![[Aug. 14th - Aug.18th] Week 8 Report](https://static.wixstatic.com/media/ea4d52_b5a09e0fd0524e53974003de0627af74~mv2.png/v1/fill/w_408,h_228,fp_0.50_0.50,q_95,enc_auto/ea4d52_b5a09e0fd0524e53974003de0627af74~mv2.png)
![[Aug. 5th] Week 7 Report](https://static.wixstatic.com/media/ea4d52_996d48877c714cdda64eb8594a342bf0~mv2.png/v1/fill/w_408,h_368,fp_0.50_0.50,q_95,enc_auto/ea4d52_996d48877c714cdda64eb8594a342bf0~mv2.png)
[Aug. 14th - Aug.18th] Week 8 Report
This week focused on mainly two parts: Performance & Integration.
Work this week:
1.Reconstruct nodes to save memory.
For the performance before is rather disappointing on memory issue. In this way, what I did to save space for each node:
a.) Remove color rgba information for every single node.
Set color as uniform. Add graduate color choice and do this step in shader.
b.) Remove indices information, instead, be more aware about the order of vertices.
2. Detect those nodes that can not be seen on screen.
Don't rely on z-culling. For z-culling take place after fragement shader. But if a bunch of nodes are not visible, they should not be seen on screen.
What I did for this issue:
a.)Use sampled-depth-query to determine invisible nodes. When network fist setup, do depth query for the whole viewport.
b.) After translation/scaling do depth query around the changing object.
So query just take place once may culling many nodes.
3. Integration to Grind
On going... in great process now
