Evandro Menezes - Austin TX, US David F. Tobias - Pflugerville TX, US Richard Russell - Austin TX, US Morrie Altmejd - Austin TX, US
Assignee:
Advanced Micro Devices, Inc. - Sunnyvale CA
International Classification:
G06F 126 G06F 132
US Classification:
713320, 713323
Abstract:
A computer system that has multiple performance states periodically obtains utilization information for a plurality of tasks operating on the processor and determines processor utilization according to the utilization information for the plurality of tasks. The system compares the processor utilization to at least one threshold and selectively adjusts a current processor performance state to another performance state according to the comparison.
Performance And Power Optimization Via Block Oriented Performance Measurement And Control
Morrie Altmejd - Austin TX, US Evandro Menezes - Austin TX, US Dave Tobias - Pflugerville TX, US
Assignee:
Advanced Micro Devices, Inc. - Sunnyvale CA
International Classification:
G06F001/32 G06F001/26
US Classification:
713324, 713320
Abstract:
An integrated circuit includes a plurality of functional blocks. Utilization information for the various functional blocks is generated. Based on that information, the power consumption and thus the performance levels of the functional blocks can be tuned. Thus, when a functional block is heavily loaded by an application, the performance level and thus power consumption of that particular functional block is increased. At the same time, other functional blocks that are not being heavily utilized and thus have lower performance requirements can be kept at a relatively low power consumption level. Thus, power consumption can be reduced overall without unduly impacting performance.
Use Of A Neutral Instruction As A Dependency Indicator For A Set Of Instructions
A dependency instruction encodes dependency information among a group of instructions. A processor decodes the dependency instruction associated with the group of instructions. The processor can then execute the group of instructions in an order based on the dependency information in the dependency instruction. The dependency information may be encoded in a neutral instruction so processors that do not support dependency instructions can execute a program containing them.
Evandro Menezes - Austin TX, US Harsha Jagasia - Austin TX, US Morrie Altmejd - Austin TX, US David Tobias - Pflugerville TX, US
Assignee:
Advanced Micro Devices, Inc. - Sunnyvale CA
International Classification:
G06F 9/44 G06F 17/30 G06F 12/00 G06F 15/16
US Classification:
707 1, 707 10, 707200, 718102, 709201
Abstract:
A task queue management technique leverages infrastructure provided by file and operating systems to manage task queues substantially without otherwise typical problems regarding management of the size of the queue and/or the state of the queue (e. g. , empty or full) while maintaining, regulating and altering the queue and the order of tasks in the queue. Task producer code is configurable to cause one or more executable files corresponding to each of one or more tasks to be stored in a queue directory responsive to receiving the one or more tasks. A file system associates each executable file with a time stamp indicating when the executable file was stored. Task consumer code is configurable to execute the one or more executable files in an order indicated by the time stamps associated with the executable files and to remove each executable file after execution of each executable file.
System And Method For Controlling An Intergrated Circuit To Enter A Predetermined Performance State By Skipping All Intermediate States Based On The Determined Utilization Of The Intergrated Circuit
David F. Tobias - Pflugerville TX, US Evandro Menezes - Austin TX, US Richard Russell - Austin TX, US Morrie Altmejd - Austin TX, US
Assignee:
Advanced Micro Devices, Inc. - Sunnyvale CA
International Classification:
G06F 1/00 G06F 1/32
US Classification:
713300, 713320, 713322
Abstract:
A computer system has multiple performance states. The computer system periodically determines utilization information for the computer system and adjusts the performance state according to the utilization information. If a performance increase is required, the computer system always goes to the maximum performance state. If a performance decrease is required, the computer system steps the performance state down to a next lower performance state.
An information processing system is configured to schedule tasks to a plurality of processors using processor performance information. For example, the maximum performance level of each of the processors, the current performance level of each of the processors, and the number of processors can be used to schedule tasks to one or more of the processors. A task distribution frequency which takes this information into account can be useful. One such task distribution frequency fis calculated so that f=D/ΣDwhere i ranges from 1-N and D=MP/CP/N, where MPis a maximum performance level for the processor i, CPis a current performance level for the processor i, and N is the number of processors. Tasks are distributed according to the task distribution frequency f.
Method And Apparatus For Improving Responsiveness Of A Power Management System In A Computing Device
David F. Tobias - Pflugerville TX, US Evandro Menezes - Austin TX, US Richard Russell - Austin TX, US Morrie Altmejd - Austin TX, US
Assignee:
Advanced Micro Devices, Inc. - Sunnyvale CA
International Classification:
G06F 1/00 G06F 1/26 G06F 1/32
US Classification:
713300, 713320, 713322
Abstract:
A computer system has multiple performance states. The computer system periodically determines utilization information for the computer system and adjusts the performance state according to the utilization information. If a performance increase is required, the computer system always goes to the maximum performance state. If a performance decrease is required, the computer system steps the performance state down to a next lower performance state.
Evandro Menezes - Austin TX, US Dave Tobias - Pflugerville TX, US Morrie Altmejd - Austin TX, US
Assignee:
Advanced Micro Devices, Inc. - Sunnyvale CA
International Classification:
H04L 12/28
US Classification:
370389, 370392, 370401
Abstract:
A transaction is sent over a communication link to a receiving node. The transaction includes data along with a tag identifying a data type of the data. The receiving node either forwards or processes the data according to whether the data type matches the type of data type processed by circuitry associated with the receiving node. Thus, a destination is determined for data transported on the communication link using a data type identifier sent with the data.
Samsung Austin R&D Center
Senior Staff Compiler Engineer
Skills:
Software Engineering Algorithms Embedded Software Firmware Multithreading Debugging Processors Microprocessors Microcontrollers X86 X86_64 X86 64 Arm Aarch64 68K Intel 8051 Z80 Embedded Systems Soc Hardware Computer Architecture System Architecture Hardware Architecture Microarchitecture High Performance Computing Performance Tuning Compilers Gcc Llvm C C++ Python Perl X86 Assembly Operating Systems Rtos Linux Linux Kernel Embedded Linux Device Drivers Testing Assembler Computer Hardware Software Development Arm Architecture High Performance Computing