Hitachi, Ltd.

Hitachi Microcomputer Engineering, Ltd.

Other

Hitachi Engineering Co., Ltd.

Makoto Hanawa

Tadahiko Nishimukai

Kunio Uchiyama

Atsushi Hasegawa

Ikuya Kawasaki

Kenji Kaneko

Kentaro Shimada

Osamu Nishii

Makoto Suzuki

Hiroaki Nambu

Noboru Masuda

Kazuo Kanetani

Takeshi Kusunoki

Kaname Yamasaki

Kazumichi Yamamoto

Katsuhiro Shimohigashi

Susumu Narita

Takuichiro Nakazawa

Yoshio Miki

Takayuki Kawahara

Masakazu Aoki

Yoshinobu Nakagome

Goro Kitsukawa

Shigeki Morinaga

Hiroshi Takeda

Masayuki M Nakamura

Takehisa Hayashi

Kouki Noguchi

Kanji Oishi

Hisayuki Higuchi

Tetsuhiko Okada

Kazuhiko Iwasaki

Kazunori Nakajima

Suguru Tachibana

Junichi Tatezaki

Hisashi Kaziwara

Tatsuya Kawashimo

Takeshi Asai

Takashi Inagawa

Noriyasu Ido

Osamu Shinbo

A data processor formed on a LSI chip has an instruction address generator, an instruction cache memory having entries each storing an instruction address and an instruction corresponding to the instruction address, an instruction decoder decoding an instruction from said cache memory corresponding to an instruction address from said instruction address generator, an operand address generator generating an operand address in response to an output signal of said instruction decoder, and an operand cache memory having entries each storing an operand address and operand data corresponding to the operand address in its entry. The data processor executes an instruction that makes entries in both of said instruction cache memory and said operand cache memory ineffective.

Data processor capable of executing an instruction that makes a cache memory ineffective

A clocked logic gate circuit is constituted so that a switch unit is constituted by a logic block and a reference MOS transistor, the source of the reference MOS transistor is connected to one output of the logic block, the gate of the reference MOS transistor is connected to the other output of the logic block, and MOS transistors (input transistors) constituting the logic block are connected in parallel. With this arrangement, complementary inputs are not required and a driving MOS transistor and an input transistor (or a driving MOS transistor and a reference MOS transistor) can be connected in series. As a result, a circuit is obtained which is simpler than the double rail logic in constitution is facilitated and can be operated at a higher speed than a CMOS logic circuit and a path transistor logic circuit.

Clocked logic gate circuit

In a cache memory of a super-scalar or VLIW processor to concurrently process a plurality of memory accesses, to provide a memory capable of multi-port access operation, there is provided a unit which subdivides the cache memory into a plurality of memory banks for concurrent operations thereof and which allocates memory ports independently to the respective memory banks. In a first cycle, the first and second memory ports are allocated to the first and second memory banks, respectively. If a hit occurs, the plural accesses are completed in one cycle. If a miss results, the first and second memory ports are allocated respectively to the second and first memory banks in a second cycle.

Multi-port memory and a data processor accessing the same

A data processor for executing instructions using operand data stored in a main memory includes an instruction control unit having a first associative memory storing instructions read out from the main. The data processor also includes an instruction controller reading out an instruction from the first associative memory when the instruction is present in the first associative memory and reading an instruction from the main memory when the instruction is not present in the first associative memory. The controller also has an output the instruction to be executed. An instruction execution unit has a second associative memory storing operand data read out from the main memory. An instruction executor executes the instruction by using operand data read out from the second associative memory when the operand data is present in the second associative memory and from the main memory when the operand data is not present in the second associative memory.

Data processor

A multiprocessor system of the present invention has an address bus, a data bus, first and second processors, four access queues, first and second arbiters, and a shared memory divided into four banks. The four access queues are constituted by first-in first-out memories for buffering a plurality of access-request addresses transmitted through the address bus. When a processor requires data from the memory bank, the processor sends a processor ID with a data access request. When the memory bank sends data in return, the memory bank outputs the processor ID of the request originator with the required data. Even if continuous access requests are addressed to one bank of the shared memory, a succeeding access requested need not wait for a previous access request to be finished. According, the throughput of the system can be improved greatly. The first and second arbiters serve to decide ownership of buses.

Multiprocessor system having distinct data bus and address bus arbiters

A method of effecting a branch operation without the need of instruction fetching is carried out according to the taken/untaken branch with respect to a program containing plural branch instructions, and the method is performed by an apparatus information processing. By detecting a branch instruction stored in an instruction buffer, determining its branch distance and branch condition, and if the branch distance is less than a predetermined positive distance, by then providing that branch condition as an execution condition for the instruction located within said predetermined distance to store in an instruction register, a series of instructions succeeding that branch instruction can be processed into a conditional instruction in the apparatus. The instructions may be continuously executed without refetching instructions, in both cases that a branch condition is taken and untaken. Also, the penalty of miss-prediction of a branch will be minimized.

Branch operation system where instructions are queued until preparations

A data processor for executing instructions using operand data stored in a main memory includes an instruction control unit having a first associative memory storing instructions read out from the main memory. The data processor also includes an instruction controller reading out an instruction from the first associative memory when the instruction is present in the first associative memory and reading an instruction from the main memory when the instruction is not present in the first associative memory. The controller also has as an output instruction to be executed. An instruction execution unit has a second associative memory storing operand data read out from the main memory. An instruction executioner executes the instruction by using operand data read out from the second associative memory when the operand data is present in the second associative memory and from the main memory when the operand data is not present in the second associative memory.

Soft errors generated at an active time are reduced by adding a small-scale circuit to a high performance LSI, such as a processor without reducing the performance of the circuit. The processor has individual logic circuits each having a plurality of stages of logic gates for outputting true signals and complement signals for the individual logic gates. A latch circuit latches the true and complement signals of the logic circuits separately and a compare circuit detects for an error by comparing the true and complement output signals of the logic circuits to determine if they are at the same logical signal level or not, just upstream of the latch in which the individual true and complement output signals of the final logic circuit stages are individually latched. When the compare circuit detects an error because the true and complement output signals are at the same logical signal level, a recovery process is executed.

Logic circuit having error detection function and processor including

To improve the efficiency of instruction branch operations, particularly in a pipeline processor, a branch reservation instruction is generated during program compile. The system includes a processor having a branch address stack 10 for storing pairs of branch point addresses and branch target addresses, a program counter (PC) 12 which holds a current instruction fetch address, a comparator 11 which compares the branch point address of the most recently entered pair stored in the stack 10 and the value of the PC 12, and a selector 14 which, when the result of comparison shows the coincidence, switches the instruction fetch address from the value of the PC 12 to the branch target address of the most recently entered pair stored in the stack 10. The branch operation is performed by reserving a pair of a branch target address and a branch point address in the stack 10 by a branch reservation instruction in advance of the branch point in the instruction sequence, and switching the next instruction fetch address to the branch target address promptly when the instruction fetch reaches the branch point.

Instruction branching method and a processor

A data processor for executing instructions using operand data stored in a main memory includes an instruction control unit having a first associative memory storing instructions read out from the main memory. The data processor also includes an instruction controller reading out an instruction from the first associative memory when the instructioon is present in the first associative memory and reading an instruction from the main memory when the instruction is not present in the first associative memory. The controller also has as an output the instruction to be executed. An instruction execution unit has a second associative memory storing operand data read out from the main memory. An instruction executioner executes the instruction by using operand data read out from the second associative memory when the operand data is present in the second associative memory and from the main memory when the operand data is not present in the second associative memory.

Purge control for ON-chip cache memory

System with reservation instruction execution to store branch target

A multiprocessor system includes an address bus 170, a data bus 180, processors 110 and 120, access queues 135 and 145, shared memories 130 and 140, and lock control circuits 500 and 510. Particularly, a lock-in indicative flag register 501 is provided in the lock control circuit 500. While an operand cache 112 in one processor 110 is making a lock access to a predetermined address of the shared memory 130, the flag register 501 is set on the basis of a lock command signal 260 so that an access of an instruction cache 122 in another processor 120 to the predetermined address of the shared memory 130 is prohibited but an access to a different address is permitted at the time of the lock access. After the lock access is released, the lock control circuit 500 accepts an access to the predetermined address.

Multiprocessor system having a processor invalidating operand cache when

A data processor for executing instructions using operand data stored in a main memory includes an instruction control unit having a first associative memory storing instructions read out from the main memory. The data processor also includes an instruction controller reading out an instruction from the first associative memory when the instruction is present in the first associative memory and reading an instruction from the main memory when the instruction is not present in the first associative memory. The controller also has as an output; and an instruction execution unit has a second associative memory storing operand data read out from the main memory. An instruction executioner executes the instruction by using operand data read out from the second associative memory when the operand data is present in the second associative memory and from the main memory when the operand data is not present in the second associative memory.

Data processor with on-chip cache memory and purge controller responsive

Growing community of inventors

Kokubunji, Japan

Makoto Hanawa