SIMDSort reads beyond end of array

[Reen]

I tried to use your AVX2 SIMDSort code to sort a list of 2,000,000 floats. In the final pass of MergePass8 with run_size = 512 the code in the for-loop accesses memory outside the array. As 2,000,000 is divisible by 64, I do not run into the assert() at the entry point.

I tried to debug it, but I'm unable to find a reasonable abort criterion for the loop. Can you provide a hint on how to fix this?

Thank you for your time.

[saatviks]

We should have added an assertion for this too(I think its explained in the report) - but the input size has to be a power of 2 as well.

Edit: Yes, we have mentioned it under the Usage section in the README.

[PatwinchIR]

For now, you might need to manually match your input size to the power of 2. We might release a version in the future to automatically handle this. Sorry about the inconvenience.