I am currently playing with simple math like
a[i] *= 1.02
Over array sizes of 4k , 8k , .... 128M etc
The 2700U CPU gives me about 28GB/s in the top cache and about 13GB/s in memory
The 7252 30GB/s Cache and 24GB/s memory
Vega 64 28GB/s 4k to 64k and 360GB/s for > 128k sizes
2700U APU... barely 4GB/s
This is with rocm-3.3 .
Any idea if the 2700U apu should be that slow with opencl ?
a[i] *= 1.02
Over array sizes of 4k , 8k , .... 128M etc
The 2700U CPU gives me about 28GB/s in the top cache and about 13GB/s in memory
The 7252 30GB/s Cache and 24GB/s memory
Vega 64 28GB/s 4k to 64k and 360GB/s for > 128k sizes
2700U APU... barely 4GB/s
This is with rocm-3.3 .
Any idea if the 2700U apu should be that slow with opencl ?