You asked about a problem with D=100um and lambda=532nm=0.532um, or D/lambda=190. Unfortunately, this is probably beyond the reach of DDSCAT.
The practical limit on size/wavelength depends on various things, including the value of the dielectric function, the detailed target shape (e.g., is the target "sparse" with lots of "holes", or is in dense (e.g., a solid rectangular block). It also depends on the "dimensionality" of the problem — in some cases periodic targets are tractable with very large ratios of size/wavelength.
Example: For an infinite column, one can go to quite large size/wavelength, We have published results for a hexagonal prism with diameter/wavelength = 70um/0.55um = 127 (Flatau & Draine 2008, Optics Express 22, 21384).
For isolated targets, practical memory and cpu time limitations limit one to smaller targets. Values of size/wavelength larger than ~20 are very challenging. A rule of thumb is that you need to have |m|*pi*d/lambda < 1, where d is the interdipole separation.
It may also be limited by the size of your computer memory. Targets with large size/wavelength require large amounts of memory. You mayneed to use a compiler that allows you to address individual arrays >2 GB. Intel fortran allows this, but you have to figure out the right "flags" to use on your system.