Imagination Technologies Limited

Simon Thomas Nield

Luca O Iuliano

Yoong-Chert Foo

Adam De Grasse

Ollie Mower

Thomas Michael Rose

James McCarthy

Jonathan Mark Redshaw

James Mccarthy

Methods and parallel processing units for avoiding inter-pipeline data hazards identified at compile time. For each identified inter-pipeline data hazard the primary instruction and secondary instruction(s) thereof are identified as such and are linked by a counter which is used to track that inter-pipeline data hazard. When a primary instruction is output by the instruction decoder for execution the value of the counter associated therewith is adjusted to indicate that there is hazard related to the primary instruction, and when primary instruction has been resolved by one of multiple parallel processing pipelines the value of the counter associated therewith is adjusted to indicate that the hazard related to the primary instruction has been resolved. When a secondary instruction is output by the decoder for execution, the secondary instruction is stalled in a queue associated with the appropriate instruction pipeline if at least one counter associated with the primary instructions from which it depends indicates that there is a hazard related to the primary instruction.

Methods and systems for inter-pipeline data hazard avoidance

A method of activating scheduling instructions within a parallel processing unit is described. The method comprises decoding, in an instruction decoder, an instruction in a scheduled task in an active state and checking, by an instruction controller, if a swap flag is set in the decoded instruction. If the swap flag in the decoded instruction is set, a scheduler is triggered to de-activate the scheduled task by changing the scheduled task from the active state to a non-active state.

Scheduling tasks using swap flags

Apparatus identifies a set of M output memory addresses from a larger set of N input memory addresses containing at least one non-unique memory address. A comparator block performs comparisons of memory addresses from a set of N input memory addresses to generate a binary classification dataset that identifies a subset of addresses from the set of input addresses, where each address in the subset identified by the binary classification dataset is unique within that subset. Combination logic units receive a predetermined selection of bits of the binary classification dataset and sort its received predetermined selection of bits into an intermediary binary string in which the bits are ordered into a first group identifying addresses belonging to the identified subset, and a second group identifying addresses not belonging to the identified subset. Output generating logic selects between bits belonging to different intermediary binary strings to generate a binary output identifying a set of output memory addresses containing at least one address in the identified subset.

Sorting memory address requests for parallel memory access using input address match masks

A method of scheduling instructions within a parallel processing unit is described. The method comprises decoding, in an instruction decoder, an instruction in a scheduled task in an active state, and checking, by an instruction controller, if an ALU targeted by the decoded instruction is a primary instruction pipeline. If the targeted ALU is a primary instruction pipeline, a list associated with the primary instruction pipeline is checked to determine whether the scheduled task is already included in the list. If the scheduled task is already included in the list, the decoded instruction is sent to the primary instruction pipeline.

Scheduling tasks using targeted pipelines

Methods of encoding and decoding are described which use a variable number of instruction words to encode instructions from an instruction set, such that different instructions within the instruction set may be encoded using different numbers of instruction words. To encode an instruction, the bits within the instruction are re-ordered and formed into instruction words based upon their variance as determined using empirical or simulation data. The bits in the instruction words are compared to corresponding predicted values and some or all of the instruction words that match the predicted values are omitted from the encoded instruction.

Encoding and decoding variable length instructions

A method of activating scheduling instructions within a parallel processing unit includes checking if an ALU targeted by a decoded instruction is full by checking a value of an ALU work fullness counter stored in the instruction controller and associated with the targeted ALU. If the targeted ALU is not full, the decoded instruction is sent to the targeted ALU for execution and the ALU work fullness counter associated with the targeted ALU is updated. If, however, the targeted ALU is full, a scheduler is triggered to de-activate the scheduled task by changing the scheduled task from the active state to a non-active state. When an ALU changes from being full to not being full, the scheduler is triggered to re-activate an oldest scheduled task waiting for the ALU by removing the oldest scheduled task from the non-active state.

Scheduling tasks using work fullness counter

A method of scheduling tasks within a GPU or other highly parallel processing unit is described which is both age-aware and wakeup event driven. Tasks which are received are added to an age-based task queue. Wakeup event bits for task types, or combinations of task types and data groups, are set in response to completion of a task dependency and these wakeup event bits are used to select an oldest task from the queue that satisfies predefined criteria.

Task scheduling in a GPU using wakeup event state data

Queues for inter-pipeline data hazard avoidance

A memory subsystem for use with a single-instruction multiple-data (SIMD) processor comprising a plurality of processing units configured for processing one or more workgroups each comprising a plurality of SIMD tasks, the memory subsystem comprising: a shared memory partitioned into a plurality of memory portions for allocation to tasks that are to be processed by the processor; and a resource allocator configured to, in response to receiving a memory resource request for first memory resources in respect of a first-received task of a workgroup, allocate to the workgroup a block of memory portions sufficient in size for each task of the workgroup to receive memory resources in the block equivalent to the first memory resources.

Allocation of memory resources to SIMD workgroups

A method for a plurality of pipelines, each having a processing element having first and second inputs and first and second lines, wherein at least one of the pipelines includes first and second logic operable to select a respective line so that data is received at the first and second inputs respectively. A first mode is selected and for the at least one pipeline, the first and second lines of that pipeline are selected such that the processing element of that pipeline receives data via the first and second lines of that pipeline, the first line being capable of supplying data that is different to the second line. A second mode is selected and for the at least one pipeline a line of another pipeline is selected, the second line of the at least one pipeline is selected and the same data at the second line is supplied as the first line.

Data selection for a processor pipeline using multiple supply lines

Methods of encoding and decoding are described which use a variable number of instruction words to encode instructions from an instruction set, such that different instructions within the instruction set may be encoded using different numbers of instruction words. To encode an instruction, the bits within the instruction are reordered and formed into instruction words based upon their variance as determined using empirical or simulation data. The bits in the instruction words are compared to corresponding predicted values and some or all of the instruction words that match the predicted values are omitted from the encoded instruction.

Growing community of inventors

Hertfordshire, United Kingdom

Simon Thomas Nield