Often the context is just 2 vertices, so it's not worth allocating large arrays to create the Weld context.
Resize the arrays in the end so that memory is better used in other areas.
Details
Details
Diff Detail
Diff Detail
- Repository
- rB Blender
- Branch
- arcpatch-D13622 (branched from master)
- Build Status
Buildable 19529 Build 19529: arc lint + arc unit
Event Timeline
| source/blender/modifiers/intern/MOD_weld.cc | ||
|---|---|---|
| 379 | Wouldn't it be possible to calculate the size of this exactly, depending on the vert_kill_len and the original number of vertices? | |
| source/blender/modifiers/intern/MOD_weld.cc | ||
|---|---|---|
| 379 | Unfortunately not. | |
Comment Actions
I am not familiar with this code, but if you are looking at reducing memory, have you tried switching to pre-increment instead of post-increment, to make sure you aren’t making a copy and destroying it every loop, in the cases that the compiler is unable to optimize away?
Comment Actions
++i vs i++ and possibly ++wvert_len vs wvert_len++ for example. Take the suggestion with a large grain of salt, as I just found out about post increment creating a copy of the value before increasing the value and returning the copy, and I don’t fully grasp where preincrement isn’t applicable.
Sources:
https://google.github.io/styleguide/cppguide.html#Preincrement_and_Predecrement
https://herbsutter.com/2013/05/13/gotw-2-solution-temporary-objects/
The section titled “The iterator increment uses postincrement.”