Out-of-bounds read in reduction_kernel.cu variant reduce6 for array size 1

[emjotde]

Hi,
your kernel reduce6 from the reduction example does an out-of-bounds read for trivial input size 1 here:

cuda-samples/Samples/reduction/reduction_kernel.cu

Line 357 in 6be5146

if (nIsPow2 || i + blockSize < n) mySum += g_idata[i + blockSize];

This is caused by determining that size 1 is also a power of 2 here:

cuda-samples/Samples/reduction/reduction.cpp

Line 88 in 6be5146

extern "C" bool isPow2(unsigned int x) { return ((x & (x - 1)) == 0); }

but incorrectly assuming that we can increase by blockSize (==1).

Best,
Marcin

[mdoijade]

@emjotde samples sources are expected to be used as is, modified source bugs are expected. closing this issue.

[emjotde]

It's not really a modified source bug. The bug is there regardless whether I modify it or not. In the very least the algorithm isn't working as advertised in this snippet. Considering that these samples have an encyclopedic role leaving it in there seems borderline irresponsible.

This call

cuda-samples/Samples/reduction/reduction_kernel.cu

Line 594 in 6be5146

reduce6<T, 1, true>

will read out of bounds for size 1.

[mdoijade]

@emjotde yes this needs to be investigated reopening. thanks for providing details.
just on a side note, size==1 is not good input for GPU.

[mdoijade]

This is fixed now on the tot, closing the issue.