thanks for the heads up!
But isn't pcie bandwidth usually more abundant than disk bandwidth? 8x pcie 2.0 lanes or 4x pcie 3.0 lanes should be around 4 GigaBytes per Second, in each direction.
freeing the cpu for other tasks would also be a huge bonus
I haven't checked throughput on igpu yet, not sure how much that would be
\edit:
I am rather new to the topic, but yes, a multithreaded solution would be great. sha256sum for instance seems to work single threaded. I could work around that by just processing multiple files in parallel, for now. just am not sure if using a gpu could make sense in the first place, so I started by poking around here
But isn't pcie bandwidth usually more abundant than disk bandwidth? 8x pcie 2.0 lanes or 4x pcie 3.0 lanes should be around 4 GigaBytes per Second, in each direction.
freeing the cpu for other tasks would also be a huge bonus
I haven't checked throughput on igpu yet, not sure how much that would be
\edit:
I am rather new to the topic, but yes, a multithreaded solution would be great. sha256sum for instance seems to work single threaded. I could work around that by just processing multiple files in parallel, for now. just am not sure if using a gpu could make sense in the first place, so I started by poking around here