1,3 → 1,25 |
View3ds 0.076 - XII 2021 |
1. Detecting manifold chunks procedure based on kind of sorted pivot |
table. Chunks are counted and this number displayed. |
2. New calculating normal vectors proc that use some data produced |
by new chunks routine. Now big object loading is fast. I load object that |
contains ~500000 vertices, ~700000 faces and ~2000 0000 unique edges |
in few seconds on i5 2cond gen. Earlier such objects calculating was |
rather above time limits. |
3. On http://board.flatassembler.net occasionaly there are some disccusions |
about optimizing. Some clever people, wich skills and competence I trust, |
claims - for CPU's manufactured last ~15 years size of code is crucial |
for speed. (Better utilize CPU cache). |
So I wrote some 'movsd' mnemonics instead 'mov [edi],sth'; 'loop' instead |
'dec ecx,jnz sth'. Moreover I come back to init some local varibles |
by 'push' (flat_cat.inc). I took effort to change divisions to |
multiplications two_tex.inc (works ok in fpu only Ext = NON mode and |
of course in Ext = SSE3 mode), grd_tex.inc (single line not parallel |
muls, whole drawing routine 4 divs instead 27 divisions), |
bump_tex.inc - 3 divs in SSE2 mode.s See sources for details. |
4. Editor button allows now editing by vertex all above 65535 vert objects. |
---------------------------------------------------------------------------------- |
|
View3ds 0.075 - XII 2021 |
1. Cusom rotate using keys and mouse scroll support by Leency. |
---------------------------------------------------------------------------------- |
9,7 → 31,6 |
set 'on' this option. Note that is non real time model, especially when |
complex object is computed. I took effort to introduce accelerating |
structure - AABB (Axis Aligned Bounding Boxes).. but it is disabled |
|
for now - seems to work incorrect(slow). |
---------------------------------------------------------------------------------- |
|