If SRCS contains any .S files then remove the corresponding .c file from
SRCS and add to LSRCS (for lint).
Change the 'string' Makefiles so that the arch/*/string/Makefile.inc
need only specify the .S files for that architecture and not the .c
files for all the files they don't override.

Change all archs so that strchr.[cS] and strrchr.[cS] exist and generate
duplicate symbols for index() and rindex().
libc and kernel build ok for affected archs (i386, amd64, vax, sun3,
evbmips-eb) all seem ok.
(vax might lose an asm index() - but comments imply it is worse than the C)

Implement a significantly improved version of memccpy(), with thanks to
Ignatios Souvatzis for correcting the SVR4 ABI code and for a few
micro-optimizations that followed from these changes.

branches: 1.2.18;
Assembly language implementation of strcat(). Derived by merging existing
strlen() and strcpy() functions.

Separate machine dependencies.

Use reach-over sources from common/lib/libc

Move UCB-licensed code from 4-clause to 3-clause licence.

Patches provided by Joel Baker in PR 22280, verified by myself.

On the 68010, we are forced to do bytewise compares
and copies if the two addresses aren't of the same
evenness.

branches: 1.11.8;
Add register prefixes.

branches: 1.10.6;
Fix typo

Change b{eq,ne,lt,le,gt,ge} -> j{eq,ne,lt,le,gt,ge}.

Faster versions of bcmp(), bcopy(), bzero(), and memset().

The implementation of these functions is losely based on the original
BSD code plus versions I was working on which would yeild code tuned
for a particular m68k or coldfire family cpu, code space or run time
efficiency, etc. depending on the values of various macros.

My original code was intended for use in OS-less embedded systems.
Because NetBSD can not benefit from the considerable complexity, I've
removed the extraneous preprocessor goop. The resulting code is tuned
for the '020 - '060. It will not function on the '000 or coldfire.

#include <machine/asm.h> instead of "DEFS.h"

Use jCC instead of bCC, as the assembler will then use the (smaller,
faster) byte-displacement form of the instruction if the offset will
fit. This happens to be all occurances in this case, so I could have
used bCCs. However, a quick survey of the rest of NetBSD's m68k code
showed that jCC was used in similar cases. I did the same to be
consistant.

Add L prefix to local labels

Fix a brain-o by me; if I'm going to use stuff in <machine/asm.h>, I better
darn well include it in the right place. "Hello, McFly!"

New style RCS ids, and employ the RCSID() macro.

Make the coding style more consistent.

Machine-dependent files for generic m68k machines (copied from hp300)

Clean up deleted files.

Machine-dependent files for generic m68k machines (copied from hp300)

Use reach-over sources from common/lib/libc

Move UCB-licensed code from 4-clause to 3-clause licence.

Patches provided by Joel Baker in PR 22280, verified by myself.

On the 68010, we are forced to do bytewise compares
and copies if the two addresses aren't of the same
evenness.

branches: 1.15.8;
Add register prefixes.

branches: 1.14.2;
When using the SVR4 ABI, return pointer results in a0, too.

Fixup reverse bcopy

Change b{eq,ne,lt,le,gt,ge} -> j{eq,ne,lt,le,gt,ge}.

Faster versions of bcmp(), bcopy(), bzero(), and memset().

The implementation of these functions is losely based on the original
BSD code plus versions I was working on which would yeild code tuned
for a particular m68k or coldfire family cpu, code space or run time
efficiency, etc. depending on the values of various macros.

My original code was intended for use in OS-less embedded systems.
Because NetBSD can not benefit from the considerable complexity, I've
removed the extraneous preprocessor goop. The resulting code is tuned
for the '020 - '060. It will not function on the '000 or coldfire.

#include <machine/asm.h> instead of "DEFS.h"

Use jCC instead of bCC, as the assembler will then use the (smaller,
faster) byte-displacement form of the instruction if the offset will
fit. This happens to be all occurances in this case, so I could have
used bCCs. However, a quick survey of the rest of NetBSD's m68k code
showed that jCC was used in similar cases. I did the same to be
consistant.

Fix a brain-o by me; if I'm going to use stuff in <machine/asm.h>, I better
darn well include it in the right place. "Hello, McFly!"

New style RCS ids, and employ the RCSID() macro.

remove single quote from comment

Make the coding style more consistent.

size_t is unsigned.

Fix typo.

Add code for memcpy() and memmove().

Machine-dependent files for generic m68k machines (copied from hp300)

Clean up deleted files.

Machine-dependent files for generic m68k machines (copied from hp300)

Use reach-over sources from common/lib/libc

Move UCB-licensed code from 4-clause to 3-clause licence.

Patches provided by Joel Baker in PR 22280, verified by myself.

Add register prefixes.

branches: 1.9.6;
Change b{eq,ne,lt,le,gt,ge} -> j{eq,ne,lt,le,gt,ge}.

Faster versions of bcmp(), bcopy(), bzero(), and memset().

The implementation of these functions is losely based on the original
BSD code plus versions I was working on which would yeild code tuned
for a particular m68k or coldfire family cpu, code space or run time
efficiency, etc. depending on the values of various macros.

My original code was intended for use in OS-less embedded systems.
Because NetBSD can not benefit from the considerable complexity, I've
removed the extraneous preprocessor goop. The resulting code is tuned
for the '020 - '060. It will not function on the '000 or coldfire.

#include <machine/asm.h> instead of "DEFS.h"

Use jCC instead of bCC, as the assembler will then use the (smaller,
faster) byte-displacement form of the instruction if the offset will
fit. This happens to be all occurances in this case, so I could have
used bCCs. However, a quick survey of the rest of NetBSD's m68k code
showed that jCC was used in similar cases. I did the same to be
consistant.

Add L prefix to local labels

Fix a brain-o by me; if I'm going to use stuff in <machine/asm.h>, I better
darn well include it in the right place. "Hello, McFly!"

New style RCS ids, and employ the RCSID() macro.

Make the coding style more consistent.

Machine-dependent files for generic m68k machines (copied from hp300)

Clean up deleted files.

Machine-dependent files for generic m68k machines (copied from hp300)

Use reach-over sources from common/lib/libc

Move UCB-licensed code from 4-clause to 3-clause licence.

Patches provided by Joel Baker in PR 22280, verified by myself.

On the 68010 only, assemble a different version of
this function, since the 68010 doesn't have bit
instructions.

branches: 1.7.8;
Add register prefixes.

branches: 1.6.6;
#include <machine/asm.h> instead of "DEFS.h"

Fix a brain-o by me; if I'm going to use stuff in <machine/asm.h>, I better
darn well include it in the right place. "Hello, McFly!"

New style RCS ids, and employ the RCSID() macro.

Use bfffo rather than a slow loop.

Changed beq's to jeq's, because (a) the jeq gets optimized to a short branch,
and (b) gas currently assembles the beq's wrongly.

Machine-dependent files for generic m68k machines (copied from hp300)

Clean up deleted files.

Changed beq's to jeq's, because (a) the jeq gets optimized to a short branch,
and (b) gas currently assembles the beq's wrongly.

Machine-dependent files for generic m68k machines (copied from hp300)

Use reach-over sources from common/lib/libc

Move UCB-licensed code from 4-clause to 3-clause licence.

Patches provided by Joel Baker in PR 22280, verified by myself.

Add register prefixes.

branches: 1.8.2;
When using the SVR4 ABI, return pointer results in a0, too.

#include <machine/asm.h> instead of "DEFS.h"

Use jCC instead of bCC, as the assembler will then use the (smaller,
faster) byte-displacement form of the instruction if the offset will
fit. This happens to be all occurances in this case, so I could have
used bCCs. However, a quick survey of the rest of NetBSD's m68k code
showed that jCC was used in similar cases. I did the same to be
consistant.

Fix a brain-o by me; if I'm going to use stuff in <machine/asm.h>, I better
darn well include it in the right place. "Hello, McFly!"

New style RCS ids, and employ the RCSID() macro.

Make the coding style more consistent.

No reason these can't share code...

Machine-dependent files for generic m68k machines (copied from hp300)

Clean up deleted files.

Machine-dependent files for generic m68k machines (copied from hp300)

Merge riastradh-drm2 to HEAD.

Adjust for coldfire

branches: 1.3.2;
Add END(func)

Convert to normal Motorola syntax for addressing.
%r@ -> (%r)
%r@- -> -(%r)
%r@+ -> (%r)+
%r@(n) n(%r)
etc.
No object differences.

branches: 1.1.2; 1.1.62; 1.1.68;
Implement a significantly improved version of memccpy(), with thanks to
Ignatios Souvatzis for correcting the SVR4 ABI code and for a few
micro-optimizations that followed from these changes.

The overall improvement was measured at 3-4x for copies of at least
4 bytes on the 68030, and 8 bytes on the 68040. For smaller copies
the improvement was still measurable, if not as dramatic.

Use reach-over sources from common/lib/libc

Move UCB-licensed code from 4-clause to 3-clause licence.

Patches provided by Joel Baker in PR 22280, verified by myself.

Add register prefixes.

branches: 1.9.6;
Change b{eq,ne,lt,le,gt,ge} -> j{eq,ne,lt,le,gt,ge}.

#include <machine/asm.h> instead of "DEFS.h"

Fix a brain-o by me; if I'm going to use stuff in <machine/asm.h>, I better
darn well include it in the right place. "Hello, McFly!"

New style RCS ids, and employ the RCSID() macro.

Make the coding style more consistent.

Fix a couple of nits.

More assembler code.

Fudged an addressing mode.

Initially cloned from bcmp.S.

Use reach-over sources from common/lib/libc

RCS Id Police.

Fix typo.

#define the appropriate symbol and #include bcopy.S.

Use reach-over sources from common/lib/libc

RCS Id Police.

Fix typo.

#define the appropriate symbol and #include bcopy.S.

Use reach-over sources from common/lib/libc

Move UCB-licensed code from 4-clause to 3-clause licence.

Patches provided by Joel Baker in PR 22280, verified by myself.

On the 68010, don't bother with the by-8-longwords loop, instead
only 16-bit align the region and use loop mode with longwords
to do the bulk of the work.

On the 68020+, fixed an optimization bug where all regions that start
on odd addresses would not be properly longword aligned.

branches: 1.12.8;
Add register prefixes.

branches: 1.11.2;
When using the SVR4 ABI, return pointer results in a0, too.

oops, bzero -> memset

Change b{eq,ne,lt,le,gt,ge} -> j{eq,ne,lt,le,gt,ge}.

Faster versions of bcmp(), bcopy(), bzero(), and memset().

The implementation of these functions is losely based on the original
BSD code plus versions I was working on which would yeild code tuned
for a particular m68k or coldfire family cpu, code space or run time
efficiency, etc. depending on the values of various macros.

My original code was intended for use in OS-less embedded systems.
Because NetBSD can not benefit from the considerable complexity, I've
removed the extraneous preprocessor goop. The resulting code is tuned
for the '020 - '060. It will not function on the '000 or coldfire.

#include <machine/asm.h> instead of "DEFS.h"

Fix a brain-o by me; if I'm going to use stuff in <machine/asm.h>, I better
darn well include it in the right place. "Hello, McFly!"

New style RCS ids, and employ the RCSID() macro.

Make the coding style more consistent.

Fix a bug, and speed up by a few cycles.

Fix a couple of nits.

More assembler code.

Use reach-over sources from common/lib/libc

Move UCB-licensed code from 4-clause to 3-clause licence.

Patches provided by Joel Baker in PR 22280, verified by myself.

Add register prefixes.

branches: 1.8.2;
When using the SVR4 ABI, return pointer results in a0, too.

#include <machine/asm.h> instead of "DEFS.h"

Use jCC instead of bCC, as the assembler will then use the (smaller,
faster) byte-displacement form of the instruction if the offset will
fit. This happens to be all occurances in this case, so I could have
used bCCs. However, a quick survey of the rest of NetBSD's m68k code
showed that jCC was used in similar cases. I did the same to be
consistant.

Fix a brain-o by me; if I'm going to use stuff in <machine/asm.h>, I better
darn well include it in the right place. "Hello, McFly!"

New style RCS ids, and employ the RCSID() macro.

Make the coding style more consistent.

No reason these can't share code...

Machine-dependent files for generic m68k machines (copied from hp300)

Clean up deleted files.

Machine-dependent files for generic m68k machines (copied from hp300)

Use reach-over sources from common/lib/libc

Move UCB-licensed code from 4-clause to 3-clause licence.

Patches provided by Joel Baker in PR 22280, verified by myself.

Add register prefixes.

branches: 1.8.2;
When using the SVR4 ABI, return pointer results in a0, too.

#include <machine/asm.h> instead of "DEFS.h"

Use jCC instead of bCC, as the assembler will then use the (smaller,
faster) byte-displacement form of the instruction if the offset will
fit. This happens to be all occurances in this case, so I could have
used bCCs. However, a quick survey of the rest of NetBSD's m68k code
showed that jCC was used in similar cases. I did the same to be
consistant.

Add L prefix on local labels

Fix a brain-o by me; if I'm going to use stuff in <machine/asm.h>, I better
darn well include it in the right place. "Hello, McFly!"

New style RCS ids, and employ the RCSID() macro.

fix stupid typo

Assembly language implementation of strcat(). Derived by merging existing
strlen() and strcpy() functions.

Use reach-over sources from common/lib/libc

RCS Id Police.

No reason these can't share code...

Use reach-over sources from common/lib/libc

netbsd.org->NetBSD.org

Use the EXTBL macro where we want the extbl instruction.
On the 68010 this expands to an extw/extl combination.

branches: 1.15.2;
Another change by Hiroshi Horimoto.

Fix spelling of contributors name

Add register prefixes.

branches: 1.12.6;
Revert last change, some folks have reported problems

Use insn sequence found by superoptimizer to compute return value. This
avoids the relatively expensive scc and ext instructions.

I removed the extra instructions which compute the return value in Hiroshi
Horitomo's implementation a bit too hastily. They are required to handle
the unsigned comparison required by the standard.

New implementations of strcmp() and strncmp().

This code uses subb instead of cmpb to compare chars. When they are
not equal, the function's return value is computed by sign extending
the difference. The basic idea was contributed by Hiroshi Horitomo
<horimoto@cs-aoi.cs.sist.ac.jp> in PR #3223, but his implementation
used additional instructions to compute the return value.

Also, the series of insns used to compare one pair of characters has
been unrolled 4 times. This helps the 680[23]0, where the cost of a
taken branch is significantly more than that of a not-taken branch.

#include <machine/asm.h> instead of "DEFS.h"

Use jCC instead of bCC, as the assembler will then use the (smaller,
faster) byte-displacement form of the instruction if the offset will
fit. This happens to be all occurances in this case, so I could have
used bCCs. However, a quick survey of the rest of NetBSD's m68k code
showed that jCC was used in similar cases. I did the same to be
consistant.

Add L prefix on local labels

Fix a brain-o by me; if I'm going to use stuff in <machine/asm.h>, I better
darn well include it in the right place. "Hello, McFly!"

Fix a typo.

New style RCS ids, and employ the RCSID() macro.

Make the coding style more consistent.

Machine-dependent files for generic m68k machines (copied from hp300)

Clean up deleted files.

Machine-dependent files for generic m68k machines (copied from hp300)

Use reach-over sources from common/lib/libc

Move UCB-licensed code from 4-clause to 3-clause licence.

Patches provided by Joel Baker in PR 22280, verified by myself.

Add register prefixes.

branches: 1.8.2;
When using the SVR4 ABI, return pointer results in a0, too.

#include <machine/asm.h> instead of "DEFS.h"

Use jCC instead of bCC, as the assembler will then use the (smaller,
faster) byte-displacement form of the instruction if the offset will
fit. This happens to be all occurances in this case, so I could have
used bCCs. However, a quick survey of the rest of NetBSD's m68k code
showed that jCC was used in similar cases. I did the same to be
consistant.

Add L prefix on local labels

Fix a brain-o by me; if I'm going to use stuff in <machine/asm.h>, I better
darn well include it in the right place. "Hello, McFly!"

New style RCS ids, and employ the RCSID() macro.

Make the coding style more consistent.

Machine-dependent files for generic m68k machines (copied from hp300)

Clean up deleted files.

Machine-dependent files for generic m68k machines (copied from hp300)

Use reach-over sources from common/lib/libc

Move UCB-licensed code from 4-clause to 3-clause licence.

Patches provided by Joel Baker in PR 22280, verified by myself.

Add register prefixes.

branches: 1.8.6;
#include <machine/asm.h> instead of "DEFS.h"

Use jCC instead of bCC, as the assembler will then use the (smaller,
faster) byte-displacement form of the instruction if the offset will
fit. This happens to be all occurances in this case, so I could have
used bCCs. However, a quick survey of the rest of NetBSD's m68k code
showed that jCC was used in similar cases. I did the same to be
consistant.

Add L prefix on local labels

Fix a brain-o by me; if I'm going to use stuff in <machine/asm.h>, I better
darn well include it in the right place. "Hello, McFly!"

New style RCS ids, and employ the RCSID() macro.

Rewrite to remove instruction in inside of loop. Instead of maintaining a
count, subtract the start address from the end address.

Make the coding style more consistent.

Machine-dependent files for generic m68k machines (copied from hp300)

Clean up deleted files.

Machine-dependent files for generic m68k machines (copied from hp300)

Use reach-over sources from common/lib/libc

netbsd.org->NetBSD.org

Use the EXTBL macro where we want the extbl instruction.
On the 68010 this expands to an extw/extl combination.

branches: 1.13.8;
Fix spelling of contributors name

Add register prefixes.

branches: 1.11.6;
Revert last change, some folks have reported problems

Use insn sequence found by superoptimizer to compute return value. This
avoids the relatively expensive scc and ext instructions.

I removed the extra instructions which compute the return value in Hiroshi
Horitomo's implementation a bit too hastily. They are required to handle
the unsigned comparison required by the standard.

New implementations of strcmp() and strncmp().

This code uses subb instead of cmpb to compare chars. When they are
not equal, the function's return value is computed by sign extending
the difference. The basic idea was contributed by Hiroshi Horitomo
<horimoto@cs-aoi.cs.sist.ac.jp> in PR #3223, but his implementation
used additional instructions to compute the return value.

Also, the series of insns used to compare one pair of characters has
been unrolled 4 times. This helps the 680[23]0, where the cost of a
taken branch is significantly more than that of a not-taken branch.

#include <machine/asm.h> instead of "DEFS.h"

Use jCC instead of bCC, as the assembler will then use the (smaller,
faster) byte-displacement form of the instruction if the offset will
fit. This happens to be all occurances in this case, so I could have
used bCCs. However, a quick survey of the rest of NetBSD's m68k code
showed that jCC was used in similar cases. I did the same to be
consistant.

Add L prefix on local labels

Fix a brain-o by me; if I'm going to use stuff in <machine/asm.h>, I better
darn well include it in the right place. "Hello, McFly!"

New style RCS ids, and employ the RCSID() macro.

Make the coding style more consistent.

Machine-dependent files for generic m68k machines (copied from hp300)

Clean up deleted files.

Machine-dependent files for generic m68k machines (copied from hp300)

Use reach-over sources from common/lib/libc

Move UCB-licensed code from 4-clause to 3-clause licence.

Patches provided by Joel Baker in PR 22280, verified by myself.

Use dbeq loops, code submitted by Hiroshi Horimoto in PR 3223, integrated
and tested by Allen Briggs.
Needs 1/3 less time on 68030, 45% less time on 68040.
XXX On 68060, 1/2 less time for the padding part, but 40% more time on the
copy part. I've stared at the 68060 User Manual, but can't understand why;
must be some wierd (branch) cache interaction.

Add register prefixes.

branches: 1.9.2;
When using the SVR4 ABI, return pointer results in a0, too.

Change b{eq,ne,lt,le,gt,ge} -> j{eq,ne,lt,le,gt,ge}.

#include <machine/asm.h> instead of "DEFS.h"

Use jCC instead of bCC, as the assembler will then use the (smaller,
faster) byte-displacement form of the instruction if the offset will
fit. This happens to be all occurances in this case, so I could have
used bCCs. However, a quick survey of the rest of NetBSD's m68k code
showed that jCC was used in similar cases. I did the same to be
consistant.

Add L prefix on local labels

Fix a brain-o by me; if I'm going to use stuff in <machine/asm.h>, I better
darn well include it in the right place. "Hello, McFly!"

New style RCS ids, and employ the RCSID() macro.

Make the coding style more consistent.

Machine-dependent files for generic m68k machines (copied from hp300)

Clean up deleted files.

Machine-dependent files for generic m68k machines (copied from hp300)

Use reach-over sources from common/lib/libc

RCS Id Police.

No reason these can't share code...

Merge riastradh-drm2 to HEAD.

Adjust for coldfire

branches: 1.9.2;
Add END(func)

Convert to normal Motorola syntax for addressing.
%r@ -> (%r)
%r@- -> -(%r)
%r@+ -> (%r)+
%r@(n) n(%r)
etc.
No object differences.

branches: 1.7.66; 1.7.72;
Add register prefixes.

branches: 1.6.6;
#include <machine/asm.h> instead of "DEFS.h"

Use jCC instead of bCC, as the assembler will then use the (smaller,
faster) byte-displacement form of the instruction if the offset will
fit. This happens to be all occurances in this case, so I could have
used bCCs. However, a quick survey of the rest of NetBSD's m68k code
showed that jCC was used in similar cases. I did the same to be
consistant.

New style RCS ids, and employ the RCSID() macro.

Make the coding style more consistent.

Need to divide count by 2.

Assembler version.