any problems to make it run on gpu to reduce time?

[Yes-Jumby]

I have test it in model consisted of triangle and texture.as showed in below:
number of vertex(million) time(s)
0.5 0.4511
1 0.9188
2 2.0026
3 4.4232
5 6.9844
10 12.0139
12 19.8880
15 25.3627
function I have used:
meshopt_simplify
meshopt_optimizeVertexCache
meshopt_optimizeVertexFetchRemap
meshopt_remapIndexBuffer
meshopt_remapVertexBuffer
and I have checked few of those function in source,I think it is feasible to save time by translate those code running in cpu to gpu（NVIDIA by cuda） to save time.
Do anyone else share my viewpoint？If anything in this way is unfeasible,point it out?