Tesla, Inc.

Advanced Micro Devices Corporation

Globalfoundries Inc.

Debjit Das Sarma

Emil Talpes

Peter Joseph Bannon

Kevin Altair Hurd

Benjamin Floering

Christopher Hsiong

Marius Evers

Ranganathan Sudhakar

Scott Hilker

David A Kaplan

Jay E Fleischman

Michael Estlick

Gregory William Smaus

Krishnan V Ramani

Kelvin Domnic Goveas

Francesco Spadini

Ganesh Venkataramanan

Nikhil Gupta

William McGee

Wei-Yu Chen

Betty Ann McDaniel

Douglas R Williams

David Oliver

Hanbing Liu

Sahil Arora

Daryl Lieu

Jonathan Choy

Chun Liu

William Mcgee

A microprocessor system comprises a vector computational unit and a control unit. The vector computational unit includes a plurality of processing elements. The control unit is configured to provide at least a single processor instruction to the vector computational unit. The single processor instruction specifies a plurality of component instructions to be executed by the vector computational unit in response to the single processor instruction and each of the plurality of processing elements of the vector computational unit is configured to process different data elements in parallel with other processing elements in response to the single processor instruction.

Instruction set architecture for a vector computational unit

A system for handling errors in a neural network includes a neural network processor for executing a neural network associated with use of a vehicle. The neural network processor includes an error detector configured to detect a data error associated with execution of the neural network and a neural network controller configured to receive a report of the data error from the error detector. In response to receiving the report, the neural network controller is further configured to signal that a pending result of the neural network is tainted, without terminating execution of the neural network.

System and method for handling errors in a vehicle neural network processor

A microprocessor system comprises a computational array and a vector computational unit. The computational array includes a plurality of computation units. The vector computational unit is in communication with the computational array and includes a plurality of processing elements. The processing elements are configured to receive output data elements from the computational array and process in parallel the received output data elements.

Vector computational unit

A system for handling errors in a neural network includes a neural network processor for executing a neural network associated with use of a vehicle. The neural network processor includes an error detector configured to detect a data error associated with execution of the neural network and a neural network controller configured to receive a report of the data error from the error detector. In response to receiving the report, the neural network controller is further configured to signal that a pending result of the neural network is tainted without terminating execution of the neural network.

Vector computational unit receiving data elements in parallel from a last row of a computational array

A microprocessor system comprises a matrix computational unit and a control unit. The matrix computational unit includes a plurality of processing elements. The control unit is configured to provide a matrix processor instruction to the matrix computational unit. The matrix processor instruction specifies a floating-point operand formatted using a first floating-point representation format. The matrix computational unit accumulates an intermediate result value calculated using the floating-point operand. The intermediate result value is in a second floating-point representation format.

Data path for scalable matrix node engine with mixed data formats

A microprocessor system comprises a matrix computational unit and a control unit. The matrix computational unit includes one or more processing elements. The control unit is configured to provide a matrix processor instruction to the matrix computational unit. The matrix processor instruction specifies a floating-point operand formatted with an exponent that has been biased with a specified bias.

Scalable matrix node engine with configurable data formats

Overhead associated with verifying function return addresses to protect against security exploits is reduced by taking advantage of branch prediction mechanisms for predicting return addresses. More specifically, returning from a function includes popping a return address from a data stack. Well-known security exploits overwrite the return address on the data stack to hijack control flow. In some processors, a separate data structure referred to as a control stack is used to verify the data stack. When a return instruction is executed, the processor issues an exception if the return addresses on the control stack and the data stack are not identical. This overhead can be avoided by taking advantage of the return address stack, which is a data structure used by the branch predictor to predict return addresses. In most situations, if this prediction is correct, the above check does not need to occur, thus reducing the associated overhead.

Using return address predictor to speed up control stack return address verification

A processor, a device, and a non-transitory computer readable medium for performing branch prediction in a processor are presented. The processor includes a front end unit. The front end unit includes a level 1 branch target buffer (BTB), a BTB index predictor (BIP), and a level 1 hash perceptron (HP). The BTB is configured to predict a target address. The BIP is configured to generate a prediction based on a program counter and a global history, wherein the prediction includes a speculative partial target address, a global history value, a global history shift value, and a way prediction. The HP is configured to predict whether a branch instruction is taken or not taken.

Bandwidth increase in branch prediction unit and level 1 instruction cache

A processor includes a physical register file having physical registers and an execution unit to perform an arithmetic operation to generate a result mapped to a physical register, wherein the processor delays a write of the result to the physical register file until the result is qualified as valid. A method includes mapping the same physical register both to store load data of a load-execute operation and to subsequently store a result of an arithmetic operation of the load-execute operation, and writing the load data into the physical register. The method further includes, in a first clock cycle, executing the arithmetic operation to generate the result, and, in a second clock cycle, providing the result as a source operand for a dependent operation. The method includes, in a third clock cycle, enabling a write of the result to the physical register file responsive to the result qualifying as valid.

Register file management for operations using a single physical register for both source and result

The present application provides a method and apparatus for supporting denormal numbers in a floating point multiply-add unit (FMAC). One embodiment of the FMAC is configurable to add a product of first and second operands to a third operand. This embodiment of the FMAC is configurable to determine a minimum exponent shift for a sum of the product and the third operand by subtracting a minimum normal exponent from a product exponent of the product. This embodiment of the FMAC is also configurable to cause bits representing the sum to be left shifted by the minimum exponent shift if a third exponent of the third operand is less than or equal to the product exponent and the minimum exponent shift is less than or equal to a predicted left shift for the sum.

Floating point multiply-add unit with denormal number support

Growing community of inventors

San Jose, CA, United States of America

Debjit Das Sarma