When running in preprocess-only mode generate the equivalent of
standard alignment using nops. This at the very least allows some
kind of reasonable output and allows for dependency generation to
proceed; the only way to *really* address this problem is to move
alignment generation into the assembler proper; this would also allow
the align/alignb distinction to be removed and handle padding with
instructions which are more than one byte.
This should resolve bug 3392319.
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
This change may have backward compatibility issue but
most probably the sane program never used sections with
base address less then instruction alignment.
Note that alignment may only increase which means if a
section is aligned on 2^5 the align 2^4 will not affect
the section.
Signed-off-by: Cyrill Gorcunov <gorcunov@gmail.com>
Remove unnecessary duplicated patterns; with indirection we can handle
lists of any length.
For 16-bit generic padding, alternate between SI and DI dependencies.
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
Use a "times" construct rather than "%rep" for higher performance.
No need to preprocess the same line over and over for no good reason.
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
Apparently the current recommendation is for a smaller threshold when
using the "generic"-style alignment macros (short jumps are cheaper on
newer CPUs.)
Also change the alignment threshold definition to reflect the maximum
number of padding instead of when to start using jumps.
