Re: Character strings

New Message Reply Date view Thread view Subject view Author view

Ryutarou OHBUCHI (ohbuchi++at++trl.ibm.co.jp)
Thu, 05 Jan 1995 10:22:16 +0900

>
> I suspect that Michael Jones only intended the billboard-versus-character
> strings in Performer as a EXAMPLE of the kinds of trade-off that he has
> to make in assigning priorities rather than suggesting that we would lose
> good billboards if we had character strings.

Yes.

>
> Given that; I believe that the priority for the Performer team should
> always be to implement things that are hard or impossible to do without their
> help.

I agree with Steve on this. However,

> In the case of character strings versus billboards there is already a
> simple mechanism to get character strings into Performer (I do it all the time)
> provided you have a reasonable modeller (eg MultiGen or DWB). Model each
> letter of the alphabet as a separate object, then instance each object as needed
> to make up a string - you could either do this by hand inside your modeller,
> or write a simple subroutine to do it inside Performer. In fact, MultiGen (at least)
> already has a couple of fonts modelled this way and a routine to let you
> just type text in. (Check out the 'T' icon in the middle of the icon bar).

My colleague did try the method you mentioned. Loading and
instancing characters gets a bit hairly if you have thousands (3k to
10k) of Kanji characters, each of which is significantly more
vertex-intensive than English alphabets. We tried texturing
approach as well, but modifying string is expensive. Both methods
are rather heavy. What is the best approach? Faster graphics
engine? (We use Onyx/RE2.)

>
> On the other hand, to do good billboards is pretty nasty - particularly if
> you have multiple channels - all looking from different directions with
> various channel offsets. Point-rotate billboards need some fairly hairy maths
> and there is no easy way to get the culling right if you 'do it yourself'.
>
> This is a clear case where we should let the Performer team do the difficult
> bits.

Character strings themselves, in fact, need to be billboards in many
applications, for example, annotating interactive 3D human anatomy.

>
> Finally - why pick text as the thing that Performer should generate automatically?
> I would personally prefer that they generate aircraft models automatically....

As Michael Jones suggested, how about implementation of character
support as a pfUtil? (Assuming this is possible...)

-----
Ryutarou Ohbuchi
IBM Japan, Ltd., SA-S76
1623-14, Shimo-tsuruma, Yamato-shi,
Kanagawa-ken 242, Japan
Phone: 0462-73-4882 Fax: 0462-73-7413
Internet: ohbuchi++at++trl.ibm.co.jp

New Message Reply Date view Thread view Subject view Author view

This archive was generated by hypermail 2.0b2 on Mon Aug 10 1998 - 17:50:49 PDT

This message has been cleansed for anti-spam protection. Replace '++at++' in any mail addresses with the '@' symbol.