horacefithers
New member
Burnout!
It's been at least a month (and probably more) since I last fired up surveyor or driver to do "something" useful on the BC&SJ (or anything else for that matter...)
I guess spending most waking hours that weren't absolutely required for other stuff working on the route caught up with me.
Right now I'm trying to resurrect a ray tracing program I wrote for the Commodore Amiga back in the early '90s -
RayDance http://s145079212.onlinehome.us/raydance/index.htm.
It's slow going. I think aging is not the friend of a programmer, at least not as I close in on 70. It feels that everything takes much longer than it used to and
this is despite Visual Studio 2017 having a much better UI, compiler, and debugger than was ever available on the Amiga (the compiler there was so bad
that all too frequently I would rewrite critical C functions in 68000 machine code and sometimes see up to a 30% speed improvement.
That would be impossible these days given how good modern compilers are with code optimization. Plus now I have a fairly fire-breathing computer with
triple monitors so having a dozen windows open isn't a problem.
The big problem I'm dealing with is trying to optimize Binary Spacial Partitioning algorithms to get make it render faster. Yeah, it's blazing fast compared with
the 25mhz 68030/68882 cpu in the Amiga, but I feel like I should be able to do better. So I'm re-writing that code. Again. Maybe this time I'll be able to
optimize the BSP trees it generates enough to see a significant speed improvement (I'd love to see a factor of 2x - the current code has problems with
tree threads to go too deep and especially with having too many primitives (triangles & spheres) in a give bounding volume. The original program used a
hierarchical bounding box optimization system. Maybe I should go back to that?
Anyway, sometimes I still think about what to do next with Trainz. I think that when the time comes I'd like to do a DEM route, but which one? Something not
too long (105 miles was a bit much!), but with interesting scenery and perhaps some ops potential. Maybe go more modern, too. Perhaps '80s instead of the '50s?
Although I've been a closet South Pacific Coast nut for decades. Doing a SPC route would be cool, but the lack of good 3' narrow gauge track would be a problem
as would the need to scratch build most structures. I've been really admiring what Jango has done with the Gilpin Tram route and that's (really) narrow gauge!
HF
It's been at least a month (and probably more) since I last fired up surveyor or driver to do "something" useful on the BC&SJ (or anything else for that matter...)
I guess spending most waking hours that weren't absolutely required for other stuff working on the route caught up with me.
Right now I'm trying to resurrect a ray tracing program I wrote for the Commodore Amiga back in the early '90s -
RayDance http://s145079212.onlinehome.us/raydance/index.htm.
It's slow going. I think aging is not the friend of a programmer, at least not as I close in on 70. It feels that everything takes much longer than it used to and
this is despite Visual Studio 2017 having a much better UI, compiler, and debugger than was ever available on the Amiga (the compiler there was so bad
that all too frequently I would rewrite critical C functions in 68000 machine code and sometimes see up to a 30% speed improvement.
That would be impossible these days given how good modern compilers are with code optimization. Plus now I have a fairly fire-breathing computer with
triple monitors so having a dozen windows open isn't a problem.
The big problem I'm dealing with is trying to optimize Binary Spacial Partitioning algorithms to get make it render faster. Yeah, it's blazing fast compared with
the 25mhz 68030/68882 cpu in the Amiga, but I feel like I should be able to do better. So I'm re-writing that code. Again. Maybe this time I'll be able to
optimize the BSP trees it generates enough to see a significant speed improvement (I'd love to see a factor of 2x - the current code has problems with
tree threads to go too deep and especially with having too many primitives (triangles & spheres) in a give bounding volume. The original program used a
hierarchical bounding box optimization system. Maybe I should go back to that?
Anyway, sometimes I still think about what to do next with Trainz. I think that when the time comes I'd like to do a DEM route, but which one? Something not
too long (105 miles was a bit much!), but with interesting scenery and perhaps some ops potential. Maybe go more modern, too. Perhaps '80s instead of the '50s?
Although I've been a closet South Pacific Coast nut for decades. Doing a SPC route would be cool, but the lack of good 3' narrow gauge track would be a problem
as would the need to scratch build most structures. I've been really admiring what Jango has done with the Gilpin Tram route and that's (really) narrow gauge!
HF