Extending documentation - #218
Conversation
pshriwise
left a comment
pshriwise
left a comment
There was a problem hiding this comment.
Just a few thoughts on wording. Should the GPRT submodule have changed as part of this?
Nope, my bad on that one. Lazy application of |
|
|
||
| .. glossary:: | ||
|
|
||
| AS |
There was a problem hiding this comment.
AS is pretty generic. Can we turn this into two entries: TLAS and BLAS? I think it would prove more useful in the documentation.
Unless perhaps you feel this is too low level to address.
Another option might be to define the term "Acceleration Structure" here, spelling it out. Terms here don't always have to be abbreviations.
There was a problem hiding this comment.
TLAS and BLAS? I think it would prove more useful in the documentation. Unless perhaps you feel this is too low level to address.
hmmm thats a good idea. I guess it depends on the audience right. We're assuming anyone reading the XDG docs is likely wanting to implement XDG in some kind of downstream simulation code. In that case, I do think its beneficial to have an understanding that there is a two-level structure of trees. And more importantly that the Volumes and Surfaces that they are interacting with correspond to TLAS and BLAS respectively.
Since we also have the global TLAS trees too, the current umbrella term of "Acceleration Structure" probably fails to capture what the RT accel structure is actually doing.
There was a problem hiding this comment.
What do you think about AS --> Acceleration Structure?
There was a problem hiding this comment.
What do you think about AS --> Acceleration Structure?
Ah yeah I forgot to do the one change that even sparked this conversation and prompted me to add all of this extra detail lol
Yep, I think that makes sense. Given the discussion about adding TLAS and BLAS to the glossary would you be okay with me adding some diagrams and a more detailed description of ray tracing and ray tracing acceleration structures as a whole to https://github.com/Waqar-ukaea/xdg/blob/d13968751daab933746a42def56990eff433242f/docs/methods/acceleration_data_structures.rst ? Something akin to slides 5 & 6 in this presentation is what I had in mind https://docs.google.com/presentation/d/1CA4E30kae5_Bdu1CwrLFjvTwSNLd1qB4/edit?slide=id.p4#slide=id.p4 |
By all means! Somewhere in there I'm thinking it would be useful to generalize that terminology for Embree where BLAS --> Geometry; TLAS --> Scene. |
…t DPRT have been removed
To clarify, you'd like to wait to merge until you've added the TLAS/BLAS discussion? |
Yep if that's okay. I'm already partially through writing this up. And I'll re-request a review once that's in place. |
…ation in XDG via Embree vs GPRT
|
I've updated the In 5e37e32 I have tried to highlight this with these two mapping tables - not sure if I am getting too in the weeds of the details now with this though. What do you think? I actually do have another PR up already which switches Embree setup to follow a more explict TLAS-BLAS instance workflow here - #130. My original plan for this PR was to revisit it once I got access to the ANL machines for benchmarking though. |
pshriwise
left a comment
pshriwise
left a comment
There was a problem hiding this comment.
More thoughts here. It's looking good!
|
|
||
| .. glossary:: | ||
|
|
||
| AS |
There was a problem hiding this comment.
What do you think about AS --> Acceleration Structure?
pshriwise
left a comment
pshriwise
left a comment
There was a problem hiding this comment.
Thanks for these improvements @Waqar-ukaea!
|
@pshriwise I see you've approved this. Happy to merge? |
2 participants
Wanted to open a draft PR to make a start on adding some more information/diagrams to the XDG documentation. I will keep adding to this as and when I find the motivation/time