US20070271440A1 - Computer processor architecture selectively using finite-state-machine for control code execution - Google Patents

Abstract

A microprocessor architecture including a finite state machine in combination with a microcode instruction cache for executing microinstructions. Microinstructions which would normally result in small sequences of high-repetition looped operations are implemented in a finite state machine (FSM). The use of the FSM is more energy-efficient than looping instructions in a cache or register set. In addition, the flexibility of a cache, or other memory oriented approach, in executing microcode instructions is still available. A microinstruction is identified as an FSM operation (as opposed to a cache operation) by an ID tag. Other fields of the microinstruction can be used to identify the type of FSM circuitry to use, direct configuration of a FSM to implement the microinstruction, indicate that certain fields are to be implemented in one or more FSMs and/or in memory-oriented operations such as in a cache or register.

Description

BACKGROUND OF THE INVENTION
• This application relates in general to digital processing and more specifically to a digital processor microarchitecture for reducing power consumption in processors.
• Reducing power consumption in microprocessor design is becoming more important as increasing numbers of portable digital processing applications, such as cellular telephones, personal digital assistants (PDAs), pocketbook computers, etc., become more popular. Also, as digital processors become faster, and as process technologies allow for increasingly dense designs, the effective capacitance, Ceff, encountered in new processor designs increases. The increased Ceff can require prohibitive amounts of power to achieve desired speeds. Recent projections indicate that traditional processor architectures that include significant random-access memory (RAM) and read-only memory (ROM) in the processor's microstore, cache, buffers, registers and other areas; requires much more power due to fast signal switching requirements encumbered by high Ceff.
• Traditionally, the liberal use of memory at the processor level has been desirable. The program store allows the processor to be re-configured with updated instruction set functionality. The design of microprocessor instructions, or the “instruction set,” can be changed to a degree without greatly affecting hardware design. The use of a microinstruction cache allows instructions that need to be repeated many times to be more quickly accessed by placing instructions within the loop to be executed in the fast cache. However, the use of large memories and memory caches also require cell arrays, control lines, decode logic, etc., whose operations are highly affected by Ceff. In general, storage accessing has a significant overhead at speeds desired for central processing unit (CPU) instruction execution.
• The prior art has tried to minimize the use of memory devices in modern processor architecture designs by making the microinstruction or instruction cache extremely small, and by using large, complex instructions, e.g., very long instruction word (VLIW) or complex instruction set computer (CISC) approaches. These approaches use large instruction words, that can perform one or more functions that would otherwise require multiple, simpler, instructions. At an extreme, a single complex microinstruction might be the entirety of a loop so that frequently-executed loops and functions can be performed by keeping the single microinstruction in a small, single-word cache and executing the same microinstruction many times. However, the mere operation of looping a register, memory or cache, requires high switching speeds of signals in many transistors in the register memory, control memory and associated support circuitry. Such an approach is prohibitive where Ceff is large. The trend of denser numbers of transistors to control more and more circuitry and chip designs is causing a rapid increase in Ceff. Other aspects of microprocessor design can further thwart register and memory oriented approaches.
• Thus, it is desirable to provide a system that alleviates one or more shortcomings in the prior art.
SUMMARY OF THE INVENTION
• One embodiment of the invention uses a finite state machine in combination with a microcode instruction (“microinstruction”) cache. Selected instructions which would normally result in small sequences of high-repetition looped operations are implemented in a finite state machine (FSM). The use of the FSM is more energy-efficient than looping instructions in a cache or register set. In addition, the flexibility of a cache, or other memory oriented approach, in executing microcode instructions is still available.
• In a preferred embodiment a microinstruction, or many microinstructions, are identified as an FSM operation (as opposed to a cache operation) by an ID tag. Other fields of the microinstruction can be used to identify the type of FSM circuitry to use, direct configuration of a FSM to implement the microinstruction, indicate that certain fields are to be implemented in one or more FSMs and/or in memory-oriented operations such as in a cache or register.
• In one embodiment, the invention provides a digital processor responsive to a microinstruction to perform an operation, the digital processor comprising, a memory for storing data associated with the operation to be performed; a finite state machine (FSM) for performing a function in association with the operation; and control circuitry for providing data to the register and for starting execution of the FSM in response to the microinstruction.
BRIEF DESCRIPTION OF THE DRAWINGS
• FIG. 1 shows basic components of the present invention;
• FIG. 2A is a first diagram of an FSM control designation in a microinstruction;
• FIG. 2B is a second diagram of an FSM control designation in a microinstruction;
• FIG. 2C is a third diagram of an FSM control designation in a microinstruction; and
• FIG. 3 is a flowchart of basic steps in a routine to determine whether FSM circuitry is invoked.
DESCRIPTION OF A PREFERRED EMBODIMENT
• FIG. 1 illustrates a preferred embodiment of the invention.
• InFIG. 1,microprocessor architecture100 includesmicrostore102 for storing microinstructions for execution by the components of the microprocessor architecture. As is known in the art, such instructions direct memory and register accessing, logical and arithmetic data operations, sequencing operations, and other operations to direct hardware in a digital processing system to perform functionality according to instructions stored in system memory (not shown).
• Microstore102 is indexed byprogram counter104 to access a location of a current microinstruction whose data is output ascurrent instruction106. Note that a separate register or latch can be used to hold the current microinstruction, or the data can simply be output directly from the microstore. Many approaches to accessing and manipulating microinstructions are possible and are within the scope of the present invention.
• The current microinstruction is used bycontrol circuitry108 to direct finite state machine (FSM)112,FSM configuration110 and memory oriented circuitry such asregister114 andloop control116 to perform functions in accordance with the current microinstruction.Control circuitry108 can also be used to direct, or control, other components in the system such asprogram counter104, arithmetic/logic hardware andsequencing hardware118, system memory accesses, etc.
• Register114 andloop control116 can be microinstruction looping components as are known in the art. Other standard architecture components can be used. In a preferred embodiment,control circuitry108 determines whether the current instruction, or a portion of the current instruction, is suitable, or has been designated, for using FSM processing. If so,control circuitry108 directs the use ofFSM112 to execute according to data in the current instruction. In applications where the FSM is configurable,FSM configuration circuit110 is used to pre-configure FSM112, as desired. Note that any number of configurable, or fixed, FSMs can be used. For ease of illustration, only one is shown inFIG. 1.
• Control circuitry108 can be implemented in various ways, as desired. For example, in a simple implementation the control circuitry need be no more than hardwired routing of control signals and data in the current instruction, or from the microstore, to the appropriate components (such as FSM112 or register114). Other implementations can use multiplexers, gates or other devices to more flexibly control, generate, process, monitor, etc., the information from a microinstruction and use the information to control components.
• Note that memory oriented circuitry can be other than a simple register and loop control. For example, multiple registers can be used. A memory bank can be employed, fast memory can be used as a cache, etc. In general, an FSM can be used in combination with any type of computer architecture, including additional FSMs.
• FSM configuration circuitry110 can be invoked by earlier instructions to prepare an FSM for later use by a different microinstruction. Alternatively, portions of an instruction can be used to direct configuration of the FSM via the configuration circuitry, while other parts of the same instruction serve to control FSM execution. In some applications, control of an FSM need not be more involved than merely invoking the FSM to begin execution.
• FIGS.2A-C illustrate different types of FSM control microinstructions.
• InFIG. 2A,microinstruction200 includestypical fields204 for directing the operation of CPU circuitry. Typical fields can be, e.g., a next address field, arithmetic logic unit (ALU) control field, register access field, memory access, access mode, etc. Additionally, an FSM control bit FSMCTL at202 is reserved. FSMCTL is set to 0 the FSM feature of the present invention is not invoked. In this case, the CPU can behave as a traditional processor. However, when FSMCTL is set to 1 then fields204, and subsequent fields in subsequent instructions, can invoke FSM processing.
• FIG. 2B illustrates the microinstruction format where FSM control is enabled.
• InFIG. 2B,microinstruction200 now hasFSM control bit202 set to 1. This enablesFSMCTL control field206 which would otherwise be used for other purposes. Depending on the microinstruction set design, additional fields such as208, etc., can be used for FSM control. In different embodiments, many different arrangements for determining when, and how, a microinstruction (or instructions) will be interpreted to control an FSM. For example, other approaches can use fixed, predetermined fields for FSM control so that an FSM control bit is not needed. The configuration information for the FSM can be stored in these fields.
• FIG. 2C illustrates an FSM control mode where multiple instructions are used.
• InFIG. 2C, the FSM control bit is set to 1.FSMCTL field212 then indicates how many microinstructionwords following word200 are to be used for FSM control. In this example, the remaining fields ofmicroinstruction200 along with the next two microinstructions,214 and216, can be used in their entirety for FSM-related purposes.
• In other embodiments, the FSM control bit can be derived from a signal generated by hardware within the digital processing system, or obtained or derived from an external source. One refinement is to providecontrol circuit108 ofFIG. 1 with a “power saving mode” signal. When on,control circuit108 will attempt to utilize FSM execution unit hardware as much as possible. When off, the emphasis will be on speed of execution which. One approach is to provide parallel instructions for different functions to be executed. One set of microinstructions is FSM oriented while the other is non-FSM. The control circuitry can selectively retrieve the desired (FSM, or non-FSM) instructions from a microstore as dictated by the power saving mode. Both the FSM and normal microstore operations could be executing in parallel.
• FIG. 3 illustrates a flowchart showing basic steps in a routine to determine whether FSM circuitry, such asFSM110 ofFIG. 1, is invoked.
• InFIG. 3, the routine offlowchart300 is entered atstep302. Atstep304, a next microinstruction to be executed is obtained. Atstep306, a determination is made as to whether FSM mode execution is desired, or directed. As discussed above, this determination can be made from information in the obtained microinstruction, itself, or from a different microinstruction or from a control signal. If not desired, then step308 is performed to execute the microinstruction's functions or operations in a non-FSM manner.
• If, however, it is desired atstep306 to use FSM mode execution then step310 is next executed to check whether FSM mode is available. Even though desired, FSM mode may not be available for a variety of reasons including lack of availability of the FSM execution unit (112 ofFIG. 1), or lack of other resources. In an embodiment where dual sets of microinstructions are used to allow FSM or non-FSM modes for same operations, not all operations may be provided with dual FSM microinstructions so that FSM mode execution will not be possible. If FSM mode is not possible then step308 is executed to execute the microinstruction's operations in a non-FSM mode.
• If FSM mode is possible then step312 is performed to check whether configuration data is needed. If so,step314 obtains the configuration data. Configuration data can be provided by microinstructions, data elsewhere in the system, or hardwired circuitry for frequently used tasks. After configuration, or if it determined that configuration is not necessary, execution proceeds to step316 where microinstruction operations are executed in an FSM mode and the routine exits atstep318.
• Note that the functionality described by the steps ofFIG. 3 may be achieved in many different ways. Steps can be added to or removed from those shown inFIG. 3. The steps can be modified and the sequencing can vary without departing from the scope of the invention. The functionality can be achieved by hardware, software or a combination of both.
• Although the invention has been described with respect to specific embodiments above, the embodiments are merely illustrative, and not restrictive, of the invention. For example, even though aspects of the invention have been described advantageously reducing power consumption, the invention can be applied to designs where use of an FSM along with other execution units is desirable for other reasons. In some applications, power consumption may actually increase. For example, reducing the area of a design can be achieved, faster implementation of certain microinstructions can be realized (resulting in faster performance), advantages in microinstruction length, complexity, organization, execution, etc., can be obtained.
• Although a distinction has been made between FSM and non-FSM execution, these terms are intended to mean either use of specialized FSM execution unit circuitry or non-use of such circuitry for a specific instruction, set of instructions, or portion of instruction. The term “non-FSM” or similar does not mean that no FSM is used. Nor does “FSM” mean that only FSM circuitry is used. For example, a non-FSM execution may still use FSM circuitry, but not the specialized FSM circuitry that is illustrated asFSM112 ofFIG. 1. In the art, FSM is a design approach that can be used to implement many different functions. For example, it is popular to use an FSM to generate control signals so that an FSM may be used incontrol circuitry108 and ALU hardware/sequencer118. The discussion herein, relating to use or non-use of FSM execution hardware includes specialized FSM circuitry that is selectable according to a microinstruction's fields, a control signal, or other component or approach discussed or implied herein. The type of specialized circuitry can vary widely. The same specialized circuitry need not be used for different instructions, or even for a same instruction executed at a later time.
• Thus, the scope of the invention is to be determined solely by the appended claims.

Claims (26)

1-8. (canceled)

9. A computer processor for executing a microinstruction, the computer processor comprising:

a configurable finite state machine (FSM);

processing circuitry; and

control circuitry for controlling the configurable finite state machine (FSM) to implement at least a portion of the microinstruction in parallel with the processing circuitry.

10. The computer processor ofclaim 9, wherein the configurable finite state machine (FSM) and processing circuitry are operated concurrently.

11. (canceled)

12. A method of performing an operation according to a microinstruction having a finite state machine control field, comprising:

storing data associated with the operation to be performed;

configuring a configurable finite state machine via a finite state machine configuration circuit to perform a function in association with the operation;

detecting a predefined value in the finite state machine control field in the microinstruction; and

providing data to enable execution of the function by the finite state machine in response to the predefined value.

13. The method ofclaim 12, further comprising configuring a second configurable finite state machine via the finite state configuration circuit to perform a second function in association with the operation.

14. The method ofclaim 12, further comprising configuring the configurable finite state machine in advance to perform the function associated with the microinstruction.

15. The method ofclaim 12, wherein a second microinstruction having the same function as the first microinstruction is executed by the finite state machine.

16. A method of processing a microinstruction using a plurality of finite state machines in a processing system, the microinstruction including a finite state machine data field and directing at least one operation, the method comprising:

reading the microinstruction;

determining the at least one operation directed by the microinstruction to be performed;

selecting one of the plurality of finite state machines to perform the at least one operation of the microinstruction; and

configuring the selected finite state machine to perform the at least one operation of the microinstruction.

17. The method ofclaim 16, wherein the finite state machine is selected by hardware within the processing system.

18. The method ofclaim 16, wherein the finite state machine is selected by a source external to the processing system.

19. The method ofclaim 16, wherein the configuration of the selected finite state machine occurs prior to reading the microinstruction.

20. The method ofclaim 16, wherein the microinstruction includes configuration data used to select the finite state machine and control data to control the performance of the operation via the finite state machine.

21. The method ofclaim 16, wherein the finite state machine is selected from the plurality of finite state machines based on the use of the plurality of finite state machines in the processing system.

22. The method ofclaim 16, wherein the finite state machine is selected based on power requirements.

23. The method ofclaim 16, wherein configuring the finite state machine is performed via a hardwired configuration circuit.

24. The method ofclaim 16, wherein configuring the finite state machine is performed via the microinstruction.

25. A method for executing software with a computer processor having a configurable finite state machine, control circuitry for configuring the configurable finite state machine to perform different operations, and processing circuitry, the method comprising:

reading the software;

deriving at least one microinstruction from the software, the microinstruction including directing an operation;

determining whether the operation should be executed using the configurable finite state machine;

providing an instruction that the operation is to be executed by the configurable finite state machine based on the determination; and

executing the operation with the configurable finite state machine.

26. A processor system for processing a microinstruction including a finite state machine data field and directing at least one operation, the processor comprising:

a plurality of finite state machines;

an input device to read the microinstruction;

a controller to determine the at least one operation directed by the microinstruction to be performed and to select one of the plurality of finite state machines to perform the at least one operation of the microinstruction; and

a configuration circuit to configure the selected finite state machine to perform the at least one operation of the microinstruction.

27. The processor system ofclaim 26, wherein controller is hardware within the processing system.

28. The processor system ofclaim 26, wherein the controller configures the selected finite state machine prior to the input device reading the microinstruction.

29. The processor system ofclaim 26, wherein the microinstruction includes configuration data used to select the finite state machine and control data to control the performance of the operation via the finite state machine.

30. The processor system ofclaim 26, wherein the controller selects the finite state machine from the plurality of finite state machines based on the use of the plurality of finite state machines in the processor system.

31. The processor system ofclaim 26, wherein the controller selects the finite state machine based on power requirements of the processor system.

32. The processor system ofclaim 26, wherein the configuration circuit is hardwired.

33. The processor system ofclaim 26, wherein the configuration circuit performs the configuration via reading the microinstruction.

Images