I am confused about the first advantage. What's the cost of extra point-in-triangle tests?
imm
is there a convention for how large the block is in this method? also in order to take advantage of early in/early out does this mean that the edge samples of the block must be processed first?
kencheng
@QianDong: This method will cause you to test more points than the incremental method, since you're testing the entire bounding box instead of samples near the triangle. These are the "extra" point-in-triangle tests. The cost of doing these tests, however, is negligible compared to the performance of running this method in parallel.
SIMD processing as mentioned in class: Single instruction, multiple data
I am confused about the first advantage. What's the cost of extra point-in-triangle tests?
is there a convention for how large the block is in this method? also in order to take advantage of early in/early out does this mean that the edge samples of the block must be processed first?
@QianDong: This method will cause you to test more points than the incremental method, since you're testing the entire bounding box instead of samples near the triangle. These are the "extra" point-in-triangle tests. The cost of doing these tests, however, is negligible compared to the performance of running this method in parallel.