optimized by post-T&L vertex cache

[ultrafly]

hi.

I use the method raised by Mike Abrash(<<Xbox Vertex Performance>>).But I find some strange things.

I suppose the post-T&L wertex should be 11 wertexs in my R9500 pro 64M.

First, after using 20,000 triangles,I got two results,optimized code is 255 fps, and no optimized code is 177 fps. But when I reduce the size of triangles(no change the numbers of triangles), I found the fps under no optimized code is faster than optimized code. why?

Second,I found use D3DFILL_SOLID is faster than D3DFILL_WIREFRAME,why?

I am sorry for my pool english.

 

[Luke Philpot]

I'm not sure about your first question, however about your second question, I am presuming that it might be because each "line" is one polygon... however I'm probably wrong, this is just a guess.