Intel Corporation

Kuck & Associates, Inc.

Arch D Robison

Alexey Kukanov

Ilan Pardo

Oren Ben-Kiki

James Henry Cownie

Cheng C Wang

Youfeng Wu

Tatiana Shpeisman

Victor W Lee

Paul M Petersen

Hongbo Rong

Alejandro Duran Gonzalez

Todd A Anderson

Peng Tu

Nadav Chachmon

Hai Liu

P G Lowney

Instructions and logic provide atomic range operations in a multiprocessing system. In one embodiment an atomic range modification instruction specifies an address for a set of range indices. The instruction locks access to the set of range indices and loads the range indices to check the range size. The range size is compared with a size sufficient to perform the range modification. If the range size is sufficient to perform the range modification, the range modification is performed and one or more modified range indices of the set of range indices is stored back to memory. Otherwise an error signal is set when the range size is not sufficient to perform said range modification. Access to the set of range indices is unlocked responsive to completion of the atomic range modification instruction. Embodiments may include atomic increment next instructions, add next instructions, decrement end instructions, and/or subtract end instructions.

Instructions and logic to provide atomic range modification operations

Technologies for multithreaded synchronization including a computing device having a many-core processor. Each processor core includes multiple hardware threads. A hardware thread executed by a processor core enters a synchronization barrier and synchronizes with other hardware threads executed by the same processor core. After synchronization, the hardware thread synchronizes with a source hardware thread that may be executed by a different processor core. The source hardware thread may be assigned using an n-way shuffle of all hardware threads, where n is the number of hardware threads per processor core. The hardware thread resynchronizes with the other hardware threads executed by the same processor core. The hardware thread alternately synchronizes with the source hardware thread and the other hardware threads executed by the same processor core until all hardware threads have synchronized. The computing device may reduce a Boolean value over the synchronization barrier. Other embodiments are described and claimed.

Technologies for fast synchronization barriers for many-core processing

A processor of an aspect includes a plurality of processor elements, and a first processor element. The first processor element may perform a user-level fork instruction of a software thread. The first processor element may include a decoder to decode the user-level fork instruction. The user-level fork instruction is to indicate at least one instruction address. The first processor element may also include a user-level thread fork module. The user-level fork module, in response to the user-level fork instruction being decoded, may configure each of the plurality of processor elements to perform instructions in parallel. Other processors, methods, systems, and instructions are disclosed.

User-level fork and join processors, methods, systems, and instructions

Technologies for generating composable library functions include a first computing device that includes a library compiler configured to compile a composable library and second computing device that includes an application compiler configured to compose library functions of the composable library based on a plurality of abstractions written at different levels of abstractions. For example, the abstractions may include an algorithm abstraction at a high level, a blocked-algorithm abstraction at medium level, and a region-based code abstraction at a low level. Other embodiments are described and claimed herein.

Technologies for low-level composable high performance computing libraries

A predicated instruction compilation system includes a control flow graph generation module to generate a control flow graph of a program code to be compiled into the predicated instructions to be executed on a processor that does not include any program counter. Each of the instructions includes a predicate guard and a predicate update. The compilation system also includes a control flow transformation module to automatically generate the predicate guard and an update to the predicate state on the processor. A computer-implemented method of compiling a program code into predicated instructions is also described.

Method and system of compiling program code into predicated instructions for execution on a processor without a program counter

Technologies for multithreaded synchronization and work stealing include a computing device executing two or more threads in a thread team. A thread executes all of the tasks in its task queue and then exchanges its associated task stolen flag value with false and stores that value in a temporary flag. Subsequently, the thread enters a basic synchronization barrier. The computing device performs a logical-OR reduction over the temporary flags of the thread team to produce a reduction value. While waiting for other threads of the thread team to enter the barrier, the thread may steal a task from a victim thread and set the task stolen flag of the victim thread to true. After exiting the basic synchronization barrier, if the reduction value is true, the thread repeats exchanging the task stolen flag value and entering the basic synchronization barrier. Other embodiments are described and claimed.

Technologies for efficient synchronization barriers with work stealing support

Implementing fair scalable reader writer mutual exclusion for access to a critical section by a plurality of processing threads is accomplished by creating a first queue node for a first thread, the first queue node representing a request by the first thread to access the critical section; setting at least one pointer within a queue to point to the first queue node, the queue representing at least one thread desiring access to the critical section; waiting until a condition is met, the condition comprising the first queue node having no preceding write requests as indicated by at least one predecessor queue node on the queue; permitting the first thread to enter the critical section in response to the condition being met; and causing the first thread to release a spin lock, the spin lock acquired by a second thread of the plurality of processing threads.

Fair scalable reader-writer mutual exclusion

A method for computing a trip count for a loop in advance of the execution of the loop is provided. The method comprises identifying the elements of a loop; returning infinity, if a first index value satisfies a first condition and that a first step size is equal to zero; modifying the first index value and the first step size, if the first index value satisfies the first condition, when the first step size is not equal to zero, and the first step size is greater than half of a first modulus; returning the result computed by applying a formula that divides the difference between a first condition value and the first index value by the first step size and rounds up to a next integer when there is a non-zero remainder; and returning a second trip count for a second loop based on the elements of the first loop.

Advance trip count computation in a concurrent processing environment

Implementing fair scalable reader writer mutual exclusion for access to a critical section by a plurality of processing threads in a processing system is accomplished by creating a first queue node for a first thread on the first thread's stack, the queue node representing a request by the first thread to access the critical section; adding the first queue node to a queue pointed to by a single word reader writer mutex for the critical section, the queue representing a list of threads desiring access to the critical section, each queue node in the queue being on a stack of a thread of the plurality of processing threads; waiting until the first queue node has no preceding write requests as indicated by predecessor queue nodes on the queue; entering the critical section by the first thread; exiting the critical section by the first thread; and removing the first queue node from the queue.

A method, device, system, and computer readable medium are disclosed. In one embodiment the method includes managing one or more threads attempting to steal task work from one or more other threads. The method will block a thread from stealing a mailed task that is also residing in another thread's task pool. The blocking occurs when the mailed task was mailed to an idle third thread. Additionally, some tasks are deferred instead of immediately spawned.

Preserving hardware thread cache affinity via procrastination

A dependence graph having a linear number of edges and one or more tie vertices is generated by constructing a tree of nodes, receiving requests to create cut and/or fan vertices corresponding to each node, adjusting a frontier of nodes up or down, and creating one or more cut or fan vertices, zero or more tie vertices, and at least one predecessor edge.

Fast tree-based generation of a dependence graph

A system and method to reduce the size of source code in a processing system are described. Multiple subgraph structures are identified within a graph structure constructed for multiple source code instructions in a program. Unifiable variables that are not simultaneously used in the source code instructions are identified within each subgraph structure. Finally, one or more unifiable instructions from a tine of a corresponding subgraph structure are transferred to a handle of the corresponding subgraph structure, each unifiable instruction containing one or more unifiable variables.

System and method to reduce the size of source code in a processing system

An arrangement is provided for eliminating partial redundancy. Original code is processed to perform run-time behavior preserving redundancy elimination. Partial redundancy is removed in a manner so that the run-time behavior of the original code is preserved.

Run-time behavior preserving partial redundancy elimination

The present invention is a method and system to reduce storage in a inter-procedural analysis solver. In one embodiment, local graphs are pruned. The local graphs represent local problems, which correspond to separately compilable components in a software program. Each of the local graphs has edges and vertices. Each edge has a transfer function. Each vertex has a value. Values of the local graph form a lattice under a partial ordering.

Growing community of inventors

Champaign, IL, United States of America

Arch D Robison