Patent Application
20100115234
The invention relates generally to vector computer processors, and more specifically in one embodiment to a configurable vector length computer processor.
A portion of the disclosure of this patent document contains material to which the claim of copyright protection is made. The copyright owner has no objection to the facsimile reproduction by any person of the patent document or the patent disclosure, as it appears in the U.S. Patent and Trademark Office file or records, but reserves all other rights whatsoever.
Most general purpose computer systems are built around a general-purpose processor, which is typically an integrated circuit operable to perform a wide variety of operations useful for executing a wide variety of software. The processor is able to perform a fixed set of instructions, which collectively are known as the instruction set for the processor. A typical instruction set includes a variety of types of instructions, including arithmetic, logic, and data instructions.
In more sophisticated computer systems, multiple processors are used, and one or more processors runs software that is operable to assign tasks to other processors or to split up a task so that it can be worked on by multiple processors at the same time. In such systems, the data being worked on is typically stored in memory that is either centralized, or is split up among the different processors working on a task.
Instructions from the instruction set of the computer's processor or processor that are chosen to perform a certain task form a software program that can be executed on the computer system. Typically, the software program is first written in a high-level language such as “C” that is easier for a programmer to understand than the processor's instruction set, and a program called a compiler converts the high-level language program code to processor-specific instructions.
In multiprocessor systems, the programmer or the compiler will usually look for tasks that can be performed in parallel, such as calculations where the data used to perform a first calculation are not dependent on the results of certain other calculations such that the first calculation and other calculations can be performed at the same time. The calculations performed at the same time are said to be performed in parallel, and can result in significantly faster execution of the program. Although some programs such as web browsers and word processors don't consume a high percentage of even a single processor's resources and don't have many operations that can be performed in parallel, other operations such as scientific simulation can often run hundreds or thousands of times faster in computers with thousands of parallel processing nodes available.
Multiple operations can also be performed at the same time using one or more vector processors, which perform an operation on multiple data elements at the same time. For example, rather than instruction that adds two numbers together to produce a third number, a vector instruction may add elements from a 64-element vector to elements from a second 64-element vector to produce a third 64-element vector, where each element of the third vector is the sum of the corresponding elements in the first and second vectors.
In this example, the vector registers each hold 64 elements, so the vector length is said to be 64. The vector processor can handle sets of data smaller than 64 by using a vector length register specifying that some number fewer than 64 elements are to be processed, or can handle sets of data larger than 64 elements by using multiple vector operations to process all elements in the data set, such as by using a program loop.
The vectors in some further examples do not operate on elements that are sequential in memory, but instead operate on elements that are spaced some distance apart, such as on certain elements of a large array for scientific computing and modeling applications. This distance between elements in a vector is referred to as the stride, such that sequential words from memory have a stride of one, whereas a vector comprising every sixteenth element in memory has a stride of 16.
Vector processing provides other benefits to program efficiency, but at the cost of significant load or startup time relative to a scalar operation. Although the vectors must be completely loaded from memory before functions can be performed on the elements, other steps such as checking for variable independence need only be performed once for an entire vector operation. Instruction and coding efficiency are also improved with vector operations, as is memory access where the vector has a known or consistent memory access pattern. Vector processor design choices such as vector length consider these efficiencies and tradeoffs in an attempt to provide both good scalar operation performance and efficient vector operation.
Some embodiments of the invention comprise a processor core that comprises one or more vector units operable to change between a fine-grained vector mode having a shorter maximum vector length and a coarse-grained vector mode having a longer maximum vector length. Changing vector modes comprises halting all instruction stream execution in the core, flushing one or more registers in a register space, reconfiguring one or more vector registers in the register space, and restarting instruction execution in the core.
In the following detailed description of example embodiments of the invention, reference is made to specific examples by way of drawings and illustrations. These examples are described in sufficient detail to enable those skilled in the art to practice the invention, and serve to illustrate how the invention may be applied to various purposes or applications. Other embodiments of the invention exist and are within the scope of the invention, and logical, mechanical, electrical, and other changes may be made without departing from the scope or subject of the present invention. Features or limitations of various embodiments of the invention described herein, however essential to the example embodiments in which they are incorporated, do not limit the invention as a whole, and any reference to the invention, its elements, operation, and application do not limit the invention as a whole but serve only to define these example embodiments. The following detailed description does not, therefore, limit the scope of the invention, which is defined only by the appended claims.
Vector processor architectures often include vector registers having a fixed number of entries, each vector register capable of holding a single vector. Vector functional units, such as an add/subtract unit, a multiply unit and a divide unit, and logic operation units are either dedicated to serving vector operations or are shared with scalar operations. Scalar registers are also used in some vector operations, such as where every element of a vector is multiplied by a scalar number. An example processor might have, for example, eight vector registers with 64 elements per register, where each element is a 64-bit word.
It is desirable in some applications to have vector lengths that are longer, while in other applications greater performance could be achieved if vector lengths were shorter or if the processor functioned more like a scalar processor. One embodiment of the invention seeks to address problems such as this by providing a reconfigurable processor core, such as where a more vectorized and a less vectorized configuration are available within the same processor core and can be selected to improve application execution efficiency.
In one such example, a processor chip contains 32 cores, where each core is capable of operating in either a vector threaded mode supporting four streams having a maximum vector length of 16, or a scalar threaded mode supporting 32 streams of a maximum vector length of one. Each mode has the same instruction set architecture, same instruction issue rate, and same instruction processing performance, but will provide different application performance based on the parallelization or vectorization that can be achieved for a given application.
In one such example illustrated in
Vector registers allocated to each of four different instruction streams of the four-stream 16-element vector configuration are shown at 101, each stream being allocated 32 registers having 16 elements each, such that there is a maximum vector length of 16. Address registers for each stream are allocated in register space 102, but only consume two elements of 32 registers per stream—the remaining register space that is crossed out is unused in this vector mode.
In
In this example embodiment, the address registers and vector registers are a part of a processor core, as shown in
To change modes between fine-grained parallel applications that benefit from running in a 32-stream mode and coarse-grained parallel applications that benefit from the longer vector length of the 4-stream mode, the processor core quiets all executing threads in the core being reconfigured, and flushes the registers. The registers and instruction pipelines are reloaded under the new vector/stream mode, and execution is restarted.
Changing modes therefore involves repartitioning the register space and reassignment of registers to different streams, or between vector and address register allocation, depending on the embodiment being practiced. The actual register space remains the same, as is illustrated in the example of
The processor of this example can therefore be configured for fine-grained or coarse-grained parallelism on the fly, even within an executing application. The ability to configure the processor core on the fly, even within a job or application, provides greater flexibility and efficiency in execution than prior systems could provide. Further, the ability to switch modes on a core-by-core basis rather than on a system-by-system basis or chip-by-chip basis enables configuration of individual cores to best suit the applications assigned to those specific cores. For example, a processor chip containing 32 cores can configure 28 cores to work on a coarse-grained parallel application using a vector length of 16, while the remaining four cores execute fine-grained threads that do not lend themselves to vector parallelization as well.
Although specific embodiments have been illustrated and described herein, it will be appreciated by those of ordinary skill in the art that any arrangement which is calculated to achieve the same purpose may be substituted for the specific embodiments shown. This application is intended to cover any adaptations or variations of the example embodiments of the invention described herein. It is intended that this invention be limited only by the claims, and the full scope of equivalents thereof.
