoretically, if you put it on a system with 16 CPUs, it would have used them all to run more smoothly. But, in order to do that, you really had to understand how this chip worked. Of course, I worked at VM Labs, and I was there throughout its development, so I had a map of that chip in my head at all times so I could do it. But, for most people coming to develop a game, unless you wanted to get beyond just using the simple C+ libraries, most people probably wouldn't have bothered, because it was really hard work to do.
GS: And in fact most people didn't.
JM: Most people didn't, and that's why most games were a bit crap!
GS: How do you feel about the time you spent at VM Labs?
JM: I enjoyed my time at VM Labs. It was interesting because I got to see the birth of a piece of silicon and work very closely with the hardware engineers. I worked with some great guys, I had some good times. I'm proud of Tempest 3000 because on that hardware, it's pretty damn good.
Llamas In Space: Catching Up with Llamasoft's Jeff Minter
Gamasutra speaks to Llamasoft founder Jeff Minter about the upcoming Space Giraffe for Xbox 360, his work ethic, his involvement with the ill-fated Nuon 'console', and his pet sheep, in this in-depth interview.
www.gamedeveloper.com
Related:
Jeff Minter's Sprite Library and the time VM Labs hired him to help with the SDK for homebrew
Back in the last days of the NUON VM Labs hired Jeff Minter to go in and help out with the SDK effort that VM Labs was doing to try and get homebrew going. I remember he came on the forums for a while and whipped up a sprite library but I never heard of anyone ever using it. Does anyone...
www.thehelper.net
Last edited: