Silicon Graphics, Incorporated

Mips Technologies, Inc.

Other

Nintendo Co., Ltd.

Nvidia Corporation

Silicon Graphics Computer Systems, Inc.

Peter Y Hsu

William A Huffman

Henry Packard Moreton

Timothy J Van Hook

Earl A Killian

Paul Kingsley Rodman

Joseph T Scanlon

Chandra Shekhar Joshi

Monica R Nofal

Joseph P Bratt

John Brennan

Man K Tang

Steve J Ciavaglia

Howard H Cheng

Peter Anthony Sandon

Yu-Chung C Liao

Manfred R Kuehnle

Monier Maher

Sanjay J Patel

Egon Matijevic

Christopher Frederick Lamb

Steve Ciavagia

John Brennen

Steven J Ciavaglia

A system for performing data-parallel operations and task-parallel operations. A first switch fabric node (SFN) includes first and second lane processing engines (LPEs). The first LPE includes a first set of lane processing units (LPUs) configured to perform data-parallel operations, where each LPU performs a set of operations, and each LPU uses a different set of data for the set of operations, and each LPU within the first set of LPUs uses a different set of data for the set of operations. The second LPE includes a second set of LPUs configured to perform task-parallel operations, where each LPU performs a different set of operations. A processing control engine (PCE) is configured to distribute instructions and data to the first LPE and the second LPE. Advantageously, data parallel operations and task parallel operations are able to be performed on the same processor simultaneously.

System of lanes of processing units receiving instructions via shared memory units for data-parallel or task-parallel operations

The present invention provides extended precision in SIMD arithmetic operations in a processor having a register file and an accumulator. A first set of data elements and a second set of data elements are loaded into first and second vector registers, respectively. Each data element comprises N bits. Next, an arithmetic instruction is fetched from memory. The arithmetic instruction is decoded. Then, the first vector register and the second vector register are read from the register file. The present invention executes the arithmetic instruction on corresponding data elements in the first and second vector registers. The resulting element of the execution is then written into the accumulator. Then, the resulting element is transformed into an N-bit width element and written into a third register for further operation or storage in memory. The transformation of the resulting element can include, for example, rounding, clamping, and/or shifting the element.

Providing extended precision in SIMD vector arithmetic operations

The present invention provides alignment and ordering of vector elements for SIMD processing. In the alignment of vector elements for SIMD processing, one vector is loaded from a memory unit into a first register and another vector is loaded from the memory unit into a second register. The first vector contains a first byte of an aligned vector to be generated. Then, a starting byte specifying the first byte of an aligned vector is determined. Next, a vector is extracted from the first register and the second register beginning from the first bit in the first byte of the first register continuing through the bits in the second register. Finally, the extracted vector is replicated into a third register such that the third register contains a plurality of elements aligned for SIMD processing. In the ordering of vector elements for SIMD processing, a first vector is loaded from a memory unit into a first register and a second vector is loaded from the memory unit into a second register. Then, a subset of elements are selected from the first register and the second register. The elements from the subset are then replicated into the elements in the third register in a particular order suitable for subsequent SIMD vector processing.

Alignment and ordering of vector elements for single instruction multiple data processing

The present invention provides extended precision in SIMD arithmetic operations in a processor having a register file and an accumulator. A first set of data elements and a second set of data elements are loaded into a first vector register and a second vector register, respectively. Each data element comprises N bits. Next, an arithmetic instruction is fetched from memory. The arithmetic instruction is decoded. Then, a first vector register and a second vector register are read from the register file. The present invention then executes the arithmetic instruction on corresponding data elements in the first and second vector registers. The result of the execution is then written into the accumulator. Then, each element in the accumulator is transformed into an N-bit width element and stored into the memory.

Method for providing extended precision in SIMD vector arithmetic operations

A microprocessor including a control unit and a cache connected with the control unit for storing data to be used by the control, wherein the cache is selectively configurable as either a single cache or as a partitioned cache having a locked cache portion and a normal cache portion. The normal cache portion is controlled by a hardware implemented automatic replacement process. The locked cache portion is locked so that the automatic replacement process cannot modify the contents of the locked cache. An instruction is provided in the instruction set that enables software to selectively allocate lines in the locked cache portion to correspond to locations in an external memory, thereby enabling the locked cache portion to be completely managed by software.

Method and apparatus for software management of on-chip cache

A computing system has first and second instruction storing circuits, each instruction storing circuit storing N instructions for parallel output. An instruction dispatch circuit, coupled to the first instruction storing circuit dispatches L instructions stored in the first instruction storing circuit, wherein L is less than or equal to N. An instruction loading circuit, coupled to the instruction dispatch circuit and to the first and second instruction storing circuits, loads L instructions from the second instruction storing circuit into the first instruction storing circuit after the L instructions are dispatched from the first instruction storing circuit and before further instructions are dispatched from the first instruction storing circuit. The instruction loading circuit loads the L instructions from the second instruction storing circuit into the positions previously occupied by the L instructions dispatched from the first instruction storing circuit. A feedback path is also provided to reload an instruction not previously dispatched.

Loading previously dispatched slots in multiple instruction dispatch buffer before dispatching remaining slots for parallel execution

A computing system contains an apparatus having an instruction memory to store a plurality of lines of a plurality of instructions, and a branch memory to store a plurality of branch prediction entries, each branch prediction entry containing information for predicting whether a branch designated by a branch instruction stored in the instruction memory will be taken when the branch instruction is executed. Each branch prediction entry includes a branch target field for indicating a target address of a line containing a target instruction to be executed if the branch is taken, a destination field indicating where the target instruction is located within the line indicated by the branch target address, and a source field indicating where the branch instruction is located within the line corresponding to the target address. A counter stores an address value used for addressing the instruction memory, and an incrementing circuit increments the address value in the counter for sequentially addressing the lines in the instruction memory during normal sequential operation. A counter loading circuit loads the target address into the counter when the branch prediction entry predicts the branch designated by the branch instruction stored in the instruction memory will be taken when the branch instruction is executed, causing the line containing the target instruction to be fetched and entered into the pipeline immediately after the line containing the branch instruction. An invalidate circuit invalidates any instructions following the branch instruction in the line containing the branch instruction and prior to the target instruction in the line containing the target instruction.

Branch prediction entry with target line index calculated using relative position of second operation of two step branch operation in a line of instructions

A computing system that contains an apparatus comprising an instruction memory to store a plurality of lines of a plurality of instructions, and a branch memory to store a plurality of branch prediction entries, each branch prediction entry containing information for predicting whether a branch designated by a branch instruction stored in the instruction memory will be taken when the branch instruction is executed. Each branch prediction entry includes a branch target field for indicating a target address of a line containing a target instruction to be executed if the branch is taken, a destination field indicating where the target instruction is located within the line indicated by the branch target address, and a source field indicating where the branch instruction is located within the line corresponding to the target address. A counter stores an address value used for addressing the instruction memory, and an incrementing circuit increments the address value in the counter for sequentially addressing the lines in the instruction memory during normal sequential operation. A counter loading circuit loads the target address into the counter when the branch prediction entry predicts the branch designated by the branch instruction stored in the instruction memory will be taken when the branch instruction is executed, causing the line containing the target instruction to be fetched and entered into the pipeline immediately after the line containing the branch instruction. An invalidate circuit invalidates any instructions following the branch instruction in the line containing the branch instruction and prior to the target instruction in the line containing the target instruction.

Invalidating instructions in fetched instruction blocks upon predicted

Alignment and ordering of vector elements for single instruction

Method for providing extended precision in SIMD vector arithmetic

A conflict resolution system for interleaved memories in processors capable of issuing multiple independent memory operations per cycle. The conflict resolution system includes an address bellow for temporarily storing memory requests, and cross-connect switches to variously route multiple parallel memory requests to multiple memory banks. A control logic block controls the address bellow and the cross-connect switches to reorder the sequence of memory requests to avoid conflicts. The reordering removes conflicts and increases the occurrence of alternating memory requests that can issue simultaneously.

Conflict resolution in interleaved memory systems with multiple parallel

A method for preventing deadlock due to the need for data exclusivity when performing forced atomic instructions in a multi-level cache in a multi-processor system. The system determines whether an aligned multi-byte word in which the data of a forced atomic instruction, such as an integer store operation, is exclusive in a first level cache. If so, the forced atomic instruction is allowed to enter a second level cache pipeline. If not, the forced atomic instruction is prevented from entering the second level cache pipeline and a cache miss and fill operation is initiated to cause the aligned word to be exclusive in the first level cache.

Method for preventing multi-level cache system deadlock in a

A computing system has first and second instruction storing circuits, each instruction storing circuit storing N instructions for parallel output. An instruction dispatch circuit, coupled to the first instruction storing circuit dispatches L instructions stored in the first instruction storing circuit, wherein L is less than or equal to N. An instruction loading circuit, coupled to the instruction dispatch circuit and to the first and second instruction storing circuits, loads L instructions from the second instruction storing circuit into the first instruction storing circuit after the L instructions are dispatched from the first instruction storing circuit and before further instructions are dispatched from the first instruction storing circuit. A bypass circuit for bypassing the second instruction storing circuit is also provided.

Growing community of inventors

Fremont, CA, United States of America

Peter Y Hsu