Electrical computers and digital data processing systems: input/ – Interrupt processing – Multimode interrupt processing
Reexamination Certificate
2000-01-20
2001-09-25
Myers, Paul R. (Department: 2181)
Electrical computers and digital data processing systems: input/
Interrupt processing
Multimode interrupt processing
Reexamination Certificate
active
06295574
ABSTRACT:
BACKGROUND OF THE INVENTION
1. Field of the Invention
The present invention is related to the field of microprocessors and, more particularly, to the interrupt handling mechanisms within microprocessors.
2. Description of the Relevant Art
Modern computer systems and the software which runs thereon demand a high performance interrupt structure in order to operate efficiently. Interrupts are used to switch between tasks, and so a multi-tasking operating system benefits from a high performance interrupt structure. A “multi-tasking” operating system is configured to run multiple programs concurrently. Additionally, interrupts provide a means for an electronic device external to the microprocessor to request attention from the operating system. Modem day computer systems are including increasing numbers of these electronic devices, prompting the need for a high performance interrupt structure.
Interrupts cause a microprocessor within the computer system to suspend execution of a task in order to execute a specific software routine (referred to as an interrupt service routine) comprising a set of instructions. The interrupt is typically unrelated to the instructions being executed by the microprocessor at the time the interrupt is signalled. Instead, the interrupt may be caused by an external device requiring software attention. For example, a buffer within an input/output device may fill with data to be transferred to another device or to memory. Many other sources for interrupts are well-known to the skilled artisan.
The instructions being executed by the microprocessor at the time the interrupt occurs are referred to herein as a “task”. A task may be a portion of a program, an operating system routine, or even another interrupt service routine.
Because the interrupt is normally unrelated to the task being performed by the microprocessor and is asynchronous to the task itself, the interrupt service routine is executed in such a way that the task may be resumed. In order to resume the task, the “context” within which the task is executing may be saved to memory. The context of a task is the state of the microprocessor at a particular moment of time in which the task is executing. The context may include register values associated with the task when the task is interrupted. In other embodiments, context may be defined to include other values as well. When the context is saved, the register portion of the context may be saved to memory. After saving the context, the interrupt service routine may be executed. Upon completion of the interrupt service routine, the context may be restored to the microprocessor and the task is resumed. Since the restored context is substantially identical to the context when the task was interrupted, the task executes normally. In other words, the interrupt had no affect on the result of executing the task if the task is unrelated to the interrupt. Instead, only the time required to execute the task is affected.
The x86 architecture defines the context of 32 bit microprocessors to be 104 bytes. Other microprocessor architectures may define contexts having more or fewer bytes. Additionally, when the segment registers (which are part of the context in the x86 architecture) are reloaded, segment reloads are initiated to translate the segments. More bytes are transferred when the reload occurs, and clock cycles are required to translate the extra bytes into a format for storing within the microprocessor.
Unfortunately, storing and retrieving a large number of bytes to and from memory (as a context save entails) often requires a relatively large number of clock cycles. A lock cycle refers to the amount of time required by portions of the microprocessor to perform their functions. At the end of the clock cycle, the results of each function are stored in a storage location (e.g. a register or memory) and may be used by another function in the next clock cycle. The bus used by a microprocessor to communicate with other electrical devices may operate according to a different clock cycle than the microprocessor itself. The clock cycle associated with the bus is often referred to as the bus clock cycle.
If the context is saved by the microprocessor when an interrupt is recognized by the microprocessor, the interrupt is being handled via a “task switch”. The interrupt service routine is isolated from the interrupted task such that any modifications the interrupt service routine performs to the microprocessor's context information will not affect the operation of the task when resumed. The context is restored prior to resuming the task. Often, an interrupt service routine will only require access to a few registers within the register set to perform its function. In this case, a full context save is not necessary since some registers will not be modified by the interrupt service routine. Instead, only those storage locations which must be changed in order to fetch the instructions within the interrupt service routine need be saved prior to beginning execution of the interrupt service routine. For example, in the x86 architecture the EIP register and CS segment register (which define the address and segment of the instructions to be fetched and executed) and the flags register (which is modified by many of the x86 instructions) are saved. These values are pushed onto the stack defined by the x86 architecture when not using the task switch method of interrupt handling.
When the task switch method of interrupt handling is not in use, an interrupt service routine must save the values stored within registers which it employs to carry out its intended function. Often, the values are stored on the stack. This method of interrupt handling is referred to as an interrupt gate or trap gate in the x86 architecture, depending on whether or not the interrupt service routine may itself be interrupted. If the interrupt service routine does not use all of the microprocessor's context, then clock cycles may be saved with respect to performing the full context save of a task switch. The interrupt service routine is entered and exited more rapidly since context save and restore is not performed. Unfortunately, at least a few registers must still be stored. Additionally, the interrupt service routines are lengthened by the number of instructions required to save and restore context values used by the interrupt service routines. Furthermore, an administrative burden is placed on the programmer of the interrupt service routine to update the save and restore portions of the routine when the routine is changed.
Since there are multiple sources of interrupts, the computer system provides a mechanism for identifying one of multiple interrupt service routines. The computer system thus provides flexibility to the programmer in that an interrupt service routine may be tailored to the needs of a particular device or interrupt source. Without the interrupt vector approach, all interrupts would fetch an interrupt service routine from the same address in memory. A relatively complicated routine stored at the address would need to be written to perform the handling of all types of interrupts from all electronic devices. As used herein, the term “fetching” refers to transferring the contents of a memory location to a destination.
One method for providing the address of the interrupt service routine for a given interrupt is for the microprocessor to request an interrupt vector from another electronic device in the system. An “interrupt vector” is a number which is indicative of a particular interrupt service routine. In the x86 microprocessor architecture, for example, the interrupt vector is an index into an interrupt vector table which provides information identifying the address of the associated interrupt service routine. The interrupt vector table is also referred to as an interrupt descriptor table.
In many computer systems, the interrupt vector is provided via a dedicated bus transaction. A “bus transaction” is a transfer of information across a bus. Bus transactions may include address
Advanced Micro Devices , Inc.
Conley Rose & Tayon PC
Merkel Lawrence J.
Myers Paul R.
LandOfFree
Real time interrupt handling for superscalar processors does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Real time interrupt handling for superscalar processors, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Real time interrupt handling for superscalar processors will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2501482