whitequark changed the topic of #solvespace to: SolveSpace--parametric 2d/3d CAD · latest version 2.3 · http://solvespace.com · code at https://github.com/solvespace/solvespace · logs at https://irclog.whitequark.org/solvespace
f8l has quit [Ping timeout: 248 seconds]
f8l has joined #solvespace
<GitHub> [solvespace] rpavlik commented on pull request #308 60f2ee8: The minus can be vectorized, or sometimes (seen in practice) auto-vectorized. https://github.com/solvespace/solvespace/pull/308#discussion_r167411402
<GitHub> [solvespace] rpavlik commented on pull request #308 60f2ee8: RelWith Deb Info, yes.... https://github.com/solvespace/solvespace/pull/308#discussion_r167411621
<GitHub> [solvespace] rpavlik commented on pull request #308 60f2ee8: I don't remember if I tested this single change independently. However, you're on the right path with intrinsic: my hunch is it gives better data to the compiler and thus can be optimized better. (It also reduces operations: one comparison, and the abs which probably is a bit manipulation, vs two comparisons, but that's down to the level where I'm not sure if intui
Radicarian has joined #solvespace