Patent Grant
7200740
This application is related to the following U.S. Patent applications, having a common filing date and a common assignee. Each of these applications is hereby incorporated by reference in its entirety for all purposes:
This invention relates in general to the field of branch prediction in microprocessors, and more particularly to branch target address caching.
Computer instructions are typically stored in successive addressable locations within a memory. When processed by a Central Processing Unit (CPU), or processor, the instructions are fetched from consecutive memory locations and executed. Each time an instruction is fetched from memory, a program counter (PC), or instruction pointer (IP), within the CPU is incremented so that it contains the address of the next instruction in the sequence. This is the next sequential instruction pointer, or NSIP. Fetching of an instruction, incrementing of the program counter, and execution of the instruction continues linearly through memory until a program control instruction is encountered.
A program control instruction, also referred to as a branch instruction, when executed, changes the address in the program counter and causes the flow of control to be altered. In other words, branch instructions specify conditions for altering the contents of the program counter. The change in the value of the program counter because of the execution of a branch instruction causes a break in the sequence of instruction execution. This is an important feature in digital computers, as it provides control over the flow of program execution and a capability for branching to different portions of a program. Examples of program control instructions include jump, conditional jump, call, and return.
A jump instruction causes the CPU to unconditionally change the contents of the program counter to a specific value, i.e., to the target address for the instruction where the program is to continue execution. A conditional jump causes the CPU to test the contents of a status register, or possibly compare two values, and either continue sequential execution or jump to a new address, called the target address, based on the outcome of the test or comparison. A call instruction causes the CPU to unconditionally jump to a new target address, but also saves the value of the program counter to allow the CPU to return to the program location it is leaving. A return instruction causes the CPU to retrieve the value of the program counter that was saved by the last call instruction, and return program flow back to the retrieved instruction address.
In early microprocessors, execution of program control instructions did not impose significant processing delays because such microprocessors were designed to execute only one instruction at a time. If the instruction being executed was a program control instruction, by the end of execution the microprocessor would know whether it should branch, and if it was supposed to branch, it would know the target address of the branch. Thus, whether the next instruction was sequential, or the result of a branch, it would be fetched and executed.
Modern microprocessors are not so simple. Rather, it is common for modern microprocessors to operate on several instructions at the same time, within different blocks or pipeline stages of the microprocessor. Hennessy and Patterson define pipelining as, “an implementation technique whereby multiple instructions are overlapped in execution.” Computer Architecture: A Quantitative Approach, 2nd edition, by John L. Hennessy and David A. Patterson, Morgan Kaufmann Publishers, San Francisco, Calif., 1996. The authors go on to provide the following excellent illustration of pipelining:
“A pipeline is like an assembly line. In an automobile assembly line, there are many steps, each contributing something to the construction of the car. Each step operates in parallel with the other steps, though on a different car. In a computer pipeline, each step in the pipeline completes a part of an instruction. Like the assembly line, different steps are completing different parts of the different instructions in parallel. Each of these steps is called a pipe stage or a pipe segment. The stages are connected one to the next to form a pipe—instructions enter at one end, progress through the stages, and exit at the other end, just as cars would in an assembly line.”
Thus, as instructions are fetched, they are introduced into one end of the pipeline. They proceed through pipeline stages within a microprocessor until they complete execution. In such pipelined microprocessors, it is often not known whether a branch instruction will alter program flow until it reaches a late stage in the pipeline. However, by this time, the microprocessor has already fetched other instructions and is executing them in earlier stages of the pipeline. If a branch instruction causes a change in program flow, all of the instructions in the pipeline that followed the branch instruction must be thrown out. In addition, the instruction specified by the target address of the branch instruction must be fetched. Throwing out the intermediate instructions and fetching the instruction at the target address creates processing delays in such microprocessors, referred to as a branch penalty.
To alleviate this delay problem, many pipelined microprocessors use branch prediction mechanisms in an early stage of the pipeline that make predictions of branch instructions. The branch prediction mechanisms predict the outcome, or direction, of the branch instruction, i.e., whether the branch will be taken or not taken. The branch prediction mechanisms also predict the branch target address of the branch instruction, i.e., the address of the instruction that will be branched to by the branch instruction. The processor then branches to the predicted branch target address, i.e., fetches subsequent instructions according to the branch prediction, sooner than it would without the branch prediction, thereby potentially reducing the penalty if the branch is taken.
A branch prediction mechanism that caches target addresses of previously executed branch instructions is referred to as a branch target address cache (BTAC), or branch target buffer (BTB). In a simple BTAC or BTB, when the processor decodes a branch instruction, the processor provides the branch instruction address to the BTAC. If the address generates a hit in the BTAC and the branch is predicted taken, then the processor may use the cached target address from the BTAC to begin fetching instructions at the target address, rather than at the next sequential instruction address.
The benefit of the BTAC over a predictor that merely predicts taken/not taken, such as a branch history table (BHT) is that the BTAC saves the time needed to calculate the target address beyond the time needed to determine that a branch instruction has been encountered. Typically, branch prediction information (e.g., taken/not taken) is stored in the BTAC along with the target address. A BTAC is historically employed at the instruction decode stages of the pipeline. This is because the processor must first determine that a branch instruction is present.
An example of a processor that employs a BTB is the Intel® Pentium® II and III processor. Referring now to
Typically, instructions are fetched from a next sequential fetch address 112, which is simply the current instruction cache 102 fetch address 122 incremented by the size of an instruction cache 102 line by an incrementer 118. However, if a branch instruction is decoded by the instruction decode logic 132, then control logic 114 selectively controls a multiplexer 116 to select the branch target address 136 supplied by the BTB 134 as the fetch address 122 for the instruction cache 102 rather than selecting the next sequential fetch address 112. The control logic 114 selects the instruction cache 102 fetch address 122 based on the pre-decode information 104 from the instruction cache 102 and whether the BTB 134 predicts the branch instruction will be taken or not taken based on an instruction pointer 138 used to index the BTB 134.
Rather than indexing the BTB 134 with the instruction pointer of the branch instruction itself, the Pentium II/III indexes the BTB 134 with the instruction pointer 138 of an instruction prior to the branch instruction being predicted. This enables the BTB 134 to lookup the target address 136 while the branch instruction is being decoded. Otherwise, the processor 100 would have to wait to branch an additional branch penalty delay of waiting to perform the BTB 134 lookup after the branch instruction is decoded. Presumably, once the branch instruction is decoded by the instruction decode logic 132 and the processor 100 knows that the target address 136 was generated based on certainty that a branch instruction is present, only then does the processor 100 branch to the target address 136 provided by the BTB 134 based on the instruction pointer 138 index.
Another example of a processor that employs a BTAC is the AMD® Athlon® processor. Referring now to
As may be observed from
Although the prior methods provide branch prediction improvements, there are disadvantages to the prior methods. A disadvantage of both the prior methods discussed above is that the instruction pre-decode information, and in the case of Athlon the branch target addresses, substantially increase the size of the instruction cache. It has been speculated that for Athlon the branch prediction information essentially doubles the size of the instruction cache. Additionally, the Pentium II/III BTB stores a relatively large amount of branch history information per branch instruction for predicting the branch direction, thereby increasing the size of the BTB.
A disadvantage of the Athlon integrated BTAC is that the integration of the BTAC into the instruction cache causes space usage inefficiency. That is, the integrated instruction cache/BTAC occupies storage space for caching branch instruction information for non-branch instructions as well as branch instructions. Much of the space taken up inside the Athlon instruction cache by the additional branch prediction information is wasted since the instruction cache has a relatively low concentration of branch instructions. For example, a given instruction cache line may have no branches in it, and thus all the space taken up by storing the target addresses and other branch prediction information in the line are unused and wasted.
Another disadvantage of the Athlon integrated BTAC is that of conflicting design goals. That is, the instruction cache size may be dictated by design goals that are different from the design goals of the branch prediction mechanism. Requiring the BTAC to be the same size as the instruction cache, in terms of cache lines, which is inherent in the Athlon scheme, may not optimally meet both sets of design goals. For example, the instruction cache size may be chosen to achieve a certain cache-hit ratio. However, it may be that the required branch target address prediction rate might have been achieved with a smaller BTAC.
Furthermore, because the BTAC is integrated with the instruction cache, the data access time to obtain the cached branch target address is by necessity the same as the access time of the cached instruction bytes. In the case of the relatively large Athlon instruction cache, the access time may be relatively long. A smaller, non-integrated BTAC might have a data access time substantially less than the access time of the integrated instruction cache/BTAC.
The Pentium II/III method does not suffer many of the Athlon integrated instruction cache/BTAC problems mentioned since the Pentium II/III BTB is not integrated with the instruction cache. However, because the Pentium II/III BTB is indexed with the instruction pointer of an already decoded instruction, rather than the instruction cache fetch address, the Pentium II/III solution potentially may not be able to branch as early as the Athlon solution, and therefore, may not reduce the branch penalty as effectively. The Pentium II/III solution potentially addresses this problem by indexing the BTB with the instruction pointer of a previous instruction, or previous instruction group, rather than the actual branch instruction pointer, as mentioned above.
However, a disadvantage of the Pentium II/III method is that some amount of branch prediction accuracy is sacrificed by using the instruction pointer of a previous instruction, rather than the actual branch instruction pointer. The reduction in accuracy is due, in part, because the branch instruction may be reached via multiple instruction paths in the program. That is, instruction pointers of multiple previous instructions to the branch instruction may be cached in the BTB for the same branch instruction. Consequently, multiple entries must be consumed in the BTB for such a branch instruction, thereby reducing the overall number of branch instructions that may be cached in the BTB. The greater the number of instructions previous to the branch instruction used, the greater the number of paths by which the branch instruction may be reached.
Additionally, because using a prior instruction pointer introduces the possibility of multiple paths to the same branch instruction, it potentially takes the Pentium II/III direction predictor in the BTB longer to “warm up”. The Pentium II/III BTB maintains branch history information for predicting the direction of the branch. When a new branch instruction is brought into the processor and cached, the multiple paths to the branch instruction potentially cause the branch history to become updated more slowly than would be the case if only a single path to the branch instruction were possible, resulting in less accurate predictions.
Therefore, what is needed is a branch prediction apparatus that makes efficient use of chip real estate, but also provides accurate branching early in the pipeline to reduce branch penalty.
The present invention provides a branch prediction method and apparatus that makes efficient use of chip real estate, but also provides accurate branching early in the pipeline to reduce branch penalty. Accordingly, in attainment of the aforementioned object, it is a feature of the present invention to provide an apparatus in a processor for speculatively performing a return instruction. The apparatus includes a first call/return stack, configured for pushing thereon a plurality of return addresses of a corresponding plurality of call instructions in response to fetching from an instruction cache a plurality of cache lines predicted to include the corresponding plurality of call instructions, and for popping therefrom a first return address in response to fetching from the instruction cache a cache line predicted to include a return instruction. The first return address is a top one of the plurality of return addresses simultaneously stored in the first call/return stack as a result of the pushing. Each of the plurality of return addresses is pushed onto the first call/return stack prior to decoding the corresponding call instruction. The apparatus also includes a second call/return stack, configured to provide a second return address in response to decoding the return instruction, subsequent to the first call/return stack popping therefrom the first return address. The apparatus also includes a comparator, coupled to the first and second call/return stacks, for comparing the first and second return addresses prior to the return instruction reaching an execution stage of a pipeline of the processor. The execution stage is configured to finally resolve the return instruction. The apparatus also includes control logic, coupled to the comparator, for controlling the processor to branch to the first return address. The control logic subsequently controls the processor to branch to the second return address if the comparator indicates the first and second return addresses do not match.
In another aspect, it is a feature of the present invention to provide a microprocessor for predicting return instruction target addresses. The microprocessor includes an instruction cache, for generating a line of instruction bytes selected by a fetch address. The fetch address is received from an address bus. The microprocessor also includes address selection logic, coupled to the address bus, for selecting the fetch address and providing the fetch address on the address bus. The microprocessor also includes a branch target address cache (BTAC), coupled to the address bus, for caching indications of previously executed return instructions and for providing one of the indications in response to the fetch address. The microprocessor also includes a first call/return stack, coupled to the BTAC, for providing a first return address to the address selection logic in response to the one of the indications. The first call/return stack is configured to simultaneously store a plurality of return addresses. The plurality of return addresses are pushed onto the first call/return stack in response to indications provided from the BTAC of previously executed call instructions in response to the fetch address. The microprocessor also includes decode logic, coupled to the instruction cache, for decoding the line of instruction bytes. The microprocessor also includes a second call/return stack, coupled to the decode logic, for providing a second return address to the address selection logic in response to the decode logic indicating that a return instruction is present in the line of instruction bytes. The second call/return stack is configured to store a plurality of return addresses. The second call/return stack is physically distinct from the first call/return stack. The microprocessor also includes an execution stage, coupled to the decode logic, for finally resolving return instructions. The first and second call/return stacks provide the first and second return addresses to the address selection logic prior to the return instruction reaching the execution stage.
In another aspect, it is a feature of the present invention to provide a method for speculatively branching a microprocessor to a target address of a return instruction. The microprocessor includes an execution stage for finally resolving the return instruction. The method includes pushing onto a first call/return stack a plurality of return addresses of a corresponding plurality of call instructions, causing the plurality of return addresses to be simultaneously stored in the first call/return stack. For each of the plurality of return addresses the pushing is performed prior to decoding of the corresponding call instruction. The method also includes generating a first target address by popping one of the plurality of return addresses off a top of the first call/return stack and branching to the first target address. The method also includes generating a second target address by a second call/return stack subsequent to the branching to the first target address. The second call/return stack is configured to store a plurality of return addresses. The second call/return stack is physically distinct from the first call/return stack. The method also includes comparing the first and second target addresses prior to the return instruction reaching the execution stage and branching to the second target address if the first and second target addresses do not match.
In another aspect, it is a feature of the present invention to provide a microprocessor for predicting return instruction target addresses. The microprocessor includes an instruction cache, for providing a line of instructions in response to a fetch address received on an address bus. The microprocessor also includes a multiplexer, having a plurality of inputs, configured to select one of the plurality of inputs for provision on the address bus as the fetch address to the instruction cache. The microprocessor also includes a speculative branch target address cache (BTAC), coupled to the address bus, for indicating a speculative presence of a return instruction in the line of instructions. The microprocessor also includes a speculative call/return stack, coupled to the speculative BTAC, for providing a speculative return address to a first of the plurality of multiplexer inputs in response to the speculative BTAC indicating the speculative presence of the return instruction. The speculative call/return stack is configured to simultaneously store a plurality of return addresses. The plurality of return addresses are pushed onto the speculative call/return stack in response to instances of the speculative BTAC indicating a speculative presence of a call instruction in the line of instructions. The microprocessor also includes decode logic, configured to receive and decode the line of instructions. The microprocessor also includes a non-speculative call/return stack, coupled to the decode logic, for providing a non-speculative return address to a second of the plurality of multiplexer inputs in response to the decode logic indicating that the return instruction is actually present in the line of instructions. The speculative call/return stack is configured to store a plurality of return addresses. The non-speculative call/return stack is physically distinct from the speculative call/return stack. The microprocessor also includes a comparator, coupled to the speculative and non-speculative call/return stacks, for comparing the speculative and non-speculative return addresses prior to the return instruction reaching an execution stage of a pipeline of the processor. The execution stage is configured to finally resolve the return instruction. The multiplexer selects the speculative return address in a first instance, and selects the non-speculative return address in a second instance subsequent to the first instance if the comparator indicates that the speculative and non-speculative return addresses do not match.
In another aspect, it is a feature of the present invention to provide a method for predicting a return address of a return instruction in a microprocessor. The method includes pushing a first return address onto a first call/return stack, in response to fetching from an instruction cache a first cache line predicted to include a first call instruction. The method also includes pushing a second return address onto the first call/return stack, in response to fetching from the instruction cache a second cache line predicted to include a second call instruction. The method also includes popping the second return address from the first call/return stack, in response to fetching from the instruction cache a cache line predicted to include a first return instruction. The method also includes branching the microprocessor to the second return address, after the popping the second return address. The method also includes popping the first return address from the first call/return stack, in response to fetching from the instruction cache a cache line predicted to include a second return instruction. The method also includes branching the microprocessor to the first return address, after the popping the first return address. The method also includes pushing a third return address onto a second call/return stack, in response to decoding the first call instruction, after the popping the first return address. The method also includes pushing a fourth return address onto the second call/return stack, in response to decoding the second call instruction. The method also includes popping the fourth return address from the second call/return stack, in response to decoding the first return instruction. The method also includes comparing the second and fourth return addresses prior to the first return instruction reaching an execution stage of a pipeline of the processor. The execution stage is configured to finally resolve the first return instruction. The method also includes branching the microprocessor to the fourth return address, after the popping the fourth return address, if the second and fourth return addresses do not match.
In another aspect, it is a feature of the present invention to provide a branch prediction apparatus in a processor. The apparatus includes a first call/return stack, configured for pushing thereon a first return address, in response to fetching from an instruction cache a first cache line predicted to include a first call instruction; pushing thereon a second return address, in response to fetching from the instruction cache a second cache line predicted to include a second call instruction; and popping therefrom the second return address, in response to fetching from the instruction cache a cache line predicted to include a first return instruction. The apparatus also includes control logic, coupled to the first call/return stack, configured to branch the microprocessor to the first return address, after the popping the first return address. The first call/return stack is further configured for popping therefrom the first return address, in response to fetching from the instruction cache a cache line predicted to include a second return instruction. The control logic is further configured to branch the microprocessor to the first return address, after the popping the first return address. The apparatus also includes a second call/return stack, configured for pushing thereon a third return address, in response to decoding the first call instruction, after the popping the first return address; pushing thereon a fourth return address, in response to decoding the second call instruction; andpopping therefrom the fourth return address, in response to decoding the first return instruction. The apparatus also includes a comparator, coupled to the first and second call/return stacks, configured to compare the second and fourth return addresses prior to the first return instruction reaching an execution stage of a pipeline of the processor, wherein the execution stage is configured to finally resolve the first return instruction. The control logic is further configured to branch the microprocessor to the fourth return address, after the popping the fourth return address, if the second and fourth return addresses do not match.
An advantage of the present invention is that it potentially reduces the branch penalty by enabling the processor to branch on a return instruction without having to wait until the return instruction is decoded in contrast to conventional approaches having only a single call/return stack that branch later in the pipeline. Furthermore, the present invention potentially provides more accurate return instruction prediction than a speculative branch target address cache providing the return address since the call/return stack accounts for the possibility of multiple return paths.
Other features and advantages of the present invention will become apparent upon study of the remaining portions of the specification and drawings.
Referring now to
The first stage is the I-stage 302, or instruction fetch stage. The I-stage 302 is the stage where the processor 300 provides a fetch address to an instruction cache 432 (see
In the present invention, the processor 300 further comprises a speculative branch target address cache (BTAC) 402 (see
Advantageously, as may be seen from
Advantageously, in most cases, the two-cycle bubble is small enough that it may be absorbed by an instruction buffer 342, F-stage instruction queue 344 and/or X-stage instruction queue 346, described below. Consequently, in many cases, the speculative BTAC 402 enables the processor 300 to achieve zero-penalty branches.
The processor 300 further comprises a speculative call/return stack 406 (see
The V-stage 308 is the stage in which instructions are written to the instruction buffer 342. The instruction buffer 342 buffers instructions for provision to an F-stage 312. The V-stage 308 also includes decode logic for providing information about the instruction bytes to the instruction buffer 342, such as x86 prefix and mod R/M information, and whether an instruction byte is a branch opcode value.
The F-stage 312, or instruction format stage 312, includes instruction format and decode logic 436 (see
The F-stage 312 also includes branch instruction target address calculation logic 416 (see
An F-stage instruction queue 344 receives the formatted instructions. Formatted instructions are provided by the F-stage instruction queue 344 to an instruction translator in the X-stage 314.
The X-stage 314, or translation stage 314, instruction translator translates x86 macroinstructions into microinstructions that are executable by the remainder of the pipeline stages. The translated microinstructions are provided by the X-stage 314 to an X-stage instruction queue 346.
The X-stage instruction queue 346 provides translated microinstructions to an R-stage 316, or register stage 316. The R-stage 316 includes the user-visible x86 register set, in addition to other non-user-visible registers. Instruction operands for the translated microinstructions are stored in the R-stage 316 registers for execution of the microinstructions by subsequent stages of the pipeline 300.
An A-stage 318, or address stage 318, includes address generation logic that receives operands and microinstructions from the R-stage 316 and generates addresses required by the microinstructions, such as memory addresses for load/store microinstructions.
A D-stage 322, or data stage 322, includes logic for accessing data specified by the addresses generated by the A-stage 318. In particular, the D-stage 322 includes a data cache for caching data within the processor 300 from a system memory. In one embodiment, the data cache is a two-cycle cache. A G-stage 324 is the second stage of the data cache access, and the data cache data is available in an E-stage 326.
The E-stage 326, or execution stage 326, includes execution logic, such as arithmetic logic units, for executing the microinstructions based on the data and operands provided from previous stages. In particular, the E-stage 326 produces a resolved target address 356 of all branch instructions. That is, the E-stage 326 target address 356 is known to be the correct target address of all branch instructions with which all predicted target addresses must match. In addition, the E-stage 326 produces a resolved direction (DIR) 481 (see
An S-stage 328, or store stage 328, performs a store to memory of the results of the microinstruction execution received from the E-stage 326. In addition, the target address 356 of branch instructions calculated in the E-stage 326 is provided to the instruction cache 432 in the I-stage 302 from the S-stage 328. Furthermore, the BTAC 402 of the I-stage 302 is updated from the S-stage 328 with the resolved target addresses of branch instructions executed by the pipeline 300 for caching in the BTAC 402. In addition, other speculative branch information (SBI) 454 (see
A W-stage 332, or write-back stage 332, writes back the result from the S-stage 328 into the R-stage 316 registers, thereby updating the processor 300 state.
The instruction buffer 342, F-stage instruction queue 344 and X-stage instruction queue 346, among other things, serve to minimize the impact of branches upon the clocks per instruction value of the processor 300.
Referring now to
Referring now to
The instruction cache 432 includes a data array 506 for caching instruction bytes. The data array 506 is arranged as a plurality of lines indexed by a portion of the virtual fetch address 495. In one embodiment, the data array 506 stores 64 KB of instruction bytes arranged in 32 byte lines. In one embodiment, the data instruction cache 432 is a 4-way set associative cache. Hence, the data array 506 comprises 512 lines of instruction bytes indexed by bits [13:5] of the fetch address 495.
The line of instruction bytes 494 selected by the virtual fetch address 495 is provided on the output of the instruction cache 432 to the instruction buffer 342 as shown in
The instruction cache 432 also includes a tag array 504 for caching tags. The tag array 504, like the data array 506, is indexed by the same bits of the virtual fetch address 495. Physical address bits are cached in the tag array 504 as physical tags. The physical tags 514 selected by the fetch address 495 bits are provided on the output of the tag array 504.
The instruction cache 432 also includes a comparator 508 that compares the physical tags 514 with the physical page number 512 provided by the TLB 502 to generate a hit signal 518 for indicating whether the virtual fetch address 495 hit in the instruction cache 432. That is, the hit signal 518 indicates whether the instructions of the task currently being executed by the processor 300 at the fetch address 495 are cached in the data array 506 of the instruction cache 432. The hit signal 518 is a true indication of whether the current task instructions are cached since the instruction cache 432 converts the virtual fetch address 495 to a physical address and uses the physical address to determine a cache hit.
The operation of the instruction cache 432 as just described is in contrast to the BTAC 402 operation, which determines a hit based only on a virtual address, i.e., the fetch address 495, not on a physical address. A consequence of the distinction in operation is that virtual aliasing may occur such that the BTAC 402 produces an erroneous target address 352, as described below.
Referring again to
In addition, the instruction buffer 342 stores a speculatively branched (SB) bit 438 for each instruction byte stored in the instruction buffer 342. If the processor 300 speculatively branches to a speculative target address 352 provided by the BTAC 402 or to a speculative return address 353 provided by the speculative call/return stack 406 based on SBI 454 cached in the BTAC 402, the SB bit 438 is set for an instruction byte indicated by the SBI 454. That is, if the processor 300 speculatively branches based on a presumption that a branch instruction for which SBI 454 is cached in the BTAC 402 is present in the line of instruction bytes 494 provided by the instruction cache 432, the SB bit 438 is set for one of the instruction bytes 494 stored in the instruction buffer 342. In one embodiment, the SB bit 438 is set for the opcode byte of the presumed branch instruction as indicated by the SBI 454.
Instruction decode logic 436 receives instruction bytes 493 from the instruction buffer 342 in order to decode the instruction bytes 493, including branch instruction bytes, to generate instruction decode information 492. The instruction decode information 492 is used to make branch instruction predictions and to detect and correct erroneous speculative branches. The instruction decode logic 436 provides the instruction decode information 492 to downstream portions of the pipeline 300. In addition, the instruction decode logic 436 generates a next sequential instruction pointer (NSIP) 466 and a current instruction pointer (CIP) 468 when decoding the current instruction. In addition, the instruction decode logic 436 provides instruction decode information 492 to the non-speculative target address calculator 416, the non-speculative call/return stack 414, and the non-speculative branch direction predictor 412. Preferably, the non-speculative call/return stack 414, the non-speculative branch direction predictor 412, and the non-speculative target address calculator 416 reside in the F-stage 312 of the pipeline 300.
The non-speculative branch direction predictor 412 generates a non-speculative prediction of the direction of a branch instruction 444, i.e., whether the branch will be taken or not taken, in response to the instruction decode information 492 received from the instruction decode logic 436. Preferably, the non-speculative branch direction predictor 412 includes one or more branch history tables for storing a history of resolved directions of executed branch instructions. Preferably, the branch history tables are used in conjunction with decode information of the branch instruction itself provided by the instruction decode logic 436 to predict a direction of conditional branch instructions. An exemplary embodiment of the non-speculative branch direction predictor 412 is described in U.S. patent application Ser. No. 09/434,984 HYBRID BRANCH PREDICTOR WITH IMPROVED SELECTOR TABLE UPDATE MECHANISM, having a common assignee and which is hereby incorporated by reference. Logic that ultimately resolves the direction of the branch instruction preferably resides in the E-stage 326 of the pipeline 300.
The non-speculative call/return stack 414 generates the non-speculative return address 355 of
In addition, the instruction decode information 492 includes a return address 488 if the instruction currently being decoded by the instruction decode logic 436 is a call instruction. Preferably, the return address 488 comprises the value of the instruction pointer of the currently decoded call instruction plus the length of the call instruction. The return address 488 is pushed onto the non-speculative call/return stack 414 when the instruction decode information 492 indicates the instruction is a call instruction so that the return address 488 can be provided as non-speculative return address 355 upon subsequent decode of a return instruction by the instruction decode logic 436.
An exemplary embodiment of the non-speculative call/return stack 414 is described in U.S. patent application Ser. No. 09/271,591 METHOD AND APPARATUS FOR CORRECTING AN INTERNAL CALL/RETURN STACK IN A MICROPROCESSOR THAT SPECULATIVELY EXECUTES CALL AND RETURN INSTRUCTIONS, having a common assignee and which is hereby incorporated by reference.
The non-speculative target address calculator 416 generates the non-speculative target address 354 of
The branch prediction apparatus 400 includes the speculative branch target address cache (BTAC) 402. The BTAC 402 is addressed with a fetch address 495 provided on the fetch address bus for indexing a line within the BTAC 402. The BTAC 402 is not integrated with the instruction cache 432, but rather, is separate and distinct from the instruction cache 432, as shown. That is, the BTAC 402 is distinct from the instruction cache 432, both physically and conceptually. The BTAC 402 is physically distinct from the instruction cache 432 in that it is spatially located in a different location within the processor 300 than the instruction cache 432. The BTAC 402 and instruction cache 432 are conceptually distinct in that they are different in size, i.e., in one embodiment they comprise a different number of cache lines. The BTAC 402 and instruction cache 432 are also conceptually distinct in that the instruction cache 432 converts the fetch address 495 to a physical address for determining a hit of a line of instruction bytes; whereas, the BTAC 402 is indexed by the virtual fetch address 495 as a virtual address, without converting to a physical address.
Preferably, the BTAC 402 resides in the I-stage 302 of the pipeline 300. The BTAC 402 caches target addresses of previously executed branch instructions. When the processor 300 executes a branch instruction, the resolved target address of the branch instruction is cached in the BTAC 402 via update signals 442. The instruction pointer (IP) 1512 (see
To generate the cached branch target address 352 of
The multiplexer 422 selects the fetch address 495 from among a plurality of addresses, including the BTAC 402 target address 352, as will be discussed below. The multiplexer 422 output provides the fetch address 495 to the instruction cache 432 and BTAC 402. If the multiplexer 422 selects the BTAC 402 target address 352, then the processor 300 will branch to the BTAC 402 target address 352. That is, the processor 300will begin fetching instructions from the instruction cache 432 at the BTAC 402 target address 352.
In one embodiment, the BTAC 402 is smaller than the instruction cache 432. In particular, the BTAC 402 caches target addresses for a smaller number of cache lines than are comprised in the instruction cache 432. A consequence of the BTAC 402 not being integrated with the instruction cache 432, yet using the instruction cache 432 fetch address 495 as an index, is that if the processor 300 branches to the target address 352 generated by the BTAC 402 it does so speculatively. The branch is speculative because there is no certainty that a branch instruction resides in the selected instruction cache 432 line at all, much less that the branch instruction for which the target address 352 was cached. A hit in the BTAC 402 only indicates that a branch instruction was previously present in the instruction cache 432 line selected by the fetch address 495. There are at least two reasons there is no certainty a branch instruction is present in the selected cache line.
A first reason there is no certainty that a branch instruction is in the instruction cache 432 line indexed by the fetch address 495 is because the fetch address 495 is a virtual address; therefore, virtual aliasing may occur. That is, two different physical addresses may alias to the same virtual fetch address 495. A given fetch address 495, which is virtual, may translate to two different physical addresses associated with two different processes or tasks of a multitasking processor such as processor 300. The instruction cache 432 performs virtual to physical translation using the translation lookaside buffer 502 of
The operating system performing a task switch provides an example of a situation in which the virtual aliasing condition may occur. After the task switch, the processor 300 will fetch instructions from the instruction cache 432 at a virtual fetch address 495 associated with the new process equal to a virtual fetch address 495 of the old process that includes a branch instruction whose target address 352 is cached in the BTAC 402. The instruction cache 432 will produce the instructions for the new process based on the physical address translated from the virtual fetch address 495, as described above with respect to
Thus, a branch to the BTAC 402 target address 352 is speculative because in some situations the processor 300 will branch to an incorrect target address 352 generated by the BTAC 402 because a branch instruction is not present in the instruction cache 432 at the fetch address 495, due to virtual aliasing, for example. In contrast, the Athlon integrated BTAC/instruction cache 202 of
In addition, the non-speculative target address calculator 416, non-speculative call/return stack 414, and non-speculative branch direction predictor 412 predictions are also non-speculative because they generate branch predictions only after the branch instruction has been fetched from the instruction cache 432 and has been decoded by the instruction decode logic 436, as will be described below.
It should be understood that although the direction prediction 444 generated by the non-speculative branch direction predictor 412 is “non-speculative,” i.e., made with the certainty that a branch instruction exists in the current instruction stream because the branch instruction has been decoded by the instruction decode logic 436, the non-speculative direction prediction 444 is a “prediction” nevertheless. That is, if the branch instruction is a conditional branch instruction, such as an x86 JCC instruction, the branch may or may not be taken in any given execution of the branch instruction.
Similarly, the target address 354 generated by the non-speculative target address calculator 416 and the return address 355 generated by the non-speculative call/return stack 414 are non-speculative since they are generated with the certainty that a branch instruction exists in the current instruction stream; but they are still predictions, nevertheless. For example, in the case of an x86 indirect jump through memory, the memory contents may have changed since the last time the indirect jump was executed. Hence, the target address may have changed accordingly. Thus, “non-speculative” in this context is not to be confused with “unconditional” as to branch direction or “certain” as to target address. Similarly, “speculative” in this context is not to be confused with “prediction” or “non-certain” as to branch direction or target address.
A second reason there is no certainty that the branch instruction is in the instruction cache 432 line indexed by the fetch address 495 is the existence of self-modifying code. Self-modifying code may change the contents of the instruction cache 432, but the change is not reflected in the BTAC 402. Hence, a BTAC 402 hit may occur for a line of the instruction cache 432 that previously included a branch instruction, but which has been modified or replaced by a different instruction.
The branch prediction apparatus 400 also includes the speculative call/return stack 406. The speculative call/return stack 406 stores speculative target addresses for return instructions. The speculative call/return stack 406 generates the speculative return address 353 of
The control logic 404 generates control signals 483 to control the speculative call/return stack 406 to provide the speculative return address 353 when the BTAC 402 indicates a return instruction may be present in a line of the instruction cache 432 specified by the fetch address 495. Preferably, the BTAC 402 indicates a return instruction may be present in a line of the instruction cache 432 specified by the fetch address 495 when the selected BTAC 402 entry 602 VALID 702 and RET 706 bits (see
The BTAC 402 generates the HIT signal 452 and speculative branch information (SBI) 454 in response to the fetch address 495. The HIT signal 452 indicates that the fetch address 495 generated a cache tag hit in the BTAC 402, described below with respect to
The SBI 454 includes a BEG 446 signal (branch instruction beginning byte offset within a line in the instruction cache 432) and a LEN 448 signal (branch instruction length). The BEG 446 value, the LEN 448 value and the fetch address 495 are added together by an adder 434 to generate a return address 491. The return address 491 is provided on the adder 434 output to the speculative call/return stack 406 so that the return address 491 can be pushed onto the speculative call/return stack 406. The control logic 404 operates the speculative call/return stack 406 in conjunction with the BTAC 402 via signals 483 to push the return address 491. The return address 491 is pushed only if the selected BTAC 402 entry 602 VALID 702 and CALL 704 bits (see
The branch prediction apparatus 400 also includes the control logic 404. The control logic 404 controls multiplexer 422 via control signals 478 to select one of the plurality of address inputs to be the fetch address 495. The control logic 404 also sets the SB bits 438 in the instruction buffer 342 via signal 482.
The control logic 404 receives the HIT signal 452, the SBI 454, the non-speculative branch direction prediction 444 from the non-speculative branch direction predictor 412, and a FULL signal 486 from the instruction buffer 342.
The branch prediction apparatus 400 also includes prediction check logic 408. The prediction check logic 408 generates an ERR signal 456, which is provided to the control logic 404 to indicate that an erroneous speculative branch was performed based on a BTAC 402 hit, as described below with respect to
The prediction check logic 408 also receives the output 485 of a comparator 489. The comparator 489 compares the speculative target address 352 generated by the BTAC 402 and the resolved target address 356 of
The prediction check logic 408 also receives the output 487 of a comparator 497. The comparator 497 compares the speculative return address 353 generated by the speculative call/return stack 406 and the resolved target address 356. The speculative return address 353 is registered and piped down the instruction pipeline 300 to the comparator 497.
The BTAC 402 speculative target address 352 is also registered and piped down the instruction pipeline 300 for comparison with the non-speculative target address calculator 416 target address 354 by a comparator 428. The comparator 428 output 476 is provided to the control logic 404. Similarly, the speculative return address 353 generated by the speculative call/return stack 406 is also registered and piped down the instruction pipeline 300 for comparison with the non-speculative return address 355 by a comparator 418. The comparator 418 output 474 is also provided to the control logic 404.
The branch prediction apparatus 400 also includes a save multiplexed/register 424. The save mux/reg 424 is controlled by a control signal 472 generated by the control logic 404. The output 498 of the save mux/reg 424 is provided as an input to the multiplexer 422. The save mux/reg 424 receives as inputs its own output 498 and the BTAC 402 speculative target address 352.
The multiplexer 422 also receives as an input the S-stage 328 branch address 356. The multiplexer 422 also receives as an input the fetch address 495 itself. The multiplexer 422 also receives as an input a next sequential fetch address 499 generated by an incrementer 426, that receives the fetch address 495 and increments it to the next sequential instruction cache 432 line.
Referring now to
Each of the data array 612 and tag array 614 is organized into four ways, shown as way 0, way 1, way 2, and way 3. Preferably, each of the data array 612 ways stores two entries for caching a branch target address and speculative branch information, designated A and B. Hence, the data array 612 generates eight entries 602 each time it is read. The eight entries 602 are provided to an 8:2 way select mux 606.
Each of the data array 612 and tag array 614 is indexed by the instruction cache 432 fetch address 495 of
The tag array 614 generates a tag 616 for each way. Preferably, each tag 616 comprises 20 bits of virtual address, and each of the four tags 616 is compared with bits [31:12] of the fetch address 495 by a block of comparators 604. The comparators 604 generate the HIT signal 452 of
In addition, the comparators 604 generate control signals 618 to control the way select mux 606. In response, the way select mux 606 selects the A and B entry, 624 and 626, respectively, of one of the four ways in the line generated by the BTAC 402. The A entry 624 and B entry 626 are provided to an A/B select mux 608 and to the control logic 404. The control logic 404 generates a control signal 622 to control the A/B select mux 608 in response to the HIT 452 signal, entry A 624 and entry B 626, the fetch address 495 and other control signals. In response, the A/B select mux 608 selects one of entry A 624 or entry B 626 as the BTAC 402 target address 352 of
Preferably, the BTAC 402 is a single-ported cache. A single-ported cache has the advantage of being smaller, and therefore able to cache more target addresses than a dual-ported cache in the same amount of space. However, a dual-ported cache is contemplated to facilitate simultaneous reads and writes of the BTAC 402. The simultaneous read and write feature of the dual-ported BTAC 402 enables faster updates of the BTAC 402 since the updating writes do not have to wait for reads. The faster updates generally result in a more accurate prediction, since the information in the BTAC 402 is more current.
In one embodiment, the instruction cache 432 lines comprise 32 bytes each. However, the instruction cache 432 provides a half-cache line of instruction bytes 494 at time. In one embodiment, each line of the BTAC 402 stores two entries 602, and therefore two target addresses 714, per half-cache line of the instruction cache 432.
Referring now to
The VALID bit 702 indicates whether the entry 602 may be used for speculatively branching the processor 300 to the associated target address 714. In particular, the VALID bit 702 is initially cleared because the BTAC 402 is empty since no valid target addresses have been cached. The VALID bit 702 is set when the processor 300 executes a branch instruction and the resolved target address and speculative branch information associated with the branch instruction is cached in the entry 602. Subsequently, the VALID bit 702 is cleared if the BTAC 402 makes an erroneous prediction based on the entry 602, as described below with respect to
The BEG field 446 specifies the branch instruction beginning byte offset within a line in the instruction cache 432. The BEG field 446 is used to calculate a return address for storage in the speculative call/return stack 406 of
The LEN 448 field specifies the length in bytes of the branch instruction. The LEN field 448 is used to calculate a return address for storage in the speculative call/return stack 406 of
The CALL bit 704 indicates whether the cached target address 714 is associated with a call instruction. That is, if a call instruction was executed by the processor 300 and the target address of the call instruction was cached in the entry 602, then the CALL bit 704 will be set.
The RET bit 706 indicates whether the cached target address 714 is associated with a return instruction. That is, if a return instruction was executed by the processor 300 and the target address of the return instruction was cached in the entry 602, then the RET bit 706 will be set.
The WRAP bit 708 is set if the branch instruction bytes span two instruction cache 432 lines. In one embodiment, the WRAP bit 708 is set if the branch instruction bytes span two instruction cache 432 helf-lines.
The BDPI (branch direction prediction information) field 712 comprises a T/NT (taken/not taken) field 722 and a SELECT bit 724. The T/NT field 722 comprises a direction prediction of the branch, i.e., it indicates whether the branch is predicted taken or not taken. Preferably, the T/NT field 722 comprises a two-bit up/down saturating counter, for specifying the four states strongly taken, weakly taken, weakly not taken, and strongly not taken. In another embodiment, the T/NT field 722 comprises a single T/NT bit.
The SELECT bit 724 is used to select between the BTAC 402 T/NT direction prediction 722 and a direction prediction made by a branch history table (BHT) 1202 (see
In one embodiment, the SELECT bit 724 comprises a two-bit up/down saturating counter, for specifying the four states strongly BTAC, weakly BTAC, weakly BHT, and strongly BHT. In this embodiment, if after execution of the branch, the selected predictor (i.e., BTAC 402 or BHT 1202) correctly predicted the direction, the saturating counters count toward the selected predictor. If the selected predictor incorrectly predicted the direction but the other predictor correctly predicted the direction, the saturating counters count toward the non-selected predictor.
Referring now to
If a BTAC 402 hit did not occur, then the control logic 404 does not speculatively branch, in step 822. That is, the control logic 404 controls the multiplexer 422 via control signal 478 of
However, if a BTAC 402 hit did occur, the control logic 404 determines whether the A entry 624 of
The control logic 404 determines the entry 624 is “valid” if the VALID bit 702 of
The control logic 404 determines the entry 624 is “taken” if the T/NT field 722 of
The control logic 404 determines the entry 624 is “seen” if the BEG field 446 of
If the A entry 624 is valid, predicted taken, and is seen, the control logic 404 examines the B entry 626 of
If the A entry 624 is valid, predicted taken, and is seen, but the B entry 626 is not valid, predicted not taken, or is not seen, the control logic 404 examines the RET field 706 of
After speculatively branching during step 814 or step 818, the control logic 404 generates an indication on control signal 482 that that a speculative branch was performed in response to the BTAC 402, in step 816. That is, regardless of which of the speculative call/return stack 406 return address 353 or BTAC 402 entry A 624 target address 352 the processor 300speculatively branched to, the control logic 404 indicates on control signal 482 that a speculative branch was performed. The control signal 482 is used to set the SB bit 438 for a byte of the instruction when it proceeds into the instruction buffer 342 of
If the A entry 624 is invalid, or is predicted not taken, or is not seen, as determined during step 804, the control logic 404 determines whether the B entry 626 is valid, seen and taken, in step 824. The control logic 404 determines whether the B entry 626 is valid, seen and taken in a manner similar to the one described with respect to step 804 for the A entry 624.
If the B entry 626 is valid, predicted taken, and is seen, the control logic 404 examines the RET field 706 to determine whether the B entry 626 has cached return instruction information, in step 832. If the RET bit 706 is not set, the control logic 404 controls A/B mux 608 of
After speculatively branching during step 834 or step 818, the control logic 404 generates an indication on control signal 482 that that a speculative branch was performed in response to the BTAC 402, in step 816.
If both the A entry 624 and the B entry 626 are invalid, predicted not taken, or are not seen, the control logic 404 does not speculatively branch, in step 822.
If both the A entry 624 and the B entry 626 are valid, predicted taken, and seen, the control logic 404 determines which of the presumed branch instructions whose information is cached in the A entry 624 and B entry 626 is the first seen of the valid and taken branch instructions in the instruction cache 432 line instruction bytes 494, in step 808. That is, if both of the presumed branch instructions are seen, valid and taken, the control logic 404 determines which of the presumed branch instructions has the smaller memory address by comparing the BEG 446 fields of the A entry 624 and B entry 626. If the B entry 626 BEG 446 value is smaller than the A entry 624 BEG 446 value, then the control logic 404 proceeds to step 832 to speculatively branch based on the B entry 626. Otherwise, the control logic 404 proceeds to step 812 to speculatively branch based on the A entry 624.
In one embodiment, the speculative call/return stack 406 is not present. Hence, steps 812, 818, and 832 are not performed.
It may be observed from
Referring now to
The example also shows portions of an A entry 602A and a B entry 602B within a line of the BTAC 402 selected by the fetch address 495. Entry A 602A contains cached information associated with the CALL instruction and entry B 602B contains cached information for the JCC instruction. Entry A 602A shows a VALID bit 702A set to 1 to indicate a valid entry A 602A, i.e., that the associated target address 714 and SBI 454 of
Entry B 602B shows a VALID bit 702B set to 1 to indicate a valid entry B 602B. Entry B 602B also shows a BEG field 446B with a value of 0x02, corresponding to the least significant bits of the instruction pointer address of the JCC instruction. Entry B 602B also shows a T/NT field 722B with a value of Taken, indicating the JCC instruction is predicted Taken. The B entry 602B is provided to the control logic 404 via signals 626 of
In addition, the BTAC 402 asserts the HIT signal 452 to indicate that the fetch address 495 caused a hit in the BTAC 402. The control logic 404 receives entry A 602A and entry B 602B and generates A/B select signal 622 of
The control logic 404 determines during step 802 that a hit occurred in the BTAC 402 based on the HIT signal 452 being asserted. The control logic 404 next determines during step 804 that entry A 602A is valid based on the VALID bit 702A being set. The control logic 404 also determines during step 804 that entry A 602A is taken, since the T/NT field 722A indicates Taken. The control logic 404 also determines during step 804 that entry A 602A is seen, since the BEG field 446A value of 0x0C is greater than or equal to the corresponding lower bits of the fetch address 495 value of 0x09. Since entry A 602A is valid, taken, and seen, the control logic 404 proceeds to step 806.
The control logic 404 determines during step 806 entry B 602B is valid based on the VALID bit 702B being set. The control logic 404 also determines during step 806 that entry B 602B is taken, since the T/NT field 722B indicates Taken. The control logic 404 also determines during step 806 that entry B 602B is not seen, since the BEG field 446B value of 0x02 is less than the corresponding lower bits of the fetch address 495 value of 0x09. Since entry B 602B is not seen, the control logic 404 proceeds to step 812.
The control logic 404 determines during step 812 that the cached instruction associated with entry A 602A is not a return instruction via a clear RET bit 706 of
Hence, as may be seen from the example of
Referring now to
Next, the prediction check logic 408 of
If a speculative branch was performed, then the prediction check logic 408 examines the currently decoded instruction to determine whether the instruction is a non-branch instruction, in step 1012. Preferably, the prediction check logic 408 determines whether the instruction is a non-branch instruction for the x86 instruction set.
If the instruction is not a branch instruction, the prediction check logic 408 asserts the ERR signal 456 of
If the instruction is not a branch instruction, the control logic 404 next controls multiplexer 422 of
If it is determined during step 1012 that the instruction is a valid branch instruction, the prediction check logic 408 determines whether the SB bit 438 is set for any of the bytes in the instruction in a non-opcode byte location within the instruction bytes of the decoded instruction, in step 1014. That is, although a byte may contain a valid opcode value for the processor 300 instruction set, the valid opcode value may be in a byte location that is not valid for the instruction format. For an x86 instruction, barring prefix bytes, the opcode byte should be the first byte of the instruction. For example, the SB bit 438 may erroneously be set for a branch opcode value in an immediate data or displacement field of the instruction, or in a mod R/M or SIB byte of an x86 instruction due to a virtual aliasing condition. If the branch opcode byte is in a non-opcode byte location, then steps 1022 and 1024 are performed to correct the erroneous speculative prediction.
If the prediction check logic 408 determines during step 1012 that the instruction is a valid branch instruction, and determines during step 1014 no SB bits 438 are set for non-opcode bytes, then the prediction check logic 408 determines whether there is a speculative and non-speculative instruction length mismatch, in step 1016. That is, the prediction check logic 408 compares the non-speculative instruction length generated by the instruction decode logic 436 during step 1002 with the speculative LEN 448 field of
If the prediction check logic 408 determines during step 1012 that the instruction is a valid branch instruction, and determines during step 1014 the SB bit 438 is set only for the opcode byte, and determines during step 1016 the instruction lengths match, then the instruction proceeds down the pipeline 300 until it reaches the E-stage 326 of
Next, the prediction check logic 408 determines whether the BTAC 402 erroneously predicted the direction of the branch instruction, in step 1034. That is, the prediction check logic 408 compares the correct direction DIR 481 resolved by the E-stage 326 with the prediction 722 of
If the BTAC 402 predicted an erroneous direction, the prediction check logic 408 asserts the ERR signal 456 to notify the control logic 404 of the error, in step 1042. In response, the control logic 404 updates the BTAC 402 direction prediction 722 via update signal 442 of
If no direction error is detected during step 1034, the prediction check logic 408 determines whether the BTAC 402 or speculative call/return stack 406 erroneously predicted the target address of the branch instruction, in step 1036. That is, if the processor 300 speculatively branched to the BTAC 402 target address 352, then the prediction check logic 408 examines the result 485 of comparator 489 of
If a target address error is detected during step 1036, the prediction check logic 408 asserts the ERR signal 456 to indicate the detection of an erroneous speculative branch, in step 1052. In addition, the control logic 404 updates the BTAC 402 via update signal 442 with the resolved target address 356 generated during step 1032 for the corresponding BTAC 402 entry 602 of
Referring now to
The previous code sequence includes an x86 JMP (unconditional jump) instruction at address location 0x00000010. The target address of the JMP is address 0x00001234. The JMP has already been executed; hence, the target address 0x00001234 is already cached in the BTAC 402 of
The current code sequence includes an ADD (arithmetic add) instruction at 0x00000010, the same virtual address of the JMP instruction in the previous code sequence. At location 0x00001234 in the current code sequence is a SUB (arithmetic subtract) instruction, and at 0x00001236 is an INC (arithmetic increment) instruction.
The table 1100 comprises eight columns and six rows. The last seven columns of the first row designate seven clock cycles, 1 through 7. The last five rows of the first column designate the first five stages of the pipeline 300, namely the I-stage 302, B-stage 304, U-stage 306, V-stage 308, and F-stage 312. The remaining cells of the table specify the contents of each of the stages during the various clock cycles while executing the current code sequence.
During clock cycle 1, the BTAC 402 and instruction cache 432 are accessed. The ADD instruction is shown in I-stage 302. The fetch address 495 of
During clock cycle 2, the ADD instruction is shown in the B-stage 304. This is the second clock of the instruction cache 432 fetch cycle. The tag array 614 provides the tags 616 and the data array 612 provides the entries 602 of
During cycle 3, the ADD instruction is shown in U-stage 306. The ADD instruction is provided by the instruction cache 432 and latched in the U-stage 306. Because of steps 802 through 814of
During clock cycle 4, the ADD proceeds to the V-stage 308, where it is written to the instruction buffer 342. Clock cycle 4 is the speculative branch cycle. That is, the processor 300 begins fetching instructions at the cached target address 352 value 0x00001234 according to step 814 of
During clock cycle 5, the error in the speculative branch is detected. The ADD instruction proceeds to the F-stage 312. The SUB instruction proceeds to the B-stage 304. The INC instruction, the instruction at the next sequential instruction pointer, is shown in the I-stage 302. The F-stage 312 instruction decode logic 436 of
During clock cycle 6, the erroneous speculative branch is invalidated. The instruction buffer 342 is flushed according to step 1022. In particular, the ADD instruction is flushed from the instruction buffer 342. Additionally, the BTAC 402 is updated to clear the VALID bit 702 associated with the entry 602 that caused the erroneous speculative branch according to step 1022. Furthermore, the control logic 404 controls multiplexer 422 to select the CIP 468 as the fetch address 495 during the next cycle.
During clock cycle 7, the erroneous speculative branch is corrected. The processor 300 begins fetching instructions from the instruction cache 432 at the instruction pointer of the ADD instruction that was being decoded by the instruction decode logic 436 when the error was detected during clock cycle 5. That is, the processor 300 branches to CIP 468 corresponding to the ADD instruction according to step 1024, thereby correcting the erroneous speculative branch performed during clock cycle 5. Hence, the ADD instruction is shown in the I-stage 302 during clock cycle 7. This time, the ADD will proceed down the pipeline 300 and execute.
Referring now to
The global branch history registers 1206 comprise a shift register for storing a global history of branch instruction direction outcomes 1212 for all branch instructions executed by the processor 300 received by the global branch history registers 1206. Each time the processor 300 executes a branch instruction, the DIR 481 bit of
The global branch history 1206 is provided via signals 1214 to the exclusive OR logic 1204 for performance of a logical exclusive OR operation with the fetch address 495 of
The branch history table 1202 comprises an array of storage elements for storing a history of branch direction outcomes for a plurality of branch instructions. The array is indexed by the output 1216 of the exclusive OR logic 1204. When the processor 300 executes a branch instruction, the array element of the branch history table 1202 indexed by the exclusive OR logic 1204 output 1216 is selectively updated via signal 1218 as a function of the resolved branch direction DIR 481.
In one embodiment, each of the storage elements in the branch history table 1202 array comprises two direction predictions: an A and B direction prediction. Preferably, the branch history table 1202 generates the A and B direction predictions on T/NT_A/B 1222 signals as shown, for specifying a direction prediction to be selected against each of the A entry 624 and B entry 626 of
In one embodiment, each of the A and B predictions comprises a single T/NT (taken/not taken) bit. In this embodiment, the single T/NT bit is updated with the value of the DIR bit 481. In another embodiment, each of the A and B predictions comprises a two-bit up/down saturating counter, for specifying the four states strongly taken, weakly taken, weakly not taken, and strongly not taken. In this embodiment, the saturating counters count in the direction indicated by the DIR bit 481.
The mux 1208 receives the two direction prediction bits T/NT_A/B 1222 from the branch history table 1202 and the T/NT direction prediction 722 of
It may be observed that if the processor 300 branches to the target address 352 generated by the BTAC 402 based, at least in part, on the direction predictions 1222 provided by the branch history table 1202, it does so speculatively. The branch is speculative because, although a hit in the BTAC 402 indicates that a branch instruction was previously present in the instruction cache 432 line selected by the fetch address 495, there is no certainty that a branch instruction resides in the selected instruction cache 432 line, as discussed above.
It may also be observed that the hybrid speculative branch direction predictor 1200 of
Referring now to
The BTAC 402 of
If a valid BTAC 402 hit occurred during step 1302, then the control logic 404 examines the CALL bit 704 of
If the CALL bit 704 is set, the control logic 404 next controls the multiplexer 422 to select the BTAC 402 target address 352 of
If the control logic 404 determines during step 1304 that the CALL bit 704 is not set, then the control logic 404 examines the RET bit 706 of
After popping the speculative return address 353, the control logic 404 controls the multiplexer 422 to select the speculative return address 353 popped off the speculative call/return stack 406 in order to speculatively branch to the return address 353, in step 1316.
The return instruction proceeds down the pipeline 300 until it reaches the F-stage 312 of
The control logic 404 examines the comparator 418 result 474 to determine if a mismatch occurred, in step 1324. If the speculative return address 353 and the non-speculative return address 355 do not match, then the control logic 404 controls multiplexer 422 to select the non-speculative return address 355 in order to cause the processor 300 to branch to the non-speculative return address 355, in step 1326.
If the control logic 404 determines during step 1304 that the CALL bit 704 is not set, and determines during step 1312 that the RET bit 706 is not set, then the control logic 404 controls multiplexer 422 to speculatively branch to the BTAC 402 target address 352 of
Thus, it may observed from
Referring now to
In particular, the instruction decode logic 436 determines whether the instruction is a branch instruction, the length of the instruction, and the type of the branch instruction. Preferably, the instruction decode logic 436 determines whether the branch instruction is a conditional or unconditional type branch instruction, a PC-relative type branch instruction, a return instruction, a direct type branch instruction, or an indirect type branch instruction.
If the instruction is a branch instruction, the non-speculative branch direction predictor 412 generates the non-speculative direction prediction 444 of
The control logic 404 determines whether the branch instruction is a conditional branch instruction, in step 1404. That is, the control logic 404 determines whether the instruction may be taken or not taken depending upon a condition, such as whether certain flag bits are set, such as a zero flag, carry flag, etc. In the x86 instruction set, the JCC instruction is a conditional type branch instruction. In contrast, the RET, CALL and JUMP instructions, for example, are unconditional branch instructions in the x86 instruction set because they always have a direction of taken.
If the branch is a conditional type branch instruction, the control logic 404 determines whether there is a mismatch between the non-speculative direction 444 predicted by the non-speculative branch direction predictor 412 and the speculative direction 722 of
If there is a direction prediction mismatch, the control logic 404 determines whether the non-speculative direction prediction 444 is taken or not taken, in step 1414. If the non-speculative direction prediction 444 is not taken, the control logic 404 controls multiplexer 422 to select the NSIP 466 of
If the non-speculative direction prediction 444 is taken, the control logic 404 controls multiplexer 422 to branch to the non-speculative target address 354, in step 1432. Again, the speculative direction prediction 722 is overridden because the non-speculative direction prediction 444 is generally more accurate.
If the control logic 404 determines during step 1412 that there is not a direction prediction mismatch, and that a speculative branch was performed for the branch instruction (i.e., if the SB bit 438 is set), the control logic 404 determines whether there is a mismatch between the speculative target address 352 and the non-speculative target address 354, in step 1428. If there is a target address mismatch for a conditional type branch, the control logic 404 controls multiplexer 422 to branch to the non-speculative target address 354, in step 1432. The speculative target address prediction 352 is overridden because the non-speculative target address prediction 354 is generally more accurate. If there is not a target address mismatch for a conditional type branch, no action is taken. That is, the speculative branch is allowed to proceed, subject to error correction as described with respect to
If during step 1404, the control logic 404 determines the branch instruction is not a conditional type branch, the control logic 404 determines whether the branch instruction is a return instruction, in step 1406. If the branch instruction is a return instruction, the control logic 404 determines whether there is a mismatch between the speculative return address 353 generated by the speculative call/return stack 406 and the non-speculative return address 355 generated by the non-speculative call/return stack 414, in step 1418.
If there is a mismatch between the speculative return address 353 and the non-speculative return address 355, the control logic 404 controls the multiplexer 422 to branch to the non-speculative return address 355, in step 1422. That is, the control logic 404 selectively overrides the speculative return address 353. The speculative return address 353 is overridden because the non-speculative return address 355 is generally more accurate. If there is not a target address mismatch for a direct type branch, no action is taken. That is, the speculative branch is allowed to proceed, subject to error correction as described with respect to
If during step 1406, the control logic 404 determines the branch instruction is not a return instruction, the control logic 404 determines whether the branch instruction is a PC-relative type branch instruction, in step 1408. In the x86 instruction set, a PC-relative type branch instruction is a branch instruction in which a signed offset specified in the branch instruction is added to the current program counter value to compute the target address.
In an alternate embodiment, the control logic 404 also determines whether the branch instruction is a direct type branch instruction, in step 1408. In the x86 instruction set, a direct type branch instruction is a branch instruction in which the target address is specified in the instruction itself. Direct type branch instructions are also referred to as immediate type branch instructions, since the target address is specified in an immediate field of the instruction.
If the branch instruction is a PC-relative type branch instruction, the control logic 404 determines whether there is a mismatch between the speculative target address 352 and the non-speculative target address 354, in step 1424. If there is a target address mismatch for a PC-relative type branch, the control logic 404 controls multiplexer 422 to branch to the non-speculative target address 354, in step 1426. The speculative target address prediction 352 is overridden because the non-speculative target address prediction 354 is generally more accurate for a PC-relative type branch. If there is not a target address mismatch for a PC-relative type branch, no action is taken. That is, the speculative branch is allowed to proceed, subject to error correction as described with respect to
If during step 1408, the control logic 404 determines the branch instruction is not a PC-relative type branch instruction, no action is taken. That is, the speculative branch is allowed to proceed, subject to error correction as described with respect to
It has been observed that for indirect type branch instructions, the BTAC 402 prediction is generally more accurate than the relatively small F-stage 312 BTB. Hence, if it is determined that the branch is an indirect type branch instruction, the control logic 404 does not override the BTAC 402 speculative prediction. That is, if a speculative branch was performed due to a BTAC 402 hit as described in
Referring now to
The selecting BTAC 402 line also includes an A/B LRU bit 1504 for indicating which of entry A 602A and entry B 602B was least recently used. In one embodiment, each time a BTAC 402 hit occurs on a given target address 714, the A/B LRU bit 1504 is updated to specify the opposite entry of the entry for which the hit occurred. That is, if the control logic 404 proceeds to step 812 of
The replacement apparatus also includes a multiplexer 1506. The mux 1506 receives as inputs the fetch address 495 of
When a branch instruction executes and is taken, the target address 714 of the branch instruction and associated SBI 454 are written into, or cached in, a BTAC 402 entry 602. That is, the BTAC 402 is updated with the new target address 714 of the executed branch instruction and associated SBI 454. The control logic 404 must decide which side, A or B, of the BTAC 402 to update for the BTAC 402 line and way selected by the update IP 1512. That is, the control logic 404 must decide whether to replace the entry A 602A or the entry B 602B of the selected line and way. The control logic 404 decides which side to replace as shown in Table 1 below.
Table 1 is a truth table having two inputs, the VALID bit 702 of entry A 602A and the VALID bit 702 of entry B 602B. The output of the truth table is the action for determining the side of the BTAC 402 to replace. As shown, if the A entry 602A is invalid and the B entry 602B is valid, then the control logic 404 replaces the A entry 602A. If the A entry 602A is valid and the B entry 602B is invalid, then the control logic 404 replaces the B entry 602B. If both the A entry 602A and B entry 602B are valid, then the control logic 404 replaces the least recently used entry as specified by the A/B LRU bit 1504 in the line and way selected by the update IP 1512.
If both the A entry 602A and B entry 602B are invalid, then the control logic 404 must decide which side to replace. One solution is to always write to one side, for example, side A. However, this solution poses a problem illustrated by Code Sequence 1 below.
In Code Sequence 1, the three instructions shown are in the same instruction cache 432 line because their instruction pointer addresses are equal except for the lower 4 address bits; accordingly, the JMP and CALL instructions select the same BTAC 402 line and way. Assume in this example both the A entry 602A and the B entry 602B in the BTAC 402 line and way selected by the instruction pointers for the JMP and CALL instructions are invalid when the instructions execute. Using the solution of “always update side A when both entries are invalid”, the JMP instruction would see that both sides are invalid and would update the A entry 602A.
However, since the CALL instruction is relatively close to the JMP instruction in the program sequence, if the pipeline is relatively long, as in processor 300, a relatively large number of cycles may pass before the VALID bit 702 of entry A 602A is updated. Hence, a high probability exists that the CALL instruction will sample the BTAC 402 before the BTAC 402 is updated by the executed JMP instruction, and in particular, before the entry A 602A VALID bit 702 and BTAC 402 way replacement status for the selected BTAC 402 line is updated by the JMP instruction. Hence, the CALL instruction will see that both sides are invalid and will also update the A entry 602A according to the “always update side A when both entries are invalid” solution. This is problematic, since the target address 714 for the JMP instruction will be needlessly clobbered since an empty, i.e., invalid B entry 602B was available for caching the target address 714 of the CALL instruction.
To solve this problem, as shown in Table 1, if both the A entry 602A and B entry 602B are invalid, then the control logic 404 advantageously selects the side which is the inverse, or not, of a side stored in a global replacement status flag register, LastWritten 1502, comprised in and updated by the replacement apparatus. The LastWritten register 1502 stores an indication of whether side A or B of the BTAC 402 was last written to an invalid entry 602 of the BTAC 402 globally. Advantageously, the method uses the LastWritten register 1502 to avoid the problem illustrated by Code Sequence 1 above as described presently with respect to
Referring now to
When the control logic 404 needs to update a BTAC 402 entry 602, the control logic 404 examines the VALID bit 702 for each of the selected A entry 602A and B entry 602B. The control logic 404 determines if both the A entry 602A and the B entry 602B are valid, in step 1602. If both entries are valid, the control logic 404 examines the A/B LRU bit 1504 bit to determine whether entry A 602A or entry B 602B was least recently used, in step 1604. If entry A 602A was least recently used, the control logic 404 replaces entry A 602A, in step 1606. If entry B 602B was least recently used, the control logic 404 replaces entry B 602B, in step 1608.
If the control logic 404 determines during step 1602 that not both entries are valid, it determines whether the A entry 602A is valid and the B entry 602B is invalid, in step 1612. If so, the control logic 404 replaces the B entry 602B, in step 1614. Otherwise, the control logic 404 determines whether the A entry 602A is invalid and the B entry 602B is valid, in step 1622. If so, the control logic 404 replaces the A entry 602A, in step 1624. Otherwise, the control logic 404 examines the LastWritten register 1502, in step 1632.
If the LastWritten register 1502 indicates the A side of the BTAC 402 was not last written to a selected line and way in which both the A entry 602A and the B entry 602B are invalid, the control logic 404 replaces the A entry 602A, in step 1634. The control logic 404 subsequently updates the LastWritten register 1502 to specify that side A of the BTAC 402 was the last side written to a selected line and way in which both the A entry 602A and the B entry 602B were invalid, in step 1636.
If the LastWritten register 1502 indicates the B side of the BTAC 402 was not last written to a selected line and way in which both the A entry 602A and the B entry 602B are invalid, the control logic 404 replaces the B entry 602B, in step 1644. The control logic 404 subsequently updates the LastWritten register 1502 to specify that side B of the BTAC 402 was the last side written to a selected line and way in which both the A entry 602A and the B entry 602B were invalid, in step 1646.
As may be observed, the method of
Referring now to
Hence, in
Although simulations have revealed no observable performance difference between the embodiment of
The two JMP instructions at instruction pointers 0x00000010 and 0x00000014 are in the same instruction cache 432 line and select the same line in the BTAC 402. The JMP instruction at instruction pointer 0x12345678 is in a different instruction cache 432 line from the other two JMP instructions and selects a different line in the BTAC 402 from the other two JMP instructions. Assume the following conditions when the JMP 0x12345678 instruction executes. The LastWritten register 1502 specifies side B. Both the A entry 602A and the B entry 602B in the BTAC 402 line and way selected by the instruction pointers for the JMP 0x12345678 and JMP 0x20000000 instructions are invalid. The BTAC 402 line and way selected by the instruction pointer for the JMP 0x00000014 instruction indicates the A entry 602A is valid and the B entry 602B is invalid. Assume the JMP 0x20000000 instruction executes before the JMP 0x12345678 instruction updates the BTAC 402. Consequently, the instruction pointers of the JMP 0x12345678 and JMP 0x20000000 instructions select the same way in the same BTAC 402 line.
According to both
Referring now to
The additional array 1812 is dual-ported; whereas, the data array 612 is single-ported. Because the A/B LRU bit 1504 and T/NT bits 722 are updated more frequently than the rest of fields in the entry 602, providing dual-ported access to the more frequently updated fields reduces the likelihood of a bottleneck being created at the BTAC 402 during periods of high traffic. However, since dual-ported cache arrays are larger than single-ported cache arrays and consume more power, the less frequently accessed fields are stored in the single-ported data array 612.
Referring now to
In addition, instead of the LastWritten register 1502, the embodiment of
Table 2 is similar to Table 1, except that it has three inputs, including the additional VALID bit 702 for entry C 702C. In the equations, “lw” corresponds to the LastWritten value and “lwp” corresponds to the LastWrittenPrev value. In one embodiment, LastWritten and LastWrittenPrev are updated only when all three entries are invalid, analogous to the method of
Referring now to
At block 2002, as a cache line is being fetched from the instruction cache 432 at the fetch address 495, the fetch address 495 also indexes the BTAC 402, as described above with respect to
At block 2004, in response to the BTAC 402 predicting the first CALL instruction, the adder 434 calculates the return address 491 of the first CALL instruction, as discussed above with respect to
At block 2006, as a subsequent cache line is being fetched from the instruction cache 432, the BTAC 402 predicts the cache line contains a second CALL instruction. Flow proceeds to block 2008.
At block 2008, in response to the BTAC 402 predicting the second CALL instruction, the adder 434 calculates the return address 491 of the second CALL instruction, which is pushed onto the speculative CALL/RET stack 406. Flow proceeds to block 2012.
At block 2012, as a subsequent cache line is being fetched from the instruction cache 432, the BTAC 402 predicts the cache line contains a first RET instruction via the SBI 454, as described above with respect to
At block 2014, in response to the BTAC 402 predicting the first RET instruction, the speculative CALL/RET stack 406 pops a return address 353, which is the second CALL instruction return address that was pushed at block 2008, according to block 1314 of
At block 2016, as a subsequent cache line is being fetched from the instruction cache 432, the BTAC 402 predicts the cache line contains a second RET instruction. Flow proceeds to block 2018.
At block 2018, in response to the BTAC 402 predicting the second RET instruction, the speculative CALL/RET stack 406 pops the first CALL instruction return address that was pushed at block 2004 and the control logic 404 branches to the popped return address 353. Flow proceeds to block 2022.
At block 2022, the first CALL instruction reaches the instruction decode logic 436 which decodes it and calculates the return address 488 of the first CALL instruction, which is pushed onto the non-speculative CALL/RET stack 414, as discussed above with respect to
At block 2024, the second CALL instruction reaches the instruction decode logic 436 which decodes it and calculates the return address 488 of the second CALL instruction, which is pushed onto the non-speculative CALL/RET stack 414. Flow proceeds to block 2026.
At block 2026, the first RET instruction reaches the instruction decode logic 436 which decodes it and indicates the presence of the first RET instruction to the non-speculative CALL/RET stack 414 via instruction decode information 492, as described above with respect to
At block 2028, the comparator 418 compares the return address 353 provided by the speculative CALL/RET stack 406 at block 2014 with the return address 355 provided by the non-speculative CALL/RET stack 414 at block 2026, as described above with respect to
At decision block 2032, the control logic 404 receives the result 474 of the comparison made by the comparator 418 at block 2028 to determine whether there is a mismatch between the return address 353 provided by the speculative CALL/RET stack 406 at block 2014 and the return address 355 provided by the non-speculative CALL/RET stack 414 at block 2026, as described above with respect to
At block 2034, the control logic 404 controls the mux 422 to cause fetching to branch to the return address 355 provided by the non-speculative CALL/RET stack 414 at block 2026, as described above with respect to
Although the present invention and its objects, features, and advantages have been described in detail, other embodiments are encompassed by the invention. For example, the BTAC may be arranged in any number of cache arrangements, including direct-mapped, fully associative, or different number of way caches. Furthermore, the size of the BTAC may be increased or decreased. Also, a fetch address other than the fetch address of the line actually containing the branch instruction being predicted may be used to index the BTAC and branch history table. For example, the fetch address of the previous fetch may be used to reduce the size of a bubble introduced before branching. Additionally, the number of target addresses stored in each way of the cache may be varied. In addition, the size of the branch history table may vary and the number of bits and form of the direction prediction information stored therein may vary as well as the algorithm for indexing the branch history table. Furthermore, the size of the instruction cache may vary and the type of virtual fetch address used to index the instruction cache and BTAC may vary.
Those skilled in the art should appreciate that they can readily use the disclosed conception and specific embodiments as a basis for designing or modifying other structures for carrying out the same purposes of the present invention without departing from the spirit and scope of the invention as defined by the appended claims.
| Number | Name | Date | Kind |
|---|---|---|---|
| 4181942 | Forster et al. | Jan 1980 | A |
| 4200927 | Hughes et al. | Apr 1980 | A |
| 4860197 | Langendorf et al. | Aug 1989 | A |
| 5142634 | Fite et al. | Aug 1992 | A |
| 5163140 | Stiles et al. | Nov 1992 | A |
| 5313634 | Eickemeyer | May 1994 | A |
| 5353421 | Emma et al. | Oct 1994 | A |
| 5355459 | Matsuo et al. | Oct 1994 | A |
| 5394530 | Kitta | Feb 1995 | A |
| 5404467 | Saba et al. | Apr 1995 | A |
| 5434985 | Emma et al. | Jul 1995 | A |
| 5513330 | Stiles | Apr 1996 | A |
| 5530825 | Black et al. | Jun 1996 | A |
| 5553246 | Suzuki | Sep 1996 | A |
| 5604877 | Hoyt et al. | Feb 1997 | A |
| 5623614 | Van Dyke et al. | Apr 1997 | A |
| 5623615 | Salem et al. | Apr 1997 | A |
| 5634103 | Dietz et al. | May 1997 | A |
| 5687349 | McGarity | Nov 1997 | A |
| 5687360 | Chang | Nov 1997 | A |
| 5706491 | McMahan | Jan 1998 | A |
| 5721855 | Hinton et al. | Feb 1998 | A |
| 5734881 | White et al. | Mar 1998 | A |
| 5752069 | Roberts et al. | May 1998 | A |
| 5761723 | Black et al. | Jun 1998 | A |
| 5768576 | Hoyt et al. | Jun 1998 | A |
| 5805877 | Black et al. | Sep 1998 | A |
| 5812839 | Hoyt et al. | Sep 1998 | A |
| 5828901 | O'Toole et al. | Oct 1998 | A |
| 5832289 | Shaw et al. | Nov 1998 | A |
| 5850532 | Narayan et al. | Dec 1998 | A |
| 5850543 | Shiell et al. | Dec 1998 | A |
| 5864707 | Tran et al. | Jan 1999 | A |
| 5867701 | Brown et al. | Feb 1999 | A |
| 5881260 | Raje et al. | Mar 1999 | A |
| 5881265 | McFarland et al. | Mar 1999 | A |
| 5931944 | Ginosar et al. | Aug 1999 | A |
| 5948100 | Hsu et al. | Sep 1999 | A |
| 5961629 | Nguyen et al. | Oct 1999 | A |
| 5964868 | Gochman et al. | Oct 1999 | A |
| 5968169 | Pickett | Oct 1999 | A |
| 5974543 | Hilgendorf et al. | Oct 1999 | A |
| 5978909 | Lempel | Nov 1999 | A |
| 6035391 | Isaman | Mar 2000 | A |
| 6041405 | Green | Mar 2000 | A |
| 6044459 | Bae et al. | Mar 2000 | A |
| 6081884 | Miller | Jun 2000 | A |
| 6085311 | Narayan et al. | Jul 2000 | A |
| 6088793 | Liu et al. | Jul 2000 | A |
| 6101595 | Pickett et al. | Aug 2000 | A |
| 6108773 | Col et al. | Aug 2000 | A |
| 6122729 | Tran | Sep 2000 | A |
| 6134654 | Patel et al. | Oct 2000 | A |
| 6151671 | D'Sa et al. | Nov 2000 | A |
| 6157988 | Dowling | Dec 2000 | A |
| 6170054 | Poplingher | Jan 2001 | B1 |
| 6175897 | Ryan et al. | Jan 2001 | B1 |
| 6185676 | Poplingher et al. | Feb 2001 | B1 |
| 6233676 | Henry et al. | May 2001 | B1 |
| 6250821 | Schwendinger | Jun 2001 | B1 |
| 6256727 | McDonald | Jul 2001 | B1 |
| 6260138 | Harris | Jul 2001 | B1 |
| 6279105 | Konigsburg et al. | Aug 2001 | B1 |
| 6279106 | Roberts | Aug 2001 | B1 |
| 6308259 | Witt | Oct 2001 | B1 |
| 6314514 | McDonald | Nov 2001 | B1 |
| 6321321 | Johnson | Nov 2001 | B1 |
| 6351796 | McCormick et al. | Feb 2002 | B1 |
| 6374350 | D'Sa et al. | Apr 2002 | B1 |
| 6457120 | Sinharoy | Sep 2002 | B1 |
| 6502185 | Keller et al. | Dec 2002 | B1 |
| 6560696 | Hummel et al. | May 2003 | B1 |
| 6601161 | Rappoport et al. | Jul 2003 | B2 |
| 6647467 | Dowling | Nov 2003 | B1 |
| 6725357 | Cousin | Apr 2004 | B1 |
| 6748441 | Gemmell | Jun 2004 | B1 |
| 6754808 | Roth et al. | Jun 2004 | B1 |
| 6823444 | Henry et al. | Nov 2004 | B1 |
| 6898699 | Jourdan et al. | May 2005 | B2 |
| 6968444 | Kroesche et al. | Nov 2005 | B1 |
| 20040030866 | McDonald | Feb 2004 | A1 |
| 20040139281 | McDonald | Jul 2004 | A1 |
| 20040139292 | McDonald | Jul 2004 | A1 |
| 20040139301 | McDonald | Jul 2004 | A1 |
| 20040143709 | McDonald | Jul 2004 | A1 |
| 20040143727 | McDonald | Jul 2004 | A1 |
| 20050044343 | Henry et al. | Feb 2005 | A1 |
| 20050076193 | Henry et al. | Apr 2005 | A1 |
| 20050198479 | Bean et al. | Sep 2005 | A1 |
| 20050198481 | Henry et al. | Sep 2005 | A1 |
| Number | Date | Country | |
|---|---|---|---|
| 20020188833 A1 | Dec 2002 | US |
