WO1999034282A1 - Improved instruction dispatch mechanism for a guarded vliw architecture - Google Patents
Improved instruction dispatch mechanism for a guarded vliw architecture Download PDFInfo
- Publication number
- WO1999034282A1 WO1999034282A1 PCT/IB1998/001521 IB9801521W WO9934282A1 WO 1999034282 A1 WO1999034282 A1 WO 1999034282A1 IB 9801521 W IB9801521 W IB 9801521W WO 9934282 A1 WO9934282 A1 WO 9934282A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- fields
- functional units
- issue
- instruction
- functional unit
- Prior art date
Links
- 230000015654 memory Effects 0.000 claims description 26
- 238000000034 method Methods 0.000 description 6
- 238000013459 approach Methods 0.000 description 3
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- 229930091051 Arenine Natural products 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 239000004065 semiconductor Substances 0.000 description 1
- 238000013519 translation Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F9/00—Arrangements for program control, e.g. control units
- G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
- G06F9/30—Arrangements for executing machine instructions, e.g. instruction decode
- G06F9/30003—Arrangements for executing specific machine instructions
- G06F9/30072—Arrangements for executing specific machine instructions to perform conditional operations, e.g. using predicates or guards
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F9/00—Arrangements for program control, e.g. control units
- G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
- G06F9/30—Arrangements for executing machine instructions, e.g. instruction decode
- G06F9/38—Concurrent instruction execution, e.g. pipeline, look ahead
- G06F9/3885—Concurrent instruction execution, e.g. pipeline, look ahead using a plurality of independent parallel functional units
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F9/00—Arrangements for program control, e.g. control units
- G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
- G06F9/30—Arrangements for executing machine instructions, e.g. instruction decode
- G06F9/30145—Instruction analysis, e.g. decoding, instruction word fields
- G06F9/3016—Decoding the operand specifier, e.g. specifier format
- G06F9/30167—Decoding the operand specifier, e.g. specifier format of immediate specifier, e.g. constants
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F9/00—Arrangements for program control, e.g. control units
- G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
- G06F9/30—Arrangements for executing machine instructions, e.g. instruction decode
- G06F9/38—Concurrent instruction execution, e.g. pipeline, look ahead
- G06F9/3836—Instruction issuing, e.g. dynamic instruction scheduling or out of order instruction execution
- G06F9/3853—Instruction issuing, e.g. dynamic instruction scheduling or out of order instruction execution of compound instructions
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F9/00—Arrangements for program control, e.g. control units
- G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
- G06F9/30—Arrangements for executing machine instructions, e.g. instruction decode
- G06F9/38—Concurrent instruction execution, e.g. pipeline, look ahead
- G06F9/3885—Concurrent instruction execution, e.g. pipeline, look ahead using a plurality of independent parallel functional units
- G06F9/3889—Concurrent instruction execution, e.g. pipeline, look ahead using a plurality of independent parallel functional units controlled by multiple instructions, e.g. MIMD, decoupled access or execute
Definitions
- the present invention relates to the architecture of very long instruction word (VLIW) processors and more particularly to a method and apparatus for dispatching instructions for VLIW processors.
- VLIW very long instruction word
- VLIW very long instruction word
- a very long instruction word (VLIW) processor comprises multiple, parallel functional units controlled on a cycle-by-cycle basis by a very lone, instruction word (i.e., 100 or more bits).
- Very long instruction words comprise a concatenation of fields or "issue slots,” each of which independently specifies the operation of a functional unit.
- VLIW processors are used in a variety of applications including super-computers and mainframes, workstations and personal computers, and dedicated processors in audio and video consumer products.
- Figure 1 illustrates a conventional VLIW processor 1 00.
- Instruction memory 1 10 which can be a random access memory (RAM) or a read only memory (ROM), is typically pipelined and supplemented by an instruction cache (not shown) to enhance execution throughput.
- Each instruction word loaded into instruction issue register 120 contains a number of issue slots 121-127, each issue slot 121-127 for controlling a corresponding functional unit 131-137 in the VLIW processor 1 00.
- a VLIW processor may comprise any useful combination of functional units, and Figure 1 depicts one such combination.
- each issue slot 121-127 within an instruction word loaded into the instruction issue register 120 specifies an operation to be started in the current clock cycle for the corresponding functional unit 131-137.
- each issue slot 121-127 typically contains an opcode and operands for the corresponding functional unit 131-137.
- the opcode is useful for functional units that perform a variety of different operations.
- Operands for the functional units 131-137 are read from a shared, multi-ported register file 140, and results from the functional unit 131-137 are written into the register file 140.
- the term "specification of an operation” as used herein refers to a combination of an opcode, if needed., and operands, if needed, employed to specify an operation of a functional unit.
- each issue slot 121-127 contains a specification of operation to be executed by a corresponding functional unit 131- 137.
- issue slot 121 contains a specification of the operation 210 of the constant generation unit 13 1, namely, a constant value CONSTANT and a register RD to hold the constant value.
- the specification of the operation 220 of arithmetic-logic units (ALU) 132 and 133 contained in respective issue slots 122 and 123 holds an ALU opcode, two source registers RA and RB, and a destination register RD for the operation of respective.
- Typical ALU opcodes indicate operations such as addition, subtraction, negation, logical and, logical or, logical exclusive or, logical complementation, and the like.
- Issue slot 124 holds a specification 240 containing a MUL opcode for multiplication, division, or square root, source registers RA and RB, and a destination register RD for the multiplier unit 134.
- Issue slot 125 includes specification 250 having an FPU opcode (e.g., addition, subtraction, and comparison), source registers RA and RB, and destination register RD for the floating point unit 135.
- a data memory unit 136 is controlled by issue slot 126, having a specification 260 including a MEM opcode, indicating a load or store operation, address registers RA and RB, and a data register RD.
- the jump control unit 13 7 with reference to specification 270 in the corresponding issue slot 137 uses register RA and RB to indicate a conditional value and a jump destination address within instruction memory 1 10; the specification 270 also holds a MP opcode specifying whether to jump always (unconditional jump), jump if the conditional value register is true, jump if the register is false, or not jump at all (NOP).
- the contents of these issue slots 121-127, the operation of the functional unit 131-137, and the format of specifications 210-270 are to be regarded as exemplary and may be adjusted to suit any useful configuration.
- VLIW processors In order for a software program to run on a VLIW machine, a "fine grained parallel" or “instruction level parallel” translation must be found. This is accomplished by a compiler that translates a conventional high-level programming language, such as ANSI-C, into VLIW instructions. Such compilers, are described in John R. Ellis, BULLDOG: A Compiler for VLIW Architectures. MIT Press 1985, ISBN 0-262-05034-X. Functional units in conventional VLIW processors are controlled by exactly one issue slot in the instruction word. For example, the VLIW processor 1 00 depicted in Figure 1 includes seven issue slots 121- 127 corresponding to the seven respective functional units 131-137. Thus, compilers for conventional VLIW processors emit instructions in which a functional unit is controlled by values in a signal issue slot.
- High performance processors including VLIW processors, are subject to the so-called "branch delay" problem caused by the latency of instruction memory, which is the time in machine cycles between the transmission of an instruction address to an instruction memory and the receipt of the corresponding instruction word for execution.
- the address of the next instruction can be predicted, allowing the instruction memory to be pipelined for an effective instruction memory latency of one machine cycle.
- the address of the next instruction cannot be predicted, because the destination address depends on the outcome of evaluating a conditional expression. Consequently, the branch delay represents a number of machine cycles in which the instruction word at the destination address is not available due to the evaluation of the conditional expression and instruction memory latency.
- VLIW processors One disadvantage with current VLIW processors is evident when two or more of the parallel program paths in the branch delay period include operations pertaining to the same functional unit, such as a floating point unit 135. Since each functional unit is controlled by exactly one issue slot in the instruction word, operations for the same functional unit along different parallel program paths must be issued in separate instructions. For example, both program paths of a numerical analysis program may employ floating point operations after a conditional branch. In a VLIW processor with one floating point unit 135, only one floating point operation can be issued in each instruction. Thus, the floating point operations along one path, e.g. the "condition true” path, must be issued in different instructions than the floating point operations along the other path, e.g. the "condition false" path.
- the conventional VLIW processor 100 includes two arithmetic-logic units 132 and 133. While a second arithmetic-logic may not be prohibitively expensive, a second floating point unit is expensive to implement again on a monolithic semiconductor device in terms of consumption of surface area and power. Other complex functional units, such as multipliers, barrel shifters, and even data memory units, are also expensive to duplicate.
- VLIW very lone, instruction word
- An instruction register stores the very long instruction word, which has multiple fields, each field including: an identifier for selecting one of the functional units, a specification of an operation to be executed by the selected functional unit, and a guard value for indicating whether or not to inhibit the operation.
- a routing circuit delivers the specification of the operation to be executed from the instruction register to the selected functional unit according to the associated guard value, e.g. only if the guard value indicates that the operation is not inhibited.
- VLIW processor for executing a sequence of very long instruction words specifying multiple operations to be executed in parallel.
- the VLIW processor comprises multiple functional units for executing the operations in parallel and an instruction register for storing the very long instruction word.
- the very long instruction word has multiple fields, the number of which is greater than the number of functional units.
- a routing circuit delivers contents of one of the fields in the instruction from the instruction register to the associated functional unit.
- each field may include a specification of an operation to be executed by an associated functional and a guard value for indicating whether or not to inhibit the operation and controlling the routing circuit.
- Still another aspect of the present invention is a method of dispatching instructions in a VLIW processor comprising the step of issuing a very long instruction word having multiple fields.
- Each field includes an identifier for selecting one of multiple functional units, a specification of an operation to be executed by the selected functional unit, and a guard value for indicating whether or not to inhibit the operation.
- the specifications within the fields are routed to corresponding functional units based on the guard values, and the operation specified by the routed specification is performed.
- Yet another aspect of the present invention is a method of dispatching instructions in a VLIW processor comprising the step of issuing a very long instruction word having multiple fields.
- the contents within the fields are routed to the associated functional units, the number of which is less than the number of fields in the very long instruction word, and the operation specified by the routed content is performed.
- Each field may include a specification of an operation to be executed by an associated functional unit from among multiple functional units, and a guard value for indicating whether or not to inhibit the operation and controlling how the field contents are routed.
- Fig. 1 depicts a conventional VLIW processor
- Fig,. 2 illustrates formats of conventional VLIW issue slots
- Fig. 3 depicts a VLIW processor according to one embodiment of the present invention
- FIG. 4 illustrates formats of VLIW issue slots according to an embodiment of the present invention
- Fig. 5 depicts a VLIW processor according to another embodiment of the present invention.
- Fig. 6 illustrates formats of VLIW issue slots according to another embodiment of the present invention.
- VLIW instruction dispatch method and mechanism An improved VLIW instruction dispatch method and mechanism are described.
- numerous specific details are set forth in order to provide a thorough understanding of the present invention. It will be apparent, however, to one skilled in the art that the present invention may be practised without these specific details. In other instances, well-known structures and devices are shown in block diagram form in order to avoid unnecessarily obscuring the present invention.
- the present invention addresses and solves the problems associated with inflexible operation scheduling during the branch delay period in conventional VLIW processors by multiplexing the inputs to functional units from more than one issue slot based on the guard values.
- the guard values indicate separate pro,-ram paths after a branch condition. Since each issue slot contains a guard value, operations in issue slots are associated with a program path after a branch condition. If the guard values of two issue slots in a single instruction word for a common functional unit indicate mutually exclusive program paths, then the two issue slots are not in contention, because one of the issue slots is inhibited by its guard value, while the other is not inhibited by its guard value. Consequently, instructions can be issued controlling the common functional unit in two or more issue slots along mutually exclusive program paths. This approach is cost-effective because an extra multiplexer is typically much cheaper to implement than a second functional unit, such as a data memory unit, a floating point unit, a multiplier, and a barrel shifter.
- additional issue slots are designated for particular functional units.
- An instruction issue register thus contains more issue slots than there are functional units.
- issue slots 321-327 are positionally associated with corresponding functional units 331-337; i.e., particular positions within the instruction issue register 320 are designated for particular functional units.
- Seven functional units 331-337 are depicted in Fig. 3, viz.
- a constant generation unit 331 a first arithmetic-logic unit 332, a second arithmetic- logic unit 333, a multiplier 335, a floating-point unit 335, a data memory unit 336, and a branch control unit 337, respectively corresponding to the functional units 131-137, described herein above.
- issue slots 321-327 Associated with each of the seven functional units 331-337 are nine issue slots 321-327, wherein floating point unit 335 is associated with two issue slots 325a and 325b and data memory unit 336 is associated with two issue slots 326a and 326b.
- issue slots 321-327 contain operation parcels 410-470 that include a guard value G and a specification of the operation (e.g. an opcode and operands, if needed).
- the guard value G indicates which program path the operation is to be executed.
- guard values is disclosed in the commonly assigned U.S. Patent No. 5, 450,556 issued on Sep.
- guard value specifies a path expression that matches path information generated by the branch control unit 137. For example, if the branch control unit 137 implement three-way branches, then a program path can be expressed in a guard value by two bits. Guard values can also express conditional program paths due to subsequent branching in the branch delay period by concatenation.
- the specification of the operation within operation parcels 410-470 contain similar subfields such as opcodes and operands as described herein above with respect to specifications 210-270 in Fig. 2.
- the contents of these issue slots 321-327, the operation of the functional units 331-337, and the format of specifications in operation parcels 410-470 are to be regarded as exemplary and may be adjusted to suit any useful configuration.
- Multiplexer 355 has inputs coupled to the instruction register 320 at issue slots 325a and 325b. The output of multiplexer 355 is coupled to the floating point unit 335. Thus, multiplexer 355 routes an opcode and operands from one of the issue slots 325a and 325b to the floating point unit 335. Similarly, a multiplexer 356 has inputs coupled to the instruction register 320 at issue slots 326a and 326b, and the output of multiplexer 356 is coupled to the data memory unit 336. Thus, multiplexer 356 routes an opcode and operands from one of the issue slots 326a and 326b via to the data memory unit 336.
- Both multiplexers 355 and 356 are switched by guard values derived from the associated issue slots 325a/325b and 326a/326b, respectively, so that the opcode and operands for the non-inhibited issue slot is routed to the associated functional unit.
- the guard values are compared against the current program path maintained by the branch control unit 337. For example, issue slot 325a contains a guard value " 1 " that indicates a true condition in a branch path, while issue slot 325b contains a guard value "O" that indicates a false condition in a branch path.
- the current program path becomes a "1" and the multiplexer 355 selects the opcode and operand from the issue slot that has a guard value that matches the current program path, namely issue slot 325a.
- the false condition branch path is chosen, then the current program path becomes a "O" and the multiplexer 325 selects the opcode and operand from the other issue slot 325b. Consequently, an instruction can be issued from instruction memory 3 1 0 into instruction issue register 320 so that a common functional unit, such as a floating point unit 335, can be controlled along either branch path.
- a compiler can emit instructions having issue slots controlling a common fimctional unit insofar as the guard values for the issue slots are mutually exclusive.
- Instruction issue register 320 contains a plurality of fields for issue slots 321-227 corresponding to a functional unit 331-337.
- the multiplexer 355 is also configured to select an opcode and operands from one of the issue slots 325a and 325b based on the respective, guard values stored therein.
- issue slots 326a and 326b correspond to the data memory unit 336
- multiplexer 356 is configured to route a specification of an operation to be performed by the data memory unit 336 from one of the issue slots 326a and 326b.
- the multiplexer 356 is also configured to select an opcode and operands from one of the issue slots 326a and 326b based on the respective guard values stored therein. Consequently, a single instruction word executed during the branch delay can effectively utilise a selected functional unit, e.g. floating point unit
- the embodiment of the present invention illustrated in Fig. 3 expands the size of the instruction word in order to designate additional issue slots for desired functional units.
- this embodiment of the present invention is subject to some disadvantages. For example, in sequential operation, only one of the issue slots for a given functional unit will be utilised, thereby increasing the bandwidth requirements to sustain the VLIW processor at or near peak performance. Further, the increased size of the instruction word raises the cost of instruction memories, caches, and buses. Finally, the size of compiled code for a VLIW compiled program is larger, thereby reducing the overall cost-performance of the program.
- VLIW processor 500 uses instruction words having only three issue slots 521-523 in instruction issue register 520, wherein each issue slot 521-523.
- the VLIW processor 500 includes routing circuitry comprising multiplexers 551-557 coupled to instruction issue register 520 at each issue slot 521-523 delivering the operands contained therein to the identified functional unit 531-537. Seven such functional units 531-537 are depicted in Fig.
- a constant generation unit 531 a constant generation unit 531, a first arithmetic-logic unit 532, a second arithmetic-logic unit 533, a multiplier 535, a floatingpoint unit 535, a data memory unit 536, and a branch control unit 537, respectively corresponding to the functional units 131-137, described herein above.
- issue slots 521-523 contain operation parcels 610-670 that include a guard value G, a functional unit identifier FU, and a specification of the operation (e.g. an opcode and operands, if needed).
- the guard value G indicates which program path the operation is to be executed.
- a functional unit identifier FU can be an integer or other scalar value, for example, a three-bit number from 0-7, wherein a value of " 1 " indicates the constant generation unit 531 of Fig. 5.
- the specification of the operation within operation parcels 610-670 contain similar subfields such as opcodes and operands as described herein above with respect to specifications 210-270 in Fig. 2.
- Each multiplexer 531 -337 selects one of the issue slots 521 -323 based on the guard values and functional unit identifiers contained therein and outputs the contents of a selected issue slot 521-523 to a respective functional unit 531-537.
- multiplexer 551 is associated with the constant generation unit 531 and outputs a selected issue slot 521- 523 thereto. More specifically, multiplexers 551-557 select the contents of only those issue slots 521-523 that contain a functional unit identifier of the associated functional unit 531-537.
- multiplexers 551-557 select the contents of only those issue slots 521-523 that contain a guard value that indicates that the current program path is not inhibited.
- the multiplexers 551-557 may be implemented as two multiplexers coupled in series: the first multiplexer selects among the issue slots that match the functional unit, based on comparing the functional unit identifiers FU, and the second multiplexer selects among the remaining issues that match of the current program path, based on comparing the guard values G.
- multiplexer 551 is configured to select among all three issue slots 521-523 for an operation parcel destined for constant generation unit 53 1. More specifically, multiplexer 551 makes the selection among issue slots 521-523 based on whether the issue slots 521-523 have a functional unit identifier that identifies the constant generation unit 5 3 1. In the example, since the constant generation unit 53 1 is identified by the number " 1 ", only those issue slots 521-523 are chosen in which the functional unit identifier, FU1, FU2, and FU3, respectively, contained therein is a "1.” Moreover, the multiplexer 551 selects from among the issue slots 521-523 based on whether the guard value indicates that the current program path is not inhibited.
- the guard values are compared against the current program path maintained by the branch control unit 537.
- issue slot 521 contains a guard value G 1 of "1" that indicates a true condition in a branch path, a functional unit identifier FU1 of "1” to indicate the constant generation unit 53 1.
- Issue slot 522 has a guard value G2 of "O” that indicates a false condition in a branch path and a functional unit identifier FU2 of "1,” and issue slot 523 has a guard value G3 of "I” and afunctionalunitidentifierFU3of '2".
- Formultiplexer551,issueslot523 willnotbe chosen since its functional unit identifier FU3 of "2" does not match the identifier " 1 " assigned to the corresponding functional unit 53 1. If the true condition branch path is chosen, then the current program path becomes a " 1 " and multiplexer 551 selects the opcode and operand from the remaining issue slot that has a guard value that matches the current program path, namely issue slot 521. On the other hand, if the false condition branch path is chosen, then the current program path becomes a "O" and the multiplexer 551 selects the opcode and operand from the other issue slot 522.
- both the functional unit identifier must correspond to the functional unit 531-537 associated with the multiplexer 551-557 and the guard value must indicate that the current program path is not inhibited.
- Two or more issue slots 521-523 can contain functional unit identifiers for the same functional unit 531 -537 if their respective guard value indicate mutually exclusive programs paths.
- both issue slots 521 and 522 may contain a functional unit identifier for the constant generation unit 531 but guard values for different and mutually exclusive program paths.
- the constant generation unit 531 can be operated by a single guarded instruction notwithstanding which program path is currently executing.
- Instruction issue register 320 contains a plurality of fields for issue slots 521-523, each field containing a functional unit identifier and a guard value. Coupled to each issue slot 521-523 of instruction issue register 320 are multiplexers 551-557, each associated with and coupled to a respective functional unit531-537. Multiplexers 551- 557 select the contents of one of the issue slots 521-523 if the functional unit identifier for the issue slot 521-523 identifies the respective functional unit 531-537 and the guard value is not inhibited.
- a single instruction word executed during the branch delay can effectively utilize a selected functional unit, e.g. constant generation unit 53 1, along every mutually exclusive program path.
- a selected functional unit e.g. constant generation unit 53 1
- data is fetched from register file 340 and results are written back into register file 540 based upon operands routed from issue slots 521-523 of the instruction issue register 520 to the corresponding functional units 531-537.
Landscapes
- Engineering & Computer Science (AREA)
- Software Systems (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Advance Control (AREA)
- Executing Machine-Instructions (AREA)
Abstract
Description
Claims
Priority Applications (4)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
DE69835425T DE69835425T2 (en) | 1997-12-29 | 1998-10-01 | IMPROVED COMMUNICATION MECHANISM FOR A PROTECTED VLIW ARCHITECTURE |
JP53467499A JP3829166B2 (en) | 1997-12-29 | 1998-10-01 | Extremely long instruction word (VLIW) processor |
KR1019997007912A KR100606397B1 (en) | 1997-12-29 | 1998-10-01 | Improved instruction dispatch mechanism for a guarded vliw architecture |
EP98944151A EP0976032B1 (en) | 1997-12-29 | 1998-10-01 | Improved instruction dispatch mechanism for a guarded vliw architecture |
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
US08/998,486 US5974537A (en) | 1997-12-29 | 1997-12-29 | Guard bits in a VLIW instruction control routing of operations to functional units allowing two issue slots to specify the same functional unit |
US08/998,486 | 1997-12-29 |
Publications (1)
Publication Number | Publication Date |
---|---|
WO1999034282A1 true WO1999034282A1 (en) | 1999-07-08 |
Family
ID=25545265
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
PCT/IB1998/001521 WO1999034282A1 (en) | 1997-12-29 | 1998-10-01 | Improved instruction dispatch mechanism for a guarded vliw architecture |
Country Status (7)
Country | Link |
---|---|
US (1) | US5974537A (en) |
EP (1) | EP0976032B1 (en) |
JP (1) | JP3829166B2 (en) |
KR (1) | KR100606397B1 (en) |
DE (1) | DE69835425T2 (en) |
TW (1) | TW405094B (en) |
WO (1) | WO1999034282A1 (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2004511043A (en) * | 2000-10-05 | 2004-04-08 | コーニンクレッカ フィリップス エレクトロニクス エヌ ヴィ | Retargetable compilation system and method |
Families Citing this family (21)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP3541669B2 (en) | 1998-03-30 | 2004-07-14 | 松下電器産業株式会社 | Arithmetic processing unit |
EP1050809A1 (en) * | 1999-05-03 | 2000-11-08 | STMicroelectronics SA | Computer instruction dependency |
EP1050810A1 (en) * | 1999-05-03 | 2000-11-08 | STMicroelectronics SA | A computer system comprising multiple functional units |
US20040158695A1 (en) | 1999-05-03 | 2004-08-12 | Laurent Ugen | Method and apparatus for handling transfer of guarded instructions in a computer system |
US7007156B2 (en) * | 2000-12-28 | 2006-02-28 | Intel Corporation | Multiple coprocessor architecture to process a plurality of subtasks in parallel |
US7093107B2 (en) * | 2000-12-29 | 2006-08-15 | Stmicroelectronics, Inc. | Bypass circuitry for use in a pipelined processor |
JP2002318687A (en) * | 2001-04-18 | 2002-10-31 | Fujitsu Ltd | Information processor and computer system |
US6886091B1 (en) * | 2001-06-29 | 2005-04-26 | Koninklijke Philips Electronics N.V. | Replacing VLIW operation with equivalent operation requiring fewer issue slots |
AU2002348842A1 (en) * | 2001-11-26 | 2003-06-10 | Koninklijke Philips Electronics N.V. | Wlim architecture with power down instruction |
KR100464406B1 (en) * | 2002-02-08 | 2005-01-03 | 삼성전자주식회사 | Apparatus and method for dispatching very long instruction word with variable length |
KR100947446B1 (en) * | 2002-03-28 | 2010-03-11 | 엔엑스피 비 브이 | Vliw processor |
CN1685310A (en) * | 2002-09-24 | 2005-10-19 | 皇家飞利浦电子股份有限公司 | Apparatus, method, and compiler enabling processing of load immediate instructions in a very long instruction word processor |
AU2003283686A1 (en) * | 2002-12-12 | 2004-06-30 | Koninklijke Philips Electronics N.V. | Modular integration of an array processor within a system on chip |
JP4283131B2 (en) * | 2004-02-12 | 2009-06-24 | パナソニック株式会社 | Processor and compiling method |
EP1751655A2 (en) * | 2004-05-13 | 2007-02-14 | Koninklijke Philips Electronics N.V. | Run-time selection of feed-back connections in a multiple-instruction word processor |
JP2006012185A (en) * | 2005-08-02 | 2006-01-12 | Matsushita Electric Ind Co Ltd | Processor |
US7523434B1 (en) * | 2005-09-23 | 2009-04-21 | Xilinx, Inc. | Interfacing with a dynamically configurable arithmetic unit |
US20070073999A1 (en) * | 2005-09-28 | 2007-03-29 | Verheyen Henry T | Hardware acceleration system for logic simulation using shift register as local cache with path for bypassing shift register |
US7822948B2 (en) * | 2008-01-03 | 2010-10-26 | International Business Machines Corporation | Apparatus, system, and method for discontiguous multiple issue of instructions |
KR101645001B1 (en) * | 2009-02-18 | 2016-08-02 | 삼성전자주식회사 | Apparatus and method for generating VLIW instruction and VLIW processor and method for processing VLIW instruction |
US10719325B2 (en) * | 2017-11-07 | 2020-07-21 | Qualcomm Incorporated | System and method of VLIW instruction processing using reduced-width VLIW processor |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
EP0605927A1 (en) * | 1992-12-29 | 1994-07-13 | Koninklijke Philips Electronics N.V. | Improved very long instruction word processor architecture |
US5450556A (en) * | 1990-09-05 | 1995-09-12 | North American Philips Corporation | VLIW processor which uses path information generated by a branch control unit to inhibit operations which are not on a correct path |
WO1996021186A2 (en) * | 1994-12-30 | 1996-07-11 | Philips Electronics N.V. | Plural multiport register file to accommodate data of differing lengths |
US5574939A (en) * | 1993-05-14 | 1996-11-12 | Massachusetts Institute Of Technology | Multiprocessor coupling system with integrated compile and run time scheduling for parallelism |
Family Cites Families (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO1991004536A1 (en) * | 1989-09-20 | 1991-04-04 | Dolphin Server Technology A/S | Instruction cache architecture for parallel issuing of multiple instructions |
US5333280A (en) * | 1990-04-06 | 1994-07-26 | Nec Corporation | Parallel pipelined instruction processing system for very long instruction word |
EP0950946B1 (en) * | 1993-11-05 | 2001-08-16 | Intergraph Corporation | Software scheduled superscaler computer architecture |
US5600810A (en) * | 1994-12-09 | 1997-02-04 | Mitsubishi Electric Information Technology Center America, Inc. | Scaleable very long instruction word processor with parallelism matching |
US5774737A (en) * | 1995-10-13 | 1998-06-30 | Matsushita Electric Industrial Co., Ltd. | Variable word length very long instruction word instruction processor with word length register or instruction number register |
-
1997
- 1997-12-29 US US08/998,486 patent/US5974537A/en not_active Expired - Lifetime
-
1998
- 1998-10-01 WO PCT/IB1998/001521 patent/WO1999034282A1/en active IP Right Grant
- 1998-10-01 DE DE69835425T patent/DE69835425T2/en not_active Expired - Fee Related
- 1998-10-01 EP EP98944151A patent/EP0976032B1/en not_active Expired - Lifetime
- 1998-10-01 KR KR1019997007912A patent/KR100606397B1/en not_active IP Right Cessation
- 1998-10-01 JP JP53467499A patent/JP3829166B2/en not_active Expired - Lifetime
- 1998-11-06 TW TW087118523A patent/TW405094B/en not_active IP Right Cessation
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US5450556A (en) * | 1990-09-05 | 1995-09-12 | North American Philips Corporation | VLIW processor which uses path information generated by a branch control unit to inhibit operations which are not on a correct path |
EP0605927A1 (en) * | 1992-12-29 | 1994-07-13 | Koninklijke Philips Electronics N.V. | Improved very long instruction word processor architecture |
US5574939A (en) * | 1993-05-14 | 1996-11-12 | Massachusetts Institute Of Technology | Multiprocessor coupling system with integrated compile and run time scheduling for parallelism |
WO1996021186A2 (en) * | 1994-12-30 | 1996-07-11 | Philips Electronics N.V. | Plural multiport register file to accommodate data of differing lengths |
Non-Patent Citations (2)
Title |
---|
IEEE COMPUT. SOC. PRESS, Sept. 1994, (USA), STEVEN F.L., STEVEN G.B., LIANG WANG, "System Architecture and Integration". * |
PATENT ABSTRACTS OF JAPAN; & JP 9091141 A (TOSHIBA KK) 4 April 1997. * |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2004511043A (en) * | 2000-10-05 | 2004-04-08 | コーニンクレッカ フィリップス エレクトロニクス エヌ ヴィ | Retargetable compilation system and method |
Also Published As
Publication number | Publication date |
---|---|
EP0976032B1 (en) | 2006-08-02 |
TW405094B (en) | 2000-09-11 |
EP0976032A1 (en) | 2000-02-02 |
JP3829166B2 (en) | 2006-10-04 |
DE69835425D1 (en) | 2006-09-14 |
DE69835425T2 (en) | 2007-03-08 |
US5974537A (en) | 1999-10-26 |
JP2001515628A (en) | 2001-09-18 |
KR100606397B1 (en) | 2006-07-28 |
KR20000075837A (en) | 2000-12-26 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
EP0976032B1 (en) | Improved instruction dispatch mechanism for a guarded vliw architecture | |
US5129067A (en) | Multiple instruction decoder for minimizing register port requirements | |
US7028170B2 (en) | Processing architecture having a compare capability | |
US6892293B2 (en) | VLIW processor and method therefor | |
KR100190738B1 (en) | Parallel processing system and method using surrogate instructions | |
US7124160B2 (en) | Processing architecture having parallel arithmetic capability | |
KR100715055B1 (en) | Vliw processor processes commands of different widths | |
US5564056A (en) | Method and apparatus for zero extension and bit shifting to preserve register parameters in a microprocessor utilizing register renaming | |
JP2500036B2 (en) | Hybrid pipeline connection processor and processing method thereof | |
US6301653B1 (en) | Processor containing data path units with forwarding paths between two data path units and a unique configuration or register blocks | |
EP0427245B1 (en) | Data processor capable of simultaneously executing two instructions | |
EP1124181B1 (en) | Data processing apparatus | |
US20060095721A1 (en) | Tightly coupled accelerator | |
US5805850A (en) | Very long instruction word (VLIW) computer having efficient instruction code format | |
EP1261914B1 (en) | Processing architecture having an array bounds check capability | |
US20010042187A1 (en) | Variable issue-width vliw processor | |
EP0982655A2 (en) | Data processing unit and method for executing instructions of variable lengths | |
US7558816B2 (en) | Methods and apparatus for performing pixel average operations | |
US5768553A (en) | Microprocessor using an instruction field to define DSP instructions | |
US7996655B2 (en) | Multiport execution target delay queue FIFO array | |
US11256516B2 (en) | Processing system with a main processor pipeline and a co-processor pipeline | |
John et al. | Improving the parallelism and concurrency in decoupled architectures |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
AK | Designated states |
Kind code of ref document: A1 Designated state(s): JP KR |
|
AL | Designated countries for regional patents |
Kind code of ref document: A1 Designated state(s): AT BE CH CY DE DK ES FI FR GB GR IE IT LU MC NL PT SE |
|
WWE | Wipo information: entry into national phase |
Ref document number: 1998944151 Country of ref document: EP |
|
ENP | Entry into the national phase |
Ref country code: JP Ref document number: 1999 534674 Kind code of ref document: A Format of ref document f/p: F |
|
WWE | Wipo information: entry into national phase |
Ref document number: 1019997007912 Country of ref document: KR |
|
121 | Ep: the epo has been informed by wipo that ep was designated in this application | ||
WWP | Wipo information: published in national office |
Ref document number: 1998944151 Country of ref document: EP |
|
WWP | Wipo information: published in national office |
Ref document number: 1019997007912 Country of ref document: KR |
|
WWG | Wipo information: grant in national office |
Ref document number: 1019997007912 Country of ref document: KR |
|
WWG | Wipo information: grant in national office |
Ref document number: 1998944151 Country of ref document: EP |