Processor and method for generating less than (LT), Greater than (GT), and equal to (EQ) condition code bits concurrent with a logical or complex operation

[0002] This application is related to application Ser. No. 09/005,471, now pat. No. 6,035,390, entitled “Method and Apparatus for Generating Less Than (LT), Greater Than (GT), and Equal To (EQ) Condition Code Bits Concurrent with an Arithmetic or Logical Operation,” which was filed on Jan. 12, 1998, and is incorporated herein by reference.

[0003] 1. Technical Field

[0004] The present invention relates in general to a method and system for data processing and, in particular, to a method and apparatus for computing condition code bits in a processor. Still more particularly, the present invention relates to a method and apparatus for computing less than (LT), greater than (GT), and equal to (EQ) condition code bits in a processor concurrent with the execution of an instruction.

[0005] 2. Description of the Related Art

[0006] A state-of-the-art superscalar processor can comprise, for example, an instruction cache for storing instructions, an instruction buffer for temporarily storing instructions fetched from the instruction cache for execution, one or more execution units for executing sequential instructions, a branch processing unit (BPU) for executing branch instructions, a dispatch unit for dispatching sequential instructions from the instruction buffer to particular execution units, and a completion buffer for temporarily storing sequential instructions that have finished execution, but have not completed.

[0007] Branch instructions executed by the branch processing unit (BPU) of the superscalar processor can be classified as either conditional or unconditional branch instructions. Unconditional branch instructions are branch instructions that change the flow of program execution from a sequential execution path to a specified target execution path and that do not depend upon a condition supplied by the occurrence of an event. Thus, the branch specified by an unconditional branch instruction is always taken. In contrast, conditional branch instructions are branch instructions for which the indicated branch in program flow may be taken or not taken depending upon a condition within the processor, for example, the state of specified condition register bits or the value of a counter. Conditional branch instructions can be further classified as either resolved or unresolved, based upon whether or not the condition upon which the branch depends is available when the conditional branch instruction is evaluated by the branch processing unit (BPU). Because the condition upon which a resolved conditional branch instruction depends is known prior to execution, resolved conditional branch instructions can typically be executed and instructions within the target execution path fetched with little or no delay in the execution of sequential instructions. Thus, it is advantageous to determine condition register bits or another condition upon which a conditional branch instruction may depend as quickly as possible so that the conditional branch instruction can be resolved prior to execution. Even if a conditional branch instruction is not resolved prior to its execution, meaning that the conditional branch is speculatively predicted, it is still advantageous to compute the condition upon which the branch instruction depends as quickly as possible because the performance penalty incurred in the event of misprediction is thereby minimized.

[0008] Condition register bits upon which conditional branch instructions may depend are set in response to predetermined architecturally defined instructions, for example, compare instructions and certain “recording” forms of add, subtract, and other arithmetic and logical instructions. The condition register bits set by compare instructions and recording instructions include a less than (LT) bit, a greater than (GT) bit, and an equal to (EQ) bit, which indicate whether the result of a particular instruction is less than, greater than, or equal to zero, respectively. Conventional processors first determine the result of an instruction (e.g., add) and then compare the result with zero in subsequent cycle(s) to produce the condition register bits. As will be appreciated, this serial architecture places an inherent limitation upon how early the condition register bits can be determined. More recently, various techniques have been employed in order to determine the value of the EQ bit in parallel with the execution of certain types of instructions. Although the early determination of the value of the EQ bit provides some performance advantages over the prior serial approach, there remains a need in the art for a method and apparatus for computing the value of all of the LT, GT, and EQ bits concurrent with arithmetic and logical operations in order to enhance branch processing performance.

[0009] A processor includes execution resources and condition code logic. The execution resources execute an arithmetic or logical instruction by arithmetically or logically combining at least two operands. Concurrent with the execution of the arithmetic or logical instruction by the execution resources, the condition code logic determines less than, greater than, and equal to condition code bits associated with the result of the arithmetic or logical instruction. In one embodiment, the condition code logic includes a single computation stage that receives as inputs individual bit values of bit positions within first and second operands and logically combines the individual bit values. The single computation stage outputs, for each bit position, propagate, generate, and kill signals that collectively indicate values for the less than, greater than, and equal to condition code bits. One or more merging stages coupled to the computation stage then merge the propagate, generate, and kill signals into output signals that set the condition code bits. The condition code logic is also capable of receiving externally computed condition code bits associated with complex instructions and utilizing such condition code bits to produce the output signals.

[0010] All objects, features, and advantages of the present invention will become apparent in the following detailed written description.

[0011] The novel features believed characteristic of the invention are set forth in the appended claims. The invention itself however, as well as a preferred mode of use, further objects and advantages thereof, will best be understood by reference to the following detailed description of an illustrative embodiment when read in conjunction with the accompanying drawings, wherein:

[0012] FIG. 1 depicts an illustrative embodiment of a data processing system in which the method and system of the present invention may advantageously be utilized;

[0013] FIG. 2 is a block diagram of the condition code unit (CCU) of the processor depicted in FIG. 1; and

[0014] FIG. 3 is a timing diagram illustrating the computation of the LT, GT, and EQ condition register bits concurrent with the execution of an instruction.

[0015] With reference now to the figures and in particular with reference to FIG. 1, there is depicted a block diagram of an illustrative embodiment of a processor, indicated generally at 10, for processing data in accordance with the present invention. In the depicted illustrative embodiment, processor 10 comprises a single integrated circuit superscalar microprocessor. Accordingly, as discussed further below, processor 10 includes various execution units, registers, buffers, memories, and other functional units, which are all formed by integrated circuitry. Processor 10 preferably comprises one of the PowerPC™ line of microprocessors available from IBM Microelectronics such as the PowerPC™ 604, which is described in detail in PowerPC™ 604 RISC Microprocessor User's Manual (available from IBM Microelectronics as Order No. MPR604UMU-01 and incorporated herein by reference); however, those skilled in the art will appreciate from the following description that the present invention can advantageously be implemented within other suitable processors, particularly those which utilize condition code bits to resolve conditional branch instructions.

[0016] As illustrated in FIG. 1, processor 10 is coupled to bus 11 via a bus interface unit (BIU) 12 within processor 10. BIU 12 controls the transfer of information between processor 10 and other devices coupled to bus 11, such as a lower level cache or main memory (not illustrated), which together with processor 10 and bus 11 form a fully functional data processing system. BIU 12 is also connected to instruction cache 14 and data cache 16 within processor 10. High-speed caches, such as instruction cache 14 and data cache 16, enable processor 10 to achieve relatively fast access times to a subset of data or instructions previously transferred from lower level memory to caches 14 and 16, thus improving the overall performance of the data processing system. Instruction cache 14 is further connected to sequential fetcher 17, which fetches up to a cache line of instructions from instruction cache 14 during each cycle and transmits the fetched instructions to both branch processing unit (BPU) 18 and instruction queue 19. Branch instructions are retained by BPU 18 for execution and are cancelled from instruction queue 19; sequential instructions, on the other hand, are cancelled from BPU 18 and buffered within instruction queue 19 for subsequent execution by sequential instruction execution circuitry within processor 10.

[0017] BPU 18 includes count register (CTR) 40, link register (LR) 42, and condition register (CR) 44, the value of which may be utilized to resolve conditional branch instructions. BPU 18 further includes CR rename buffers 46, which temporarily store renamed copies of CR 44 generated by the execution of compare instructions or concurrent with the execution of certain recording instructions. In a preferred embodiment, CR 44 (and each of CR rename buffers 46) contains a number of distinct fields that each comprise one or more bits. Among these fields is a condition code field containing LT, GT, and EQ bits, which respectively indicate if a value (typically the result of an instruction) is less than zero, greater than zero, or equal to zero. Conditional branch instructions that cannot be resolved prior to execution by reference to CR 44, LR 42 or CTR 40 are preferably predicted utilizing conventional branch processing circuitry within BPU 18 such as a branch history table (BHT) or branch target address cache (BTAC).

[0018] In the depicted illustrative embodiment, in addition to BPU 18, the execution circuitry of processor 10 comprises multiple execution units for sequential instructions, including one or more integer units (IUs) 22, a load-store unit (LSU) 28, and a floating-point unit (FPU) 30. As is well-known to those skilled in the computer arts, each of execution units 22, 28, and 30 typically executes one or more instructions of a particular type of sequential instructions during each processor cycle. For example, IU(s) 22 perform integer mathematical and logical operations such as addition, subtraction, ANDing, ORing, and XORing, utilizing source operands received from specified general purpose registers (GPRs) 32 or GPR rename buffers 33. Following the execution of an integer instruction, IU 22 outputs the data results of the instruction to GPR rename buffers 33, which provide temporary storage for the result data until the result data is written from GPR rename buffers 33 to one or more of GPRs 32. FPU 30 typically performs single and double-precision floating-point arithmetic and logical operations, such as floating-point multiplication and division, on source operands received from floating-point registers (FPRs) 36 or FPR rename buffers 37. FPU 30 outputs data resulting from the execution of floating-point instructions to selected FPR rename buffers 37, which temporarily store the result data until the result data is written from FPR rename buffers 37 to selected FPRs 36. As its name implies, LSU 28 typically executes floating-point and fixed-point instructions which either load data from memory (i.e., either data cache 16 or main memory) into selected GPRs 32 or FPRs 36 or which store data from a selected one of GPRs 32, GPR rename buffers 33, FPRs 36, or FPR rename buffers 37 to memory.

[0019] In addition to execution units 22, 28, and 30, the execution circuitry of processor 10 includes condition code unit (CCU) 24, which directly executes signed and unsigned compare instructions and computes condition code bits for certain recording instructions processed by other execution units. As illustrated, CCU 24 receives control signals 21 from dispatch unit 20 that indicate the instruction type of compare and recording instruction as they are dispatched. CCU 24 also receives from GPRs 32 and/or GPR rename buffers 33 (and possibly other sources) the operands specified by compare and recording instructions. Compare and recording instructions in the PowerPC™ instruction set architecture for which CCU 24 computes condition code bits are summarized below in Table I.

[0020] In Table I, the period (“.”) following an instruction mnemonic designates that mnemonic as a recording form of an instruction that would not otherwise set the condition code bits; an exclamation point (“!”) preceding an operand name indicates the one's complement of the operand; CA is an architecturally defined carry-in bit; and the control signals ADD, SUB, S_CMP, and US_CMP, AND, NAND, OR, NOR, XOR and EQV indicate add, subtract, signed compare, unsigned compare, logical AND, logical NAND, logical OR, logical NOR, logical exclusive OR, and equivalence operations, respectively. In accordance with the present invention, CCU 24 directly executes compare instructions to obtain the resulting LT, GT, and EQ condition code bits in a single cycle. CCU 24 also computes, in a single cycle, the LT, GT, and EQ condition code bits for the recording instructions listed in Table I in parallel with the execution of the recording instructions by IUs 22. Furthermore, CCU 24 merges in externally computed condition code bits associated with complex instructions (e.g., multiply, divide, square root, and arithmetic shift) so that the outputs of CCU 24 are always utilized to set the condition code bits of CR 44. In this manner, the conditions upon which conditional branch instructions depend are computed as early as possible, resulting in less speculation and quicker recovery from branch misprediction.

[0021] Processor 10 employs both pipelining and out-of-order execution of instructions to further improve the performance of its superscalar architecture. Accordingly, instructions can be executed opportunistically by IUs 22, CCU 24, LSU 28, and FPU 30 in any order as long as data dependencies and antidependencies are observed. In addition, as is typical of many high-performance processors, each instruction is processed at a number of pipeline stages, including fetch, decode/dispatch, execute, finish, and completion/writeback.

[0022] During the fetch stage, sequential fetcher 17 retrieves up to a cache line of instructions from instruction cache 14. As noted above, sequential instructions fetched from instruction cache 14 are buffered within instruction queue 19, while branch instructions are removed (folded out) from the sequential instruction stream. Branch instructions are processed at each of the remaining pipeline stages by circuitry within BPU 18.

[0023] During the decode/dispatch stage, dispatch unit 20 decodes and dispatches one or more sequential instructions from instruction queue 19 to execution units 22, 24, 28, and 30. During the decode/dispatch stage, dispatch unit 20 also allocates a rename buffer within GPR rename buffers 33 or FPR rename buffers 37 for each dispatched instruction's result data, if necessary. In the depicted illustrative embodiment, instructions dispatched by dispatch unit 20 are also passed to a completion buffer within completion unit 40. Processor 10 tracks the program order of the dispatched instructions during out-of-order execution utilizing unique instruction identifiers associated with the instructions in the completion buffer of completion unit 40.

[0024] During the execute stage, execution units 22, 24, 28, and 30 execute sequential instructions received from dispatch unit 20 opportunistically as operands and execution resources for the indicated operations become available. Each of execution units 22, 24, 28, and 30 are preferably equipped with a reservation station that stores instructions dispatched to that execution unit until operands or execution resources become available. After execution of an instruction has terminated, execution units 22, 28, and 30 store data results of the instruction within either GPR rename buffers 33 or FPR rename buffers 37, depending upon the instruction type. CCU 24, on the other hand, stores the condition code bits it generates within the condition code field of one of CR rename buffers 46.

[0025] Next, execution units 22, 24, 28, and 30 notify completion unit 40 which instructions stored within the completion buffer of completion unit 40 have finished execution. Instructions are then completed by completion unit 40 in program order by marking the instructions as complete in the completion buffer. Thereafter, during the writeback stage, which is preferably merged with the completion stage, the results of the instructions are transferred from a CR rename buffer 46 to CR 44 and from GPR rename buffers 33 and FPR rename buffers 37 to GPRs 32 and FPRs 36, respectively.

[0026] Referring now to FIG. 2, there is depicted a more detailed block diagram of CCU 24. As illustrated, CCU 24 includes a single computation stage 60 and three merging stages 62, 64, and 66. Computation stage 60 includes one combination element 68 for each operand bit position (e.g., 64 in the illustrated embodiment). The combination elements 68 for bit positions 0-62 each receive as inputs the bits of the corresponding and next less significant bit positions of operands A and B, as well as the control signals 21 generated by dispatch unit 20. The combination element 68 for bit position 63 (the least significant bit (LSB)) receives as inputs bit 63 of operands A and B, control signals 21, and the CA bit and its complement. Importantly, the CA bit is not a carry such as may be generated by the execution of an add instruction, but is instead a bit within an architecturally defined register. From these inputs, computation stage 60 computes propagate (P), generate (G), and kill (Z) outputs for each of bit positions 0-63 utilizing the following equations:

[0027] Propagate, generate, and kill outputs, bit 0:

[0028] P[0]=(((!A[0] or !B[0])) and ((A[0] xor !B[0])) xor

[0029] (A[0+1] or B[0+1]))) and ADD) or

[0030] (((A[0] or !B[0]) and ((A[0] xor B[0]) xor

[0031] (!A[0+] or B[0+1]))) and SUB) or

[0032] (((A[0] or !B[0]) xor (A[1] and B[1]) and AND) or

[0033] (((A[0] and B[0]) xor (!A[1] or !B[1])) and NAND) or

[0034] (((!A[0] and !B[0]) xor (A[1] and B[1])) and OR) or

[0035] (((A[0] or B[0]) xor (!A[1] and !B[1])) and NOR) or

[0036] (((A[0] xor !B[0]) xor (A[1] xor B[1])) and XOR) or

[0037] (((A[0] xor B[0]) xor (A[1] xor !B[1])) and EQV) or

[0038] (((!A[0] or B[0]) and ((A[0] xor B[0]) xor

[0039] (A[0+1] or !B[0+1]))) and S_CMP) or

[0040] (((A[0] or !B[0])) and ((A[0] xor B[0]) xor

[0041] (A[0+1] or !B[0+1]))) and US_CMP)

[0042] G[0]=(((!A[0] or !B[0]) and ((A[0] xor !B[0]) and

[0043] (A[0+1] or B[0+1]))) and ADD) or

[0044] (((A[0] or !B[0]) and ((A[0] xor B[0]) and

[0045] (!A[0+1] or B[0+1]))) and SUB) or

[0046] (((!A[0] or !B[0]) and (A[1] and B[1])) and AND) or

[0047] (((A[0] and B[0]) and (!A[1] or !B[1])) and NAND) or

[0048] (((!A[0] and !B[0]) and (A[1] and B[1])) and OR) or

[0049] (((A[0] or B[0]) and (!A[1] and !B[1])) and NOR) or

[0050] (((A[0] xor !B[0]) and (A[1] xor B[1])) and XOR) or

[0051] (((A[0] xor B[0]) and (A[1] xor !B[1])) and EQV) or

[0052] (((!A[0] or B[0]) and ((A[0] xor B[0]) and

[0053] (A[0+1] or !B[0+1]))) and S_CMP) or

[0054] (((A[0] or !B[0]) and ((A[0] xor B[0]) and

[0055] (A[0+1] or !B[0+1]))) and US_CMP)

[0056] Z[0]=(((A[0] and B[0]) or ((A[0] xor B[0]) and

[0057] (!A[0+1] and !B[0+1]))) and ADD) or

[0058] (((!A[0] and B[0]) or ((!A[0] xor B[0]) and

[0059] (A[0+1] and !B[0+1]))) and SUB) or

[0060] (((A[0] and B[0]) and (!A[1] or !B[1]) and AND) or

[0061] (((!A[0] or !B[0]) and (A[1] and B[1])) and NAND) or

[0062] (((A[0] or B[0]) and (!A[1] or !B[1])) and OR) or

[0063] (((!A[0] and !B[0]) and (A[1] or B[1])) and NOR) or

[0064] (((A[0] xor B[0]) and (A[1] xor !B[1])) and XOR) or

[0065] (((A[0] xor !B[0]) and (A[1] xor B[1])) and EQV) or

[0066] (((A[0] and !B[0]) or ((!A[0] xor B[0]) and

[0067] (!A[0+1] and B[0+1]))) and S_CMP) or

[0068] (((!A[0] and B[0]) or ((!A[0] xor B[0]) and

[0069] (!A[0+1] and B[0+1]))) and US_CMP)

[0070] Propagate, generate, and kill outputs, bits 1-62:

[0071] P[i]=

[0072] (((A[i] xor !B[i]) xor (A[i+1] or B[i+1])) and ADD) or

[0073] (((A[i] xor B[i]) xor (!A[i+1] or B[i+1])) and SUB) or

[0074] (((!A[i] or !B[i]) xor (A[i+1] and B[i+1]) and AND) or

[0075] (((A[i] and B[i]) xor (!A[i+1] or !B[i+1])) and NAND) or

[0076] (((!A[i] and !B[i]) xor (A[i+1] and B[i+1])) and OR) or

[0077] (((A[i] or B[i]) xor (!A[1] and !B[1])) and NOR) or

[0078] (((A[i] xor !B[i])) xor (A[i+1] xor B[i+1])) and XOR) or

[0079] (((A[i] xor B[i]) xor (A[i+1] xor !B[i+1])) and EQV) or

[0080] (((A[i] xor B[i]) xor (!A[i+1] or B[i+1])) and S_CMP) or

[0081] (((A[i] xor B[i]) xor (!A[i+1] or B[i+1])) and US_CMP)

[0082] G[i]=

[0083] (((A[i] xor !B[i]) and (A[i+1] or B[i+1]) and ADD) or

[0084] ((A[i] xor B[i]) and (!A[i+1] or B[i+1]) and SUB) or

[0085] (((!A[i] or !B[i]) and (A[i+1] and B[i+1])) and AND) or

[0086] (((A[i] and B[i]) and (!A[i+1] or !B[i+1])) and NAND) or

[0087] (((!A[i] and !B[i]) and (A[i+1] and B[i+1])) and OR) or

[0088] (((A[i] or B[i]) and (!A[i+1] and !B[i+1])) and NOR) or

[0089] (((A[i] xor !B[i]) and (A[i+1] xor B[i+1])) and XOR) or

[0090] (((A[i] xor B[i]) and (A[i+1] xor !B[i+1])) and EQV) or

[0091] ((A[i] xor B[i]) and (A[i+1] or !B[i+1]) and S_CMP) or

[0092] ((A[i] xor B[i]) and (A[i+1] or !B[i+1]) and US_CMP)

[0093] Z[i]=

[0094] ((A[i] xor B[i]) and (!A[i+1] and !B[i+1]) and ADD) or

[0095] ((A[i] xor !B[i]) and (A[i+1] and !B[i+1]) and SUB) or

[0096] (((A[i] and B[i]) and (!A[i+1] or !B[i+1]) and AND) or

[0097] (((!A[i] or !B[i]) and (A[i+1] and B[i+1])) and NAND) or

[0098] (((A[i] or B[i]) and (!A[i+1] or !B[i+1])) and OR) or

[0099] (((!A[i] and !B[i]) and (A[i+1] or B[i+1])) and NOR) or

[0100] (((A[i] xor B[i]) and (A[i+1] xor !B[i+1])) and XOR) or

[0101] (((A[i] xor !B[i]) and (A[i+1] xor B[i+1])) and EQV) or

[0102] ((A[i] xor !B[i]) and (!A[i+1] and B[i+1]) and S_CMP) or

[0103] ((A[i] xor !B[i]) and (!A[i+1] and B[i+1]) and US_CMP)

[0104] Propagate, generate, and kill outputs, bit 63:

[0105] P[63]=(((A[63] xor !B[63]) xor CA) and ADD) or

[0106] (((A[63] xor B[63]) xor CA) and SUB) or

[0107] ((!A[63] or !B[63]) and AND) or

[0108] ((A[63] and B[63]) and NAND) or

[0109] ((!A[63] and !B[63]) and OR) or

[0110] ((A[63] or B[63]) and NOR) or

[0111] ((A[63] xor !B[63]) and XOR) or

[0112] ((A[63] xor B[63]) and EQV) or

[0113] (((A[63] xor B[63]) xor CA) and S_CMP) or

[0114] (((A[63] xor B[63]) xor CA) and US_CMP)

[0115] G[63]=(((A[63] xor !B[63]) and CA) and ADD) or

[0116] (((A[63] xor B[63]) and CA) and SUB) or

[0117] (((A[63] xor B[63]) and CA) and S_CMP) or

[0118] (((A[63] xor B[63]) and CA) and US_CMP)

[0119] Z[63]=(((A[63] xor B[63]) and (!CA)) and ADD) or

[0120] (((A[63] xor !B[63]) and (!CA)) and SUB) or

[0121] ((A[63] and B[63]) and AND) or

[0122] ((!A[63] or !B[63]) and NAND) or

[0123] ((A[63] or B[63]) and OR) or

[0124] ((!A[63] and !B[63]) and NOR) or

[0125] ((A[63] xor B[63]) and XOR) or

[0126] ((A[63] xor !B[63]) and EQV) or

[0127] (((A[63] xor !B[63]) and (!CA)) and S_CMP) or

[0128] (((A[63] xor !B[63]) and (!CA)) and US_CMP)

[0129] The 64 sets of P, G, and Z signals produced by computation stage 60 are then passed to merging stage 62, which merges each group of four sets into a single set of P, G, and Z signals utilizing conventional merging circuitry, such as that commonly found in carry look-ahead adders (CLAs). Thus, merging stage 62 outputs 16 sets of P, G, and Z signals to merging stage 64. Merging stage 64 similarly utilizes CLA merge circuitry to merge each group of four input sets of P, G, and Z signals into a single output set of P, G, and Z signals. As illustrated, the four sets of P, G, and Z signals output by merging stage 64 form inputs of merging stage 66.

[0130] In contrast to merging stages 62 and 64, merging stage 66 is not formed of merely conventional CLA merge circuitry, but contains additional logic to support the inclusion of condition code results computed externally for complex instructions such as multiply, divide, square root, and arithmetic shift. In the processor embodiment shown in FIG. 1, IUs 22 and FPU 30 are capable of executing complex instructions and each contain conventional condition code logic for serially computing greater than, less than, and equal to condition code bits after obtaining the result of a complex instruction. As shown in FIGS. 1 and 2, these externally generated condition code bits and a control input (EXTOP) indicating the execution of a complex instruction also form inputs to merge stage 66 of CCU 24. Merging stage 66 produces its P, G, and Z output signals according to the following equations:

[0131] P_grp=P[i] and P[i+1] and P[i+2] and P[i+3] or

[0132] (EXTOP and ExtEQ)

[0133] G_grp=G[i] or P[i] and G[i+1] or P[i] and P[i+1] and

[0134] G[i+2] or P[i] and P[i+1] and P[i+2] and G[i+3] or

[0135] (EXTOP and ExtGT)

[0136] Z_grp=Z[i] or P[i] and Z[i+1] or P[i] and P[i+1] and

[0137] Z[i+2] or P[i] and P[i+1] and P[i+2] and Z[i+3] or

[0138] (EXTOP and ExtLT)

[0139] In the above equations, the assertion of EXTOP, which is orthogonal to ADD, SUB, AND, NAND, OR, NOR, XOR, EQV, S_CMP, and US_CMP, signifies that the condition code for the operation has been computed externally to CCU 24. In addition, ExtEQ, ExtGT, and ExtLT respectively indicate that the externally generated condition code result is equal to, greater than, and less than zero. Advantageously, the inclusion of condition code bits computed externally from CCU 24 as inputs of the merge tree of CCU 24 eliminates the need for a multiplexer to select between condition code bits output by CCU 24 and externally generated condition code results for complex instructions.

[0140] The P signal output by merging stage 66 yields the value of the EQ condition code bit for all instruction types. For arithmetic and logical operations other than compare, the assertion of EQ indicates that the result is zero. For compare instructions, the assertion of EQ indicates that the operands A and B are equal. The G signal output by the merge tree provides the value of the GT bit for all instruction types. For arithmetic and logical instructions other than compare, the assertion of GT indicates that the result is positive. For compare instructions, the assertion of GT indicates that operand A is greater than operand B. Conversely, the Z signal provides the value of the LT bit for all instruction types. For arithmetic and logical instructions other than compare, the assertion of LT indicates that the result is negative. For compare instructions, the assertion of LT indicates that operand A is less than operand B. Once the P, G, and Z signals have been utilized to set the values of the EQ, GT, and LT condition code bits within one of CR rename buffers 46, the condition code bits may then be utilized to resolve conditional branch instructions.

[0141] With reference now to FIG. 3, a timing diagram is shown that illustrates the advantage of calculating the LT, GT, and EQ condition code bits in parallel with the execution of an instruction in accordance with the present invention. As depicted, a recording form of an AND (“and.”) instruction and a load (“ld”) instruction are fetched from instruction cache 14 during cycle 1, and a branch less than (“blt”) instruction is fetched from instruction cache 14 during cycle 2. The branch less than instruction is a conditional branch instruction that depends upon the state of the LT bit generated by the AND instruction.

[0142] The AND instruction is decoded by dispatch unit 20 during cycle 2 and is dispatched to an IU 22 during cycle 3. In response to dispatch of the AND instruction, the AND control signal among control signals 21 is asserted, and both the IU 22 and CCU 24 receive the operands specified by the AND instruction from one or more of GPRs 32 and GPR rename buffers 33. During cycle 4, IU 22 performs the AND operation, storing the result in one of GPR rename buffers 33. In parallel with the execution of the AND instruction, CCU 24 computes the LT, GT, and EQ condition code bits, storing them in one of CR rename buffers 46. Although the branch less than (blt) instruction is executed by BPU 18 during the following cycle (i.e., cycle 5), the availability of the condition code bits in one of CR rename buffers 46 permits the branch instruction to be executed without speculation. In contrast, a conventional processor can determine the condition code bits during cycle 5 at the earliest, meaning that the branch instruction would be executed speculatively by prediction. Thus, the present invention advantageously reduces speculative execution of branch instructions, thereby reducing the attendant performance penalty incurred in the event of misprediction of a branch.

[0143] As has been described, the present invention provides an improved method and apparatus for computing LT, GT, and EQ condition code bits. In accordance with the present invention, a processor is equipped with a condition register and a condition code unit that directly executes signed and unsigned compare instructions and computes condition code bits associated with certain other recording instructions in parallel with the execution of those instructions. The circuitry utilized to compute the condition code bits is compact and includes only a small number of logic levels in order to permit the condition code bits to be determined in a single cycle. Advantageously, merge circuitry within the condition code unit receives as inputs externally generated condition code bits associated with complex instructions. As a result, the condition code bits output by the condition code unit are always utilized to set the corresponding bits of the condition register.

[0144] While the invention has been particularly shown and described with reference to a preferred embodiment, it will be understood by those skilled in the art that various changes in form and detail may be made therein without departing from the spirit and scope of the invention.