ray direction calculation about bvh_article

[jbikker]

Article and code has been updated; should be better now. Let me know if you find anything else. - Jacco.

…

On Mon, Aug 28, 2023 at 9:54 AM Jacco Bikker ***@***.***> wrote: Hi, It's indeed a bit weird that in chapter 2 ray.O is added to the pixel position. It is best to calculate all coordinates in 'world space'; assuming the screen corners p0, p1 and p2 are in world space, the pixel position should be calculated as P(u,v) = p0 + u(p1 - p0) + v(p2 - p0). I am not sure why I did this different in chaper 2, it looks like the pixel position is now relative to the ray origin but then the ray.D calculation doesn't make sense... Sorry for the confusion! - Jacco. On Sun, Aug 6, 2023 at 8:24 AM XiaotianM ***@***.***> wrote: > Hi, I am a little confused about the way to calculate orientation > calculation. > in chapter2: > > https://github.com/jbikker/bvh_article/blob/89bf112b7ca8c277ac0a311119aa2e29d5ee41f1/faster.cpp#L277C1-L279C57 > > ray.O = float3( -1.5f, -0.2f, -2.5f ); > float3 pixelPos = ray.O + p0 + (p1 - p0) * ((x + u) / (float)SCRWIDTH) + (p2 - p0) * ((y + v) / (float)SCRHEIGHT); > ray.D = normalize( pixelPos - ray.O ), ray.t = 1e30f; > > Why need to add ray.O for pixelPos, and minus the ray.O to calculate > ray,D. It seems the rays are emitted from (0,0,0), but the ray origin point > is still ( -1.5f, -0.2f, -2.5f ) > > and in chapter1: > In chapter1: > > https://github.com/jbikker/bvh_article/blob/89bf112b7ca8c277ac0a311119aa2e29d5ee41f1/basics.cpp#L182C1-L184C41 > > float3 pixelPos = p0 + (p1 - p0) * (x / (float)SCRWIDTH) + (p2 - p0) * (y / (float)SCRHEIGHT); > ray.D = normalize( pixelPos - ray.O ); > > there is different calculation for pixelPos (no ray.O added) > > — > Reply to this email directly, view it on GitHub > <#5>, or unsubscribe > <https://github.com/notifications/unsubscribe-auth/AFH3PEEKHVELVOLLLN2TYZTXT42DBANCNFSM6AAAAAA3FWQPP4> > . > You are receiving this because you are subscribed to this thread.Message > ID: ***@***.***> >

[jbikker]

Hi, It's indeed a bit weird that in chapter 2 ray.O is added to the pixel position. It is best to calculate all coordinates in 'world space'; assuming the screen corners p0, p1 and p2 are in world space, the pixel position should be calculated as P(u,v) = p0 + u(p1 - p0) + v(p2 - p0). I am not sure why I did this different in chaper 2, it looks like the pixel position is now relative to the ray origin but then the ray.D calculation doesn't make sense... Sorry for the confusion! - Jacco.

…

On Sun, Aug 6, 2023 at 8:24 AM XiaotianM ***@***.***> wrote: Hi, I am a little confused about the way to calculate orientation calculation. in chapter2: https://github.com/jbikker/bvh_article/blob/89bf112b7ca8c277ac0a311119aa2e29d5ee41f1/faster.cpp#L277C1-L279C57 ray.O = float3( -1.5f, -0.2f, -2.5f ); float3 pixelPos = ray.O + p0 + (p1 - p0) * ((x + u) / (float)SCRWIDTH) + (p2 - p0) * ((y + v) / (float)SCRHEIGHT); ray.D = normalize( pixelPos - ray.O ), ray.t = 1e30f; Why need to add ray.O for pixelPos, and minus the ray.O to calculate ray,D. It seems the rays are emitted from (0,0,0), but the ray origin point is still ( -1.5f, -0.2f, -2.5f ) and in chapter1: In chapter1: https://github.com/jbikker/bvh_article/blob/89bf112b7ca8c277ac0a311119aa2e29d5ee41f1/basics.cpp#L182C1-L184C41 float3 pixelPos = p0 + (p1 - p0) * (x / (float)SCRWIDTH) + (p2 - p0) * (y / (float)SCRHEIGHT); ray.D = normalize( pixelPos - ray.O ); there is different calculation for pixelPos (no ray.O added) — Reply to this email directly, view it on GitHub <#5>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AFH3PEEKHVELVOLLLN2TYZTXT42DBANCNFSM6AAAAAA3FWQPP4> . You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>

[XiaotianM]

HI Jacco,

Thank you very much for your replay.
In fact, in other chapter's code, there also adds ray.O for pixel position calculation.
Anyway, I got your points now. This issue can be closed.