The old foldall strategy used BLOCK_2D_X even in the 1 d case, resulting in 32 (on most systems) which in case of 1D is clearly too small :-) Now, we use BLOCK_1D_X in the 1 d case :-)
The old foldall strategy used BLOCK_2D_X even in the 1 d case, resulting in 32 (on most systems) which in case of 1D is clearly too small :-) Now, we use BLOCK_1D_X in the 1 d case :-)