Patent Application
20100332804
1. Field of the Invention
This invention relates to computing systems, and more particularly, to efficient out-of-order picking of instructions in a processor.
2. Description of the Relevant Art
Modern microprocessors typically have increasing pipeline depth in order to support higher clock frequencies and increased microarchitectural complexity. Despite improved device speed, higher clock frequencies of next-generation processors allow fewer levels of logic to fit within a single clock cycle compared to previous generations. Certain loops of logic, or paths, within a processor may experience difficulty in fitting within a single pipeline stage. One such path is a wide instruction issue selection path that resides within an issue queue (IQ). The length of this logic path may be set by several factors including the size of the IQ, instruction dependencies, instruction latencies, the number and functionality of pipeline stages within a corresponding microarchitecture, and speculative instruction effects such as misprediction and recovery.
A modern IQ selects multiple dispatched instructions out of program order to enable more instruction level parallelism, which yields higher performance. Also, out-of-order (o-o-o) issue and execution of instructions helps hide instruction latencies. However, several instructions do not have single-cycle latencies. These multi-cycle instructions complicate the selection logic within an IQ. In addition, a latency between two instruction types, such as a floating-point arithmetic type and an integer arithmetic type, may not be consistent. For example, five pairings of these instruction types may yield a latency of 12 clock cycles between the producer generating a result and a consumer receiving the result. However, an additional sixth pairing may yield a latency of 13 clock cycles. This latter latency may set a final latency of 13 clock cycles for each of the six pairings, which decreases performance.
In addition to the above, one or more instructions may have a nondeterministic latency. For example, within an IQ, a load instruction that misses a cache may have an indeterminate time for generating a result. Within the IQ, a divide operation may have a source data dependent latency that is unknown at the time of instruction issue. Therefore, scheduling the issue of younger instructions dependent on these types of instructions with a nondeterministic latency is made more difficult. Additionally, the delay to generate a result from these instructions with a nondeterministic latency may increase as the pipeline depth increases.
By predicting a hit in the cache, a load instruction may be treated as a speculative instruction. In such a case, a known latency may be predicted. Younger instructions, Is including dependent instructions, may then issue early assuming that the load instruction hits in the cache. However, during a load miss, or misprediction of the speculative load, recovery occurs. During recovery, any younger instructions dependent on the mispredicted load instruction may then re-execute. One approach for recovery includes deallocating the younger dependent instructions from the IQ at the time of their early issue before a load miss. Then after a load miss, all younger instructions are re-fetched. While this approach eliminates storing post-issue instructions in the IQ, performance may suffer due to the overhead associated with re-fetching the instructions.
A second approach maintains storage of all younger instructions in the IQ until the older load hit status is known. In the case of a load miss, during recovery the younger dependent instructions may subsequently re-issue according to a predetermined policy. As pipeline depth increases, the speculative window of the load instruction increases. Accordingly, the size of the IQ increases as the number of instructions in the IQ waiting to be re-issued increases. These instructions fill a larger portion of the entries of the IQ. Unless a cache miss occurs, these post-issue instructions are not candidates for selection and they add complexity to the issue selection logic.
In addition to the above, parasitic capacitances and wire route delays continue to increase with each newer processor generation. Therefore, wire delays still limit the dimension of many processor structures such as an IQ. Within an IQ, the delay of a wide o-o-o issue selection path is proportional to the number of entries of the IQ. As stated earlier, higher clock frequencies allow fewer levels of logic to fit within a single clock cycle. In order for a processor to achieve high performance, the IQ needs to supply a sufficient number of instructions to functional units each clock cycle despite the various constraints mentioned above.
In view of the above, efficient methods and mechanisms for efficient out-of-order picking of instructions in a processor are desired.
Systems and methods for efficient picking of instructions for out-of-order issue and execution in a processor are contemplated. In one embodiment, a processor comprises a unified pick queue that includes an array configured to dynamically allocate an entry of a plurality of entries for a decoded instruction. Each entry is configured to store age information and dependency information relative to other decoded instructions. Also, each entry stores a picked field and a trigger field.
The picked field, when asserted, indicates the decoded instruction has already been picked for out-of-order issue and execution. Therefore, the decoded instruction will not be re-picked. The trigger field, when asserted in a first clock cycle, indicates a result of the decoded instruction will be available M clock cycles from the first clock cycle, wherein M is a nonzero integer. Control logic determines a first instruction dependent on a result of a second instruction is ready to be picked before the result of the second instruction is available. In this case, the second instruction has asserted picked and trigger fields.
The trigger field of a decoded and picked instruction is asserted a predetermined number of clock cycles after the clock cycle in which the instruction is picked. The assertion of the trigger field after a predetermined number of clock cycles allows a younger dependent instruction to be picked. Subsequently, the younger dependent instruction begins execution in the same clock cycle the generated results become available. Therefore, the younger dependent instruction is able to perform just-in-time execution.
The latency of a producer-consumer pair may be inconsistent due to processor cores having result forwarding hardware between a restricted set of functional units. The unified pick queue is able to pick instructions despite these inconsistencies without penalizing instructions with a late conservative pick. The unified pick queue also is able to both determine which instructions are ready to be picked and determine from this pool of ready instructions which ones are not gated by older ready instructions assigned to a same hardware resource or hardware resource group. In one embodiment, the unified pick queue is able to perform these determinations within a single clock cycle despite a high clock frequency demand.
These and other embodiments will become apparent upon reference to the following description and accompanying drawings.
While the invention is susceptible to various modifications and alternative forms, specific embodiments are shown by way of example in the drawings and are herein described in detail. It should be understood, however, that drawings and detailed description thereto are not intended to limit the invention to the particular form disclosed, but on the contrary, the invention is to cover all modifications, equivalents and alternatives falling within the spirit and scope of the present invention as defined by the appended claims.
In the following description, numerous specific details are set forth to provide a thorough understanding of the present invention. However, one having ordinary skill in the art should recognize that the invention may be practiced without these specific details. In some instances, well-known circuits, structures, signals, computer program instruction, and techniques have not been shown in detail to avoid obscuring the present invention.
Referring to
In one embodiment, processor core 100 may support execution of multiple threads. Multiple instantiations of a same processor core 100 that is able to concurrently execute multiple threads may provide high throughput execution of server applications while maintaining power and area savings. A given thread may include a set of instructions that may execute independently of instructions from another thread. For example, an individual software process may consist of one or more threads that may be scheduled for execution by an operating system. Such a core 100 may also be referred to as a multithreaded (MT) core. In one embodiment, core 100 may be configured to concurrently execute instructions from a variable number of threads, such as up to eight concurrently executing threads.
In one embodiment, core 100 may be configured for multithreaded, out-of-order execution. More specifically, in one embodiment, core 100 may be configured to perform dynamic multithreading. Generally speaking, under dynamic multithreading, the execution resources of core 100 may be configured to efficiently process varying types of computational workloads that exhibit different performance characteristics and resource requirements. Such workloads may vary across a continuum that emphasizes different combinations of individual-thread and multiple-thread performance.
At one end of the continuum, a computational workload may include a number of independent tasks, where completing the aggregate set of tasks within certain performance criteria (e.g., an overall number of tasks per second) is a more significant factor in system performance than the rate at which any particular task is completed. For example, in certain types of server or transaction processing environments, there may be a high volume of individual client or customer requests (such as web page requests or file system accesses). In this context, individual requests may not be particularly sensitive to processor performance. For example, requests may be input/output (I/O)-bound rather than processor-bound-completion of an individual request may need I/O accesses (e.g., to relatively slow memory, network, or storage devices) that dominate the overall time needed to complete the request, relative to the processor effort involved. Thus, a processor that is capable of concurrently processing many such tasks (e.g., as independently executing threads) may exhibit better performance on such a workload than a processor that emphasizes the performance of only one or a small number of concurrent tasks.
At the other end of the continuum, a computational workload may include individual tasks whose performance is highly processor-sensitive. For example, a task that involves significant mathematical analysis and/or transformation (e.g., cryptography, graphics processing, scientific computing) may be more processor-bound than I/O-bound. Such tasks may benefit from processors that emphasize single-task performance, for example through speculative execution and exploitation of instruction-level parallelism.
Dynamic multithreading represents an attempt to dynamically allocate processor resources in a manner that flexibly adapts to workloads that vary along the continuum described above. In one embodiment, core 100 may be configured to implement fine-grained multithreading, in which core 100 may select instructions to execute from among a pool of instructions corresponding to multiple threads, such that instructions from different threads may be scheduled to execute adjacently. For example, in a pipelined embodiment of core 100 employing fine-grained multithreading, instructions from different threads may occupy adjacent pipeline stages, such that instructions from several threads may be in various stages of execution during a given core processing cycle. Through the use of fine-grained multithreading, core 100 may be configured to efficiently process workloads that depend more on concurrent thread processing than individual thread performance.
In one embodiment, core 100 may also be configured to implement out-of-order processing, speculative execution, register renaming and/or other features that improve the performance of processor-dependent workloads. Moreover, core 100 may be configured to dynamically allocate a variety of hardware resources among the threads that are actively executing at a given time, such that if fewer threads are executing, each individual thread may be able to take advantage of a greater share of the available hardware resources. This may result in increased individual thread performance when fewer threads are executing, while retaining the flexibility to support workloads that exhibit a greater number of threads that are less processor-dependent in their performance. In various embodiments, the resources of core 100 that may be dynamically allocated among a varying number of threads may include branch resources (e.g., branch predictor structures), load/store resources (e.g., load/store buffers and queues), instruction completion resources (e.g., reorder buffer structures and commit logic), instruction issue resources (e.g., instruction selection and scheduling structures), register rename resources (e.g., register mapping tables), and/or memory management unit resources (e.g., translation lookaside buffers, page walk resources).
In the illustrated embodiment, core 100 includes an instruction fetch unit (IFU) 202 that includes an L1 instruction cache 205. IFU 202 is coupled to a memory management unit (MMU) 270, L2 interface 265, and trap logic unit (TLU) 275. IFU 202 is additionally coupled to an instruction processing pipeline that begins with a select unit 210 and proceeds in turn through a decode unit 215, a rename unit 220, a pick unit 225, and an issue unit 230. Issue unit 230 is coupled to issue instructions to any of a number of instruction execution resources: an execution unit 0 (EXU0) 235, an execution unit 1 (EXU1) 240, a load store unit (LSU) 245 that includes a L1 data cache 250, and/or a floating point/graphics unit (FGU) 255. These instruction execution resources are coupled to a working register file 260. Additionally, LSU 245 is coupled to L2 interface 265 and MMU 270.
In the following discussion, exemplary embodiments of each of the structures of the illustrated embodiment of core 100 are described. However, it is noted that the illustrated partitioning of resources is merely one example of how core 102 may be implemented. Alternative configurations and variations are possible and contemplated.
Instruction fetch unit (IFU) 202 may be configured to provide instructions to the rest of core 100 for execution. In one embodiment, IFU 202 may be configured to select a thread to be fetched, fetch instructions from instruction cache 205 for the selected thread and buffer them for downstream processing, request data from L2 cache 105 in response to instruction cache misses, and predict the direction and target of control transfer instructions (e.g., branches). In some embodiments, IFU 202 may include a number of data structures in addition to instruction cache 205, such as an instruction translation lookaside buffer (ITLB), instruction buffers, and/or structures configured to store state that is relevant to thread selection and processing.
In one embodiment, virtual to physical address translation may occur by mapping a virtual page number to a particular physical page number, leaving the page offset unmodified. Such translation mappings may be stored in an ITLB or a DTLB for rapid translation of virtual addresses during lookup of instruction cache 205 or data cache 250. In the event no translation for a given virtual page number is found in the appropriate TLB, memory management unit 270 may be configured to provide a translation. In one embodiment, MMU 270 may be configured to manage one or more translation tables stored in system memory and to traverse such tables (which in some embodiments may be hierarchically organized) in response to a request for an address translation, such as from an ITLB or DTLB miss. (Such a traversal may also be referred to as a page table walk or a hardware table walk.) In some embodiments, if MMU 270 is unable to derive a valid address translation, for example if one of the memory pages including a necessary page table is not resident in physical memory (i.e., a page miss), MMU 270 may be configured to generate a trap to allow a memory management software routine to handle the translation.
During the course of operation of some embodiments of core 100, exceptional events may occur. For example, an instruction from a given thread that is selected for execution by select unit 210 may not be a valid instruction for the ISA implemented by core 100 (e.g., the instruction may have an illegal opcode), a floating-point instruction may produce a result that needs further processing in software, MMU 270 may not be able to complete a page table walk due to a page miss, a hardware error (such as uncorrectable data corruption in a cache or register file) may be detected, or any of numerous other possible architecturally-defined or implementation-specific exceptional events may occur.
In one embodiment, trap logic unit 275 may be configured to manage the handling of such events. For example, TLU 275 may be configured to receive notification of an exceptional event occurring during execution of a particular thread, and to cause execution control of that thread to vector to a supervisor-mode software handler (i.e., a trap handler) corresponding to the detected event. Such handlers may include, for example, an illegal opcode trap handler configured to return an error status indication to an application associated with the trapping thread and possibly terminate the application, a floating-point trap handler configured to fix up an inexact result, etc. In one embodiment, TLU 275 may be configured to flush all instructions from the trapping thread from any stage of processing within core 100, without disrupting the execution of other, non-trapping threads.
In the illustrated embodiment, L2 interface 265 may be configured to provide a centralized interface to the L2 cache associated with a particular core 102, on behalf of the various functional units that may generate L2 accesses. In one embodiment, L2 interface 265 may be configured to maintain queues of pending L2 requests and to arbitrate among pending requests to determine which request or requests may be conveyed to L2 cache during a given execution cycle.
In one embodiment, during each execution cycle of core 100, IFU 202 may be configured to select one thread that will enter the IFU processing pipeline. Thread selection may take into account a variety of factors and conditions, some thread-specific and others IFU-specific. For example, certain instruction cache activities (e.g., cache fill), i-TLB activities, or diagnostic activities may inhibit thread selection if these activities are occurring during a given execution cycle. Additionally, individual threads may be in specific states of readiness that affect their eligibility for selection. For example, a thread for which there is an outstanding instruction cache miss may not be eligible for selection until the miss is resolved.
In some embodiments, those threads that are eligible to participate in thread selection may be divided into groups by priority, for example depending on the state of the thread or of the ability of the IFU pipeline to process the thread. In such embodiments, multiple levels of arbitration may be employed to perform thread selection: selection occurs first by group priority, and then within the selected group according to a suitable arbitration algorithm (e.g., a least-recently-fetched algorithm). However, it is noted that any suitable scheme for thread selection may be employed, including arbitration schemes that are more complex or simpler than those mentioned here.
Once a thread has been selected for fetching by IFU 202, instructions may actually be fetched for the selected thread. To perform the fetch, in one embodiment, IFU 202 may be configured to generate a fetch address to be supplied to instruction cache 205. In various embodiments, the fetch address may be generated as a function of a program counter associated with the selected thread, a predicted branch target address, or an address supplied in some other manner (e.g., through a test or diagnostic mode). The generated fetch address may then be applied to instruction cache 205 to determine whether there is a cache hit.
In some embodiments, accessing instruction cache 205 may include performing fetch address translation (e.g., in the case of a physically indexed and/or tagged cache), accessing a cache tag array, and comparing a retrieved cache tag to a requested tag to determine cache hit status. If there is a cache hit, IFU 202 may store the retrieved instructions within buffers for use by later stages of the instruction pipeline. If there is a cache miss, IFU 202 may coordinate retrieval of the missing cache data from L2 cache 105. In some embodiments, IFU 202 may also be configured to prefetch instructions into instruction cache 205 before the instructions are actually requested to be fetched.
In many ISAs, instruction execution proceeds sequentially according to instruction addresses (e.g., as reflected by one or more program counters). However, control transfer instructions (CTIs) such as branches, call/return instructions, or other types of instructions may cause the transfer of execution from a current fetch address to a nonsequential address. As mentioned above, IFU 202 may be configured to predict the direction and target of CTIs (or, in some embodiments, a subset of the CTIs that are defined for an ISA) in order to reduce the delays incurred by waiting until the effect of a CTI is known with certainty.
Through the operations discussed above, IFU 202 may be configured to fetch and maintain a buffered pool of instructions from one or multiple threads, to be fed into the remainder of the instruction pipeline for execution. Generally speaking, select unit 210 may be configured to select and schedule threads for execution. In one embodiment, during any given execution cycle of core 102, select unit 210 may be configured to select up to one ready thread out of the maximum number of threads concurrently supported by core 102 (e.g., 8 threads), and may select up to two instructions from the selected thread for decoding by decode unit 215, although in other embodiments, a differing number of threads and instructions may be selected. In various embodiments, different conditions may affect whether a thread is ready for selection by select unit 210, such as branch mispredictions, unavailable instructions, or other conditions. To ensure fairness in thread selection, some embodiments of select unit 210 may employ arbitration among ready threads (e.g. a least-recently-used algorithm).
The particular instructions that are selected for decode by select unit 210 may be subject to the decode restrictions of decode unit 215; thus, in any given cycle, fewer than the maximum possible number of instructions may be selected. Additionally, in some embodiments, select unit 210 may be configured to allocate certain execution resources of core 102 to the selected instructions, so that the allocated resources will not be used for the benefit of another instruction until they are released. For example, select unit 210 may allocate resource tags for entries of a reorder buffer, load/store buffers, or other downstream resources that may be utilized during instruction execution.
Generally, decode unit 215 may be configured to prepare the instructions selected by select unit 210 for further processing. Decode unit 215 may be configured to identify the particular nature of an instruction (e.g., as specified by its opcode) and to determine the source and sink (i.e., destination) registers encoded in an instruction, if any. In some embodiments, decode unit 215 may be configured to detect certain dependencies among instructions, to remap architectural registers to a flat register space, and/or to convert certain complex instructions to two or more simpler instructions for execution. Additionally, in some embodiments, decode unit 215 may be configured to assign instructions to slots for subsequent scheduling. In one embodiment, two slots 0-1 may be defined, where slot 0 includes instructions executable in load/store unit 245 or execution units 235-240, and where slot 1 includes instructions executable in execution units 235-240, floating point/graphics unit 255, and any branch instructions. However, in other embodiments, other numbers of slots and types of slot assignments may be employed, or slots may be omitted entirely.
Register renaming may facilitate the elimination of certain dependencies between instructions (e.g., write-after-read or “false” dependencies), which may in turn prevent unnecessary serialization of instruction execution. In one embodiment, rename unit 220 may be configured to rename the logical (i.e., architected) destination registers specified by instructions by mapping them to a physical register space, resolving false dependencies in the process. In some embodiments, rename unit 220 may maintain mapping tables that reflect the relationship between logical registers and the physical registers to which they are mapped.
Once decoded and renamed, instructions may be ready to be scheduled for execution. In the illustrated embodiment, pick unit 225 may be configured to pick instructions that are ready for execution and send the picked instructions to issue unit 230. In one embodiment, pick unit 225 may be configured to maintain a pick queue that stores a number of decoded and renamed instructions as well as information about the relative age and status of the stored instructions. During each execution cycle, this embodiment of pick unit 225 may pick up to one instruction per slot. For example, taking instruction dependency and age information into account, for a given slot, pick unit 225 may be configured to pick the oldest instruction for the given slot that is ready to execute.
In some embodiments, pick unit 225 may be configured to support load/store speculation by retaining speculative load/store instructions (and, in some instances, their dependent instructions) after they have been picked. This may facilitate replaying of instructions in the event of load/store misspeculation. Additionally, in some embodiments, pick unit 225 may be configured to deliberately insert “holes” into the pipeline through the use of stalls, e.g., in order to manage downstream pipeline hazards such as synchronization of certain load/store or long-latency FGU instructions.
Issue unit 230 may be configured to provide instruction sources and data to the various execution units for picked instructions. In one embodiment, issue unit 230 may be configured to read source operands from the appropriate source, which may vary depending upon the state of the pipeline. For example, if a source operand depends on a prior instruction that is still in the execution pipeline, the operand may be bypassed directly from the appropriate execution unit result bus. Results may also be sourced from register files representing architectural (i.e., user-visible) as well as non-architectural state. In the illustrated embodiment, core 102 includes a working register file 260 that may be configured to store instruction results (e.g., integer results, floating point results, and/or condition code results) that have not yet been committed to architectural state, and which may serve as the source for certain operands. The various execution units may also maintain architectural integer, floating-point, and condition code state from which operands may be sourced.
Instructions issued from issue unit 230 may proceed to one or more of the illustrated execution units for execution. In one embodiment, each of EXU0235 and EXU1240 may be similarly or identically configured to execute certain integer-type instructions defined in the implemented ISA, such as arithmetic, logical, and shift instructions. In the illustrated embodiment, EXU0235 may be configured to execute integer instructions issued from slot 0, and may also perform address calculation and for load/store instructions executed by LSU 245. EXU1240 may be configured to execute integer instructions issued from slot 1, as well as branch instructions. In one embodiment, FGU instructions and multi-cycle integer instructions may be processed as slot 1 instructions that pass through the EXU1240 pipeline, although some of these instructions may actually execute in other functional units.
In some embodiments, architectural and non-architectural register files may be physically implemented within or near execution units 235-240. It is contemplated that in some embodiments, core 100 may include more or fewer than two integer execution units, and the execution units may or may not be symmetric in functionality. Also, in some embodiments execution units 235-240 may not be bound to specific issue slots, or may be differently bound than just described.
Floating point/graphics unit 255 may be configured to execute and provide results for certain floating-point and graphics-oriented instructions defined in the implemented ISA. For example, in one embodiment FGU 255 may implement single- and double-precision floating-point arithmetic instructions compliant with the IEEE 754-1985 floating-point standard.
Load store unit 245 may be configured to process data memory references, such as integer and floating-point load and store instructions and other types of memory reference instructions. LSU 245 may include a data cache 250 as well as logic configured to detect data cache misses and to responsively request data from L2 cache 105. In one embodiment, data cache 250 may be configured as a set-associative, write-through cache in which all stores are written to L2 cache 105 regardless of whether they hit in data cache 250. As noted above, the actual computation of addresses for load/store instructions may take place within one of the integer execution units, though in other embodiments, LSU 245 may implement dedicated address generation logic. In some embodiments, LSU 245 may implement an adaptive, history-dependent hardware prefetcher configured to predict and prefetch data that is likely to be used in the future, in order to increase the likelihood that such data will be resident in data cache 250 when it is needed.
In various embodiments, LSU 245 may implement a variety of structures configured to facilitate memory operations. For example, LSU 245 may implement a data TLB to cache virtual data address translations, as well as load and store buffers configured to store issued but not-yet-committed load and store instructions for the purposes of coherency snooping and dependency checking. LSU 245 may include a miss buffer configured to store outstanding loads and stores that cannot yet complete, for example due to cache misses. In one embodiment, LSU 245 may implement a store queue configured to store address and data information for stores that have committed, in order to facilitate load dependency checking. LSU 245 may also include hardware configured to support atomic load-store instructions, memory-related exception detection, and read and write access to special-purpose registers (e.g., control registers).
Regarding the pick queue in the pick unit 225, in order to maintain high performance, the selection of multiple decoded and renamed instructions to send to the issue unit 230 occurs in a single pipeline stage. As described above, a pick queue within the pick unit 225 may store a number of decoded and renamed instructions as well as information about the relative age and status of the stored instructions. During a single clock cycle Pick pipeline stage, instructions may be picked from a number of instructions stored in the pick queue within the pick unit 225. Picked instructions are sent to the issue unit 230.
Logic within the pick unit 225 may split the instruction pick operation into two tasks, wherein a first task is responsible for dataflow ordering, such as instruction dependencies. A subsequent second task is responsible for resource allocation and age tracking. A dependency cycle formed between logic to perform the first task which identifies ready instructions, and logic to perform the second task which selects a set of ready instructions for execution, forms a tight loop which is well known as critical to performance as described by E. Borch, et al. Loose loops sink chips, in Proceedings of HPCA-8, February 2002.
Typically, during the first task, which is responsible for dataflow ordering, an instruction is considered ready if it is not dependent on any other instructions in the pick queue. In order to determine an instruction stored in the pick queue is now able to obtain source operands from completed older (in program order) instructions, a broadcast of the results may be used to notify younger (in program order) dependent instructions. Generally, a broadcast observes results being generated and notifies waiting dependent instructions that are now ready for execution. Further details of the pick operation, including both first and second tasks, are provided shortly.
Referring now to
Also, there may be multiple execution pipelines, such as one for integer operations, one for floating-point operations, a third for memory operations, another for graphics processing and/or cryptographic operations, and so forth. The embodiment of pipe stages 300 shown in
In one embodiment, when results are generated by older (in the pipeline) instructions, such as the completion of pipe stage Execute 308, a broadcast of this completion may occur. In one embodiment, the result tags may be broadcast. For example, during the Write Back 310 pipe stage, the results of an integer add instruction may be presented on a results bus. Control logic may detect a functional unit has completed its task (such as in pipe stage Write Back 310). Accordingly, certain control signals may be asserted, such as signal broadcast.d, to indicate to other processor resources that this particular instruction has results available for use. For example, the control signal broadcast.d may be routed to an input of a flip-flop, wherein the output of the flip-flop is an asserted broadcast.q in the subsequent pipe stage.
This broadcast signal (broadcast.d) and storage in a flip-flop is provided as an example. In other embodiments, other control signals may be used in addition to or in place of this broadcast signal. Other control signals may include a corresponding valid signal and results tags routed to comparators within the pick queue, a corresponding valid signal and decoded pick queue entry number input to a word line driver, or otherwise. In the example shown, the signaling of available results occurs in pipe stage Write Back 310 in clock cycle 8. This assertion occurs following a last execution clock cycle of an execution pipeline. In this example, the execute pipeline (Execute 308) is a single pipe stage shown in clock cycle 7.
In the clock cycle following the pipe stage Write Back 310, clock cycle 9, younger (in program order) instructions may verify their source operands are ready, since a results broadcast has been conveyed. The logic within the pick unit 225 may pick these younger instructions that were previously waiting for the results. In one embodiment, a single cycle pipe stage, Issue 306, follows for these younger instructions and then they may begin execution in clock cycle 11.
It is noted that the timing may differ in other embodiments. For example, both the broadcast of results and the pick of instructions may occur in the pipe stage Write Back 310 in clock cycle 8. Regardless, the broadcast may be gated by the actual completion of an older (in the pipeline) instruction's operation. Additionally, the earliest start of execution of a younger (in program order) instruction may be dependent on the number of pipe stages between the pipe stage Pick 304 and the pipe stage Execute 308. In this example, there is only a single cycle Issue 306 pipe stage. However, processors may have deep pipelines, and latencies may increase significantly.
In order to improve throughput and to begin the execution of younger (in program order) dependent instructions at an earlier time, a prediction may be made regarding the completion of a corresponding older (in program) order instruction. Alternatively, a reduction may be made in the number of pipe stages between Pick 304 and Execute 308. However, it is noted that modern microarchitectures may have increasing complexity used to achieve higher performance and a reduction in pipeline depth may reduce performance. On the other hand, a prediction may increase performance when the prediction is correct. For example, when an instruction is picked in pipe stage Pick 304, the number of clock cycles until a result is produced by this particular instruction at Write Back 310 may be determinate. In the example of
Turning now to
This prediction signal, labeled “Trigger” in
An entry, or an instruction, in the pick queue may be considered ready if it is not dependent on any other instructions in the pick queue. In addition, as shown in
One task of an instruction pick operation corresponding to dataflow ordering may be to provide a prediction, such as the above discussed Trigger bit. This Trigger bit may be used to drive just-in-time execution of a subsequent younger dependent instruction. Each instruction that produces a result may be referred to as a producer. Each younger dependent instruction with respect to the producer may be referred to as a consumer. Control logic corresponding to the producer may determine the latency between the clock cycle the producer is picked and the clock cycle the Trigger bit is asserted. This determination may be performed at any of a variety of times. For example, in one embodiment, the determination may be made in any clock cycle between the clock cycle the producer is decoded and the clock cycle the producer is picked.
For example, referring to
Turning now to
A single entry of the pick queue 400 may comprise corresponding entries in each of the instruction array 410, age matrix 430, and dependency matrix 440. For example, a first entry in the pick queue 400 may comprise entries 432a, 412a, and 442a. A second entry in the store queue 400 may comprise entries 432b, 412b, and 442b, and so on. In general, the pick queue 400 may be responsible for scheduling multiple instructions each clock cycle for execution.
One task of an instruction pick operation regarding dataflow ordering utilizes the instruction array 410 and the dependency matrix 440. The instruction array 410 in
In one embodiment, the entry number 420 corresponds to the position of an entry in the pick queue 400, wherein entry 0 may be configured to be at the top of instruction array 410 or at the bottom depending on logic preferences. The entry field 420 may be implied rather than an actual stored number. In one embodiment, pick queue entries may be dynamically allocated in a previous (e.g., renaming) pipe stage. The valid field 422 for a queue entry is updated with a value to indicate a valid entry when the entry is allocated. As will be discussed further, a pick queue entry may be deallocated upon instruction completion if, for example, no load/store hazards exist for the corresponding pick queue entry. The valid field 422 may be reset to a value indicating an empty entry when the entry is deallocated.
Picked field 424 may be used to indicate a corresponding instruction has been picked. Also, the picked field 424 may be used to ensure the corresponding instruction is not subsequently re-picked. Between a first clock cycle when a particular instruction is picked and a second clock cycle when the particular entry is deallocated, the particular instruction is not re-picked due to a set value in the picked field 424. In various embodiments, between the first and second clock cycles, the particular instruction may be disqualified as being ready during pipe stage Pick 304. The trigger field 426 may allow a younger dependent instruction to be picked prior to an older instruction that it depends on completing its operation as shown in
Referring to
Subsequently, when the pick unit 225 picks the particular entry in the pick queue 400, state diagram 500 transitions to state 506 which corresponds to a valid and picked state. The corresponding Boolean value for the two fields is {11}. In this state, the corresponding instruction may not be re-picked due to instruction-ready logic that receives this Boolean value as an input and disqualifies the corresponding instruction. In one embodiment, the instruction-ready logic is implemented with dynamic logic to be further described shortly. When the corresponding instruction has completed execution and has no existing impediment to deallocation (e.g., it has no load or store operation hazard), pick unit 225 may deallocate the instruction from the pick queue 400. In this case, state diagram 500 returns to initial state 502.
Returning back to
In one embodiment, the pick unit 225 may handle multi-cycle instructions in the FGU 255 pipeline differently from multi-cycle load operations. Recall that an instruction in the pick queue may be considered ready if it is not dependent on any other instructions in the pick queue or if instructions it is dependent on have been picked and triggered. The former case applies to multi-cycle FGU 255 instructions. The latter case applies to multi-cycle load instructions. A multi-cycle FGU 255 instruction that has been picked may generate results at a predetermined clock cycle. Multi-cycle FGU 255 instructions may include integer and floating-point (FP) instructions executed on the FGU 255 pipeline. Unless a pipeline flush occurs, such as for an exception, latency for the generation of the result of the multi-cycle FGU 255 instruction is known. Therefore, a corresponding entry within the pick queue 400 may be deallocated between the cycle the multi-cycle FGU 255 instruction is picked and the cycle of the generation of the corresponding result.
The deallocation described above may allow younger dependent instructions to be picked and begin execution just-in-time to receive the corresponding generated result data of the multi-cycle FGU 255 instruction. In contrast, a multi-cycle load instruction is not deallocated from the pick queue 400 when it is picked. This load instruction may later miss in the data cache. Therefore, a multi-cycle load instruction utilizes the trigger field 426, and not deallocation from the pick queue 400, in order to allow younger dependent instructions to be picked and to begin execution just-in-time to receive the corresponding generated result data.
In one embodiment, the pick queue 400 supports the speculative pick and issue of certain instructions such as a load operation. For load instructions, the corresponding trigger bit 426 may be set a predetermined number of cycles after the load is picked. This predetermined number may be based on the assumption that the load instruction hits in the data cache. The setting of this trigger bit 426 may allow younger dependent instructions to be subsequently picked. If the load instruction misses in the data cache, all of the corresponding dependent instructions may replay, or be re-picked by the pick queue 400 for execution at a later time, such as when the load miss data is returned from the memory subsystem. Upon detecting a load miss, the corresponding picked 424 and trigger 426 bits of the load instruction may be set to {1,0}. The corresponding picked 424 and trigger 426 bits of the dependent instructions may be set to {0,0}. In this state, the dependent instructions may not be picked to issue and execute since the original load instruction is picked and not yet triggered. Once the load miss completes, the corresponding fields 424 and 426 of the load instruction may be set to {1,1} allowing the dependent instructions to be possibly picked.
For certain cases such as data translation look-aside buffer (d-TLB) misses, it may be desirable to replay the load or store instruction that caused the TLB miss as well as any dependent instructions. In this case, the corresponding bits 424 and 426 of the original load or store instruction may be set to {0,0} upon completion of the d-TLB miss.
Returning to
The stalled bit allows the pick unit 225 to inject “holes” into the pipeline for execution units as required. In certain cases, the LSU 245 may need “holes” in order to service cache line fill transactions and complete load misses. In order to generate a “hole” for the LSU 245, the pick unit 225 may generate a 40-bit stalled mask for a 40-entry pick queue 400 during a prior clock cycle before the “hole” generation. Other mask and pick queue sizes other than 40 bits are possible and contemplated. This 40-bit stalled mask for a 40-entry pick queue 400 may have set bits for stored instructions in the pick queue 400 and for instructions being written to the pick queue 400 in the current cycle corresponding to particular hardware resources or a group of hardware resources, such as LSU 245 or a group of resources containing LSU 245.
This 40-bit mask may ensure no memory reference instruction will be picked the following cycle. The MMU 270 may generate a 3-cycle stall when it is reloading the data translation-lookaside-buffer (d-TLB). Also, the FGU 255 may need “holes” to be injected in order to finish long latency operations such as a divide operation. In this case, a stalled mask may be generated with logic 1's in locations of the mask corresponding to floating-point (FP) instructions in the pick queue 400. The generated results of these FP instructions may be for both instructions already stored in the pick queue 400 and for instructions being written to the pick queue 400 in the current cycle. Therefore, it may be ensured that no FGU instructions with a valid destination will be picked the following cycle. The described conditions and other contemplated conditions for stalls may be combined in a logical OR operation to produce a final stalled bit to be stored in the instruction status field 428.
Another example of information stored in the instruction status field 428 may be a single bit that is asserted when it is known the corresponding instruction is an instruction type that produces results in a single cycle. For example, it may be known that an integer arithmetic instruction is able to produce a result in a single execution clock cycle.
A third example of a sub-field within the instruction status field 428 may be an enable bit that enables the use of the trigger bit. This enabled bit may be used to avoid forwarding of results from a cryptographic operation. A resource tag corresponding to the instruction stored in the entry may be also held in the instruction status field 428.
In one embodiment, pick queue 400 may be implemented as a unified queue. As stated earlier, a processor core 100 comprising a pick queue 400 may be configured to perform dynamic multithreading. It may not be desirable to instantiate on-chip a separate pick queue 400 for each thread due to on-chip real estate constraints. Also, a single pick queue 400 divided evenly among the maximum number of supported threads, wherein the subdivisions incorporate static allocation, may be inefficient.
For example, if processor core 100 is configured to concurrently execute 8 threads and pick queue 400 has 40 total entries (or 40 entries 412, 40 entries 432, and 40 entries 442), then, in one embodiment, 5 pick queue entries support each thread. Alternatively, a separate 5-entry pick queue may exist for each thread. However, each of the 5 entries in these examples for a particular thread may not be fully utilized or may not be utilized at all, as the particular thread may not be scheduled with work. Another thread, on the other hand, may be utilizing all 5 of its pick queue entries and may need more in order to not experience stalls. Therefore, it may be more efficient and provide higher performance to use dynamic allocation of a unified pick queue 400. In one embodiment, a design tradeoff with dynamic allocation of pick queue entries is the entries do not provide an indication of an ordering among the instructions.
The purpose of maintaining dataflow ordering among the instructions stored in the pick queue 400 is to observe when results are generated and identify instructions that are ready for execution. In one embodiment, content addressable memories (CAMs) may be used within the pick queue to maintain dataflow ordering. The result tags, which may simply be a physical register number, of older (in the pipeline) instructions that have completed execution are broadcasted on a set of result buses, one for each functional unit with a writeback port. Each set of result buses may be connected to comparators, such as xnor gates, at each entry within the pick queue. These comparators allow instructions to determine if their sources match the results being generated. When all of an instruction's sources are ready, the instruction as a whole is considered ready. However, CAMs consume an appreciable amount of both on-chip real estate and electrical power.
In another embodiment, rather than use CAMs to broadcast results, a dependency matrix 440 such as that shown in
The deallocation of speculative instructions, such as load and store instructions, and dependent instructions of the speculative instructions may be handled differently from other instructions. For example, each store operation within processor core 100 may be considered to be a speculative operation in the sense that a miss to a data translation look-aside buffer (d-TLB) may arise during execution of the store instruction by the LSU 245. Similarly, each load operation within processor core 100 may be considered to be a speculative operation in the sense that a miss to the d-TLB, a miss to the data cache, or a read-after-write (RAW) data hazard may arise during execution of the instruction by the LSU 245. When a miss or hazard arises during execution of a load or a store instruction, the pick unit 225 may support various levels of replay capability for dependent instructions after the load instruction. The pick unit 225 may undo a predetermined number of instruction pick clock cycles and subsequently replay the dependent instructions. In one embodiment, independent instructions may not be replayed.
Due to load and store instruction speculation, corresponding entries 442 may not be immediately deallocated in the pick queue 400 once an instruction is picked. The picked bit 424 is used to ensure that the corresponding entry 442 is not re-picked between the cycle an instruction is picked and the cycle an instruction is deallocated from the pick queue 400. Younger dependent instructions in the pick queue 400 may not be picked unless the trigger bit 426 of the corresponding older (in program order) instruction is set.
Referring now to
Each nmos transistor 604 may be coupled via signal 606 to a separate column of a corresponding dependency matrix entry 442. Dynamic circuit 600 may be used to determine whether a particular instruction stored in the pick queue 400 may be placed in a pool of ready instructions. Initially, the output ready 630 of inverter 626 conveys a binary logic value of 1 during the end of a precharge phase of circuit 600. In another embodiment, only inverter 624 may be used to convey ready 630 and an initial logic value of 0, which indicates the entry is not ready.
During an evaluation phase of circuit 600, when a row is entirely clear of dependencies, or a wired-logical-OR on node n 610 provides no discharge, ready 630 remains high. Therefore, the particular instruction is to be placed in a pool of ready instructions for the second half of the instruction pick process. Here, it is noted circuit 600 corresponds to entry N of the dependency matrix 440. An instruction cannot be dependent on itself, so a logic 0 for the corresponding column N is the same as not placing corresponding transistors for column N. For entry N-1 of the dependency matrix 440, column N-1 would be empty and column N would have two nmos transistors 604 and 608.
Turning now to
For each column N-1 to 0 in
An asserted input value Picked0648a corresponds to the instruction in entry 0 of the pick queue 400 having already been picked in an earlier cycle. In one embodiment, this value may be from the stored value in Picked field 424 of entry 0 of pick queue 400. In this case the instruction in entry 0 has already been picked in an earlier cycle. An asserted input value Trigger0648b corresponds to the instruction in entry 0 of the pick queue 400 having been triggered in an earlier cycle. In one embodiment, this value may be from the stored value in Trigger field 426 of entry 0 of pick queue 400.
An asserted input Enable0648c corresponds to the instruction in entry 0 of the pick queue 400 having its corresponding Trigger0648b bit enabled. In one embodiment, each Trigger 648b for each column N-1 to 0 in the example shown is qualified with a corresponding Enable 648c bit. This qualification allows suppression of normal triggering for cases a source operand for a store operation or a floating-point operation comes from a cryptographic operation. In one embodiment, the cryptographic pipeline is 3 clock cycles and overlays the FGU 255 pipeline. There may be internal forwarding buses to allow cryptographic operations to forward to other cryptographic operations. These types of dependencies may be triggered every 3 cycles. These forwarding buses may not be routed to the rest of the FGU 255 or to the LSU 245.
In the case Picked0648a, Trigger0648b, and Enable0648c are all asserted, a logic value 0 is conveyed to the gate input of transistor 604a. If the signal Dependent0606a is asserted, which means the instruction in entry N of pick queue 400 is dependent on the instruction in entry 0 of the pick queue 400, then the dynamic node n 610 is not discharged by column 0. In other words, with respect to the dependency of the instruction in entry N on the instruction in entry 0, if the instruction in entry 0 has already been picked and triggered, then the instruction in entry N may be determined to be ready. As shown with examples regarding the pipe stages in
However, if the instruction in entry N is also dependent on the instruction in entry N-2 and the instruction in entry N-2 has not been picked yet or has not been triggered yet, then a logic value 1 may be asserted on both nodes 606g and 646g. Accordingly, the dynamic node n 610 is discharged and the instruction in entry N is not a candidate to be picked in this current clock cycle. Although determination circuit 700 has an extra nmos transistor 646 in each column N-1 to 0 compared to determination circuit 600, circuit 700 may allow instructions to be candidates for being determined ready in an earlier pipe stage.
In addition, determination circuit 700 has nmos transistors 604j and 608j in column N, whereas determination circuit 600 has no circuitry for column N. In circuit 700, the corresponding instruction stored in entry N of pick queue 400 may be disqualified for being a candidate for being determined ready for other reasons other than being dependent on other instructions. For example, the corresponding instruction may have already been picked. Therefore, the value stored in Picked field 424 of entry N may hold a logic value of 1. This value may be asserted on PickedN 650, which causes BlockReady 642 to be asserted. Then the series combination of nmos transistors 604j and 608j discharge dynamic node n 610 and Ready 630 transitions from a logic high value to a logic low value.
It is noted that the disqualification provided by logic OR gate 640 and its corresponding inputs does not add a series gate delay to the determination of the Ready 630 value. Each entry of pick queue 400 may have corresponding circuitry as logic OR gate 640 and nmos transistors 604j and 608j in a column number matching the entry number. Alternatively, the Ready 630 value may be disqualified further down the logic for picking instructions. However, a series gate delay may need to be added to this logic path that may be already setting the critical path of processor core 100.
The signal StallN 652 may also be used to disqualify the instruction in entry N from being a candidate for being determined ready for an instruction pick. As described above regarding
In addition, a timing condition may be also combined in the logical OR operation that may disqualify the instruction in entry N from being ready. Alternatively, this timing condition may assert a separate signal that is input to logic OR gate 640. This timing condition may be affected by a consumer instruction in a producer-consumer pair as further described below.
Generally speaking, for a particular producer-consumer pair, a corresponding latency may be determined based upon the functional unit chosen for each of the producer and the consumer instructions. The chosen functional unit may be also referred to as the instruction execution resource. Other factors, such as bypass wiring, may further determine the latency. For example, an integer arithmetic producer instruction may be issued to functional unit EXE0235 and an integer arithmetic consumer instruction may be issued to functional unit EXE1240. Selection of the particular functional units may be based, at least in part, on an opcode of the corresponding instruction. Decoding of a portion or all of the opcode may further indicate that the producer instruction receives integer source operands and conveys an integer result. A similar determination may be made for the integer consumer instruction. Based upon a knowledge of the processor design, a deterministic latency between EXE0235 and EXE1240 may be determined. This deterministic latency, such as 1 clock cycle in one embodiment, may be used for the timing of the trigger field 426 as shown in
Similar to the integer instructions discussed above, floating point instructions may have deterministic latencies for particular producer-consumer pairs. For example, a floating-point producer instruction may be issued to FGU 255 based upon a determination the producer instruction receives floating-point source operands and conveys a floating-point result. Therefore, the functional unit FGU 255 may be chosen as the corresponding instruction execution resource. A similar determination may be performed for one or more floating-point consumer instructions. Accordingly, with respect to the floating-point producer instruction, the floating-point consumer instructions may be grouped into a particular class of instructions, wherein this class has a common latency with this same producer.
However, in some cases, two consumers may be dependent on a given producer but the latency timings could be different for both. As an example, most floating point operations may be executed on FGU 255 and read their sources in a particular floating point pipe stage. However, some floating point operations may be executed by an integer execution unit (e.g., EXE0235 or EXE1240) in which case the sources must be read in a particular integer pipe stage. In such cases, it may not be possible to propagate the floating point result to the integer execution unit for consumption in the particular integer pipe stage.
For example, in one embodiment, certain graphics-oriented instructions may not be performed by a coprocessor, but by functional units within core 100. One or more of these graphics-oriented instructions may move data between integer and floating-point registers without conversion. As an example, one such instruction may copy 32 bits from a floating-point source register to the least significant 32 bits of a destination integer register and place a copy of the sign bit in each bit of the 32 most-significant bits of the destination integer register (i.e., sign-extend the destination result from 32 bits to 64 bits). A second instruction may perform the same copy as the first instruction, but place a logic 0 in each bit of the most-significant 32 bits of the integer destination register. A third instruction may copy all 64 bits of the floating-point source register to the integer destination register. Each of these instructions receives a floating-point source operand and conveys a floating-point result. However, rather than issuing to a floating point functional unit (e.g., FGU 255), each of these instructions may be issued to EXE0235 or EXE1240. Accordingly, the latency between these types of floating-point consumer instructions and a corresponding floating-point producer instruction may differ from the latency between floating-point arithmetic consumer instructions and the same floating-point producer instruction which both issue to floating point functional units. In other words, this same floating-point producer instruction may have a first class of consumer instructions (i.e. receive one or more floating-point source operands and convey a floating-point result) with two types of consumers. The first type of consumer instructions (i.e. floating-point arithmetic instructions executed in FGU 255) have a first deterministic latency value and the second type of consumer instructions (i.e. floating-point move instructions executed in EXE0235) have a second deterministic latency value.
In one example, the second latency value described above may be greater than the first latency value due to a lack of similar bypass wire routes in EXE0235 versus FGU 255, physical on-chip placements of EXE0235 and FGU 255, or otherwise. One solution to addressing these differing latencies is to use the longer second latency value to determine which pipe stage to assert the producer's trigger field 426. However, each consumer instruction of the first type would then experience a penalty for waiting one or more extra clock cycles before being a candidate to be picked. An alternate solution is to disqualify only consumer instructions of the second type from being picked during a predetermined number of clock cycles once the producer's trigger field 426 is asserted. For example, the BlockReady 642 signal of
In another example, five combinations of producer-consumer pairs may include 14 clock cycles of latency. Referring to
However, an additional sixth combination of this producer-consumer pair may include 16 clock cycles of latency, since it may not be able to forward results in a manner similar to the other five instructions within the instruction type. All six consumer instructions may be in a same class of consumer instructions (i.e. they may all receive one or more same type source operands and convey a same type result). However, the sixth instruction may be of a different type that is executed in a different functional unit, or instruction execution resource. In other words, the latency of a producer-consumer pair may be a function of both the producer and consumer instructions. This inconsistency may be due to processor cores with result forwarding hardware between a restricted set of functional units. In these processors the latencies from the same source instruction may be different, depending on the type of producer functional unit and the consumer functional unit.
Referring again to
In order that a consumer not be picked too early and may not arrive at a functional hardware resource without available source operands, each of the six combinations of this producer-consumer pair may be set to have a latency of 13 clock cycles, rather than 11 clock cycles. Processor performance may significantly decrease in such a case.
Alternatively, logic for a consumer may determine the consumer has a different timing from other consumers dependent on a particular producer—such as the sixth combination in the above example. Opcodes of the producer and consumer and allocation comparators used to set corresponding bits in the dependency matrix 440 may be used to determine which consumer has the longer latency. Also, a table may be used to store producer-consumer pair timings. The difference in latency, such as 2 clock cycles in the above example, may be stored. Then, in the above example, the producer may use the reduced value (12 clock cycles of latency instead of 14 clock cycles) to assert its corresponding Trigger bit following being picked. The first five consumers may be determined to be ready for an instruction pick in the same clock cycle the corresponding Trigger bit is asserted, since both the Picked 424 and Trigger 426 of the producer are asserted. Referring again to
In one embodiment, the 2 cycle difference between the sixth consumer and the other five consumers may be counted down. The above logic operations may repeat and the sixth consumer is disqualified again due to the asserted BlockReady 642 signal. The difference may be decremented again, and in the following clock cycle, the BlockReady 642 signal may not be asserted and the sixth consumer may be determined to be ready for an instruction pick. Now the sixth consumer may not be picked early and the other five consumers may not be penalized by 2 additional clock cycles before they are determined to be ready for an instruction pick.
Turning now to
These latencies may be used to determine the clock cycle the generated result of the producer instruction is available for use. For the first three consumer instructions, the calculated clock cycle may be used to work backwards and determine the appropriate clock cycle the producer's corresponding Trigger bit 426 may be asserted for just-in-time execution. Here, clock cycle 15 is determined after subtracting the number of cycles used by the Pick and Issue stages. A determination is not calculated for the fourth consumer instruction. Rather, the producer may assert the Trigger bit 426 in the same clock cycle as for the other three consumer instructions. The known predetermined Is difference in clock cycles between the two groups of consumer instructions may be used to delay the fourth consumer instruction from being picked. For example, an asserted delay signal may be conveyed to the logic OR gate 640 for two clock cycles aligned with the assertion of the producer's trigger bit 426. Therefore, the first three consumer instructions are not penalized by the two cycle difference with the fourth consumer instruction, Consumer3. Also, the fourth consumer instruction is not picked too early.
Once a pool of instructions is determined to be ready for an instruction pick, the next step is to pick instructions from this pool based on hardware resource groupings, or slots. A slot may correspond to one or more functional units such as a load/store unit (LSU) 245, an integer execution unit 235 or 240, a branch unit, cryptographic unit, a floating-point unit 255, or other. Each instruction may be assigned by decode unit 215 to a slot prior to being written into the pick queue 400.
In one embodiment, a slot 0 assignment may correspond to both memory reference address resolutions and integer arithmetic or logical instructions. A slot 1 assignment may correspond to integer arithmetic or logical instructions, branch instructions, floating-point instructions, cryptographic instructions, and integer instructions executed by the floating-point graphics unit (FGU), such as integer multiply and divide instructions. A slot 2 assignment may correspond to store data acquisition operations. Other assignments and number of slots are possible and contemplated. Instruction picking within a slot may be independent of instruction picking within other slots. In order to be picked, an instruction must be the oldest ready instruction for a particular slot.
Each entry in the pick queue 400 has an age vector 432. The age vector 432 may be generated for each instruction during a renaming clock cycle before the instruction is written into the pick queue 400. The age vector 432 may be a bit vector that represents all instructions currently in the pick queue 400 that are older than the instruction being written on a slot basis.
A bit vector block with N bits may be generated to ensure that the oldest ready instruction for each slot is picked each cycle. Each bit of this block vector may represent an entry in the pick queue 400. Each bit of this block vector may be constructed from a logical AND-OR reduction of a corresponding ready vector, which is made up of the values of Ready 630 for each entry, with the corresponding age vector for the entry. In one embodiment, a final pick bit for each entry of the pick queue 400 may be a combination in a logical AND operation of a corresponding Ready 630 value and an inverted corresponding block value. In another embodiment, the valid field 422 may be additionally combined if it hasn't already been combined. In some cases, timing considerations may make stalling a corresponding instruction via the StallN 652 value impractical. For these cases, a corresponding inverted “late_stall” value may be applied as a fourth term in the logical AND operation to determine a final pick bit for each entry in the pick queue 400.
Turning now to
In the age matrix 430, instructions older than a given instruction are stored as a logic value 1 in the corresponding column of the pick queue 500 entry for that instruction. For example, instruction 2 is mapped to entry 1 of the pick queue 500, which has bits set for entries 2 and 3 in the age matrix 430. These entries correspond to instruction 0 and instruction 1. Hence, instructions 0 and 1 are older than instruction 2. For simplicity, the information in Table 520 assumes that all instructions are assigned to the same slot. In practice, multiple slots may be active in the pick queue 500 at the same time. The age matrix 430 is maintained on a slot basis and is used to track the age relationships between all instructions on a slot basis. Bits set in the age vector for a given entry of the pick queue 500 may be instructions that are older than the respective instruction and also mapped to the same slot.
The dependency matrix 440 functions in a similar manner to the age matrix 430. For a given instruction, other instructions it depends on have a logic value 1 stored in the corresponding column of the pick queue 500 entry for that instruction. For example, instruction 2 is mapped to entry 1 of the pick queue 500, which has a logic value 1 stored in column 2. Column 2 corresponds to entry 2, which corresponds to instruction 0. Hence, instruction 2 is dependent on instruction 0. Instruction 2 may not be picked before instruction 0 is picked. Instruction array 410 holds the valid 422, picked 424, trigger 426, and single 530 fields for each of the four instructions.
Referring now to
A processor core 100 may be executing instructions of one or more software applications. In block 802, core 100 fetches instructions of one or more software applications. These fetched instructions are decoded and renamed in block 804. Renamed instructions are allocated in pick queue 400 in block 806.
If a particular instruction stored in pick queue 400 has source operands ready (conditional block 808), then control flow of method 800 moves to block A. Otherwise, a determination is made regarding when the source operands will be ready for the particular instruction. If the particular instruction is dependent on a nondeterministic instruction (conditional block 810), then in block 812 the particular instruction waits for an indication or indications that the corresponding source operands are ready. A nondeterministic instruction may be an instruction wherein the clock cycle of the generation of the corresponding result is not known or able to be predetermined. An example is a load instruction that misses the data cache and is experiencing a cache line fill transaction. Another example is a floating-point divide instruction wherein the generation of the result is dependent on the source data.
If the particular instruction is not dependent on a nondeterministic instruction (conditional block 810), then in block 814 the particular instruction waits for an indication to begin pipeline traversal, such as being picked in the pick queue 400, for just-in-time execution. One example of an indication is the series combination of the picked 424 and trigger bit 426 fields of a producer, wherein the particular instruction is a consumer.
As discussed in an example above regarding producer-consumer pairs, a sixth consumer of a particular instruction type may have a longer latency than five other consumers of the particular instruction type. If this is the case (conditional block 816), then in block 818, the consumer waits the appropriate number of additional delay cycles before control flow of method 800 moves to block A.
Referring now to
Control flow of method 850 moves from block A to conditional block 852. If there are instructions stored in the pick queue 400 that are older than a particular instruction previously determined to be ready for an instruction pick, and these older instructions are assigned to a same hardware resource or a same hardware resource group, then control flow of method 850 moves to block 854. In block 854, the particular instruction waits for assigned hardware resources to become available.
Otherwise, in block 856, the particular instruction may be picked for issue and execution. However, it is still possible for the instruction to be delayed due to stall signals. In block 858, a determination is made regarding the particular instruction for the latency to be used for just-in-time execution of younger dependent instructions. This value may be used to assert a corresponding trigger bit 426 in a subsequent clock cycle.
It is noted that the above-described embodiments may comprise software. In such an embodiment, the program instructions that implement the methods and/or mechanisms may be conveyed or stored on a computer readable medium. Numerous types of media which are configured to store program instructions are available and include hard disks, floppy disks, CD-ROM, DVD, flash memory, Programmable ROMs (PROM), random access memory (RAM), and various other forms of volatile or non-volatile storage.
Although the embodiments above have been described in considerable detail, numerous variations and modifications will become apparent to those skilled in the art once the above disclosure is fully appreciated. It is intended that the following claims be interpreted to embrace all such variations and modifications.
