Aug '24 • Questions (stack overflow)
Rationale behind not SIMDing v3 operations?
I was reading through the engine code and saw that the linmath.c file got changed and many of the v3 and v2 simple math operations don't use SIMD anymore. I was just wondering what the reason for this change is. I suspect the reason for the change is that the overhead of setting up the SIMD operations is slower than simply doing the component-wise operations, which themselves would use ILP (instruction level parallelism) to parallelize the operation anyways. Is this correct? I've also heard that on modern CPUs, most floating point operations are done on SIMD registers anyways, does this play into the reason why? or is it just that this is a temporary change and it'll go back to using SIMD?
1
1 comment
Rationale behind not SIMDing v3 operations?
powered by
skool.com/game-dev-grinders-8513
A group for game devs who want to ship great games from scratch in C, while mastering the ancient art of Caveman Programming
Suggested communities
Powered by