Ex Parte Biles

Board of Patent Appeals and Interferences

Jan 13, 2010

10413579 (B.P.A.I. Jan. 13, 2010)

UNITED STATES PATENT AND TRADEMARK OFFICE
________________

BEFORE THE BOARD OF PATENT APPEALS
AND INTERFERENCES
________________

Ex parte STUART DAVID BILES
________________

Appeal 2009-007907
Application 10/413,579
Technology Center 2100
________________

Decided: January 13, 2010
________________

Before JAMES D. THOMAS, JOSEPH L. DIXON, and
CAROLYN D. THOMAS, Administrative Patent Judges.

J. THOMAS, Administrative Patent Judge.

DECISION ON APPEAL

Appeal 2009-007907
Application 10/413,579

2
STATEMENT OF THE CASE
This is an appeal under 35 U.S.C. § 134(a) from the Examiner’s final
rejection of claims 1-19. The administrative record of this application
reveals an Administrative Remand from the Board to the Examiner in a
communication dated November 3, 2008 ordering the Examiner to cancel
claims 3, 4, 10, 13, and 14. These claims were cancelled by the Examiner in
a paper dated December 15, 2008. Therefore, only claims 1, 2, 5-9, 11, 12,
and 15-19 remain on appeal. We have jurisdiction under 35 U.S.C. § 6(b).
We reverse.1

Invention
The present invention relates to a branch prediction logic and method
for generating a branch bias providing a prediction as to whether execution
of a branch instruction will result in a branch being taken or not taken.
(Spec. 36, ll. 3-5, Figs. 1, 2A, 2B, and 6.)

Representative Claim
1. Branch prediction apparatus for use in an instruction
prefetch unit in a data processor for generating a branch bias
providing a prediction as to whether execution of a branch
instruction will result in a branch being taken or not taken, the
branch prediction apparatus comprising:

a first branch bias storage operable to store for each of a
plurality of first address portions a corresponding first branch

1 An oral hearing of this appeal was conducted on January 7, 2010.
Appeal 2009-007907
Application 10/413,579

3
bias value, the first branch bias storage being operable upon
receipt of a first address portion of the branch instruction’s
address to output the corresponding first branch bias value from
the first branch bias storage;

a history storage operable to store history data identifying
an outcome for a number of preceding branch instructions;

a single second branch bias storage having a number of
entries, each entry being operable to store a replacement branch
bias value and a TAG value indicative of an instance in which
the replacement branch bias value should be used as the
generated branch bias in preference to the first branch bias
value, all replacement branch bias values stored in said branch
prediction apparatus being stored in said single second branch
bias storage;

index generation logic operable to generate an index
identifying a corresponding entry in the single second branch
bias storage, the index being derived from the history data;

the single second branch bias storage being operable
upon receipt of the index to cause the content of the
corresponding entry to be output from the single second branch
bias storage; and

selection logic operable to select as the branch bias
generated by the branch prediction logic the first branch bias
value, unless the TAG value output from the single second
branch bias storage corresponds to a comparison TAG value
derived from a second address portion of the branch
instruction’s address, in which event, the selection logic is
operable, regardless of the first branch bias value, to select as
the branch bias generated by the branch prediction logic the
Appeal 2009-007907
Application 10/413,579

4
replacement branch bias value output from the single second
branch bias storage.

Prior Art and Examiner’s Rejection
The Examiner relies on the following references as evidence of
unpatentability:
Kacevas US 7,058,795 B2 June 6, 2006
(filed June 25, 2002)

Applicant’s Admitted Prior Art Figure 1 and associated text.

All claims remaining on appeal, claims 1, 2, 5-9, 11, 12, and 15-19
stand rejected under 35 U.S.C. §103. As evidence of obviousness, the
Examiner relies upon Appellant’s admitted prior art Figure 1 in view of
Kacevas. 2

CLAIM GROUPING
As reflected at page 5 of the principal Brief on appeal, Appellant
considers independent claim 1 as representative of the subject matter of
independent claims 1 and 11. We do so likewise. No other remaining claim
on appeal is argued.

2 Because of the cancellation of dependent claims 3, 4, 13, and 14 and of
independent claim 10, the separately restated rejections of these claims in
the Final Rejection and Answer have been rendered moot.
Appeal 2009-007907
Application 10/413,579

5
FINDINGS OF FACT
1. Appellant argues in the principal Brief at pages 11 and 12:
But Kacevas also fails to disclose the claim feature of
“the selection logic is operable, regardless of the first branch
bias value, to select as the branch bias generated by the branch
prediction logic the replacement branch bias value output from
the single second branch bias storage.” Instead, Kacevas
discloses a second branch bias storage (global predictor 360 in
Figure 2), where each entry includes two replacement bias
values (counters 362 and 363 in Figure 2). Col. 3, lines 21- 40.
Kacevas is similar to the YAGS scheme in Figure 1
except that the two separate predictors are shown in one global
predictor box 360 and the TAG/VALID bits are shared. When
the TAG comparison matches (see 380 in Figure 2 of Kacevas),
one of the alternative predictions is selected by multiplexer 377
to override the first branch prediction generated by the local
predictor 350. The value of the first branch prediction from the
local predictor 350 controls the selection made by the
multiplexer 375 in Figure 2 (see the multiplexer control line
355) and determines which alternative prediction is used. Thus,
like the YAGS scheme in Figure 1 described above, Kacevas
does not select a replacement branch bias output value
“regardless of the first branch bias value.”
Accordingly, the combination of the YAGS scheme in
Figure 1 and Kacevas teaches selecting between alternative
branch bias output values based on (having regard for) the first
branch bias value. But in the independent claims, the
replacement branch bias output value is selected “regardless of
the first branch bias value.” Thus, even if the proposed
combination could be made, it fails to teach all the features
recited in the independent claims.

(App. Br. 11-12.)

Appeal 2009-007907
Application 10/413,579

6
2. A substantive responsive argument by the Examiner at page 15 of
the Answer is reproduced below:
[A]pplicant’s admission in view of Kacevas clearly disclosed
the claimed limitation. Applicant admits that Kacevas
disclosed a single replacement branch bias storage shown in
element 360. Applicant further admits that when element 380
generates a hit on the single replacement bias, this overrides the
first branch prediction from element 350. The examiner agrees
and this is the reason that Kacevas disclosed the claimed
limitation. The fact that element 355 selects a certain
replacement bias doesn’t mean that Kacevas doesn’t teach the
claimed limitation. Kacevas still in fact does use the
replacement bias value from the global predictor when a hit
occurs instead of the local predictor.

(Ans. 15.)

3. The Reply Brief’s pages 2 and 3 directly address these Examiner
views in this manner:
Each cache entry in Kacevas has two possible counter values
C0 (362) and C1 (363), which are selected between by
multiplexer 375, depending on (rather than regardless of) the
first prediction value supplied over path 355. (Kacevas’s
multiplexer 375 corresponds to the multiplexer 80 in Figure 1
of the present application.) So combining the T and NT caches
does not eliminate the need for this multiplexing functionality
that depends on the first branch bias value.
Appellant also disagrees with the Examiner’s analysis at
paragraph 10-B(i) in the Answer. Here, the Examiner equates
“the single second branch bias storage” with two different
components in the two different examples given, which is
inconsistent. Even accepting this inconsistent interpretation,
claim 1 requires that “the replacement branch bias value output
Appeal 2009-007907
Application 10/413,579

7
from the single second branch bias storage” be taken as the
generated branch bias. This is a unique value, identified by the
definite article “the”; yet, the Examiner equates it with two
different values (“the T cache counter” and “the NT cache
counter”) in each of the examples given.
The Examiner’s analysis at paragraph 10-B(iii) is
similarly flawed. In discussing how “Kacevas [uses] the
replacement bias value from the global predictor,” the
Examiner overlooks the earlier definition of the single second
branch bias storage in claim 1, “each entry being operable to
store a replacement branch bias value and a TAG value
indicative of an instance in which the replacement branch bias
value should be used as the generated branch bias in preference
to the first branch bias value.” Even adopting the Examiner’s
interpretation, Kacevas’ TAG value is insufficient to indicate
when “the replacement branch bias value should be used as the
generated branch bias” because the multiplexer 375 selection
between the two counter values C0 and C1 must be done. That
selection depends on the first branch bias value ─ something
that is excluded by the last three lines of claim 1.

(Reply Br. 2-3.)

ANALYSIS
The end of representative independent claim 1 recites a clause relating
to selection logic, the significant part of which indicates that “the selection
logic is operable, regardless of the first branch bias value, to select as the
branch bias generated by the branch prediction logic the replacement branch
bias value output from the single second branch bias storage.” This
language is the focus of Appellant’s continued arguments, best stated in the
Appeal 2009-007907
Application 10/413,579

8
Reply Brief, that the applied prior art rejection does not teach all the features
of representative independent claim 1 on appeal. We agree with this view.
The reproduced portions in Findings of Fact 1 through 3 from the
principal Brief, the Answer, and the Reply Brief respectfully focus upon the
pertinent features of representative independent claim 1 on appeal as well as
the alleged corresponding structure in Figure 2 of Kacevas. For the purpose
of our analysis, we consider the Appellant’s admitted prior art Figure 1 to be
properly combinable with Kacevas within 35 U.S.C. § 103. Therefore, no
governing case law is cited in this opinion to that affect.
As generally explained by Appellant in Findings of Fact 1 and 3, the
output of the local predictor 350 in Kacevas does feed, by means of line 355,
not only the multiplexer 375 as a switching input, but also provides in
parallel a corresponding input to the subsequent multiplexer 377. Therefore,
the artisan would understand that the selection logic does not select, as the
branch bias value generated by the branch prediction logic, based upon the
replacement branch bias value output from this global predictor 360 of
Kacevas “regardless of the first branch bias value” as recited at the end of
representative independent claim 1 on appeal. As essentially argued by
Appellant, the output of the branch prediction circuits as a whole is
dependent upon or otherwise based upon the first branch bias value that may
be provided by the local predictor 350.

Appeal 2009-007907
Application 10/413,579

9
OTHER MATTERS
At the end of the oral hearing in this appeal, Appellant’s
representative argued that the Examiner’s cancellation of certain claims
violated Appellant’s due process rights. Cancelled claims are not before us
for decision.
We do not agree with this view. The administrative remand, noted
earlier in this opinion, indicated that Appellant had chosen not to appeal all
finally rejected claims. This is indicated at the top of page 2 of the principal
Brief. The Examiner’s communication stating that the claims had been
cancelled also noted this. This situation is considered an authorization from
Appellant to cancel the rejected claims that are not on appeal. The basis of
this action was Appellant’s own actions/inactions, Ex parte Ghuman, and
MPEP §1215.03.
Following the noted Examiner’s communication to Appellant, the
record in this application and appeal does not show any petition to the
Examiner, and no Supplemental Brief was filed addressing the subject. The
intent expressed during the oral hearing of having the now cancelled claims
stand/fall with those claims argued is not expressly stated in the principal
brief.
To the extent Appellant had any due process rights that were abridged
by the Examiner’s cancellation of the disputed claims, it is our view that the
issue was not raised in a timely manner before the Examiner nor the Board.
Therefore, the issue has been waived.
Appeal 2009-007907
Application 10/413,579

10

CONCLUSION AND DECISION
The rejection of representative independent claim 1 on appeal under
35 U.S.C. § 103 is reversed. Correspondingly, the rejection of independent
claim 11 is also reversed, as is the rejection of their respective dependent
claims. Therefore, the decision of the Examiner is reversed.

REVERSED

llw

NIXON & VANDERHYE, PC
901 NORTH GLEBE ROAD, 11TH FLOOR
ARLINGTON, VA 22203