Read performance drop when using stride > 1


I am using the following hyperslab call to read 2D arrays of data:

    dsp.selectHyperslab(H5S_SELECT_SET, count, offset, stride, block);

Benchmarking tests revealed that using stride > 1 dropped performance
by a factor of 5. The performance did not vary among stride values
greater that 1.

Is this to be expected or is there something I should be doing to
avoid this drop?

I am using HDF5 1.8.1 on a 64bit Linux platform.




This mailing list is for HDF software users discussion.
To subscribe to this list, send a message to
To unsubscribe, send a message to