Qualcomm Incorporated

Hongjiang Shang

Yun Du

Andrew Evan Gruber

Chun Yu

Eric Demers

Fei Wei

Alexei Vladimirovich Bourd

Chihong Zhang

Gang Zhong

Zilin Ying

Elina Kamenetskaya

Minjie Huang

Jing Han

Yibin Zhang

Guofang Jiao

Lin Chen

Jing Wu

Haikun Zhu

Liang Han

The present disclosure relates to methods and apparatus for compute processing. For example, disclosed techniques facilitate improving performance of matrix multiplication in streaming processor. Aspects of the present disclosure can execute, with a load control unit, a first load instruction to load a set of input data of an input matrix from a first memory to a second memory. Aspects of the present disclosure can also execute, with the load control unit, a second load instruction to load a set of weight data of a weight matrix from the first memory to the second memory. Additionally, aspects of the present disclosure can perform, with an ALU component, a matrix multiplication operation using the set of input data and the set of weight data to generate an output matrix. Further, aspects of the present disclosure can store the output matrix at a general purpose register accessible to the ALU component.

Performing matrix multiplication in a streaming processor

Methods and apparatus to perform matrix multiplication in a streaming processor

Example techniques are described for generating graphics content by obtaining texture operation instructions corresponding to a texture operation, in response to determining at least one of insufficient general purpose register space is available for the texture operation or insufficient wave slots are available for the texture operation, generating an indication that the texture operation corresponds to a deferred wave, executing the texture operation, sending, to a texture processor, initial texture sample instructions corresponding to the texture operation that was executed, and receiving texture mapped data corresponding to the initial texture sample instructions.

General purpose register and wave slot allocation in graphics processing

Systems and techniques are disclosed for general purpose register dynamic allocation based on latency associated with of instructions in processor threads. A streaming processor can include a general purpose registers configured to stored data associated with threads, and a thread scheduler configured to receive allocation information for the general purpose registers, the information describing general purpose registers that are to be assigned as persistent general purpose registers (pGPRs) and volatile general purpose registers (vGPRs). The plurality of general purpose registers can be allocated according to the received information. The streaming processor can include the general purpose registers allocated according to the received information, the allocated based on execution latencies of instructions included in the threads.

General purpose register allocation in streaming processor

Techniques are described for determining whether execution of an instruction would require reading more values from a memory cell of a general purpose register (GPR) than a read port of the memory cell would allow. In such a case, the techniques may store, prior to execution of the instruction, one or more values from the memory cell in a separate conflict queue. During execution of the instruction to implement an operation defined by the instruction, one value that is an operand of the operation would be read from the memory cell and another value that is an operand of the operation other would be read from the conflict queue.

Operand conflict resolution for reduced port general purpose register

Growing community of inventors

San Diego, CA, United States of America

Hongjiang Shang