[fixed-point] Support saturating constant folding

["Fu, Chao-Ying" <fu at mips dot com>]

Hello,

  This patch adds the support of saturating the result if the 
fixed-point type is saturating.  Also, multiply and divide are 
extended to support fixed-point modes which are wider than 
HOST_BITS_PER_WIDE_INT.  Please check the attached example.

  Thanks a lot!

Regards,
Chao-ying

2007-01-02  Chao-ying Fu  <fu@mips.com>

	* fold-const.c (const_binop): Add satp for saturating types.
	Pass satp to fixed_arithmetic.
	(fold_negate_const): Pass saturating info to negate a fixed value.
	* double-int.c (double_int_scmp): We should use unsigned compare
	for a.low and b.low.
	* fixed-value.c (get_fixed_sign_bit): Declare.
	(do_fixed_add, do_fixed_multiply, do_fixed_divide, do_fixed_shift):
	Add satp as the last parameter.
	(do_fixed_neg, fixed_saturate1, fixed_saturate2): Declare.
	(fixed_saturate1, fixed_saturate2): New functions to saturate the
	results.
	(get_fixed_sign_bit): New function to return the sign bit.
	(do_fixed_add): Saturate the result if satp is true.
	(do_fixed_multiply): Saturate the result if satp is true.  Support all
	modes now.
	(do_fixed_divide): Likewise.
	(do_fixed_shift): Saturate the result if satp is true.
	(do_fixed_neg): New function.
	(fixed_arithmetic): Add a parameter of satp.  Call do_fixed_neg
	for negation.  Pass satp to do_fixed_*.
	(fixed_arithmetic2): Add a parameter of satp.
	* fixed-value.h (fixed_arithmetic, fixed_arithmetic2): Add a parameter
	of satp.
	(FIXED_VALUE_NEGATE): Likewise.

Attachment: gcc.diff
Description: gcc.diff

Attachment: satconst.c
Description: satconst.c

Attachment: satconst.s
Description: satconst.s