am79c850 Advanced Micro Devices, am79c850 Datasheet - Page 64
![no-image](/images/manufacturer_photos/0/0/14/advanced_micro_devices_sml.jpg)
am79c850
Manufacturer Part Number
am79c850
Description
Supernet-r 3
Manufacturer
Advanced Micro Devices
Datasheet
1.AM79C850.pdf
(97 pages)
1. The stuck-at 1 (s@1) condition in the mask bits for a
2. The stuck-at 0 (s@0) condition in the mask bits for a
Part #1
a) Write each data entry with all 0’s.
b) Write the comparand register with all 0’s.
c) Perform a match operation—should get an all match
The individual match lines for each entry must be OR’d
together to determine if a match has occurred. The indi-
vidual exact match lines must also be OR’d together to
determine if an exact match has occurred. The pattern
shown in the first line of the table will detect all s@0
faults on the logic that generates each match and exact
line for an entry. It will also check for a s@0 fault on the
final match and exact outputs. The pattern shown in the
second line of the table will check for all s@1 faults aris-
ing from the SAX personality bits and will also detect a
s@1 fault on the final exact output. The pattern shown in
the third line of the table will check for s@1 faults on the
VALID personality bits as well as on the final match and
exact outputs. The final pattern shown in the fourth line
64
comparator can be checked by first writing all mask
bits to 0’s, all data bits to 0’s and the comparand
register to all 1’s. If any single mask bit is s@1, no
match will occur otherwise a match will result as can
be seen from the first line of table 1. To speed the
testing for this condition, all entries can be tested in
parallel if an all match indication is provided. All mask
and data bits are written with 0’s and compared
simultaneously with the comparand. Each entry
should match resulting in the all match condition. If
any mask bit is s@1, the all match condition will not
occur and the fault can be detected.
comparator are tested by the data patterns shown in
the third and fourth lines of table 1. These lines are a
subset of the patterns needed to test the XNOR
faults that are represented by lines 2 through 5 of
table 1. The outputs of the XNOR gates for each bit
are all AND’ed together to form the match output of
the comparator. The following test description will
cover all stuck-at faults for the remainder of the
comparator logic. These tests assume that all mask
bits are written with 1’s and assume the existence of
an all match indication. The test is described in
two parts:
indication.
AMD
SAX
1
0
1
1
VALID
Table 2. Patterns necessary for match and exact indicators
1
1
0
1
SA
1
1
1
0
P R E L I M I N A R Y
SUPERNET 3
MATCH[i]
d) Repeat steps a through c but this time use all 1’s. At
Part #2
a) Write each data entry with all 0’s.
b) Do 48 matches while walking a 1 through a field of 0’s
c) Repeat steps a and b with each data entry being all
Source address exact/inexact logic:
The SRAM portion of this logic, namely the storage for
the SA and SAX personality bits will be tested by the
SRAM tests presented earlier. What remains to be
tested is the logic that generates the source address
match and the source address match exact logic.
The test pattern description that follows does not yet
take into account the impact of the SKIP bit since its use
hasn’t been completely determined as yet. The patterns
also assume the existence of an all match indication as
described earlier. The following table summarizes the
patterns that need to be applied to check each entry
match and exact indicators:
of the table will check for s@1 faults on the SA personal-
ity bits as well as on the final match and exact outputs.
One could argue that the first pattern is not sufficient to
cover all possible s@0 faults in the final OR’ing of the
individual match and exact lines since these faults are
detected on the assumption of the existence of an all
match indication (an all exact indication is also neces-
sary). The all match indicator is used to speed the
testing of the AF since all entries can be written with the
same values. If someone is uncomfortable with this
potential loss in coverage, then an additional 32
patterns are needed that only apply the first pattern of
the table to each entry by itself so that only one entry
1
1
0
0
this point half of the XNOR tests (lines 2 and 5 in
table 1) have been completed and the AND circuitry
has been tested for s@0 faults.
in the comparand register. Since a single bit is in
error for each match, a match condition should never
occur.
1’s and walking a 0 through a field of 1’s in the
comparand. At this point, all the XNOR tests are
completed. This also checks for s@1 faults in the
AND circuitry since all but a single bit match. This
test also tests for the mask bit s@0.
EXACT[i]
1
0
0
0
ALL MATCH?
YES
YES
NO
NO