After porting some code from using oneAPI.jl GPU kernels to CUDA.jl I got timings in CUDA which were higher than the ones from running in an Intel GPU. Something like getting 20ms on the on an Nvidia Gtx 1080 Ti and 14ms on an Intel UHD Graphics 620, which was really not expected given that memory …

CUDA.jl not inlining "simple" functions

jpsamaroo January 30, 2021, 3:27pm 5

KernelAbstractions forcibly inlines functions on its own: Home · KernelAbstractions.jl

1 Like

Topic		Replies	Views
What's the benefit of inlining? New to Julia question	7	16534	December 13, 2017
Julia equivalent of C compiler intrinsics? General Usage	23	3206	November 8, 2018
Benchmark showing 2x slower result between nearly identical functions on Julia 0.6 General Usage	4	646	January 27, 2017
Small functions - best practices New to Julia question , compilation , function	5	851	February 8, 2021
Exponentially long compile times in generated functions that disappear after manual inlining General Usage performance , simd , array	4	1101	July 8, 2017