<- RFC Index (1901..2000)
RFC 1936
Network Working Group J. Touch
Request For Comments: 1936 B. Parham
Category: Informational ISI
April 1996
Implementing the Internet Checksum in Hardware
Status of This Memo
This memo provides information for the Internet community. This memo
does not specify an Internet standard of any kind. Distribution of
this memo is unlimited.
Abstract
This memo presents a techniques for efficiently implementing the
Internet Checksum in hardware. It includes PLD code for programming a
single, low cost part to perform checksumming at 1.26 Gbps.
Introduction
The Internet Checksum is used in various Internet protocols to check
for data corruption in headers (e.g., IP) [4] and packet bodies (e.g,
UDP, TCP) [5][6]. Efficient software implementation of this checksum
has been addressed in previous RFCs [1][2][3][7].
Efficient software implementations of the Internet Checksum algorithm
are often embedded in data copying operations ([1], Section 2). This
copy operation is increasingly being performed by dedicated direct
memory access (DMA) hardware. As a result, DMA hardware designs are
beginning to incorporate dedicated hardware to compute the Internet
Checksum during the data transfer.
This note presents the architecture of an efficient, pipelined
Internet Checksum mechanism, suitable for inclusion in DMA hardware
[8]. This design can be implemented in a relatively inexpensive
programmable logic device (PLD) (1995 cost of $40), and is capable of
supporting 1.26 Gbps transfer rates, at 26 ns per 32-bit word.
Appendix A provides the pseudocode for such a device. This design has
been implemented in the PC-ATOMIC host interface hardware [8]. We
believe this design is of general use to the Internet community.
Touch & Parham Informational [Page 1]
RFC 1936 Implementing the Internet Checksum in Hardware April 1996
The remainder of this document is organized as follows:
Review of the Internet Checksum
One's Complement vs. Two's Complement Addition
Interfaces
Summary
Appendix A - PLD source code
Touch & Parham Informational [Page 2]
RFC 1936 Implementing the Internet Checksum in Hardware April 1996
A Review of the Internet Checksum
The Internet Checksum is used for detecting corruption in a block of
data [1]. It is initialized to zero, and computed as the complement
of the ones-complement sum of the data, taken in 16-bit units. A
subsequent checksum of the data and checksum together should generate
a zero checksum if no errors are detected.
The checksum allows [1]:
- byte order "independence"
reordered output is equivalent to reordered input
- 16-bit word-order independence
reordering 16-bit words preserves the output
- incremental computation
- deferred carries
- parallel summation
a result of deferred carries, incremental
computation, and 16-bit word order independence
This note describes an implementation that computes two partial
checksums in parallel, over the odd and even 16-bit half-words of
32-bit data. The result is a pair of partial checksums (odd and
even), which can be combined, and the result inverted to generate the
true Internet Checksum. This technique is related to the long-word
parallel summation used in efficient software implementations [1].
+------------------+ +------------------+
| high half-word | | low half-word |
| ones-complement | | ones-complement |
| partial checksum | | partial checksum |
+------------------+ +------------------+
\ /
* (ones-complement sum)
|
+------------------+
| partial checksum |
+------------------+
|
* (ones-complement negative)
|
+-------------------+
| final |
| Internet Checksum |
+-------------------+
Touch & Parham Informational [Page 3]
RFC 1936 Implementing the Internet Checksum in Hardware April 1996
One's Complement vs. Two's Complement Addition
The Internet Checksum is composed of a ones-complement lookahead
adder and a bit-wise inverter. A ones-complement adder can be built
either using twos-complement components, or natively.
A twos-complement implementation of a ones-complement adder requires
either two twos-complement adders, or two cycles per add. The sum is
performed, then the high-bit carry-out is propagated to the carry-in,
and a second sum is performed. (ones-complement addition is {+1s} and
twos-complement is {+2s})
a {+1s} b == (a {+2s} b) + carry(a {+2s} b)
e.g.,
halfword16 a,b;
word32 c;
a {+1s} b == r
such that:
c = a {+2s} b; # sum value
r = (c & 0xFFFF) {+2s} (c >> 16); # sum carry
Bits of a twos-complement lookahead adder are progressively more
complex in carry lookahead. (OR the contents of each row, where terms
are AND'd or XOR'd {^})
4-bit carry-lookahead 2's complement adder:
a,b : input data
p : carry propagate, where pi = ai*bi = (ai)(bi)
g : carry generate, where gi = ai + bi
Out0 := a0 ^ b0 ^ ci
Out1 := a1 ^ b1 ^ (cip0 + g0)
Out2 := a2 ^ b2 ^ (cip0p1 + g0p1 + g1)
Out3 := a3 ^ b3 ^ (cip0p1p2 + g0p1p2 + g1p2 + g2)
Cout := cip0p1p2p3 + g0p1p2p3 + g1p2p3 + g2p3 + g3
The true ones-complement lookahead adder recognizes that the carry-
wrap of the twos-complement addition is equivalent to a toroidal
carry-lookahead. Bits of a ones-complement lookahead adder are all
the same complexity, that of the high-bit of a twos-complement
lookahead adder. Thus the ones-complement sum (and thus the Internet
Checksum) is bit-position independent. We replace `ci' with the `co'
expression and reduce. (OR terms in each row pair).
Touch & Parham Informational [Page 4]
RFC 1936 Implementing the Internet Checksum in Hardware April 1996
4-bit carry-lookahead 1's complement ring adder:
Out0 = a0 ^ b0 ^ (g3 + g2p3 + g1p2p3 + g0p1p2p3)
Out1 = a1 ^ b1 ^ (g3p0 + g2p3p0 + g1p2p3p0 + g0)
Out2 = a2 ^ b2 ^ (g3p0p1 + g2p3p0p1 + g1 + g0p1)
Out3 = a3 ^ b3 ^ (g3p0p1p2 + g2 + g1p2 + g0p1p2)
A hardware implementation can use this toroidal design directly,
together with conventional twos-complement fast-adder internal
components, to perform a pipelined ones-complement adder [8].
A VLSI implementation could use any full-lookahead adder, adapted to
be toroidal and bit-equivalent, as above. In our PLD implementation,
we implement the adders via 2- and 3-bit full-lookahead sub-
components. The adder components are chained in a ring via carry bit
registers. This relies on delayed carry-propagation to implement a
carry pipeline between the fast-adder stages.
Full-lookahead adders in a toroidal pipeline
+-+-+-+ +-+-+-+ +-+-+ +-+-+-+ +-+-+-+ +-+-+
|i|i|i| |i|i|i| |i|i| |i|i|i| |i|i|i| |i|i|
|F|E|D| |C|B|A| |9|8| |7|6|5| |4|3|2| |1|0|
+-+-+-+ +-+-+-+ +-+-+ +-+-+-+ +-+-+-+ +-+-+
"+" "+" "+" "+" "+" "+"
+-+-+-+ +-+-+-+ +-+-+ +-+-+-+ +-+-+-+ +-+-+
|s|s|s| |s|s|s| |s|s| |s|s|s| |s|s|s| |s|s|
|F|E|D| |C|B|A| |9|8| |7|6|5| |4|3|2| |1|0|
+-+-+-+ +-+-+-+ +-+-+ +-+-+-+ +-+-+-+ +-+-+
v | v | v | v | v | v | +--+
| ^ | ^ | ^ | ^ | ^ | ^ v |
| +-+ +-+ +-+ +-+ +-+ +-+ |
| |c| |c| |c| |c| |c| |c| |
| |5| |4| |3| |2| |1| |0| |
| +-+ +-+ +-+ +-+ +-+ +-+ |
+----------------------------------------------------------+
Implementation of fast-adders in PLD hardware is currently limited to
3-bits, because an i-bit adder requires 4+2^i product terms, and
current PLDs support only 16 product terms. The resulting device
takes at most 5 "idle" clock periods for the carries to propagate
through the accumulation pipeline.
Touch & Parham Informational [Page 5]
RFC 1936 Implementing the Internet Checksum in Hardware April 1996
Interfaces
The above device has been installed in a VL-Bus PC host interface
card [8]. It has a hardware and software interface, defined as
follows.
Hardware Interface
The Internet Checksum hardware appears as a single-port 32-bit
register, with clock and control signals [8]:
+----------------------+
CLR--->| |
OE---->| 32-bit register as |
CLK--->| 2 adjacent 16-bit |<---/---> 32-bit data bus
ICLK-->| ones-complement sums |
ADD--->| |
+----------------------+
CLR = zero the register
OE = write the register onto the data bus
CLK = clock to cycle the pipeline operation
ICLK = input data latch clock
ADD = initiating an add of latched input data
CLR causes the contents of the checksum register and input latch to
be zeroed. There is no explicit load; a CLR followed by a write of
the load value to a dummy location is equivalent.
The OE causes the register to be written to the data bus, or tri-
stated.
The CLK causes the pipeline to operate. If no new input data is
latched to be added (via ICLK, ADD), a virtual "zero" is summed into
the register, to permit the pipeline to empty.
The ICLK (transparently) latches the value on the data bus to be
latched internally, to be summed into the accumulator on the next ADD
signal. The ADD signal causes the latched input data (ICLK) to be
accumulated into the checksum pipeline. ADD and ICLK are commonly
tied together. One 32-bit data value can be latched and accumulated
into the pipeline adder every 26-ns clock, assuming data is stable
when the ADD/ICLK signal occurs.
The internal 32-bit register is organized as two 16-bit ones-
complement sums, over the even and odd 16-bit words of the data
stream. To compute the Internet Checksum from this quantity, ones-
complement add the halves together, and invert the result.
Touch & Parham Informational [Page 6]
RFC 1936 Implementing the Internet Checksum in Hardware April 1996
Software Interface
The device is used as a memory-mapped register. The register is read
by performing a read on its equivalent memory location.
The device is controlled via an external memory-mapped register. Bits
in this control register clear the device (set/clear the CLR line),
and enable and disable the device (set/clear the ADD line). The CLR
line can alternatively be mapped to a memory write, e.g., such that
reading the location is a non-destructive read of the checksum
register, and a write of any value clears the checksum register. The
enable/disable control must be stored in an external register.
The device is designed to operate in background during memory
transfers (either DMA or programmed I/O). Once enabled, all transfers
across that bus are summed into the checksum register. The checksum
is available 5 clocks after the last enabled data accumulation. This
delay is often hidden by memory access mechanisms and bus
arbitration. If required, "stall" instructions can be executed for
the appropriate delay.
For the following example, we assume that the device is located at
CKSUMLOC. We assume that reading that location reads the checksum
register, and writing any value to that location clears the register.
The control register is located at CTLLOC, and the checksum
enable/disable bit is CKSUMBIT, where 1 is enabled, and 0 is
disabled. To perform a checksum, a programmer would clear the
register, (optionally initialize the checksum), initiate a series of
transfers, and use the result:
/******* initialization *******/
*(CTLLOC) &= ~((ctlsize)(CKSUMBIT)); /* disable sum */
(word32)(*(CKSUMLOC)) = 0; /* clear reg */
*(CTLLOC) |= CKSUMBIT; /* enable sum */
{ (optional) write initial value to a dummy location }
/***** perform a transfer *****/
{ do one or more DMA or PIO transfers - read or write }
/***** gather the results *****/
*(CTLLOC) &= ~((ctlsize)(CKSUMBIT)); /* disable sum */
sum = (word32)(*(CKSUMLOC)); /* read sum */
sum = (sum & 0xFFFF) + (sum >> 16); /* fold halves */
sum = (sum & 0xFFFF) + (sum >> 16); /* add in carry */
ipcksum = (halfword16)(~(sum & 0xFFFF)); /* 1's negative */
Touch & Parham Informational [Page 7]
RFC 1936 Implementing the Internet Checksum in Hardware April 1996
Summary
This note describes the design of a hardware Internet Checksum that
can be implemented in an inexpensive PLD, achieving 1.26 Gbps. This
design has been implemented in the PC-ATOMIC host interface hardware
[8]. We believe this design is of general use to the Internet
community.
Security Considerations
Security considerations are not addressed here. The Internet Checksum
is not intended as a security measure.
Acknowledgements
The authors would like to thank the members of the "High-Performance
Computing and Communications", notably Mike Carlton, and "Advanced
Systems" Divisions at ISI for their assistance in the development of
the hardware, and this memo.
References
[1] Braden, R., Borman, D., and Partridge, C., "Computing the
Internet Checksum," Network Working Group RFC-1071, ISI, Cray
Research, and BBN Labs, Sept. 1988.
[2] Mallory, T., and Kullberg, A., "Incremental Updating of the
Internet Checksum," Network Working Group RFC-1141, BBN Comm.,
Jan. 1990.
[3] Plummer, W., "TCP Checksum Function Design," IEN-45, BBN, 1978,
included as an appendix in RFC-1071.
[4] Postel, Jon, "Internet Protocol," Network Working Group RFC-
791/STD-5, ISI, Sept. 1981.
[5] Postel, Jon, "User Datagram Protocol," Network Working Group
RFC-768/STD-6, ISI, Aug. 1980.
[6] Postel, Jon, "Transmission Control Protocol," Network Working
Group RFC-793/STD-7, ISI, Sept. 1981.
[7] Rijsinghani, A., "Computation of the Internet Checksum via
Incremental Update," Network Working Group RFC-1624, Digital
Equipment Corp., May 1994.
[8] Touch, J., "PC-ATOMIC", ISI Tech. Report. SR-95-407, June 1995.
Touch & Parham Informational [Page 8]
RFC 1936 Implementing the Internet Checksum in Hardware April 1996
Authors' Addresses
Joe Touch
University of Southern California/Information Sciences Institute
4676 Admiralty Way
Marina del Rey, CA 90292-6695
USA
Phone: +1 310-822-1511 x151
Fax: +1 310-823-6714
URL: http://www.isi.edu/~touch
EMail: touch@isi.edu
Bruce Parham
University of Southern California/Information Sciences Institute
4676 Admiralty Way
Marina del Rey, CA 90292-6695
USA
Phone: +1 310-822-1511 x101
Fax: +1 310-823-6714
EMail: bparham@isi.edu
Touch & Parham Informational [Page 9]
RFC 1936 Implementing the Internet Checksum in Hardware April 1996
Appendix A: PLD source code
The following is the PLD source code for an AMD MACH-435 PLD. The
MACH-435 is composed of 8 22V10-equivalent PLD blocks, connected by a
configurable internal matrix.
---- (PLD source code follows) ----
TITLE PC-ATOMIC IP Sum Accelerator - 1-clock 2- and 3-bit 26 ns version
PATTERN ip_sum
REVISION 1.01
AUTHOR J. Touch & B. Parham
COMPANY USC/ISI
DATE 06/21/94
CHIP ip_sum MACH435
; accumulates in 1 clock (1 level of logic)
;
; resources allocated to reduce fitting time
;
; uses an input register "dl" to latch the data bus values on rising edge
; accumulates a hi/lo ones-complement sum in register "q"
; the input and output are accessed via bidirectional pins "dq"
;
; uses 2 groups of 6 carry bit registers "cy"
;
; use 3-bit full-adders with carry lookahead (settles in 6 clocks)
; group 16 bits as [000102 030405 0607 080910 111213 1415]
; [161718 192021 2223 242526 272829 3031]
;
; locking the pins down speeds up fitting and is designed to force
; 4-bit components into single "segments" of the PLD.
; we could have indicated the same thing via:
; GROUP MACH_SEG_A dq[6..0]
; GROUP MACH_SEG_B dq[14..8]
; GROUP MACH_SEG_C dq[22..16]
; GROUP MACH_SEG_D dq[30..24]
;
; control pins:
;
PIN 20 clk ; adder clock
PIN 62 ip_add ; add current data to sum
PIN 83 ip_sum_ena ; output current sum
PIN 41 ip_clr ; clear current sum
PIN 23 ip_dclk ; input data latch (tied to clk, or not)
Touch & Parham Informational [Page 10]
RFC 1936 Implementing the Internet Checksum in Hardware April 1996
;
; dq are data bus pins
; dl is the input register
;
PIN [9..3] dq[6..0] IPAIR dl[6..0] ; IO port
PIN [18..12] dq[14..8] IPAIR dl[14..8] ; IO port
PIN [30..24] dq[22..16] IPAIR dl[22..16] ; IO port
PIN [39..33] dq[30..24] IPAIR dl[30..24] ; IO port
PIN ? dq[31,23,15,7] IPAIR dl[31,23,15,7] ; IO port
;
; q is the partial checksum register
; dl is the input register
; dq are the data bus pins
;
NODE ? q[31..0] OPAIR dq[31..0] ; internal data in reg
NODE ? dl[31..0] REG ; input reg
;
; cy are the carry register bits
;
NODE ? cy[31,29,26,23,21,18,15,13,10,7,5,2] REG
;1-bit internal carry bits
EQUATIONS
;
; .trst is the tri-state control, 0 means these are always inputs
;
ip_add.trst = 0
ip_clr.trst = 0
ip_sum_ena.trst = 0
;
; grab data to the input register on every clock (irrelevant if invalid)
;
dl[31..0].clkf = ip_dclk ; grab data all the time
; don't use setf, rstf, or trst for dl
; we want dl to map to input registers, not internal cells
; besides, input registers don't need setf, rstf, or trst
;
; control of the checksum register
;
dq[31..0].clkf = clk ; clk clocks everything
dq[31..0].setf = gnd ; never preset registers
dq[31..0].rstf = ip_clr ; clear on reset
dq[31..0].trst = ip_sum_ena ; ena outputs sum - read
Touch & Parham Informational [Page 11]
RFC 1936 Implementing the Internet Checksum in Hardware April 1996
;
; control for the carry register
;
cy[31,29,26,23,21,18,15,13,10,7,5,2].clkf = clk
cy[31,29,26,23,21,18,15,13,10,7,5,2].setf = gnd ; never preset
cy[31,29,26,23,21,18,15,13,10,7,5,2].rstf = ip_clr ; clear on reset
;
; INPUT DATA LATCH
; nothing fancy here - grab all inputs when ip_add signal is high
; i.e., grab data in input register
;
dl[31..0] := dq[31..0]
;
; COMBINATORIAL ADDER
;
; built as a series of 2-bit and 3-bit (carry-lookahead) full-adders
; with carries sent to the carry register "pipeline"
;
; sum[n] are sum bits
; cy[m] are carry bits
; ":+:" is XOR
;
; SUM[0] = (A0 :+: B0 :+: CARRY_IN)
;
; CY[0] = ((A0 * B0) + ((A0 :+: B0) * CARRY_IN))
;
; actually, the latter can be rewritten as
;
; CY[0] = ((A0 * B0) + ((A0 + B0) * CARRY_IN))
;
; because the XOR won't be invalidated by the AND case, since the
; result is always 1 from the first term then anyway
; this helps reduce the number of XOR terms required, which are
; a limited resource in PLDs
;
Touch & Parham Informational [Page 12]
RFC 1936 Implementing the Internet Checksum in Hardware April 1996
; SUM THE LOW-ORDER WORD
;
;
; the first 5 bits [0..4] of the low-order word
;
q[0] := (q[0] :+: (ip_add * dl[0]) :+: cy[15])
q[1] := (q[1] :+: (ip_add * dl[1]) :+:
((ip_add *
(q[0] * dl[0] +
dl[0] * cy[15])) +
(q[0] * cy[15])))
q[2] := (q[2] :+: (ip_add * dl[2]) :+:
((ip_add *
(q[1] * dl[1] +
q[1] * q[0] * dl[0] +
dl[1] * q[0] * dl[0] +
q[1] * dl[0] * cy[15] +
dl[1] * dl[0] * cy[15] +
dl[1] * q[0] * cy[15])) +
(q[1] * q[0] * cy[15])))
cy[2] := ((ip_add *
(q[2] * dl[2] +
q[2] * q[1] * dl[1] +
dl[2] * q[1] * dl[1] +
q[2] * q[1] * q[0] * dl[0] +
q[2] * dl[1] * q[0] * dl[0] +
dl[2] * q[1] * q[0] * dl[0] +
dl[2] * dl[1] * q[0] * dl[0] +
q[2] * q[1] * dl[0] * cy[15] +
q[2] * dl[1] * q[0] * cy[15] +
q[2] * dl[1] * dl[0] * cy[15] +
dl[2] * q[1] * q[0] * cy[15] +
dl[2] * q[1] * dl[0] * cy[15] +
dl[2] * dl[1] * q[0] * cy[15] +
dl[2] * dl[1] * dl[0] * cy[15])) +
(q[2] * q[1] * q[0] * cy[15]))
q[3] := (q[3] :+: (ip_add * dl[3]) :+: cy[2])
q[4] := (q[4] :+: (ip_add * dl[4]) :+:
((ip_add *
(q[3] * dl[3] +
dl[3] * cy[2])) +
(q[3] * cy[2])))
Touch & Parham Informational [Page 13]
RFC 1936 Implementing the Internet Checksum in Hardware April 1996
;
; the next 3 bits [5..7] of the low-order word
;
q[5] := (q[5] :+: (ip_add * dl[5]) :+:
((ip_add *
(q[4] * dl[4] +
q[4] * q[3] * dl[3] +
dl[4] * q[3] * dl[3] +
q[4] * dl[3] * cy[2] +
dl[4] * dl[3] * cy[2] +
dl[4] * q[3] * cy[2])) +
(q[4] * q[3] * cy[2])))
cy[5] := ((ip_add * (
q[5] * dl[5] +
q[5] * q[4] * dl[4] +
dl[5] * q[4] * dl[4] +
q[5] * q[4] * q[3] * dl[3] +
q[5] * dl[4] * q[3] * dl[3] +
dl[5] * q[4] * q[3] * dl[3] +
dl[5] * dl[4] * q[3] * dl[3] +
q[5] * q[4] * dl[3] * cy[2] +
q[5] * dl[4] * q[3] * cy[2] +
q[5] * dl[4] * dl[3] * cy[2] +
dl[5] * q[4] * q[3] * cy[2] +
dl[5] * q[4] * dl[3] * cy[2] +
dl[5] * dl[4] * q[3] * cy[2] +
dl[5] * dl[4] * dl[3] * cy[2])) +
(q[5] * q[4] * q[3] * cy[2]))
q[6] := (q[6] :+: (ip_add * dl[6]) :+: cy[5])
q[7] := (q[7] :+: (ip_add * dl[7]) :+:
((ip_add *
(q[6] * dl[6] +
dl[6] * cy[5])) +
(q[6] * cy[5])))
cy[7] := ((ip_add *
(q[7] * dl[7] +
q[7] * q[6] * dl[6] +
dl[7] * q[6] * dl[6] +
q[7] * dl[6] * cy[5] +
dl[7] * dl[6] * cy[5] +
dl[7] * q[6] * cy[5])) +
(q[7] * q[6] * cy[5]))
Touch & Parham Informational [Page 14]
RFC 1936 Implementing the Internet Checksum in Hardware April 1996
;
; the next 5 bits [8..12] of the low-order word
;
q[8] := (q[8] :+: (ip_add * dl[8]) :+: cy[7])
q[9] := (q[9] :+: (ip_add * dl[9]) :+:
((ip_add *
(q[8] * dl[8] +
dl[8] * cy[7])) +
(q[8] * cy[7])))
q[10] := (q[10] :+: (ip_add * dl[10]) :+:
((ip_add *
(q[9] * dl[9] +
q[9] * q[8] * dl[8] +
dl[9] * q[8] * dl[8] +
q[9] * dl[8] * cy[7] +
dl[9] * dl[8] * cy[7] +
dl[9] * q[8] * cy[7])) +
(q[9] * q[8] * cy[7])))
cy[10] := ((ip_add *
(q[10] * dl[10] +
q[10] * q[9] * dl[9] +
dl[10] * q[9] * dl[9] +
q[10] * q[9] * q[8] * dl[8] +
q[10] * dl[9] * q[8] * dl[8] +
dl[10] * q[9] * q[8] * dl[8] +
dl[10] * dl[9] * q[8] * dl[8] +
q[10] * q[9] * dl[8] * cy[7] +
q[10] * dl[9] * q[8] * cy[7] +
q[10] * dl[9] * dl[8] * cy[7] +
dl[10] * q[9] * q[8] * cy[7] +
dl[10] * q[9] * dl[8] * cy[7] +
dl[10] * dl[9] * q[8] * cy[7] +
dl[10] * dl[9] * dl[8] * cy[7])) +
(q[10] * q[9] * q[8] * cy[7]))
q[11] := (q[11] :+: (ip_add * dl[11]) :+: cy[10])
q[12] := (q[12] :+: (ip_add * dl[12]) :+:
((ip_add *
(q[11] * dl[11] +
dl[11] * cy[10])) +
(q[11] * cy[10])))
Touch & Parham Informational [Page 15]
RFC 1936 Implementing the Internet Checksum in Hardware April 1996
;
; the final 3 bits [13..15] of the low-order word
;
q[13] := (q[13] :+: (ip_add * dl[13]) :+:
((ip_add *
(q[12] * dl[12] +
q[12] * q[11] * dl[11] +
dl[12] * q[11] * dl[11] +
q[12] * dl[11] * cy[10] +
dl[12] * dl[11] * cy[10] +
dl[12] * q[11] * cy[10])) +
(q[12] * q[11] * cy[10])))
cy[13] := ((ip_add * (
q[13] * dl[13] +
q[13] * q[12] * dl[12] +
dl[13] * q[12] * dl[12] +
q[13] * q[12] * q[11] * dl[11] +
q[13] * dl[12] * q[11] * dl[11] +
dl[13] * q[12] * q[11] * dl[11] +
dl[13] * dl[12] * q[11] * dl[11] +
q[13] * q[12] * dl[11] * cy[10] +
q[13] * dl[12] * q[11] * cy[10] +
q[13] * dl[12] * dl[11] * cy[10] +
dl[13] * q[12] * q[11] * cy[10] +
dl[13] * q[12] * dl[11] * cy[10] +
dl[13] * dl[12] * q[11] * cy[10] +
dl[13] * dl[12] * dl[11] * cy[10])) +
(q[13] * q[12] * q[11] * cy[10]))
q[14] := (q[14] :+: (ip_add * dl[14]) :+: cy[13])
q[15] := (q[15] :+: (ip_add * dl[15]) :+:
((ip_add *
(q[14] * dl[14] +
dl[14] * cy[13])) +
(q[14] * cy[13])))
cy[15] := ((ip_add *
(q[15] * dl[15] +
q[15] * q[14] * dl[14] +
dl[15] * q[14] * dl[14] +
q[15] * dl[14] * cy[13] +
dl[15] * dl[14] * cy[13] +
dl[15] * q[14] * cy[13])) +
(q[15] * q[14] * cy[13]))
Touch & Parham Informational [Page 16]
RFC 1936 Implementing the Internet Checksum in Hardware April 1996
; SUM THE HIGH-ORDER WORD
;
;
; the first 5 bits [16..20] of the high-order word
;
q[16] := (q[16] :+: (ip_add * dl[16]) :+: cy[31])
q[17] := (q[17] :+: (ip_add * dl[17]) :+:
((ip_add *
(q[16] * dl[16] +
dl[16] * cy[31])) +
(q[16] * cy[31])))
q[18] := (q[18] :+: (ip_add * dl[18]) :+:
((ip_add *
(q[17] * dl[17] +
q[17] * q[16] * dl[16] +
dl[17] * q[16] * dl[16] +
q[17] * dl[16] * cy[31] +
dl[17] * dl[16] * cy[31] +
dl[17] * q[16] * cy[31])) +
(q[17] * q[16] * cy[31])))
cy[18] := ((ip_add *
(q[18] * dl[18] +
q[18] * q[17] * dl[17] +
dl[18] * q[17] * dl[17] +
q[18] * q[17] * q[16] * dl[16] +
q[18] * dl[17] * q[16] * dl[16] +
dl[18] * q[17] * q[16] * dl[16] +
dl[18] * dl[17] * q[16] * dl[16] +
q[18] * q[17] * dl[16] * cy[31] +
q[18] * dl[17] * q[16] * cy[31] +
q[18] * dl[17] * dl[16] * cy[31] +
dl[18] * q[17] * q[16] * cy[31] +
dl[18] * q[17] * dl[16] * cy[31] +
dl[18] * dl[17] * q[16] * cy[31] +
dl[18] * dl[17] * dl[16] * cy[31])) +
(q[18] * q[17] * q[16] * cy[31]))
q[19] := (q[19] :+: (ip_add * dl[19]) :+: cy[18])
q[20] := (q[20] :+: (ip_add * dl[20]) :+:
((ip_add *
(q[19] * dl[19] +
dl[19] * cy[18])) +
(q[19] * cy[18])))
Touch & Parham Informational [Page 17]
RFC 1936 Implementing the Internet Checksum in Hardware April 1996
;
; the next 3 bits [21..23] of the high-order word
;
q[21] := (q[21] :+: (ip_add * dl[21]) :+:
((ip_add *
(q[20] * dl[20] +
q[20] * q[19] * dl[19] +
dl[20] * q[19] * dl[19] +
q[20] * dl[19] * cy[18] +
dl[20] * dl[19] * cy[18] +
dl[20] * q[19] * cy[18])) +
(q[20] * q[19] * cy[18])))
cy[21] := ((ip_add * (
q[21] * dl[21] +
q[21] * q[20] * dl[20] +
dl[21] * q[20] * dl[20] +
q[21] * q[20] * q[19] * dl[19] +
q[21] * dl[20] * q[19] * dl[19] +
dl[21] * q[20] * q[19] * dl[19] +
dl[21] * dl[20] * q[19] * dl[19] +
q[21] * q[20] * dl[19] * cy[18] +
q[21] * dl[20] * q[19] * cy[18] +
q[21] * dl[20] * dl[19] * cy[18] +
dl[21] * q[20] * q[19] * cy[18] +
dl[21] * q[20] * dl[19] * cy[18] +
dl[21] * dl[20] * q[19] * cy[18] +
dl[21] * dl[20] * dl[19] * cy[18])) +
(q[21] * q[20] * q[19] * cy[18]))
q[22] := (q[22] :+: (ip_add * dl[22]) :+: cy[21])
q[23] := (q[23] :+: (ip_add * dl[23]) :+:
((ip_add *
(q[22] * dl[22] +
dl[22] * cy[21])) +
(q[22] * cy[21])))
cy[23] := ((ip_add *
(q[23] * dl[23] +
q[23] * q[22] * dl[22] +
dl[23] * q[22] * dl[22] +
q[23] * dl[22] * cy[21] +
dl[23] * dl[22] * cy[21] +
dl[23] * q[22] * cy[21])) +
(q[23] * q[22] * cy[21]))
Touch & Parham Informational [Page 18]
RFC 1936 Implementing the Internet Checksum in Hardware April 1996
;
; the next 5 bits [24..28] of the high-order word
;
q[24] := (q[24] :+: (ip_add * dl[24]) :+: cy[23])
q[25] := (q[25] :+: (ip_add * dl[25]) :+:
((ip_add *
(q[24] * dl[24] +
dl[24] * cy[23])) +
(q[24] * cy[23])))
q[26] := (q[26] :+: (ip_add * dl[26]) :+:
((ip_add *
(q[25] * dl[25] +
q[25] * q[24] * dl[24] +
dl[25] * q[24] * dl[24] +
q[25] * dl[24] * cy[23] +
dl[25] * dl[24] * cy[23] +
dl[25] * q[24] * cy[23])) +
(q[25] * q[24] * cy[23])))
cy[26] := ((ip_add *
(q[26] * dl[26] +
q[26] * q[25] * dl[25] +
dl[26] * q[25] * dl[25] +
q[26] * q[25] * q[24] * dl[24] +
q[26] * dl[25] * q[24] * dl[24] +
dl[26] * q[25] * q[24] * dl[24] +
dl[26] * dl[25] * q[24] * dl[24] +
q[26] * q[25] * dl[24] * cy[23] +
q[26] * dl[25] * q[24] * cy[23] +
q[26] * dl[25] * dl[24] * cy[23] +
dl[26] * q[25] * q[24] * cy[23] +
dl[26] * q[25] * dl[24] * cy[23] +
dl[26] * dl[25] * q[24] * cy[23] +
dl[26] * dl[25] * dl[24] * cy[23])) +
(q[26] * q[25] * q[24] * cy[23]))
q[27] := (q[27] :+: (ip_add * dl[27]) :+: cy[26])
q[28] := (q[28] :+: (ip_add * dl[28]) :+:
((ip_add *
(q[27] * dl[27] +
dl[27] * cy[26])) +
(q[27] * cy[26])))
Touch & Parham Informational [Page 19]
RFC 1936 Implementing the Internet Checksum in Hardware April 1996
;
; the final 3 bits [29..31] of the high-order word
;
q[29] := (q[29] :+: (ip_add * dl[29]) :+:
((ip_add *
(q[28] * dl[28] +
q[28] * q[27] * dl[27] +
dl[28] * q[27] * dl[27] +
q[28] * dl[27] * cy[26] +
dl[28] * dl[27] * cy[26] +
dl[28] * q[27] * cy[26])) +
(q[28] * q[27] * cy[26])))
cy[29] := ((ip_add * (
q[29] * dl[29] +
q[29] * q[28] * dl[28] +
dl[29] * q[28] * dl[28] +
q[29] * q[28] * q[27] * dl[27] +
q[29] * dl[28] * q[27] * dl[27] +
dl[29] * q[28] * q[27] * dl[27] +
dl[29] * dl[28] * q[27] * dl[27] +
q[29] * q[28] * dl[27] * cy[26] +
q[29] * dl[28] * q[27] * cy[26] +
q[29] * dl[28] * dl[27] * cy[26] +
dl[29] * q[28] * q[27] * cy[26] +
dl[29] * q[28] * dl[27] * cy[26] +
dl[29] * dl[28] * q[27] * cy[26] +
dl[29] * dl[28] * dl[27] * cy[26])) +
(q[29] * q[28] * q[27] * cy[26]))
q[30] := (q[30] :+: (ip_add * dl[30]) :+: cy[29])
q[31] := (q[31] :+: (ip_add * dl[31]) :+:
((ip_add *
(q[30] * dl[30] +
dl[30] * cy[29])) +
(q[30] * cy[29])))
cy[31] := ((ip_add *
(q[31] * dl[31] +
q[31] * q[30] * dl[30] +
dl[31] * q[30] * dl[30] +
q[31] * dl[30] * cy[29] +
dl[31] * dl[30] * cy[29] +
dl[31] * q[30] * cy[29])) +
(q[31] * q[30] * cy[29]))
Touch & Parham Informational [Page 20]
RFC 1936 Implementing the Internet Checksum in Hardware April 1996
;
; output nodes onto output pins (pending enable..)
;
dq[0] := {q[0]}
dq[1] := {q[1]}
dq[2] := {q[2]}
dq[3] := {q[3]}
dq[4] := {q[4]}
dq[5] := {q[5]}
dq[6] := {q[6]}
dq[7] := {q[7]}
dq[8] := {q[8]}
dq[9] := {q[9]}
dq[10] := {q[10]}
dq[11] := {q[11]}
dq[12] := {q[12]}
dq[13] := {q[13]}
dq[14] := {q[14]}
dq[15] := {q[15]}
dq[16] := {q[16]}
dq[17] := {q[17]}
dq[18] := {q[18]}
dq[19] := {q[19]}
dq[20] := {q[20]}
dq[21] := {q[21]}
dq[22] := {q[22]}
dq[23] := {q[23]}
dq[24] := {q[24]}
dq[25] := {q[25]}
dq[26] := {q[26]}
dq[27] := {q[27]}
dq[28] := {q[28]}
dq[29] := {q[29]}
dq[30] := {q[30]}
dq[31] := {q[31]}
;
; end.
;
Touch & Parham Informational [Page 21]