In my implementation of a parallel path-tracer I divided up the image into four threads in an interleaved manner to maximize the spatial coherence of the...
yeehector@...
Jan 2, 2001 5:57 pm
202
Thanks for the info, but I still have a question. If the threads terminate at the end of each frame and I start 4 new threads at the start of a frame how do I...
Paul Kahler
phkahler@...
Jan 4, 2001 5:56 pm
203
Hi, You don't want to create/destroy threads on every frame, it's really expensive. It generaly makes sense to create a work queue and just have all the...
Peter-Pike Sloan
ppsloan@...
Jan 4, 2001 7:28 pm
204
Hi, this is my first post here, so a quick hello to everyone! ... You can use the NT/2000 specific API SetThreadAffinity() to lock a thread to a particular...
Rex Deathstar
deathstr@...
Jan 5, 2001 9:21 pm
205
Hiho, I didn't follow the thread very much, so don't hit me if the idea already came up :D What would be wrong with having (n=number of processors) n threads...
Shiva
shivak@...
Jan 5, 2001 11:06 pm
206
There are a couple of things wrong with this: 1) You can't interact - you'll have a huge latency 2) There can be a huge perf difference between frames -...
Peter-Pike Sloan
ppsloan@...
Jan 5, 2001 11:42 pm
207
Please forgive me for this un-helpful post :) ... Dude, it's great to see you're still in the scene! (I'm sure most of us remember "pureness" :) Keep going...
Mark Webster
mark@...
Jan 6, 2001 1:40 pm
208
Peter's right about not using a static frame distribution. Note: I know this is _not_ real time raytracing, but it's some things I learnt while writing a...
yeehector@...
Jan 8, 2001 1:04 am
209
... Quick question here. Do multiprocessor machines share any cache between processors? If so, you'd want to interleave pixels to maximize the benefits of...
Paul Kahler
phkahler@...
Jan 9, 2001 11:37 am
210
With global illumination and frame rates in terms of 1/minute and below, the cost of creating and terminating threads is insignificant =). Re: Shiva's point of...
yeehector@...
Jan 9, 2001 11:39 am
211
... edition ... identical ... availability). ... is ... times ... that's ... I own the student edition of MSV C/C++ 6.0 and this version has the same...
jesko@...
Jan 9, 2001 5:27 pm
212
Hi, I did a bit of research a few years ago on this topic (in 1996), by the way, there were published a lot of papers about it. There are different ...
havran
havran@...
Jan 10, 2001 8:51 am
213
Hello, it took me a little time for my decision to answer the question of automated load balancing with threads because it took reasonable time to develop it,...
Jesko Schwarzer
jesko@...
Jan 10, 2001 8:51 am
214
Just posting updated numbers for my 500K triangle terrain. I am now getting 140K rays per second all the time (perhaps a dip to 130K) and going up to 215K per...
Paul Kahler
phkahler@...
Jan 10, 2001 10:06 pm
215
... If you mean processor cache, each processor has its own local cache and use the MESI protocol for cache coherency (I'm no expert on hardware, but that's as...
Hector Yee
yeehector@...
Jan 11, 2001 6:42 pm
216
... I'd include "it works" and eliminate "deadlock free". I'd also include "fast" and eliminate the "wasted time" objective. But that's just me :-) ... So are...
Paul Kahler
phkahler@...
Jan 11, 2001 6:43 pm
217
Maybe revise your octree code and go deeper (to level 16 to 18) ... One solution could be to do the octree code in integer and do float calculation with...
Jesko Schwarzer
jesko@...
Jan 11, 2001 6:44 pm
218
... include ... me ... I suppose to show only things that work with the only exception if one would bring some thoughts (and mark them as thoughts) ... :-) ...
Jesko Schwarzer
jesko@...
Jan 12, 2001 6:33 pm
219
... what's the difference between a 'critical section' and a semaphore ? is there any other (more low-level) synchronization-mechanism in linux other than...
wald
Ingo.Wald@...
Jan 12, 2001 6:52 pm
220
Critical sections are like mutexes, but they are significantly faster on PC's (but can't be used between different processes, only different threads...) There...
Peter-Pike Sloan
ppsloan@...
Jan 13, 2001 7:09 am
221
... linux ... Semaphores are more complex than CriticalSections because they provide a counter and "count" threads that are trying to get into the critical...
Jesko Schwarzer
jesko@...
Jan 14, 2001 5:23 pm
222
There is a quite impresive rt intro, Fresnel 2 by Kolor. You can get it at http://www.kaoz.org/kolor. It has two versions, one for piii and the other for less...
mep
mep_cpt@...
Jan 19, 2001 12:33 pm
223
Hello, can anyone tell me how can one change the processor privilege level to 0 under windows98 and/or NT? From what i recall on win95 there were many...
wally@...
Jan 31, 2001 4:41 pm
224
Hello, there are some problems in moderating the list, so for some time messages wont be approved. We apologize for the inconvenience and will try to fix it as...
wally@...
Feb 27, 2001 11:40 am
225
Hi there, sorry for the long period of inactivity, now the list should be running fine again now. I'll soon update the FAQ, since there are a lot of papers...
wally@...
May 30, 2001 2:32 pm
226
... Yes and No. I have always had dynamic scenes in mind for RTRT. My older software supported a scene graph with transform nodes in a tree structure. see...
Paul Kahler
phkahler@...
May 31, 2001 12:03 pm
227
Wahoo! Glad to see it back. I joined up after it had died. I come from a virtual reality direction where I build worlds in VRML. It is a pretty neat language...
Miriam English
miriame@...
May 31, 2001 12:04 pm
228
Hi everyone, We have recently had two papers on our RTRT research accepted on graphics conferences. If you are interested you find them now online at ...
Philipp Slusallek
slusallek@...
Jun 1, 2001 12:23 pm
229
Hi I subscriped to this mailing list one month ago as nothing happened since, I think I make my introduction anyway. As a normal introduction is boring, I...
Joachim Diepstraten
diepstjm@...
Jun 1, 2001 12:24 pm
230
Hi ... I hope so because I posted an introduction mail a few weeks ago but it seems it didn't get through. So I'll try it again maybe it works. And of course I...
