| Copyright (c) 1988 Regents of the University of California.
| Redistribution and use in source and binary forms, with or without
| modification, are permitted provided that the following conditions
| 1. Redistributions of source code must retain the above copyright
| notice, this list of conditions and the following disclaimer.
| 2. Redistributions in binary form must reproduce the above copyright
| notice, this list of conditions and the following disclaimer in the
| documentation and/or other materials provided with the distribution.
| 3. All advertising materials mentioning features or use of this software
| must display the following acknowledgement:
| This product includes software developed by the University of
| California, Berkeley and its contributors.
| 4. Neither the name of the University nor the names of its contributors
| may be used to endorse or promote products derived from this software
| without specific prior written permission.
| THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
| ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
| IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
| ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
| FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
| DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
| OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
| HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
| LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
| OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
| @(#)oc_cksum.s 7.2 (Berkeley) 11/3/90
| oc_cksum: ones complement 16 bit checksum for MC68020.
| oc_cksum (buffer, count, strtval)
| Do a 16 bit one's complement sum of 'count' bytes from 'buffer'.
| 'strtval' is the starting value of the sum (usually zero).
| It simplifies life in in_cksum if strtval can be >= 2^16.
| This routine will work as long as strtval is < 2^31.
| This routine is intended for MC 68020s but should also work
| for 68030s. It (deliberately) doesn't worry about the alignment
| of the buffer so will only work on a 68010 if the buffer is
| aligned on an even address. (Also, a routine written to use
| 68010 "loop mode" would almost certainly be faster than this
| We don't worry about alignment because this routine is frequently
| called with small counts: 20 bytes for IP header checksums and 40
| bytes for TCP ack checksums. For these small counts, testing for
| bad alignment adds ~10% to the per-call cost. Since, by the nature
| of the kernel's allocator, the data we're called with is almost
| always longword aligned, there is no benefit to this added cost
| and we're better off letting the loop take a big performance hit
| in the rare cases where we're handed an unaligned buffer.
| Loop unrolling constants of 2, 4, 8, 16, 32 and 64 times were
| tested on random data on four different types of processors (see
| list below -- 64 was the largest unrolling because anything more
| overflows the 68020 Icache). On all the processors, the
| throughput asymptote was located between 8 and 16 (closer to 8).
| However, 16 was substantially better than 8 for small counts.
| (It's clear why this happens for a count of 40: unroll-8 pays a
| loop branch cost and unroll-16 doesn't. But the tests also showed
| that 16 was better than 8 for a count of 20. It's not obvious to
| me why.) So, since 16 was good for both large and small counts,
| the loop below is unrolled 16 times.
| The processors tested and their average time to checksum 1024 bytes
| Sun 3/50 (15MHz) 190 us/KB
| Sun 3/180 (16.6MHz) 175 us/KB
| Sun 3/60 (20MHz) 134 us/KB
| Sun 3/280 (25MHz) 95 us/KB
| The cost of calling this routine was typically 10% of the per-
| kilobyte cost. E.g., checksumming zero bytes on a 3/60 cost 9us
| and each additional byte cost 125ns. With the high fixed cost,
| it would clearly be a gain to "inline" this routine -- the
| subroutine call adds 400% overhead to an IP header checksum.
| However, in absolute terms, inlining would only gain 10us per
| packet -- a 1% effect for a 1ms ethernet packet. This is not
| enough gain to be worth the effort.
.asciz "@(#)$Header: oc_cksum.s,v 1.1 90/07/09 16:04:43 mike Exp $"
movl sp@(4),a0 | get buffer ptr
movl sp@(8),d1 | get byte count
movl sp@(12),d0 | get starting value
movl d2,sp@- | free a reg
| test for possible 1, 2 or 3 bytes of excess at end
| of buffer. The usual case is no excess (the usual
| case is header checksums) so we give that the faster
| 'not taken' leg of the compare. (We do the excess
| first because we're about the trash the low order
| bits of the count in d1.)
jne L5 | if one or three bytes excess
jne L7 | if two bytes excess
lsrl #6,d1 | make cnt into # of 64 byte chunks
andl #0x3c,d2 | then find fractions of a chunk
dbra d1,L2 | (NB- dbra doesn't affect X)
movl d0,d1 | fold 32 bit sum to 16 bits
swap d1 | (NB- swap doesn't affect X)
L5: | deal with 1 or 3 excess bytes at the end of the buffer.
movw a0@(-3,d1:l),d2 | add in last full word then drop
addl d2,d0 | through to pick up last byte
projects / unix-history / blob