Qualcomm Incorporated

Moore Threads Technology Co., Ltd.

Jing Wu

Ying Duan

Andrew Evan Gruber

Shih-Wei Chou

Zhi Zhu

Abhay S Dixit

Harry Huy Dang

Pramod Vasant Argade

Yun Du

Chun Yu

Guofang Jiao

Lin Chen

Chihong Zhang

Eric Demers

Zilin Ying

Yasser Ahmed

Hongjiang Shang

Liang Han

Suolong Dong

Zhiwei Tang

Systems and methods to dynamically adjust operating conditions of a graphics processing unit (GPU) are disclosed. A machine learning model is trained to determine operating voltages and frequencies to be provided to a GPU core of the GPU to execute a workload comprising a plurality of commands. The trained machine learning model is deployed to firmware of the GPU. A command in the workload to be executed by the GPU core is received. The trained machine learning model determines operating voltage and frequency for the GPU core to execute the command.

Machine learning based power and performance optimization system and method for graphics processing units

Methods, apparatus, and systems for communication over a multi-wire, multi-phase interface are disclosed. A clock recovery method includes generating a combination signal that includes transition pulses, each transition pulse being generated responsive to a transition in a difference signal representative of a difference in signaling state of a pair of wires in a three-wire bus. The combination signal is provided to a logic circuit that is configured to provide a clock signal as its output, where pulses in the combination signal cause the clock signal to be driven to a first state. The logic circuit receives a reset signal that is derived from the clock signal by delaying transitions to the first state while passing transitions from the first state without added delay. The clock signal is driven from the first state after passing a transition of the clock signal to the first state.

Small loop delay clock and data recovery block for high-speed next generation C-PHY

A receiver including a first differential sense amplifier configured to amplify an input differential data signal to generate an output differential data signal; a first set of one or more differential decision feedback equalizer (DFE) taps configured to modify the output differential data signal based on a set of one or more differential tap signals, wherein the first set of one or more differential DFE taps affect an output common mode voltage associated with the output differential data signal; and a compensation circuit configured to adjusts the output common mode voltage to compensate for the effect on the output common mode voltage by the set of one or more differential DFE taps. The compensation circuit includes reference and replica receivers to generate reference and replica output common mode voltages, and a feedback circuit to adjust the output common mode voltage based on the reference and replica output common mode voltages.

Compensation of common mode voltage drop of sensing amplifier output due to decision feedback equalizer (DFE) taps

Certain aspects of the present disclosure generally relate to techniques and apparatus for doubling the frequency of a signal. For example, certain aspects are directed to a phase frequency detector (PFD)-based rising-edge-delay-only frequency doubling circuit. One example frequency doubler circuit generally includes a first delay stage, a second delay stage, a first PFD, a first rising-edge-only adjustable delay cell, a second PFD, a second rising-edge-only adjustable delay cell a logic gate, and a comparator configured to compare a direct-current (DC) voltage value of an output of the logic gate with a reference voltage and control the first and second rising-edge-only adjustable delay cells based on the comparison.

Frequency doubler based on phase frequency detectors using rising edge delay

A method for processing data in a graphics processing unit including receiving an indication that all threads of a warp in a graphics processing unit (GPU) are to execute a same branch in a first set of instructions, storing one or more predicate bits in a memory as a single set of predicate bits, wherein the single set of predicate bits applies to all of the threads in the warp, and executing a portion of the first set of instructions in accordance with the single set of predicate bits. Executing the first set of instructions may include executing the first set of instruction in accordance with the single set of predicate bits using a single instruction, multiple data (SIMD) processing core and/or executing the first set of instruction in accordance with the single set of predicate bits using a scalar processing unit.

Uniform predicates in shaders for graphics processing units

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

Methods, apparatus, and systems for data communication over a multi-wire, multi-phase interface are disclosed. A method of calibration includes configuring a 3-phase signal to include a high frequency component and a low frequency component during a calibration period, and transmitting a version of the 3-phase signal on each wire of a 3-wire interface. The version of the 3-phase signal transmitted on each wire is out-of-phase with the versions of the 3-phase signal transmitted on each of the other wires of the 3-wire interface. The 3-phase signal may be configured to enable a receiver to determine certain operating parameters of the 3-wire interface.

C-PHY training pattern for adaptive equalization, adaptive edge tracking and delay calibration

Methods, apparatus, and systems for data communication over a multi-wire, multi-phase interface are disclosed. A method for calibrating a clock recovery circuit includes recovering a first clock signal from transitions between pairs of symbols representative of successive signaling states of a 3-wire interface, where each pair of symbols includes a first symbol and a second symbol, generating a second clock signal by delaying the first clock signal by a first delay value, generating a third clock signal by delaying the second clock signal, calibrating the second clock signal and the third clock signal by initializing the first delay value such that the first sampling circuit, the second sampling circuit and the third sampling circuit capture the same symbol in a first pair of symbols, and incrementally increasing the first delay value until the second sampling circuit and the third sampling circuit capture different symbols from each pair of symbols.

Growing community of inventors

San Diego, CA, United States of America

Jing Wu