International Business Machines Corporation

Sony Computer Entertainment Inc.

Kabushiki Kaisha Toshiba

Takeshi Yamazaki

Charles Ray Johns

Michael Norman Day

Thuong Quang Truong

Peichum Peter Liu

Harm Peter Hofstee

James Allan Kahle

Matthew Edward King

Peichun Peter Liu

David James Shippy

Michael K Gschwind

Shigehiro Asano

Jeffrey Douglas Brown

Ravi Nair

Masakazu Suzuoki

Sumedh Wasudeo Sathaye

Atsushi Kunimatsu

David Mui

Nobuo Sasaki

Scott Douglas Clark

John-David Wellman

Erik Richter Altman

Peter George Capek

Asano Shigehiro

Martin E Hopkins

An apparatus and a computer program product are provided for completing a plurality of (direct memory access) DMA commands in a computer system. It is determined whether the DMA commands are chained together as a list DMA command. Upon a determination that the DMA commands are chained together as a list DMA command, it is also determined whether a current list element of the list DMA command is fenced. Upon a determination that the current list element is not fenced, a next list element is fetched and processed before the current list element has been completed.

Apparatus, computer program product, and system for completing a plurality of chained list DMA commands that include a fenced list DMA command element

The present invention provides a method and apparatus for creating memory barriers in a Direct Memory Access (DMA) device. A memory barrier command is received and a memory command is received. The memory command is executed based on the memory barrier command. A bus operation is initiated based on the memory barrier command. A bus operation acknowledgment is received based on the bus operation. The memory barrier command is executed based on the bus operation acknowledgment. In a particular aspect, memory barrier commands are direct memory access sync (dmasync) and direct memory access enforce in-order execution of input/output (dmaeieio) commands.

Memory barriers primitives in an asymmetric heterogeneous multiprocessor environment

The present invention provides for a system comprising a DMA queue configured to receive a DMA command comprising a tag, wherein the tag belongs to one of a plurality of tag groups. A counter couples to the DMA queue and is configured to increment a tag group count of the tag group to which the tag belongs upon receipt of the DMA command by the DMA queue and to decrement the tag group count upon execution of the DMA command. A tag group count status register couples to the counter and is configured to store the tag group count for each of the plurality of tag groups. And the tag group count status register is further configured to receive a request for a tag group status and to respond to the request for the tag group status.

System for asynchronous DMA command completion notification wherein the DMA command comprising a tag belongs to a plurality of tag groups

A method, an apparatus, and a computer program product are provided for completing a plurality of (direct memory access) DMA commands in a computer system. It is determined whether the DMA commands are chained together as a list DMA command. Upon a determination that the DMA commands are chained together as a list DMA command, it is also determined whether a current list element of the list DMA command is fenced. Upon a determination that the current list element is not fenced, a next list element is fetched and processed before the current list element has been completed.

Method for completing a plurality of chained list DMA commands that include a fenced list DMA command element

A DMA controller (DMAC) for handling a list DMA command in a computer system is provided. The computer system has at least one processor and a system memory, the list DMA command relates to an effective address (EA) of the system memory, and the at least one processor has a local storage. The DMAC includes a DMA command queue (DMAQ) coupled to the local storage and configured to receive the list DMA command from the local storage and to enqueue the list DMA command. An issue logic is coupled to the DMAQ and configured to issue an issue request to the DMAQ. A request interface logic (RIL) is coupled to the DMAQ and configured to read the list DMA command based on the issue request. The RIL is further coupled to the local storage and configured to send a fetch request to the local storage to initiate a fetch of a list element of the list DMA command from the local storage to the DMAQ. Each list element comprises a stall bit indicating whether the list element is fenced and a DMA completion logic (DCL) is coupled to the at least one processor, the issue logic, and the RIL, and configured to indicate completion of all outstanding bus requests relating to the list element.

Non-fenced list DMA command mechanism

A method, an apparatus, and a computer program are provided for controlling memory access. Direct Memory Access (DMA) units have become commonplace in a number of bus architectures. However, managing limited system resources has become a challenge with multiple DMA units. In order to mange the multitude of commands generated and preserve dependencies, embedded flags in commands or a barrier command are used. These operations then can control the order in which commands are executed so as to preserve dependencies.

Establishing command order in an out of order DMA command queue

A system and method are provided for setting up a direct memory access for a first processor. The system includes the first processor and a local memory. The local memory is coupled to the first processor. A first direct memory access controller (DMAC) is coupled to the first processor and the local memory. A system memory is in communication with the first DMAC. A second processor is in communication with the first DMAC such that the second processor sets up the first DMAC to handle data transfer between the local memory and the system memory. The second processor is interrupted when the first DMAC finishes handling the data transfer.

Proxy direct memory access

A method and an apparatus are provided for handling a list DMA command in a computer system. The list DMA command relates to an effective address (EA) of a system memory. At least one processor in the system has a local storage. The list DMA command is queued in a DMA queue (DMAQ). A list element is fetched from the local storage to the DMAQ. The list DMA command is read from the DMAQ. A bus request is issued for the list element. If the bus request is a last request, it is determined whether a current list element is a last list element. If the current list element is not the last list element, it is determined whether the current list element is fenced. If the current list element is not fenced, a next list element is fetched regardless of whether all outstanding requests are completed.

The present invention provides for asynchronous DMA command completion notification in a computer system. A command tag, associated with a plurality DMA command is generated. A DMA data movement command having the command tag is grouped with another DMA data movement command having the command tag. DMA commands belonging to the same tag group are monitored to see whether all DMA commands of the same tag group are completed.

System and method asynchronous DMA command completion notification by accessing register via attached processing unit to determine progress of DMA command

The present invention provides a system for managing cache replacement eligibility. A first address register is configured to request an address from an L1 cache. An L1 cache is configured to determine whether a requested address is in the L1 cache and, in response to a determination that a requested address is not in the L1 cache, is further configured to transmit the requested address to a range register coupled to the L1 cache. The range register is configured to generate a class identifier in response to a received requested address and to transmit the requested address and class identifier to a replacement management table coupled to the range register. The replacement management table is configured to generate L2 tag replacement control indicia in response to a received requested address and class identifier. An L2 address register is coupled to the first address register and configured to request an address from an L2 cache. An L2 cache is coupled to the L2 address register and the replacement management table and is configured to determine whether a requested address is in the L2 cache and is further configured to assign replacement eligibility of at least one set of cache lines in the L2 cache in response to received L2 tag replacement control indicia. In response to a determination that a requested address is not in the L2 cache, the L2 cache is further configured to overwrite a cache line within a set of the L2 cache as a function of the replacement eligibility.

Software-controlled cache set management

The present invention provides for selectively overwriting sets of a cache as a function of a replacement management table and a least recently used function. A class identifier is created as a function of an address miss. A replacement management table is employable to read the class identifier to create a tag replacement control indicia. The cache, comprising a plurality of sets, is employable to disable the replacement of at least one of the plurality of sets as a function of the tag replacement control indicia.

Software-controlled cache set management with software-generated class identifiers

The present invention discloses, in one aspect, a microprocessor. In one embodiment, the microprocessor includes a processing element configured to process an application using a bandwidth. The microprocessor also includes an access shaper coupled to the processing element and configured to shape storage requests for the processing of the application. In this embodiment, the microprocessor further includes bandwidth management circuitry coupled to the access shaper and configured to track the bandwidth usage based on the requests. A method of coordinating bandwidth allocation and a processor assembly are also disclosed.

Microprocessor having bandwidth management for computing applications and related method of managing bandwidth allocation

An improved method and apparatus for resource arbitration. Four priority classes, managed high (MH), managed low (ML), opportunistic high (OH) and opportunistic low (OL), are defined. A priority class is assigned to each resource access request. An access request concentrator (ARC) is created for each resource, through which the resource is accessed. An access request is chosen at each ARC using the priority order MH, ML, OH, and OL, in decreasing order of priority. If OH priority class resource access requests are locked out, the priority order is temporarily changed to OH, OL, MH, and ML, in decreasing order of priority. If OL priority class resource access requests are locked out, the priority order is temporarily changed to MH, OL, OH, and ML, in decreasing order of priority.

Method of resource arbitration

A method and system for optimizing the processing of graphics is disclosed. The system may comprise at least one geometry processor and at least one graphics processor. A communication channel permits communication between the geometry and graphics processors. A control processor may communicate with the geometry and graphics processor through the communications channel. A method of processing graphics data in a computer system is provided to determine whether the geometry and graphics processors are being efficiently utilized. If necessary, one or more of the geometry and graphics processors are selectively assigned or unassigned to improve the efficiency of the graphics processing circuitry in performing the graphics task.

System and method of optimizing graphics processing

A method and system for attached processing units accessing a shared memory in an SMP system. In one embodiment, a system comprises a shared memory. The system further comprises a plurality of processing elements coupled to the shared memory. Each of the plurality of processing elements comprises a processing unit, a direct memory access controller and a plurality of attached processing units. Each direct memory access controller comprises an address translation mechanism thereby enabling each associated attached processing unit to access the shared memory in a restricted manner without an address translation mechanism. Each attached processing unit is configured to issue a request to an associated direct memory access controller to access the shared memory specifying a range of addresses to be accessed as virtual addresses. The associated direct memory access controller is configured to translate the range of virtual addresses into an associated range of physical addresses.

Symmetric multi-processing system with attached processing units being able to access a shared memory without being structurally configured with an address translation mechanism

Growing community of inventors

Austin, TX, United States of America

Takeshi Yamazaki