Integrated Device Technology, Inc.

Ampere Computing LLC

Applied Micro Circuits Corporation

Mitrajit Chatterjee

Peter Onufryk

Inna Levit

Ming Tang

Babji Vallabhaneni

Steven Chau

Sanjay Patel

Sarthak Raina

Vijay Parmar

Hoan Tran

Abhishek Niraj

Anuradha Raghunathan

Waseem Kraipak

Sandeep Badida

A system and method are described herein for estimating power usage of various components of a CPU and controlling voltage regulators based on the estimated power usage. The power estimates may be based on digital power meter readings at each component, on voltage information from a voltage regulator, and on other power information. This power information is transmitted over a mesh interconnect disposed throughout the CPU such that power estimation can be accurately calculated and used to control voltage regulators without being limited by external bus speeds. More of the power management processes and components may be disposed on the CPU and connected to the mesh interconnect. These power management processes include various calibrations, adjustments, and limits so as efficiently manage and use the more rapidly processed power estimations.

Method, apparatus, and system for calibrating a processor power level estimate

An error injection system of a built-in self-repairable memory system renders the redundant spare columns of the repairable memory accessible to built-in self-test (BIST) read and write operations. To this end, the error injection system selectively injects fault data at one or more locations of the main memory during a BIST sequence, causing the BIST controller to issue a repair instruction that allocates one or more spare columns as replacement memory areas for the presumed faulty main memory locations. Thereafter, BIST read/write operations directed to the main memory locations will be performed on the allocated spare columns, thereby allowing the spare columns to be validated as part of the BIST. Injection of fault data to the main memory locations in this manner can also facilitate validation of the built-in self-repair logic by verifying the repair instruction codes that are generated in response to the injected faults.

Method to dynamically inject errors in a repairable memory on silicon and a method to validate built-in-self-repair logic

A circuit analysis tool is provided, enabled as computer software instructions, for budgeting time delays between integrated circuit (IC) modules. The instructions accept a command enabling an IC floor-plan including a first module and a second module. The first module includes a first circuit element having a signal output interface, and an output port. The second module includes an input port, and a second circuit element having a signal input interface. A command is accepted defining a maximum delay value, and a first delay value is estimated between the first circuit element signal output interface and the first module output port. A second delay value is estimated between the second circuit element signal input interface and the second module input port, and a third delay value is estimated between the first module output port and the second module input port. The first, second, and third delay values are summed, creating a time budget estimate. The time budget estimate is approved if it is less than the maximum delay value.

Integrated circuit module time delay budgeting

A system and method for scheduling a thread identifies runnable threads based on precycle signals determined before the scheduling cycle. The thread indexes of the runnable threads are grouped according to the thread priorities of the runnable threads, and the thread indexes are ranked within each group. The runnable threads that will be runnable in the next scheduling cycle are identified based on same cycle signals determined during the scheduling cycle. The highest ranked thread index of the runnable threads that will also be runnable in the next scheduling cycle is selected as the scheduled thread. In another configuration, a round robin ranking and a priority ranking are determined for the thread indexes. The thread indexes are then ranked according to the round robin ranking and the priority ranking and the highest ranked thread index of a runnable thread is selected as the scheduled thread.

Scheduler for a multiprocessing computing system

A memory unit includes a memory organized into protected and non-protected areas. A controller manages access to the memory so that the protected area can be written to through a JTAG or CPU interface. Once written to, the protected area is only accessible to particular logic and cannot be over-written until the entire memory is erased. The controller is configured to allow a BCV to be stored in the memory through either the JTAG or CPU interface. The controller is also configured to allow writing to the protected area and boot configuration vector in memory before CPU boot-up by using a JTAG clock signal provided through an external pin when a system clock signal is not available. A reset circuit generates one or more initialization signals using either the BCV from memory or another BCV provided on external BCV pins, depending upon whether another external BCV pin is asserted.

Memory unit with controller managing memory access through JTAG and CPU interfaces

An apparatus and method for transferring multi-byte words having arbitrary start and end byte addresses are described. Data transfers between a memory and a PCI-bus pass through a PCI-side aligner unit, PCI FIFO, Endian swap logic, and PCI-bus interface unit under the control of a PCI FIFO controller. The PCI-side aligner unit properly aligns the data while communicating data with the memory's bus on a word-at-a-time basis, and communicating data with the PCI FIFO managed on a byte-at-a-time basis synchronous with a clock signal provided by the memory's bus. The Endian swap logic properly orients the data in big or little Endian orientation. The PCI-bus interface unit communicates data with the PCI-bus on a word-at-a-time basis, and communicates data with the PCI FIFO managed on a byte-at-a-time basis synchronous with a clock signal provided by the PCI-bus. To keep track of unread stored data in the PCI FIFO, the PCI FIFO controller includes a #Bytes logic unit that automatically accounts for wrap-around of write and read pointers for the PCI FIFO.

Growing community of inventors

San Jose, CA, United States of America

Mitrajit Chatterjee