[r15417]

fixed new_tarval_from_long() (added missing sign extend), classify_tarval(), unsigned integer modes have minus_one_tarval()

[r15391]

[r15334]

[r15325]

- removed the *_calc functions
- simplified floating point handling
- renamed functions

[r15291]

[r15237]

constant folding for floating point tarvals is disabled for now as it calculates in the precision of the values and not of the target cpu (x86=80bit)

[r15231]

[r15081]

[r15057]

[r14847]

[r14780]

- All float operations depend on fpu mode now
- Fix bestate switcher
- Add and fix fehler22 (unknown phi operands in x87 simulator)

[r14749]

[r14601]

[r14042]

[r13853]

removed ugly defined that create simple declarartions
more doxygen docu added

[r13512]

[r13510]

Currently non-symmetric!

[r13443]

Move includes for alloca() to xmalloc.h, so not everyone and his dog has to use the ugly #ifdef macros.

[r12647]

indentation changed

[r8745]

sign_extend does zero extension for unsigned integer modes

[r8359]

[r8217]

[r8134]

- get_tarval_overflow(): when in warp mode, the strcalc module expects that ALL
  bits are set (and this includes the HIGH bits)

[r8132]

[r7954]

[r7897]

[r7896]

[r7878]

[r7728]

[r7480]

which is created from the null_value given at initialization.

This also solves the problem the new_tarval_from_str() and new_tarval_from_long() returned different
tarvals for the same values.

Reference modes are handled like integer modes, however, convert is only supported for NULL pointer (which might be wrong if null_value != 0)

[r7465]

[r6911]

[r6891]

[r6890]

[r6638]

[r6561]

some cleanup

[r6476]

removed old assert() in tarval_cmp() float case

[r6126]

as needed for IEEE-754

[r6072]

[r5973]