System and method for controlling conditional branching utilizing a control instruction having a reduced word length

The present invention relates to controlling the execution of a program in an information processor for processing instructions by pipelining.

In an information processor such as microprocessor or digital signal processor (DSP) for executing instructions by pipeline processing, a pipeline hazard resulting from conditional branching, i.e., a branch hazard, will happen. Particularly when there is a great number of pipeline stages, the branch hazard causes a serious problem.

To minimize the issuance of conditional branch instructions, conditionally executable instructions are adoptable effectively. For example, according to the technique disclosed in Japanese Laid-Open Publication No. 10-49368, each conditionally executable instruction contains a condition field specifying its own execution condition. That is to say, the instruction is selectively executed depending on whether or not the execution condition specified by its own condition field is satisfied. This technique is hard to apply to an information processor of the type using a short instruction word length, because each instruction to be executed by the processor cannot afford including such an additional condition field.

As for the short-word-length information processor, an instruction issued exclusively to control the conditional execution of succeeding instructions, i.e., an execution control instruction, is applicable effectively. For instance, according to the technique disclosed in Japanese Laid-Open Publication No. 7-253882, the execution control instruction contains a condition field specifying multiple registers. The number of registers specified in the condition field is always equal to the number of instructions, which succeed the execution control instruction and will be executed under controlled conditions, and is a fixed number. And it is determined based on the value of each of these registers whether or not a succeeding instruction corresponding to the register should be executed. As a result, conditional branch instructions do not have to be used so frequently.

According to the technique disclosed in the latter publication, however, if the number of instructions to be executed under controlled conditions should be increased, then the number of registers specified in the condition field of the execution control instruction has to be increased. As a result, the length of each instruction word must be increased unintentionally. Thus, the short-word-length information processor cannot control the conditional execution of so many succeeding instructions using the execution control instruction.

An object of the present invention is controlling the conditional execution of as many succeeding instructions as possible using an execution control instruction of a short word length to suppress the branch hazard in an information processor for processing the instructions by pipelining.

To achieve this object, the present invention introduces an instruction-specifying field, which is used to define, in binary code, the number of instructions that should be executed under controlled conditions, into the execution control instruction.

More specifically, the present invention uses an execution control instruction, which contains: a condition field specifying an execution condition; and an instruction-specifying field defining, in binary code, the number of instructions to be executed only conditionally. A decision is made as to whether or not the execution condition that has been specified by the condition field is satisfied. Based on the outcome of this decision, it is determined whether or not that number of instructions, which has been defined by the instruction-specifying field for instructions succeeding the execution control instruction, should be nullified.

FIG. 1 is a block diagram schematically illustrating an exemplary configuration for a program execution control system according to the present invention.

FIG. 2 illustrates a basic format for an execution control instruction according to the present invention.

FIG. 3 illustrates a format for an execution control instruction according to a first embodiment of the present invention.

FIG. 4 is a flowchart illustrating a program execution control procedure according to the first embodiment.

FIGS. 5A and 5B illustrate specific examples of the program execution control according to the first embodiment.

FIG. 6 illustrates a format for an execution control instruction according to a second embodiment of the present invention.

FIG. 7 is a flowchart illustrating a program execution control procedure according to the second embodiment.

FIGS. 8A and 8B illustrate specific examples of the program execution control according to the second embodiment.

FIG. 9 illustrates a format for an execution control instruction according to a third embodiment of the present invention.

FIG. 10 is a flowchart illustrating a program execution control procedure according to the third embodiment.

FIGS. 11A and 11B illustrate specific examples of the program execution control according to the third embodiment.

FIG. 12 illustrates a format for an execution control instruction according to a fourth embodiment of the present invention.

FIG. 13 is a flowchart illustrating a program execution control procedure according to the fourth embodiment.

FIGS. 14A , 14 B, 14 C and 14 D illustrate specific examples of the program execution control according to the fourth embodiment.

FIG. 15 illustrates a format for an execution control instruction according to a fifth embodiment of the present invention.

FIG. 16 is a flowchart illustrating a program execution control procedure according to the fifth embodiment.

FIGS. 17A , 17 B, 17 C and 17 D illustrate specific examples of the program execution control according to the fifth embodiment.

FIG. 1 schematically illustrates an exemplary configuration for a program execution control system according to the present invention. The system shown in FIG. 1 is used for controlling the execution of a program in an information processor of the type processing instructions by pipelining. As shown in FIG. 1 , the system includes instruction providing section 10 , instruction executing section 20 and nullification controller 30 . The instruction providing section 10 includes program counter 11 , memory 12 and instruction register 13 . The program counter 11 sequentially specifies respective addresses of instructions to be fetched and executed. The memory 12 reads out the instructions, which together constitute a program (i.e., instruction set) including an execution control instruction, one after another in accordance with the addresses specified. And the instruction register 13 stores thereon the instructions read out one by one. An instruction set INST is provided from the instruction register 13 to the instruction executing section 20 and to the nullification controller 30 . The instruction executing section 20 includes: an arithmetic logic unit (ALU) 21 for performing arithmetic or logic operations in response to the instruction set INST provided; and a flag register 22 for holding multiple flags indicating the results of the computations. The nullification controller 30 receives control flags CF from the flag register 22 . In the illustrated embodiment, the control flags CF include two flags F 1 and F 2 , which have been set equal to zero or one by the ALU 21 .

FIG. 2 illustrates a basic format for an execution control instruction according to the present invention. As shown in FIG. 2 , the execution control instruction contains instruction field 41 , condition field 42 and instruction-specifying field 43 . The instruction field 41 specifies the type of the instruction as an execution control instruction. The condition field 42 specifies an execution condition EC. And the instruction-specifying field 43 defines, in binary code, the number N (where N is a natural number) of instructions that will be executed only conditionally, i.e., just when the execution condition EC is satisfied. In response to the execution control instruction, the nullification controller 30 decides based on the control flags CF whether or not the execution condition EC specified by the condition field 42 is met. And based on the outcome of this decision, the nullification controller 30 determines whether or not the number N of instructions, which number has been defined by the instruction-specifying field 43 as the number of instructions succeeding the execution control instruction, should be nullified. Suppose the nullification controller 30 has determined that the number N of instructions should be nullified since the execution condition EC is not met, the controller 30 asserts a nullification signal NUL to be supplied to the ALU 21 . In that case, even if the number N of instructions following the execution control instruction have already been executed to a midway point of the pipeline, these instructions are nullified in the ALU 21 . Accordingly, the same results are attained as if NOP (no operation) instructions had been executed instead of these instructions. Alternatively, if the nullification controller 30 has determined that the number N of succeeding instructions should not be nullified since the execution condition EC is met, then the nullification signal NUL is not asserted. Thus, the number N of instructions succeeding the execution control instruction are enabled and executed by the ALU 21 .

FIG. 3 illustrates a 16-bit format of an execution control instruction according to a first embodiment of the present invention. As shown in FIG. 3 , the execution control instruction contains a 12-bit instruction field 41 , a 2-bit condition field 42 and a 2-bit instruction-specifying field 43 . The instruction field 41 sets an operation code OP specifying the instruction type of the operation to be performed, i.e., execution control instruction. The condition field 42 specifies any of four types of execution conditions. In the illustrated embodiment, the execution condition EC to be specified is a binary code representing “F 1 =1”, “F 1 =0”, “F 2 =1” or “F 2 =0”. The instruction-specifying field 43 specifies, in binary code, the number N (where N=1 through 4) of instructions that will be executed under controlled conditions. In the illustrated embodiment, codes “00”, “01”, “10” and “11” represent “N=1”, “N=2”, “N=3” and “N=4”, respectively.

FIG. 4 illustrates a program execution control procedure according to the first embodiment. First, in Step S 1 , the execution condition EC and the number N of instructions to be executed under controlled conditions are defined for the condition field 42 and instruction-specifying field 43 , respectively. Next, in Step S 2 , the execution control instruction is executed. Then, in Step S 3 , it is determined based on the control flags CF whether or not the execution condition EC is satisfied. If the execution condition EC is not met, then the number N of instructions succeeding the execution control instruction are nullified in Step S 4 . Alternatively, if the execution condition EC is met, then the number N of instructions following the execution control instruction are executed in Step S 5 .

FIGS. 5A and 5B illustrate specific examples of the program execution control according to the first embodiment. In the illustrated examples, “F 1 =1” is the execution condition EC and “N=4”. As shown in FIG. 5A , if the execution condition is not met (i.e., when F 1 =0), then the four instructions succeeding the execution control instruction (i.e., Succeeding Instructions 1 through 4) are nullified. On the other hand, if the execution condition is met (i.e., when F 1 =1), then the four instructions following the execution control instruction (i.e., Succeeding Instructions 1 through 4) are executed, not nullified, as shown in FIG. 5 B.

In the first embodiment, the number of bits in the condition field 42 is two. Optionally, the number of execution conditions may be increased by increasing this bit number. The number of bits in the instruction-specifying field 43 is also two in the first embodiment. Alternatively, the number of instructions to be executed under controlled conditions may be increased by increasing this bit number. These modified examples will be equally applicable to all of the following second through fifth embodiments.

FIG. 6 illustrates a 16-bit format for an execution control instruction according to a second embodiment of the present invention. The format shown in FIG. 6 is the same as that shown in FIG. 3 .

FIG. 7 illustrates a program execution control procedure according to the second embodiment. First, in Step S 11 , the execution condition EC and the number N of instructions to be executed under controlled conditions are defined for the condition field 42 and instruction-specifying field 43 , respectively. Next, in Step S 12 , the execution control instruction is executed. Then, in Step S 13 , it is determined based on the control flags CF whether or not the execution condition EC is satisfied. If the execution condition EC is not met, then the number N of instructions succeeding the execution control instruction are nullified in Step S 14 a. Thereafter, the next number N of instructions following the former instruction set are executed in Step S 14 b. Alternatively, if the execution condition EC is met, then the number N of instructions succeeding the execution control instruction are executed in Step S 15 a. Thereafter, the next number N of instructions following the former instruction set are nullified in Step S 15 b.

FIGS. 5A and 8B illustrate specific examples of the program execution control according to the second embodiment. In the illustrated examples, “F= 1 ” is the execution condition EC and “N=2”. As shown in FIG. 8A , if the execution condition is not met (i.e., when F 1 =0), then the first two instructions succeeding the execution control instruction (i.e., Succeeding Instructions 1 and 2) are nullified, and the next two instructions (i.e., Succeeding Instructions 3 and 4) are executed. On the other hand, if the execution condition is met (i.e., when F 1 =1), then the first two instructions following the execution control instruction (i.e., Succeeding Instructions 1 and 2) are executed, and the next two instructions (i.e., Succeeding Instructions 3 and 4) are nullified as shown in FIG. 8 B. That is to say, according to this embodiment, instructions can be executed conditionally in accordance with the if-then-else construct.

FIG. 9 illustrates a 16-bit format for an execution control instruction according to a third embodiment of the present invention. As shown in FIG. 9 , the execution control instruction contains a 10-bit instruction field 41 , a 2-bit condition field 42 and a 4-bit instruction-specifying field 43 . The instruction field 41 defines an operation code OP specifying the instruction type of the operation to be performed, i.e., execution control instruction. The condition field 42 specifies any of four types of execution conditions. The instruction-specifying field 43 consists of first and second instruction-specifying sub-fields 43 a and 43 b, each defining, in binary code, the number of instructions that will be executed only conditionally. The first instruction-specifying sub-field 43 a is a 2-bit field for defining, in binary code, the first number N 1 (where N 1 =1 through 4) of instructions to be executed under controlled conditions. The second instruction-specifying sub-field 43 b is also a 2-bit field for defining, in binary code, the second number N 2 (where N 2 =1 through 4) of instructions to be executed under controlled conditions.

FIG. 10 illustrates a program execution control procedure according to the third embodiment. First, in Step S 21 , the execution condition EC and the first and second numbers N 1 and N 2 of instructions to be executed under controlled conditions are defined for the condition field 42 and instruction-specifying sub-fields 43 a and 43 b, respectively. Next, in Step S 22 , the execution control instruction is executed. Then, in Step S 23 , it is determined based on the control flags CF whether or not the execution condition EC is satisfied. If the execution condition EC is not met, then the first number N 1 of instructions succeeding the execution control instruction are nullified in Step S 24 a. Thereafter, the second number N 2 of instructions following the first instruction set are executed in Step S 24 b. Alternatively, if the execution condition EC is met, then the first number N 1 of instructions succeeding the execution control instruction are executed in Step S 25 a. Thereafter, the second number N 2 of instructions following the first instruction set are nullified in Step S 25 b.

FIGS. 11A and 11B illustrate specific examples of the program execution control according to the third embodiment. In the illustrated examples, “F 1 =1” is the execution condition EC, “N 1 =1” and “N 2 =3”. As shown in FIG. 11A , if the execution condition is not met (i.e., when F 1 =0), then the first instruction succeeding the execution control instruction (i.e., Succeeding Instruction 1) is nullified, and the next three instructions (i.e., Succeeding Instructions 2, 3 and 4) are executed. On the other hand, if the execution condition is met (i.e., when F 1 =1), then the first instruction following the execution control instruction (i.e., Succeeding Instruction 1) is executed, and the next three instructions (i.e., Succeeding Instructions 2, 3 and 4) are nullified as shown in FIG. 11 B. That is to say, according to this embodiment, instructions can be executed conditionally in accordance with the if-then-else construct. In addition, the respective numbers of instructions corresponding to the THEN and ELSE statements can be defined independently.

FIG. 12 illustrates a 16-bit format for an execution control instruction according to a fourth embodiment of the present invention. As shown in FIG. 12 , the execution control instruction contains a 10-bit instruction field 41 , a 4-bit condition field 42 and a 2-bit instruction-specifying field 43 . The instruction field 41 defines an operation code OP specifying the instruction type of the operation to be performed, i.e., execution control instruction. The condition field 42 consists of first and second condition sub-fields 42 a and 42 b, each specifying a single execution condition. The first condition sub-field 42 a is a 2-bit field for specifying any of four types of execution conditions as a first execution condition EC 1 . The second condition sub-field 42 b is also a 2-bit field for specifying any of four types of execution conditions as a second execution condition EC 2 . The instruction-specifying sub-field 43 defines, in binary code, the number N (where N=1 through 4) of instructions to be executed under controlled conditions.

FIG. 13 illustrates a program execution control procedure according to the fourth embodiment. First, in Step S 31 , the first and second execution conditions EC 1 and EC 2 and the number N of instructions to be executed under controlled conditions are defined for the first and second condition sub-fields 42 a and 42 b and instruction-specifying field 43 , respectively. Next, in Step S 32 , the execution control instruction is executed. Then, in Step S 33 a, it is determined based on the control flags CF whether or not the first execution condition EC 1 is satisfied. Next, in Step S 33 b, it is determined based on the control flags CF whether or not the second execution condition EC 2 is satisfied. If neither the first nor second execution condition EC 1 nor EC 2 is met, then the number N of instructions succeeding the execution control instruction are nullified in Step S 34 a. Thereafter, the next number N of instructions succeeding the former instruction set are nullified in Step S 34 b. Alternatively, if the first execution condition EC 1 is not met and the second execution condition EC 2 is met, then the number N of instructions succeeding the execution control instruction are nullified in Step S 35 a. Thereafter, the next number N of instructions succeeding the former instruction set are executed in Step S 35 b. As another alternative, if the first execution condition EC 1 is met and the second execution condition EC 2 is not met, then the number N of instructions succeeding the execution control instruction are executed in Step S 36 a. Thereafter, the next number N of instructions succeeding the former instruction set are nullified in Step S 36 b. As further alternative, if both the first and second execution conditions EC 1 and EC 2 are met, then the number N of instructions succeeding the execution control instruction are executed in Step S 37 a. Thereafter, the next number N of instructions succeeding the former instruction set are executed in Step S 37 b.

FIGS. 14A through 14D illustrate specific examples of the program execution control according to the fourth embodiment. In the illustrated examples, “F 1 = 1 ” is the first execution condition EC 1 , “F 2 = 0 ” is the second execution condition EC 2 and “N= 2 ”. As shown in FIG. 14A , if neither the first nor second execution condition is met (i.e., when F 1 =0 and F 2 =1), then four instructions succeeding the execution control instruction (i.e., Succeeding Instructions 1 through 4) are nullified. Alternatively, if the first execution condition is not met and the second execution condition is met (i.e., when F 1 =0 and F 2 =0), then two instructions succeeding the execution control instruction (i.e., Succeeding Instructions 1 and 2) are nullified and next two instructions succeeding the former instruction set (i.e., Succeeding Instructions 3 and 4) are executed as shown in FIG. 14 B. As another alternative, if the first execution condition is met and the second execution condition is not met (i.e., when F 1 =1and F 2 =1), then two instructions succeeding the execution control instruction (i.e., Succeeding Instructions 1 and 2) are executed and next two instructions succeeding the former instruction set (i.e., Succeeding Instructions 3 and 4) are nullified as shown in FIG. 14 C. As further alternative, if both the first and second execution conditions are met (i.e., when F 1 =1 and F 2 =0), then four instructions succeeding the execution control instruction (i.e., Succeeding Instructions 1 through 4) are executed, not nullified, as shown in FIG. 14 D.

It should be noted that the number M of condition sub-fields in the execution control instruction may be equal to or larger than three. In such a case, a number N of instructions, which number has been defined by the instruction-specifying field 43 for instructions succeeding the execution control instruction, are regarded as instructions to be executed under controlled conditions. If an execution condition ECm (where m=1 through M), which has been specified by associated one of the number M of condition sub-fields, is not met, then the step of nullifying the number N of instructions at a location corresponding to the specified execution condition ECm is performed. Furthermore, if an execution condition ECm (where m=1 through M), which has been specified by associated one of the number M of condition sub-fields, is met, then the step of executing the number N of instructions at a location corresponding to the specified execution condition ECm is performed.

FIG. 15 illustrates a 16-bit format for an execution control instruction according to a fifth embodiment of the present invention. As shown in FIG. 15 , the execution control instruction contains an 8-bit instruction field 41 , a 4-bit condition field 42 and a 4-bit instruction-specifying field 43 . The instruction field 41 defines an operation code OP specifying the instruction type of the operation to be performed, i.e., execution control instruction. The condition field 42 consists of first and second condition sub-fields 42 a and 42 b, each specifying a single execution condition. The first condition sub-field 42 a is a 2-bit field for specifying any of four types of execution conditions as a first execution condition EC 1 . The second condition sub-field 42 b is also a 2-bit field for specifying any of four types of execution conditions as a second execution condition EC 2 . The instruction-specifying field 43 consists of first and second instruction-specifying sub-fields 43 a and 43 b, each defining, in binary code, the number of conditionally executable instructions. The first instruction-specifying sub-field 43 a is a 2-bit field for defining, in binary code, the first number N 1 (where N 1 =1 through 4) of instructions to be executed under controlled conditions. The second instruction-specifying sub-field 43 b is also a 2-bit field for defining, in binary code, the second number N 2 (where N 2 =1 through 4) of instructions to be executed under controlled conditions.

FIG. 16 illustrates a program execution control procedure according to the fifth embodiment. First, in Step S 41 , the first and second execution conditions EC 1 and EC 2 and first and second numbers N 1 and N 2 of conditionally executable instructions are defined for the first and second condition sub-fields 42 a and 42 b and first and second instruction-specifying sub-fields 43 a and 43 b, respectively. Next, in Step S 42 , the execution control instruction is executed. Then, in Step S 43 a, it is determined based on the control flags CF whether or not the first execution condition EC 1 is satisfied. Next, in Step S 43 b, it is determined based on the control flags CF whether or not the second execution condition EC 2 is satisfied. If neither the first nor second execution condition EC 1 nor EC 2 is met, then the number N 1 of instructions succeeding the execution control instruction are nullified in Step S 44 a. Thereafter, the next number N 2 of instructions succeeding the former instruction set are nullified in Step S 44 b. Alternatively, if the first execution condition EC 1 is not met and the second execution condition EC 2 is met, then the number N 1 of instructions succeeding the execution control instruction are nullified in Step S 45 a. Thereafter, the next number N 2 of instructions succeeding the former instruction set are executed in Step S 45 b. As another alternative, if the first execution condition EC 1 is met and the second execution condition EC 2 is not met, then the number N 1 of instructions succeeding the execution control instruction are executed in Step S 46 a. Thereafter, the next number N 2 of instructions succeeding the former instruction set are nullified in Step S 46 b. As further alternative, if both the first and second execution conditions EC 1 and EC 2 are met, then the number N 1 of instructions succeeding the execution control instruction are executed in Step S 47 a. Thereafter, the next number N 2 of instructions succeeding the former instruction set are executed in Step S 47 b.

FIGS. 17A through 17D illustrate specific examples of the program execution control according to the fifth embodiment. In the illustrated examples, “F 1 = 1 ” is the first execution condition EC 1 , “F 2 = 0 ” on is the second execution condition EC 2 , “N 1 = 1 ” and “N 2 = 3 ”. As shown in FIG. 17A , if neither the first nor second execution condition is met (i.e., when F 1 =0 and F 2 =1), then four instructions succeeding the execution control instruction (i.e., Succeeding Instructions 1 through 4) are nullified. Alternatively, if the first execution condition is not met and the second execution condition is met (i.e., when F 1 =0 and F 2 =0), then one instruction succeeding the execution control instruction (i.e., Succeeding Instruction 1) is nullified and the next three instructions following the former instruction (i.e., Succeeding Instructions 2, 3 and 4) are executed as shown in FIG. 17 B. As another alternative, if the first execution condition is met and the second execution condition is not met (i.e., when F 1 =1 and F 2 =1), then one instruction succeeding the execution control instruction (i.e., Succeeding Instruction 1) is executed and the next three instructions succeeding the former instruction (i.e., Succeeding Instructions 2, 3 and 4) are nullified as shown in FIG. 17 C. As further alternative, if both the first and second execution conditions are met (i.e., when F 1 =1 and F 2 =0), then four instructions succeeding the execution control instruction (i.e., Succeeding Instructions 1 through 4) are executed, not nullified, as shown in FIG. 17 D.

It should be noted that the number M of condition sub-fields of the execution control instruction may be equal to or larger than three. In such a case, the same number M of instruction-specifying sub-fields, which correspond to the number M of condition sub-fields, respectively, are included in the instruction-specifying field of the execution control instruction. A number Nm of instructions (where m=1 through M), which number has been defined by associated one of the number M of instruction-specifying sub-fields for instructions succeeding the execution control instruction, are regarded as instructions to be executed under controlled conditions. If an execution condition ECm (where m=1 through M), which is specified by associated one of the number M of condition sub-fields, is not met, then the step of nullifying the number Nm of instructions at a location corresponding to the specified execution condition ECm is performed. Furthermore, if the execution condition ECm (where m=1 through M), which has been specified by associated one of the number M of condition sub-fields, is met, then the step of executing the number Nm of instructions at a location corresponding to the specified execution condition ECm is performed.