utah rcsid 1.9 87/08/24 01:48:41: Preserve FLOAT to DOUBLE conversions --
on balance this seems to save some trouble down the line.

utah rcsid 1.8 87/07/17 17:09:10: Cosmetic change to tlen() -- use
#define'd constants.

utah rcsid 1.7 87/07/14 20:27:12: A problem with the new clocal conversion
code -- in double-only floating arithmetic, delete all float to double
conversions (reclaim and the table will handle them).

(6) 'word' has type CONSZ; adjusted printf()s accordingly. (7) Added
private prtdcon() routine; just like standard one but substitutes INT 0 for
floating so that the code table can use 'clrd'.

utah rcsid 1.6 87/07/11 21:50:58: Several fixes/enhancements from Sam
Leffler's tahoe compiler. (1) Delete redundant cast() and isitlong()
functions. (2) Make argument p to clocal() have class regi

utah rcsid 1.6 87/07/11 21:50:58: Several fixes/enhancements from Sam
Leffler's tahoe compiler. (1) Delete redundant cast() and isitlong()
functions. (2) Make argument p to clocal() have class register (speed).
(3) Use unsigned comparisons with pointers. (4) Greatly clean up
conversion code in clocal(); I made some changes of my own here. (5)
Pretend fields have INT or UNSIGNED type, not UCHAR or CHAR, in clocal().

show more ...

utah rcsid 1.5 86/11/28 21:33:38: Added some ARGSUSED comments. Perhaps
isitlong() ought to have just one parameter anyway? So it goes.

utah rcsid 1.4 86/08/03 01:00:36: Don't generate code after errors.

utah rcsid 1.7 87/07/17 17:30:26: Some fixes from the tahoe compiler. (1)
Fix 'sizeof (int)' to SZINT/SZCHAR. (2) Don't call toreg() unless we've
fixed the bugs with register chars. (3) Sort the t

utah rcsid 1.7 87/07/17 17:30:26: Some fixes from the tahoe compiler. (1)
Fix 'sizeof (int)' to SZINT/SZCHAR. (2) Don't call toreg() unless we've
fixed the bugs with register chars. (3) Sort the toreg() table by
frequency (if we ever do call it, it'll be faster!).

show more ...

utah rcsid 1.6 87/01/15 01:35:32: Some minor lint fixes.

utah rcsid 1.5 86/11/28 21:29:44: Add some ARGSUSED comments to shut lint
up.

utah rcsid 1.4 86/08/03 00:59:14: Speed hacks -- turn some routines into
macros, don't generate any code after an error.

utah rcsid 1.5 86/08/03 00:58:14: Move configuration flags to macdefs.h.
utah rcsid 1.7 87/01/15 01:34:57: Make 'lintall' entry more reasonable.
utah rcsid 1.8 87/07/07 21:08:55: Move localdefs.h ups

utah rcsid 1.5 86/08/03 00:58:14: Move configuration flags to macdefs.h.
utah rcsid 1.7 87/01/15 01:34:57: Make 'lintall' entry more reasonable.
utah rcsid 1.8 87/07/07 21:08:55: Move localdefs.h upstairs so that
lint can get at it.

show more ...

utah rcsid 1.5 87/12/11 00:28:45: Move localdefs.h upstairs so that
lint can get at it.

utah rcsid 1.8 87/12/10 20:24:11: A fix analogous to the fix from Jim McKie
for the vax -- add implicit float to double conversion for fortran
comparisons.

DIV, MOD, ASG DIV, ASG MOD templates with new stuff derived from vax table.
(12) Delete MUL and ASG MUL oddities; can't see what use they are. (13)
Fold some MUL templates. (14) Simplify ASG PLUS a

DIV, MOD, ASG DIV, ASG MOD templates with new stuff derived from vax table.
(12) Delete MUL and ASG MUL oddities; can't see what use they are. (13)
Fold some MUL templates. (14) Simplify ASG PLUS and ASG MINUS if REG_CHAR
is not enabled. (15) Add templates for mixed float/double assignment ops.

show more ...

char/short equality/inequality between a memory operand and a constant, so
we cull these out. (6) Fold some INCR and DECR templates together. (7)
Move the most frequently used ASSIGN template to th

char/short equality/inequality between a memory operand and a constant, so
we cull these out. (6) Fold some INCR and DECR templates together. (7)
Move the most frequently used ASSIGN template to the top. (8) Sigh -- add
several ASSIGN templates to handle various combinations with float and
double. (9) Delete redundant SIREG template for OPLTYPE doubles. (10)
Properly handle complements of unsigned char and short. (11) Replace old

show more ...

utah rcsid 1.7 87/11/29 20:51:53: (1) We don't need the first PCONV
templates if we don't have REG_CHAR. (2) Take care of redundant
conversions introduced by reclaim(). (3) Move frequently used SCO

utah rcsid 1.7 87/11/29 20:51:53: (1) We don't need the first PCONV
templates if we don't have REG_CHAR. (2) Take care of redundant
conversions introduced by reclaim(). (3) Move frequently used SCONV
templates to the top; fold some mutually redundant SCONV templates
together. (4) If only Fortran uses these GOTO templates, don't compile
them if we don't need them. (5) Ack -- tahoe won't handle unsigned

show more ...

utah rcsid 1.6 87/05/02 15:48:47: A much cheaper way to get the value of a
bitfield assignment -- call zzzcode().

utah rcsid 1.5 87/04/18 01:43:18: Extensive hacking to weed out bugs in
unsigned/float and float/unsigned conversions. Unbalanced assignments are
now handled the same way as conversions.

utah rcsid 1.4 87/04/17 03:57:21: The value of a bitfield assignment is the
bitfield, not the rhs. (Lucky thing we didn't implement signed bitfields
on the tahoe...)

utah rcsid 1.3 87/03/23 04:38:33: Add zzzcode calls for unsigned => double,
double => unsigned and i *= f.

utah rcsid 1.2 87/03/18 02:21:30: Extend sconv() usage to integer
assignments and OPLEAF promotions as well as to conversions; add tab and
newline to SCONV use of ZU and ZV; split special comparison

utah rcsid 1.2 87/03/18 02:21:30: Extend sconv() usage to integer
assignments and OPLEAF promotions as well as to conversions; add tab and
newline to SCONV use of ZU and ZV; split special comparison entries for
TCHAR/TUCHAR and TSHORT/TUSHORT so that integer promotions are performed
when the signedness of the operands differs.

show more ...

utah rcsid 1.3 87/01/15 01:40:28: Fixed bug in outstruct() which caused it
to refer to stab[-1] when dealing with unnamed structures. Also some minor
lint fixes.

utah rcsid 1.2 86/03/27 20:50:15: A problem with dbx and pointers to arrays
was fixed -- if the outermost type of an object was not an array, then when
the compiler tried to generate the subscript ra

utah rcsid 1.2 86/03/27 20:50:15: A problem with dbx and pointers to arrays
was fixed -- if the outermost type of an object was not an array, then when
the compiler tried to generate the subscript range for the array it got
confused and used the range of the most recently declared array instead of
the one it wanted.

show more ...

utah rcsid 1.4 87/11/29 20:23:26: Fix a syntax error in sucomp() (sigh...).