Question:
Is there a more elegant way to use the alignment pragma with an “alignopt” value that works for both “-char-size 8” and “-char-size 32”?
Answer:
One possibility would be to write:
#include <limits.h>
#pragma align (1024 * 32 / CHAR_BIT)
float input[1024];
Since CHAR_BIT is 8 with -char-size-8, and 32 with -char-size-32, this would produce the alignment you require.