Apr 22, 2011 - Pat. No. 6,237,084, which is a divisional of application Ser. No. ... of application software that can in
USO0RE43 729E
(19) United States (12) Reissued Patent
(10) Patent Number: US (45) Date of Reissued Patent:
Morikawa et al. (54)
(56)
PROCESSOR WHICH CAN FAVORABLY EXECUTE A ROUNDING PROCESS COMPOSED OF POSITIVE CONVERSION AND SATURATED CALCULATION PROCESSING
RE43,729 E Oct. 9, 2012
References Cited U.S. PATENT DOCUMENTS 4,935,890 A 6/1990 Funyu 4,945,507 A
7/1990 Ishida et a1.
(Continued)
(75) Inventors: Toru MorikaWa, Kadoma (JP); Nobuo
Higaki, Kadoma (JP); Akira Miyoshi,
FOREIGN PATENT DOCUMENTS
Kadoma (JP); Keizo Sumida, Kadoma
EP
0 657 804 A1
(JP)
6/1995
(Continued)
(73) Assignee: Panasonic Corporation, Osaka (JP)
OTHER PUBLICATIONS
(21) App1.No.: 13/092,453
Nadehara, Kouhei, et al., “Low-Power Multimedia RISC”, IEEE
Micro, US, IEEE Inc., NeWYork, v01. 15, N0. 6, (Dec. 1, 1995), pp.
(22) Filed:
Apr. 22, 2011
20-29, XP538227, ISSN: 0272-1732.
Related US. Patent Documents
(Continued)
Reissue of:
(64)
Patent No.:
Primary Examiner * Richard Ellis
6,237,084 May 22, 2001 09/399,577 Sep. 20, 1999
Issued:
Appl. No.: Filed:
(74) Attorney, Agent, or Firm * McDermott Will & Emery LLP
U.S. Applications:
(57)
(62)
21, 2004, noW Pat. No. Re. 43,145, Which is a division
A processor Which executes positive conversion processing, Which converts coded data into uncoded data, and saturation
of application No. 10/366,502, ?led on Feb. 13, 2003,
calculation processing, Which rounds a value to an appropri
noW Pat. No. Re. 39,121, Which is a division of appli cation No. 08/980,676, ?led on Dec. 1, 1997, noW Pat.
ate number of hits, at high speed. When a positive conversion saturation calculation instruction “MCSST D1” is decoded, the sum-product result register 6 outputs its held value to the path P1. The comparator 22 compares the magnitude of the held value of the sum-product result register 6 With the coded
Division of application No. 11/016,920, ?led on Dec.
No. 5,974,540.
(30)
Foreign Application Priority Data
ABSTRACT
32-bit integer “0x0000_00FF”. The polarity judging unit 23 Nov. 29, 1996
(JP) ..................................... .. 8-320423
judges Whether the eighth bit of the value held by the sum product result register 6 is “ON”. The multiplexer 24 outputs one of the maximum value “0x0000_00FF” generated by the
(51)
Int. Cl. G06F 9/302
(52)
US. Cl. ...................................... .. 708/551; 708/552
(58)
Field of Classi?cation Search ................ .. 708/550,
constant generator 21, the Zero value “0x0000_0000” gener ated by the Zero generator 25, and the held value of the sum-product result register 6 to the data bus 18.
708/551, 552, 203, 204, 208 See application ?le for complete search history.
2 Claims, 17 Drawing Sheets
(2006.01)
116
7
3 g"“'nsmvrcoméromsnowman"1 y8 " \E CALCULATION CIRCUIT /c2 /C3
1 /23 1 /22 i $000000 ICOMPARATORI i
a____1 5
J2
/1
i_a
,25
,21
; REGISTERFILE
/B 5 1 ZEROGENERATOR IICONSTANTGENERATORI i :
MGR
1
:I
24
MULTIPLEXER
:
j
:------------------------------------ --:
/ KCl
,04 e
)
/18
s
US RE43,729 E Page 2 US. PATENT DOCUMENTS 5,235,533 A 8/1993 Sweedler . 5,251,166 A 10/1993 Ish1da 5 402 368 A 3/l995 Y d ’
’
m ‘1
5,448,509 A
9/1995 Lee et a1.
5,504,697 A
4/1996 Ishida
5,508,951 A 5,537,562 A
4/1996 IshlkaWa 7/1996 Gallup et al.
JP JP JP
KR
8-272591 09-97178
10/1996 4/1997
10-55274
2/1998
1995-0010571
9/1995
WO
9617292
6/1996
W0
W0 9617292
6/l996
OTHER PUBLICATIONS
5,684,728 A
11/1997 Okayama et a1~
Lee, RubyB., “Subword Parallelismwith MAX-2”, IEEE Micro, US,
5,696,709 A 5,801,977 A
12/1997 Smlth, 5‘ 9/1998 Karp et al.
IEEE Inc., New York, V01. 16, N0. 1. , (Aug. 1, 1996), pp. 51-59, XP000596513
5,812,439
A
9/1998
Hansen
'
_
_
_
_
5,847,978 A 5,889,980 A 5,915,109 A
12/1998 OguIa 3/1999 Smith, Jr, 6/ 1999 Nakakimura et 31,
Japanese Of?ce Act1on, issued in correspond1ng Japanese Patent Application No. 9-327866, dated on Oct. 11, 2007. Korean Of?ce Action, issued in Corresponding Korean Patent Appli
5,917,740 A 5,974,540 A
6/1999 Volkonsky 10/1999 MOfik?Wa et a1.
cation No. 10-1997-0064288, dated on Feb. 26, 2004. TMS32010 User’s Guide, Digital Signal Processor Products, Texas
6,029,184 A
2/2000 He
6,058,410 A
5/2000 Shamngpanl
_
FOREIGN PATENT DOCUMENTS
Instruments, 1983, pp. 3-7.
Dictionary.com, de?nition of “speci?ed”, http://dictionary.reference. com/browse/ speci?ed, accessed Jun. 16, 2010.
Dictionary.com, de?nition of “de?nitely”, http://dictionaryrefer EP EP
657804 0 766 169 Al
g3
22330000823 A
JP
5866032
JP
JP JP JP JP
58-056032 A
07482141 7-210368 A 7210368 7-334346
6/1995 4/1997
ence.com/broWse/de?nitely, accessed Jun. 16, 2010. Dictionary.com, de?nition of “unambiguosly”, http://dictionaryref
4/1983
Patterson et al., AVLSI RISC, IEEE Computer, 1982, pp. 8-18 and
4/1983
5041'
erence.com/broWse/unambiguously, accessed Jun. 16, 2010. 7 / 1995 8/1995 8/1995 12/1995
.
.
,,
.
Low-Power Mult1med1a RISC, by K. Nadehara, 8207 IEEE M1cro 15 (1995) D99, N9 6 “Subword Parallelism With MAX-2,” by R. Lee, IEEE Micro Aug. 1, 1996, vol. 16, No. 4.
US. Patent
0a. 9, 2012
Sheet 1 0f 17
US RE43,729 E
FIG. 1 PRIOR ART
111
ARITHMETIC LOGIC UNIT
/62 SUM-PRODUCT RESULT REGISTER
US. Patent
OEMNP05?“
0a. 9, 2012
EUmSdHQMNTo;
Sheet 2 0f 17
US RE43,729 E
US. Patent
0a. 9, 2012
Sheet 4 0f 17
M .EaNEZQE
m H
w.05
0\r?mio .2579: .
_
_
- -
(0,
D
-
-
-
_ -
US RE43,729 E
-
_
_
.
_
-
-
l-I
US. Patent
Oct. 9, 2012
US RE43,729 E
Sheet 6 0f 17
FIG. 6 MACCB INSTRUCTION
|||||
_|_: F_____
MULTIPLIER READ ADDRESS
INDICATION
T
]
MULTIPLICAND READ ADDRESS
INDICATION
11' ' ' 'MCR
11- - - -MCR
OO- ' ' ‘REGISTER DO
0O‘ ' - 'REGISTER DO
01' ' ' 'REGISTER D1 10' ' ' ‘REGISTER D2
01 ' ' ' 'REGISTER D1
10- ' ' 'REGISTER D2
—INDICATION OF CONTENT OF ELEMENTAL OPERATION 1' - * ~MULTIPLICATION O- - - -NONE
‘—INDICATION OF CALCULATED CONTENT OF ALGEBRAIC SUM l - ‘ - 'ADDITION
0' ' ' "NONE
INDICATION OF STORAGE ADDRESS FOR SUM-PRODUCT RESULT
1""MCR 0""NONE
US. Patent
0a. 9, 2012
Sheet 7 0f 17
US RE43,729 E
FIG. 7 MCSST INSTRUCTION
POSITIVE CONVERSION STORAGE ADDRESS SATURATION CALCULATION INDICATED WIDTH INDICATION 00"“24bit POSITIVE CONVERSION 00' ' ' ' REGISTER DO 01 ""I?bit POSITIVE CONVERSION OI ' ' ' ' REGISTER D1
1 1 -"'8bilPOSITIVE CONVERSION
10- ' ' 'REGISTER D2 1 1 - ' - 'REGISTER D3
US. Patent
Oct. 9, 2012
Sheet 8 0f 17
US RE43,729 E
FIG. 8A 25 24
1716
llllllllllllllllllllllll 2524
1
8 7
§§§§§§ ~ I
CODE BIT MULTIPLIE?. MULTIELICAND ji F1]
1716
1
9 8
% w \. R
l\\
R I5
n G.1
H 2U I
1WM+
mv/ m1.
wmm n H % nS
“nn“5JRU\w|9.Ow/|muu,. W vmf .
m m w w
/A, m M 6|! 4‘k
?///// /% 1mHM
WWW W HO A M!M| m WWW H
/ uI 7.7
HL2“l
u I WU
US. Patent
0a. 9, 2012
Sheet 9 0f 17
US RE43,729 E
FIG. 9
LOGIC VALUE X LOGIC VALUE Y SELECTED INPUT VALUE 1 0 OXOOOO__OOFF 1 1 OXOO0O_OOOO O 1 0x0O0O_0OOO 0 0 STORED VALUE OF SUM-PRODUCT RESULT REGISTER
US. Patent
Oct. 9, 2012
Sheet 10 0f 17
US RE43,729 E
FIG. 10 EXAMPLE OPERATION: DO >< D1 (Ox7f X 0x70) MEMORY STORED VALUE
0X7f
0X70
REGISTER STORED VALUE D0
D1 32
4
32
/
5
/
CODE EXTENSION CODE EXTENSION CIRCUIT CIRCUIT 0x0000007f
32
,454 OXOOOOOOOOOOOO379O V
6
MCR
OUTPUT OF LOWER-ORDER
32 BITS
#1
‘:32 0x00003592 MSB I O
0x00003790>0x000000ff CIRCUIT
a0X000000ff 32 OXOOOOOOff
REGISTER STORED VALUE
D1 OxOOOOOUff |
MEMORY STORED VALUE
()xff
US. Patent
0a. 9, 2012
Sheet 11 0f 17
US RE43,729 E
FIG. 11 EXAMPLE OPERATION: D0 >< D1(Ox7f >< 0x80) MEMORY STORED REGISTER STORED 32
32
rJ4
/5
CODE EXTENSION CODE EXTENSION‘ CIRCUIT
CIRCUIT
0x0000007f
32
32 Oxffffff‘80
’ 64 OxffffffffffffcOSO \
MCR OUTPUT OF LOWER-ORDER
32 BITS
'16
32 OxffffcO/S/O3
POSITIVE CONVERSION
MSB: 1-»0x00000000
SATURATION CALCULATION CIRCUIT
/32 0x00000000 REGISTER VALUE STORED
V
D1 OXOOOOOOff |
MEMORY STORED VALUE
0X00
US. Patent
0a. 9, 2012
Sheet 14 or 17
US RE43,729 E
FIG. 13 MCSST INSTRUCTION
TORAGE ADDRESS ERIIIIIIEE‘IINXEESRJTATION WIDTH INDICATION INDICATED 11-- ~MCR 00~~24bit POSITIVE CONVERSION 01~~~16bit POSITIVECONVERSION
1 1~~~8bil POSITIVE CONVERSION
—READ ADDRESS INDICATION 11- - - 'MCR
00' ' ' ‘REGISTER D0 01' ' ' ‘REGISTER D1 10' - - 'REGISTER D2
00- - - - REGISTER D0 01- - - ' REGISTER D1 10' ' ' ' REGISTER D2
US. Patent
0a. 9, 2012
Sheet 17 0f 17
US RE43,729 E
FIG. 16 MULBSST INSTRUCTION
MULTIPLIER READ MULTIPLICAND READ ADDRESS INDICATION ADDRESS INDICATION 11----MCR
11~~McR
OO----REGISTERDO
OO~---REGISTERDO
01----REGISTERD1 O1----REGISTERD1 10' -~REG1STERD2
10~ - - REGISTER D2
- POSITIVE CONVERSION SATURATION CALCULATION
WIDTH INDICATION 01' ' ' '24bit POSITIVE VALUE
10' ' ' '16bit POSITIVE VALUE
II' ' ' '8bit POSITIVE VALUE
———-——CALCULATION CONTENT INDICATION I- ' - 'MULTIPLICATION
O‘ ' ' ‘NONE
US RE43,729 E 1
2
PROCESSOR WHICH CAN FAVORABLY EXECUTE A ROUNDING PROCESS COMPOSED OF POSITIVE CONVERSION AND SATURATED CALCULATION PROCESSING
hardware, and is provided with an instruction set that includes a “MOV MCR, * *” transfer instruction for transferring a sum
product value. An example of the hardware construction of a conventional multimedia-oriented processor is shown in FIG. 1. As shown
in FIG. 1, the arithmetic logic unit (hereinafter, “ALU”) 61 performs the multiplication of an element Fij that forms part of the compressed data and an element Gji that forms part of
Matter enclosed in heavy brackets [ ] appears in the original patent but forms no part of this reissue speci?ca
the coe?icient matrix in accordance with a multiplication instruction. The ALU 61 also reads the sum-product value
tion; matter printed in italics indicates the additions made by reissue.
stored in the sum-product result register 62, adds the multi plication result of Gji*Fij to the read sum-product value, and has the result of this addition stored in the sum-product result
[This is a divisional application of US. Ser. No. 08/980, 676 now US. Pat. No. 5,974,540 ?led Dec. 1, 1997.] More than one reissue application has been filedfor the reissue of
register 62. By repeating the above calculation, a sum-prod
(reissued as Re. 39,121 on Jun. 6, 2006) and 11/016,920 (reissued as Re. 43,145 on Jan. 24, 2012), all ofwhich are
uct value is accumulated in the sum-product result register 62. Once the multiplication has been performed a predetermined number of times, the programmer issues a sum-product value transfer instruction. By issuing a transfer instruction, the accumulated value in the sum-product result register 62 is transferred to the general registers, and is used as the matrix
divisional reissues ofU.S. Pat. No. 6,237,084. This applica
multiplication result for one row and one column. By per
tion is a divisional reissue ofapplication Ser. No. 11/016,920, filed on Dec. 21, 2004, now US. Pat. No. Re. 43,145, which is a divisional reissue ofapplication Ser. No. 10/366,502?led Feb. 13, 2003, now US. Pat. No. Re. 39,121, which isa reissue
forming N*N iterations of the above processing, the matrix multiplication of N*N compressed data and an N*N coef? cient matrix can be completed.
of 09/399,577 filed on Sep. 20, 1999, now US. Pat. No. 6,237,084, which is a divisional of application Ser. No. 08/980,676?led Dec. 1, 1997, now US. Pat. No. 5,974,540.
used, however, positive correction saturation operations for
5
US. Pat. No. 6,237,084. The reissue applications are the
present application and application Ser. Nos. 10/366,502
When a conventional multimedia-oriented processor is
amending the sum-product value pose many di?iculties for programmers. Positive conversion processing refers to the conversion of a sum-product value that is a negative value into either Zero or
BACKGROUND OF THE INVENTION
a positive value. Normally, compressed data is expressed as a coded relative value that re?ects the relation of the present value to the preceding and succeeding values. As a result,
1. Field of the Invention The present invention relates to a processor that performs processing according to instruction sequences that are stored in a ROM or the like.
2. Background of the Invention In recent years, there has been a visible increase in the use
of application software that can interactively reproduce vari ous kinds of data, such as video data, still image data, and audio data, that have been compressed according to tech niques such as frame encoding, ?eld encoding, or motion compensation. As such software has been developed, there has been increasing demand for multimedia-oriented proces sors that can ef?ciently execute the software. These multime dia-oriented processors are processors designed with a spe cial architecture to facilitate programming, such as the
there are many cases when the sum of products for each 35
element in the compressed data and the corresponding coef ?cients is a negative value. Most reproduction-related hard ware, such as displays and speakers, however is only able to process uncoded data, so that when the sum-product values are to be reproduced, it is ?rst necessary to perform positive
conversion processing. Saturation calculation processing refers to processing that sets all values that exceed a given range (or, in other words, which are “saturated”) at a predetermined value. This is to say, when an element that includes an erroneous bit generated during transfer is used in a sum-product calculation as part of
compression and decompression of video and audio data. The
the sum-product processing for compressed data, there is an increase in the probability of the sum-product value exceed
high-speed processing required for handling video data is the
ing a value that can be expressed by the stated number of bits.
matrix multiplication of compressed data that has N*N
Since most reproduction-related hardware is only physically
matrix elements with coe?icient data that also has N*N
capable of reproducing uncoded data with a ?xed valid num
matrix elements. Representative examples of compressed
ber of bits, such as eight bits, saturation processing is required
data that has N*N matrix elements are the luminescence
to convert the sum-product value into a value that can be
block composed of 16*16 luminescence elements, the blue color difference block (Cb block) composed of 8*8 color difference elements, and the red color difference block (Cr block) composed of 8*8 color difference elements used in
expressed using the valid number of bits. It has been conventional practice to perform this kind of positive value conversion processing and saturation calcula
tion processing by converting the-sum-product value using a
MPEG (Moving Pictures Experts Group) techniques. The
subroutine that corrects the sum-product value. An example
matrix multiplication for compressed data referred to here is
of a subroutine that corrects the sum-product value is
performed very frequently when executing the approxima
explained below. In this example, the register width and the
tion calculations for an inverse DCT (Discrete Cosine Trans
calculation width of the calculation unit are 32 bits, with the
form) in image compression methods such as MPEG and
width of the MCR being 32 bits, and the sum-product value
JPEG (Joint Photographic Experts Group).
being expressed as a coded 16-bit integer. The data that can be
The following is a description of conventional multimedia oriented processors that can perform high-speed matrix mul tiplication. The basic architecture of conventional multime dia-oriented processors is provided with a sum-product result register (hereinafter simply referred to as an MCR register) as
handled by the reproduction-related hardware needs to be expressed using uncoded 8-bit integers. This subroutine is set as using the data register D0 for storing the calculation result. Each instruction is expressed using two operands, with the
left and right operands being respectively called the ?rst and
