* config/ns32k/ns32k.h: Remove obsolete comment.
From-SVN: r73720
This commit is contained in:
parent
9a95dbe435
commit
e2657a6707
@ -1,3 +1,7 @@
|
||||
2003-11-18 Nathanael Nerode <neroden@gcc.gnu.org>
|
||||
|
||||
* config/ns32k/ns32k.h: Remove obsolete comment.
|
||||
|
||||
2003-11-18 Roger Sayle <roger@eyesopen.com>
|
||||
|
||||
PR middle-end/11821
|
||||
|
@ -383,14 +383,7 @@ while (0)
|
||||
/* Value is 1 if it is a good idea to tie two pseudo registers
|
||||
when one has mode MODE1 and one has mode MODE2.
|
||||
If HARD_REGNO_MODE_OK could produce different values for MODE1 and MODE2,
|
||||
for any hard reg, then this must be 0 for correct output.
|
||||
|
||||
Early documentation says SI and DI are not tieable if some reg can
|
||||
be OK for SI but not for DI. However other ports (mips, i860, mvs
|
||||
and tahoe) don't meet the above criterion. Evidently the real
|
||||
requirement is somewhat laxer. Documentation was changed for gcc
|
||||
2.8 but was not picked up by egcs (at least egcs 1.0). Having all
|
||||
integer modes tieable definitely generates faster code. */
|
||||
for any hard reg, then this must be 0 for correct output. */
|
||||
|
||||
#define MODES_TIEABLE_P(MODE1, MODE2) \
|
||||
((FLOAT_MODE_P(MODE1) && FLOAT_MODE_P(MODE2) \
|
||||
|
Loading…
x
Reference in New Issue
Block a user