I’m trying to benchmark different methods of calling the same in-place function on subsets of a large data block using CuArrays. I’m using fft as an example function because I can baseline against planning an fft over one dimension of a large array. Using this as a minimum working example, I created…

OK, that example clarifies a lot. We don’t efficiently support what you’re doing there: broadcasting a function that does its own broadcasting over the inner 1D dataset. If that inner broadcast is coarse enough, i.e. if it does enough work to saturate the GPU, a mapslices-like approach might work (a…

Map Performance with CuArrays

Specific Domains GPU

maleadt January 4, 2021, 7:29am 16

They definitely are, why do you assume they aren’t? Kernel calls aren’t free, of course, so if your kernel execution time is shorter than the launch time you’ll just scale the time it takes to launch.

Topic		Replies	Views
Performance of view with cuArrays GPU	11	2705	November 11, 2018
Mapping functions using CuArrays GPU cuarrays	2	1932	September 9, 2020
GPU Map without reduction on multiple arrays indices GPU	1	706	February 8, 2019
Calculate FFT on GPU for every row of a 2D array Performance gpu	2	1042	August 14, 2018
cuArrays vs CUDANative GPU	3	1381	November 14, 2018

Map Performance with CuArrays

Related topics