With SIMD the CPU may execute the same instructions on e.g. 4x differents datasets just as fast as on 1x dataset. Assuming the user has a SIMD-enabled objective function - are there optimization algos that can exploit that? E.g. in Optim, Roots, SciML…? The prime example would be the BiSection alg…

The tools don’t have an interface for it yet. But I was just chatting with @Elrod about doing this just last week. We’ll probably start doing some of this for all SciML problems once the GPU push is done this summer (there’s new algorithms in each area to be made for this). If you’re interested and …

Supporting SIMD-enabled objective functions in optimization APIs

Specific Domains Optimization (Mathematical)

odow February 5, 2024, 6:59pm 3

Topic		Replies	Views
[ANN] SIMDMathFunctions.jl : Fast vectorized mathematical functions for SIMD.jl , using SLEEFPirates.jl Package Announcements package , announcement	10	1094	March 18, 2024
SciML Optimization - algorithms and convergence speed Optimization (Mathematical)	10	559	May 29, 2023
Vectorization heuristics are hard! Is there a way to ask/get more information to/from the compiler/LLVM about loop vectorization Performance question	2	297	February 10, 2023
How to make Optim's optimize work for scalars Optimization (Mathematical) question	3	1292	February 8, 2020
Is the triple `@inbounds @fastmath @simd` necessary for absolute peak performance? Performance	7	456	October 21, 2024

Supporting SIMD-enabled objective functions in optimization APIs

Related topics