f6f38dd61e
Changed the test that determines whether or not source and destination
...
memory regions are "potentially overlapping" to a test that determines
that the regions are actually overlapping. Because the code for the
overlapping case is seven instructions longer, this signifcantly
improves performance in the average case.
1996-11-12 00:50:06 +00:00
c9ec05f4b2
Treat reboot() like a normal system call. From PR 2458.
1996-11-11 13:41:08 +00:00
9d7f517756
Added new macros:
...
SYSTRAP(x): Expands to the code used to call syscall x.
This is used to simplify other macros.
SYSCALL_NOERROR(x): Like SYSCALL except that "x" is a syscall
that can never fail.
RSYSCALL_NOERROR(x): Like RSYSCALL except that "x" is a syscall
that can never fail.
1996-10-17 18:14:56 +00:00
daa40d9a50
Add explicit return type
1996-02-05 04:30:51 +00:00
43d7567da3
Removed unused CALL macro
1995-12-28 22:35:54 +00:00
051ae6c48a
Remove #include "SYS.h"; it is not needed (by normal assembly functions)
1995-12-28 21:32:25 +00:00
3f78e35cca
Use <machine/asm.h> instead of DEFS.h
1995-12-28 20:03:51 +00:00
24a6d94c0d
Added macros definitions for use with ANSI preprocessors.
1995-12-28 19:44:11 +00:00
02cf7515b7
Make PIC safe.
1995-06-20 22:45:46 +00:00
7b4ed4f858
Winning Strategies has placed this code into the Public Domain
1995-04-28 22:59:46 +00:00
390695fb9c
$Id: $ -> $NetBSD: $
1995-04-28 22:57:54 +00:00
954b796156
Winning Strategies has placed this code into the Public Domain
1995-04-28 22:53:59 +00:00
10cfe03e6d
Stylistic change to match fpsetsticky().
1995-04-09 21:55:50 +00:00
65a1322cd9
Reverse sense of exception bits.
1995-04-09 21:55:20 +00:00
42d9c8af6c
Oops; need to use fnstenv/fldenv for this.
1995-04-09 21:00:57 +00:00
a94ab51d20
Fix typo.
1995-04-09 20:44:02 +00:00
10b402aa60
Add fpset{mask,round,sticky}().
1995-04-09 20:33:13 +00:00
ddbc579aa7
Another optimization.
1995-04-09 19:31:33 +00:00
14e246dc4a
Optimize.
1995-04-09 19:18:49 +00:00
907e54c9e9
ENTRY(fpgetmask) -> ENTRY(fpgetsticky)
1995-04-09 19:14:28 +00:00
80bba6a1e1
Added flt_rounds, fpget{round,mask,sticky}.
1995-04-07 14:57:16 +00:00
ce36f636ab
Separate machine dependencies.
1995-03-20 14:45:27 +00:00
4c34be6b52
Clean up deleted files.
1995-03-01 00:00:00 +00:00
572e3f3188
don't forget inifinity.c
1995-02-25 16:24:22 +00:00
ea1ca14ecb
need these for new gen/Makefile
1995-02-25 16:20:12 +00:00
6929ad87a5
need these now, with new net Makefile.inc
1995-02-25 14:58:43 +00:00
e7d917e6eb
Remove extra "u" constraint that caused gcc 2.6.3 not to compile this file.
...
Note that gcc 2.4.5, the compiler we are currently using, still compiles it
correctly after this change.
1995-02-17 06:01:47 +00:00
d04155c293
Optimize better.
1995-02-05 14:58:44 +00:00
b4bd7fea24
Replaced pops with adds.
1994-12-24 01:16:44 +00:00
26eccb0a9b
Exchange offsets of "saved signal mask" and "signal mask saved" so
...
that the layout of sigjmp_bufs and jmp_bufs are the same (except
for the "signal mask saved" appended to the end).
1994-12-24 01:05:44 +00:00
fcc40a849d
If len == 0, then the result must be 0 regardless of the other arguments.
1994-11-24 16:36:35 +00:00
2db41a19e7
Use trap gate to enter kernel.
1994-10-26 19:49:42 +00:00
837f99d25c
Clean up deleted files.
1994-10-06 04:38:25 +00:00
345c8ef85c
Clean up deleted files.
1994-04-02 05:36:54 +00:00
da34d8a406
have cerror DTRT on returns, w.r.t. quads and -1. it needs to be done
...
here (think of syscall(SYS_lseek,...)). also, kill bogus lseek thang.
1994-04-02 05:36:52 +00:00
b3086b2299
Need to set %edx to -1 before cerror, though perhaps we should have something
...
more generic for this.
1994-03-31 18:38:57 +00:00
a3450ab6e8
Add a .text and .align to SYSCALL().
1994-03-31 18:31:23 +00:00
98cbb217ac
mem -> reg & reg->reg moves take the same time; so save a cycle by reading
...
the block length from the stack instead of caching it in a register.
1994-03-22 01:15:47 +00:00
1f29a3d910
Smaller & Faster.
1994-03-16 17:40:23 +00:00
5ac78ccb34
[hn]to[nh][sl] functions used xchgb %h1,%b1 instead of the faster (on the
...
i486 & i586) rorw $8, %w1.
1994-03-16 17:34:14 +00:00
915f7d4d3d
Use RCSID macro.
1994-03-12 01:48:26 +00:00
c95cd47ba2
Use RCSID macro for RCS ID's.
1994-03-12 01:39:55 +00:00
90ff28fef4
Style police. Use `je' and `jne' only on the result of an equality test;
...
use `jz' and `jnz' in all other cases.
1994-02-14 20:05:52 +00:00
7f582f99d4
Fix spelling mistake: cashe -> cache.
1994-02-14 17:45:16 +00:00
01284ad877
Performance improvement from Bruce Evans.
...
Removes one test instruction for each character.
1994-02-14 17:44:39 +00:00
f28455f783
Use "ror $8,%ax" instead of "xchg %al,%ah": equal on a i386, 1 cycle faster
...
on a i486, and 2 cycles faster on a iPentium.
1994-02-14 17:43:19 +00:00
5cffad2151
Use movzbl, not movsbl, to zero extend for unsigned comparison.
...
Fix from Bruce Evans.
1994-02-07 20:57:05 +00:00
e2c220f94b
Fix spelling error in Copyright notice
1994-01-28 20:34:33 +00:00
1fe07f186d
Separate curbrk and mincrk.
1994-01-27 15:08:03 +00:00
3463cb4eef
Trivial speedup inspired by SPARC code.
1994-01-27 15:00:49 +00:00
ef980f76fe
One last tweak.
1994-01-27 06:53:25 +00:00
22e614ce10
Random cleanup.
1994-01-27 06:34:59 +00:00
9c8862a970
Pedanticism.
1994-01-27 06:34:02 +00:00
3cf1bd9939
SYS_brk --> SYS_break. Where's my axe?
1994-01-27 04:30:34 +00:00
17b2a501cd
Stylistic changes.
1994-01-27 04:16:50 +00:00
0496df7164
More stylistic changes.
1994-01-27 04:14:50 +00:00
e84d8b28f3
Stylistic changes.
1994-01-27 04:10:20 +00:00
b6b0ec3424
Fix bug jtc introduced.
1994-01-27 04:07:13 +00:00
7835a85a78
Optimization tweaks:
...
Use testl %reg,%reg instead of cmpl $0,%reg
Use xorl %reg,%reg instead of movl $0,%reg
1994-01-24 19:24:36 +00:00
39c0326588
PIC_*() macros moved to <machine/asm.h>
1994-01-07 00:48:22 +00:00
0e09f939b2
Fix bug #24 by renaming _logname_valid to __logname_valid.
1993-12-18 01:02:47 +00:00
cc81da7447
Shorten setup for longword operations a little.
1993-12-10 10:51:11 +00:00
dca61f6f90
Share code.
1993-12-10 10:45:18 +00:00
a9e5cc4f83
No reason these can't share code...
1993-12-08 22:10:14 +00:00
f3f7594fc9
Use PIC_PLT for error returns.
...
Correct operand order in accessing `minbrk' and `curbrk'.
1993-12-06 23:28:08 +00:00
417e7a6fff
Remove spurious PIC_EPILOGUE and PIC_GOTOFF macros.
1993-12-06 23:27:04 +00:00
3581f42619
Use PIC_PLT() macro for error returns
1993-12-06 23:26:04 +00:00
0ada107897
update for .S and MACHINE_ARCH
1993-12-04 02:33:35 +00:00
448e711c78
Clean up deleted files.
1993-12-04 01:28:42 +00:00
32d96ccd3c
Remove special case for len == 0 --- It wasn't needed. By dropping the
...
conditional, the general case (where we are actually swapping bytes) is
a tiny bit faster.
1993-11-25 01:18:24 +00:00
2be44f5352
copy SYS.h and DEFS.h, too.
1993-11-05 22:44:00 +00:00
4418616857
Move the old berkeley sccs id into the header comment like it is done
...
for other source files.
1993-10-21 02:19:25 +00:00
6be8b83abc
Gcc helper functions should still use the ENTRY() macro so they will be
...
show up in profiled code. (The gcc helper functions in the other ports
do this allready).
1993-10-21 02:03:37 +00:00
34d1aad0c1
Use PIC_PLT macro with the function calls (as was done with setjmp()).
1993-10-21 01:59:44 +00:00
79591d04bc
So many files in arch/i386 have been changed I might as well run my
...
remove unecessary trailing whitespace script.
1993-10-21 01:48:14 +00:00
823c36577c
Change instances of LIBC_RCS back to LIBC_SCCS. If we are going to change
...
th LIBC_RCS, the entire source tree should be updated at once.
Remove && !defined(lint) from conditionals around rcs id's. No one is
ever going to pass a *.s file through lint.
1993-10-21 01:39:55 +00:00
97ad548bc0
Replaced instructions with immediate operands with compatible alternatives
...
saving a few bytes while the cycle count remains the same --- could
actually be faster, depending on the cache, prefetch queue, etc.
1993-10-20 18:50:15 +00:00
6802033480
sigsetjmp() for the i386.
1993-10-19 20:44:00 +00:00
39ff8227a8
Makefile.inc, basically a place to put KMSRC definition
1993-10-15 01:24:47 +00:00
479a1c33c9
Provide i386 versions of div and ldiv -- they are simpler than the C versions,
...
since the idiv instruction does the right thing wrt truncating towards zero.
1993-10-08 23:57:13 +00:00
e09d2f42db
Use "testb %cl,%cl" instead of "cmpb $0,%cl".
...
This saves three cycles per character, and reduces the size of the
function by eight bytes (the loop is unrolled eight times).
Thanks to davidg for pointing this out.
1993-10-08 01:25:51 +00:00
b3f9f8f6f4
Provide an assembly language version of labs().
...
Since ints and longs are both 32 bits on a i386, we could alias abs()
and labs() together as mycroft did with memmove()/memcpy(), but I'm
waiting on an interpretation ruling to see if it is legal.
1993-10-08 00:01:04 +00:00
0b9f50897e
make an arch directory in libc, and move all machine dirs into it
1993-10-07 00:17:24 +00:00
jtc