CA1242282A - Instruction prefetch buffer control - Google Patents
Instruction prefetch buffer controlInfo
- Publication number
- CA1242282A CA1242282A CA000502803A CA502803A CA1242282A CA 1242282 A CA1242282 A CA 1242282A CA 000502803 A CA000502803 A CA 000502803A CA 502803 A CA502803 A CA 502803A CA 1242282 A CA1242282 A CA 1242282A
- Authority
- CA
- Canada
- Prior art keywords
- instruction
- buffer
- pointer
- instructions
- register
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Expired
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F9/00—Arrangements for program control, e.g. control units
- G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
- G06F9/30—Arrangements for executing machine instructions, e.g. instruction decode
- G06F9/38—Concurrent instruction execution, e.g. pipeline, look ahead
- G06F9/3802—Instruction prefetching
- G06F9/3808—Instruction prefetching for instruction reuse, e.g. trace cache, branch target cache
- G06F9/381—Loop buffering
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F9/00—Arrangements for program control, e.g. control units
- G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
- G06F9/30—Arrangements for executing machine instructions, e.g. instruction decode
- G06F9/38—Concurrent instruction execution, e.g. pipeline, look ahead
- G06F9/3802—Instruction prefetching
Abstract
ABSTRACT
An instruction prefetch buffer control (20) is provided for an instruction prefetch buffer array (10) which stores the code for a number of instructions that have already been executed as well as the code for a number of instructions yet to be executed. The instruction prefetch buffer control includes a register (201) for storing an instruction fetch pointer, this pointer being supplied to the buffer array (10) as a write pointer which points to the location in the array where a new word is to be written from main memory. A second register (205) stores an instruction execution pointer which is supplied to the buffer array (10) as a read pointer. A first adder (203) increments the first register to increment the instruction fetch pointer for sequential instructions and calculates a new instruction fetch pointer for branch instructions.
A second adder (215) increments the second register to increment the instruction execution pointer for sequential instructions and calculates a new instruction execution pointer for branch instructions. Incrementing of the second register is variable depending on the length of the instruction. A third adder (221) is responsive to the output of the first adder and a branch target address to calculate whether the target instruction is contained in the array (10) and, if it is, causes the new- instruction execution pointer calculated by the second adder (215) to be loaded into the second register (205).
An instruction prefetch buffer control (20) is provided for an instruction prefetch buffer array (10) which stores the code for a number of instructions that have already been executed as well as the code for a number of instructions yet to be executed. The instruction prefetch buffer control includes a register (201) for storing an instruction fetch pointer, this pointer being supplied to the buffer array (10) as a write pointer which points to the location in the array where a new word is to be written from main memory. A second register (205) stores an instruction execution pointer which is supplied to the buffer array (10) as a read pointer. A first adder (203) increments the first register to increment the instruction fetch pointer for sequential instructions and calculates a new instruction fetch pointer for branch instructions.
A second adder (215) increments the second register to increment the instruction execution pointer for sequential instructions and calculates a new instruction execution pointer for branch instructions. Incrementing of the second register is variable depending on the length of the instruction. A third adder (221) is responsive to the output of the first adder and a branch target address to calculate whether the target instruction is contained in the array (10) and, if it is, causes the new- instruction execution pointer calculated by the second adder (215) to be loaded into the second register (205).
Description
~2~Z~2 INSTRUCTION PREFE;TCH E~t~FFER
CONTROL
DESCRIPTION
Cross Reference to Related Application S . The invention disclosed herein is related in subject matter to the invention disclosed in Canadian copending patent application S.N. 490,347, filed ~eptember 10, 19~5, by Vojin G. Oklobdzija which is assigned to the assignee of this application.
Technical Field The subject invention is directed to an instruction prefetch buffer for a microprocessor and, more particularly, to such a prefetch buffer control which handles short loops and short forward references.
Description of the Prior Art The usual implementation of an inst~uction prefetch buffer (IPB) can degrade processor performance in a number of important situations~
Consider for example a pipelined processor. As normally implemented, each successful branch will cause the IPB to be cleared and a new instruction fetch to start from the target address. This leads to a penalty of several cycles depending on the depth of the pipe and the length of the target instruction being loaded and executed. Since ~2'9~
~2 Y098~-056 sucsessful branches constitute about 15~ of all instructions, the impact on overall system performance can be significant~ Fur~her, consider a processor where most instructions execute in one cycle and where there is only a single data path to memory for both data and instructions. Code which contains a large percentage of load and store instruc~ions will tend to empty the IPB
because the memory bandwidth is used to move data.
Cycles will thus be lost since instruction fetches cannot be overlapped with other operationsl This situation i~ most apt to occur in loops of the very important move character subroutine.
In the instance where a program is looping in a short loop n-times and having k-cycles lost during each passage through the loop, a total of k*n cycles will be lost, where the asterisk denotes multiplication. Such a`loop is diagramatically illustrated in Figure 1 of the drawings. Statistics obtained from a simulator executing typical instruction streams show that short branches are quite frequent. Specifically, 20 to 30~ of the successful branches are associated with short backward loops (less than 64 bytes long), and 10 to 20% are associated with short orward branches (less than 16 bytes).
Summary of the Invention It is therefore an object of the present invention to provide an instruction prefetch buffer control capable of handling short loops and short forward references to statistically lower the miss rate of instructions stored in the instruction prefetch buffer.
r ,
CONTROL
DESCRIPTION
Cross Reference to Related Application S . The invention disclosed herein is related in subject matter to the invention disclosed in Canadian copending patent application S.N. 490,347, filed ~eptember 10, 19~5, by Vojin G. Oklobdzija which is assigned to the assignee of this application.
Technical Field The subject invention is directed to an instruction prefetch buffer for a microprocessor and, more particularly, to such a prefetch buffer control which handles short loops and short forward references.
Description of the Prior Art The usual implementation of an inst~uction prefetch buffer (IPB) can degrade processor performance in a number of important situations~
Consider for example a pipelined processor. As normally implemented, each successful branch will cause the IPB to be cleared and a new instruction fetch to start from the target address. This leads to a penalty of several cycles depending on the depth of the pipe and the length of the target instruction being loaded and executed. Since ~2'9~
~2 Y098~-056 sucsessful branches constitute about 15~ of all instructions, the impact on overall system performance can be significant~ Fur~her, consider a processor where most instructions execute in one cycle and where there is only a single data path to memory for both data and instructions. Code which contains a large percentage of load and store instruc~ions will tend to empty the IPB
because the memory bandwidth is used to move data.
Cycles will thus be lost since instruction fetches cannot be overlapped with other operationsl This situation i~ most apt to occur in loops of the very important move character subroutine.
In the instance where a program is looping in a short loop n-times and having k-cycles lost during each passage through the loop, a total of k*n cycles will be lost, where the asterisk denotes multiplication. Such a`loop is diagramatically illustrated in Figure 1 of the drawings. Statistics obtained from a simulator executing typical instruction streams show that short branches are quite frequent. Specifically, 20 to 30~ of the successful branches are associated with short backward loops (less than 64 bytes long), and 10 to 20% are associated with short orward branches (less than 16 bytes).
Summary of the Invention It is therefore an object of the present invention to provide an instruction prefetch buffer control capable of handling short loops and short forward references to statistically lower the miss rate of instructions stored in the instruction prefetch buffer.
r ,
2 !32 It is anothèr object of the invention to provide an instruction prefetch buffer control which allows a processor to execute short loops directly from the buffer, the size of the buffer being determined by a statistical analysis of typical instruction streams.
According to this invention, the instruction prefe~ch buffer is implemented as a one-port-write, two-port-read register stack as disclosed in the aforementioned copending pa~ent application S.N. 490,347, filed by Vojin G.
Oklobdzija. The key to the present invention is the control of two IPB pointers referred to herein as the instruction fetch pointer (IFP) and the lS instruction execution pointer ~IEP). The IEP
points to the location in the IPB where the next instructi~ fetched will be placed, while the IEP
points to the the location in the IPB where the next instruction to be executed is located. The pointers allow the detection of branch targets within the IPB and permit the execution of a target instruction located within the IP~ without first clearing the IPB or initiating a new instruction fetch. When a backward or forward reference is encountered, it is determined by simple hardware whether this instruction is contained in the IPB or not. If the instruction referenced is outside the IPB, both pointers, IEP
and IEP, are set to the same location and the target instruction referenced is fetched from memory into the IPB. The pointers are also controlled so that in a steady state condition the buffer contains the code for a number of instructions that have already been executed as well as the code for a number of instructions yet ~29~22~3~
Y0984-~56 to be executed. For example, in the calse of a 64-byte I.PB macro, 48 bytes of code or approximately sixteen instructions that have been executed and 16 bytes of code or approximately five instructions yet to be executed are stored in the IPB in a steady state condition. For the purposes of the description of the invention, it is assumed that the rate of instruction fetch into the IPB always exceeds the rate of instruction 10 ~ execution. beca~se there will be multi-cycle instructions as well as half-word instructions~
Every fetch brings ~our bytes to the IPB which is 1.5 instructions. on the average. In this case, the number of instructions yet to be executed will tend to grow and, in a circular IPB, the instructions already executed will be steadily overwritten.. In order to preserve thf recently executed portions of the program to be used for short backward references, the rate at which the pointer IFP advances should be limited and coupled to the pointer IEP. This is achieved by hardware according to a preferred embodiment of the invention. Additional hardware also detects whether the IPB has recently been cleared by the occurance of a branch out of the IPB. In that case, the IEP register is cleared to zero.
Brief Descript~on of the Drawings The foregoing and other objects, aspects and advantages of the invention will be better understood from the following detailed description of the invention with reference to the drawings, in which:
~2~ 2 Y0984~056 Figure 1 is a diagramatical illustration of a short loop in a program indicating the loss of cycles when using conventional instruction prefetch buffers;
~ IFigure 2 is a block diagram showing the rel~ationship of the instruction prefetch buffer array,~.the instruction prefetch buffer control and the cen~ral processing unit of the microprocessor; --Figure 3 is a block diagram showing the register selection mechanism and organization of the intruction prefetch buffer array shown in _ Figure 2;
Figure 4 is a block diagram showing the instruction prefetch buffer control according to the present invention;
Figure 5 is a simplified version of thc block diagram of Figure 4 showing the instruction sequencing logic;
Figure 6 is a diagram representing a circular instruction prefetch buffer showing the relative location of the instruction fetch pointer and the instruction execution pointer;
Yigure 7 is a block and logic diagram of the branch local address generation logic;
Figure 8 is a flow chart illustrating the function of the instruction pref~tch buffer control according to the invention; and Fig~re 9 is a flow chart of the process called in the flow chart of Figure 7 to determine whether the target is in the instruction prefetch buffer.
lZ~
Y098~-056 Detailed Description of a Preferred Embodiment of the Invention Referring now tG the drawings, and more particularly to Figure 2, there is shown a portion of a microprocessor includins an instruction prefetch buffer (IPB) array 10, an IPB control 20 and the interface connections to the central processing unity (CPU) of the microprocessor.
These latter connections are delineated by the dotted line bearing the reference numeral 30.
Thus, it will be ~nderstood that while the CPU
forms no part of the invention, enough of the interface is shown below the dotted line 30 to teach one skilled in the art how the instruction lS prefetch buffer and its control function with the CPU .
Explaining in greater detail, in a preferred embodiment, the IPB array 10 is connected to a 32-bit wide processor bus 12 over which instructions are supplied to the IPB array 10 from a random access memory (RAM), not shown. In the drawing, the thirty-two lines of the bus 12 are indicated with a slash and the numeral "32"
adjacent, and this convention is used throughout the various figures. I~structions are written into the IPB array 10 under the control of the IPB
control 20 by means of a 4-bit write pointer, and instructions are read out of the IPB array 10 also under the control of the IPB control 20 by means of 5-bit read pointer. The 32 bits which are read out of the IPB array 10 include 8 bits of operation (OP) code, the remaining 24 bits comprising branch instruction (BI) and jump instruction (JI) fields. The 8-bit OP code is ~2~
~984-056 ~upplied to a microinstruction read on:Ly memory ROM) 301. This 8-bit OP code serves as an ~ddress to select a 32-bit microins~ruction which is transmitted to the CP~. This is conventional 2nd well understood in the art. The 8-bit OP code is also supplied to an instruction length decoder 303 t the output of which is supplied to both the cPU and the IPB control 20. As will be explained in more detail, the IPB array 10 is capable of h~ndling variable length instructions, 16-bits or 32-bits in the preferred embodiment. Thus, it is n2cessary to provide an instruction length decoder
According to this invention, the instruction prefe~ch buffer is implemented as a one-port-write, two-port-read register stack as disclosed in the aforementioned copending pa~ent application S.N. 490,347, filed by Vojin G.
Oklobdzija. The key to the present invention is the control of two IPB pointers referred to herein as the instruction fetch pointer (IFP) and the lS instruction execution pointer ~IEP). The IEP
points to the location in the IPB where the next instructi~ fetched will be placed, while the IEP
points to the the location in the IPB where the next instruction to be executed is located. The pointers allow the detection of branch targets within the IPB and permit the execution of a target instruction located within the IP~ without first clearing the IPB or initiating a new instruction fetch. When a backward or forward reference is encountered, it is determined by simple hardware whether this instruction is contained in the IPB or not. If the instruction referenced is outside the IPB, both pointers, IEP
and IEP, are set to the same location and the target instruction referenced is fetched from memory into the IPB. The pointers are also controlled so that in a steady state condition the buffer contains the code for a number of instructions that have already been executed as well as the code for a number of instructions yet ~29~22~3~
Y0984-~56 to be executed. For example, in the calse of a 64-byte I.PB macro, 48 bytes of code or approximately sixteen instructions that have been executed and 16 bytes of code or approximately five instructions yet to be executed are stored in the IPB in a steady state condition. For the purposes of the description of the invention, it is assumed that the rate of instruction fetch into the IPB always exceeds the rate of instruction 10 ~ execution. beca~se there will be multi-cycle instructions as well as half-word instructions~
Every fetch brings ~our bytes to the IPB which is 1.5 instructions. on the average. In this case, the number of instructions yet to be executed will tend to grow and, in a circular IPB, the instructions already executed will be steadily overwritten.. In order to preserve thf recently executed portions of the program to be used for short backward references, the rate at which the pointer IFP advances should be limited and coupled to the pointer IEP. This is achieved by hardware according to a preferred embodiment of the invention. Additional hardware also detects whether the IPB has recently been cleared by the occurance of a branch out of the IPB. In that case, the IEP register is cleared to zero.
Brief Descript~on of the Drawings The foregoing and other objects, aspects and advantages of the invention will be better understood from the following detailed description of the invention with reference to the drawings, in which:
~2~ 2 Y0984~056 Figure 1 is a diagramatical illustration of a short loop in a program indicating the loss of cycles when using conventional instruction prefetch buffers;
~ IFigure 2 is a block diagram showing the rel~ationship of the instruction prefetch buffer array,~.the instruction prefetch buffer control and the cen~ral processing unit of the microprocessor; --Figure 3 is a block diagram showing the register selection mechanism and organization of the intruction prefetch buffer array shown in _ Figure 2;
Figure 4 is a block diagram showing the instruction prefetch buffer control according to the present invention;
Figure 5 is a simplified version of thc block diagram of Figure 4 showing the instruction sequencing logic;
Figure 6 is a diagram representing a circular instruction prefetch buffer showing the relative location of the instruction fetch pointer and the instruction execution pointer;
Yigure 7 is a block and logic diagram of the branch local address generation logic;
Figure 8 is a flow chart illustrating the function of the instruction pref~tch buffer control according to the invention; and Fig~re 9 is a flow chart of the process called in the flow chart of Figure 7 to determine whether the target is in the instruction prefetch buffer.
lZ~
Y098~-056 Detailed Description of a Preferred Embodiment of the Invention Referring now tG the drawings, and more particularly to Figure 2, there is shown a portion of a microprocessor includins an instruction prefetch buffer (IPB) array 10, an IPB control 20 and the interface connections to the central processing unity (CPU) of the microprocessor.
These latter connections are delineated by the dotted line bearing the reference numeral 30.
Thus, it will be ~nderstood that while the CPU
forms no part of the invention, enough of the interface is shown below the dotted line 30 to teach one skilled in the art how the instruction lS prefetch buffer and its control function with the CPU .
Explaining in greater detail, in a preferred embodiment, the IPB array 10 is connected to a 32-bit wide processor bus 12 over which instructions are supplied to the IPB array 10 from a random access memory (RAM), not shown. In the drawing, the thirty-two lines of the bus 12 are indicated with a slash and the numeral "32"
adjacent, and this convention is used throughout the various figures. I~structions are written into the IPB array 10 under the control of the IPB
control 20 by means of a 4-bit write pointer, and instructions are read out of the IPB array 10 also under the control of the IPB control 20 by means of 5-bit read pointer. The 32 bits which are read out of the IPB array 10 include 8 bits of operation (OP) code, the remaining 24 bits comprising branch instruction (BI) and jump instruction (JI) fields. The 8-bit OP code is ~2~
~984-056 ~upplied to a microinstruction read on:Ly memory ROM) 301. This 8-bit OP code serves as an ~ddress to select a 32-bit microins~ruction which is transmitted to the CP~. This is conventional 2nd well understood in the art. The 8-bit OP code is also supplied to an instruction length decoder 303 t the output of which is supplied to both the cPU and the IPB control 20. As will be explained in more detail, the IPB array 10 is capable of h~ndling variable length instructions, 16-bits or 32-bits in the preferred embodiment. Thus, it is n2cessary to provide an instruction length decoder
3~3 to recognize the length of the instruction wnich has been read out of the IPB array 10.
Also, the 24-bit portion of the 32 bits read out O~ the IPB array 10 which comprises the BI and JI
fields is supplied tc both the CPU and the IPB
C~?ntrol 20. The 32-bit instruction address from tne instruction address register (IAR) 305 is 5applied to both the CPU and the Ips control 20.
Thus, the IPB control 20 receives as inputs 24 bits frorn the 32 bits read out of the IPB array 13, the decoded instruction length and the i~lStruction address. From this information, the IæB control 20 generates the 4-bit write and the s-bit read pointers.
Before describing the IPs control 20, a brief d@Scription of the IPB array 10 is first presented 5~ that a better understanding of the function of t~e IPB control may be had. The IPB array 10 is the subject of the above referenced copending ; patent application S.N. 490,347, of Vojin G.
o~lobdzija. In the preferred embodiment, the IPB
a~ray 10 is implemented as a 32-bit one-port-write and 16~bit two-port-read array. More ~2~ 8;~
specifically, the array 10 comprises and even array 101, referred to herein as the A array, and an odd array 102, referred to herein as the B
array. Associated with the A array 101 is an address decoding networ~ 103. A corresponding address decoding network 104 is associated with the B array 102. the A array 101 is addressed with even address values, while the B array 102 is addressed with odd address values. Each array 101 and 102 is shown as capahle of storing 16 half words (16 bits) for a total of 32 half words, but those skilled in the art will understand that this is but one specific example~ The principles are equally applicable to larger or smaller arrays and could be extended to byte fields (8 bits) and double-word fields (64 bits)O
As indicated in Figure 3, instruction I0 is a 32-bit instruction beginning at an even address in the A array 101 and continuing over to an odd address in the ~ array 102. Instruction Il is another 32-bit instruction, but this instruction begins at an odd address in the B array 102 and continues to an even address in the A array 101.
Instruction I is an example of a 16-bit instruction which begins at an odd address and is totally contained in the B array 102, while instruction I3 is another 16-bit instruction contained in the A array 101. Specific examples of instruction formats are given in the above referenced copending patent application S.N.
490,347. What should be understood here is ; that one word (32 bits) is fetched at a time from main memory or RAM and written into the IPB array 10 via the processor ~us 12. Since different length instructions are used, an instruction can be anywhere within 2-byte boundaries, and the problem is to extract the instructions one at a time however they might be distributed within the boundaries of the array.
The 32-bit bus 12 is split into two 16-bit buses 121 and 122 in the array 10 with bus 121 supplying the A array 101 with the first two bytes of the word fetched from RAM to be written in that array and bus 122 supplying the ~ array 102 with IO the second two bytes of the same word to be written in that array. Register 14 temporarily stores the wri~e pointer W from the IPB control 20 which points to the 32-bit }ocation where a new word is to be written. The 32-bit location is obtained by combining both 16-bit registers A and B from the array 10 and addressing them as one 32-bit register. Thus, write address W0 addresses the A array address A0 and the B array address B
and so forth. Therefore, the W pointer is only four bits long.
Register 15 temporarily stores the read pointer ~ from the IPB control 20 which points to the 32-bit location from where the ins~ruction is to be read. The 5-bit read pointer R is supplied to the address generation and control logic 16.
This logic generates the 4-bit addresses for each of the arrays 101 and 102. Specifically, the
Also, the 24-bit portion of the 32 bits read out O~ the IPB array 10 which comprises the BI and JI
fields is supplied tc both the CPU and the IPB
C~?ntrol 20. The 32-bit instruction address from tne instruction address register (IAR) 305 is 5applied to both the CPU and the Ips control 20.
Thus, the IPB control 20 receives as inputs 24 bits frorn the 32 bits read out of the IPB array 13, the decoded instruction length and the i~lStruction address. From this information, the IæB control 20 generates the 4-bit write and the s-bit read pointers.
Before describing the IPs control 20, a brief d@Scription of the IPB array 10 is first presented 5~ that a better understanding of the function of t~e IPB control may be had. The IPB array 10 is the subject of the above referenced copending ; patent application S.N. 490,347, of Vojin G.
o~lobdzija. In the preferred embodiment, the IPB
a~ray 10 is implemented as a 32-bit one-port-write and 16~bit two-port-read array. More ~2~ 8;~
specifically, the array 10 comprises and even array 101, referred to herein as the A array, and an odd array 102, referred to herein as the B
array. Associated with the A array 101 is an address decoding networ~ 103. A corresponding address decoding network 104 is associated with the B array 102. the A array 101 is addressed with even address values, while the B array 102 is addressed with odd address values. Each array 101 and 102 is shown as capahle of storing 16 half words (16 bits) for a total of 32 half words, but those skilled in the art will understand that this is but one specific example~ The principles are equally applicable to larger or smaller arrays and could be extended to byte fields (8 bits) and double-word fields (64 bits)O
As indicated in Figure 3, instruction I0 is a 32-bit instruction beginning at an even address in the A array 101 and continuing over to an odd address in the ~ array 102. Instruction Il is another 32-bit instruction, but this instruction begins at an odd address in the B array 102 and continues to an even address in the A array 101.
Instruction I is an example of a 16-bit instruction which begins at an odd address and is totally contained in the B array 102, while instruction I3 is another 16-bit instruction contained in the A array 101. Specific examples of instruction formats are given in the above referenced copending patent application S.N.
490,347. What should be understood here is ; that one word (32 bits) is fetched at a time from main memory or RAM and written into the IPB array 10 via the processor ~us 12. Since different length instructions are used, an instruction can be anywhere within 2-byte boundaries, and the problem is to extract the instructions one at a time however they might be distributed within the boundaries of the array.
The 32-bit bus 12 is split into two 16-bit buses 121 and 122 in the array 10 with bus 121 supplying the A array 101 with the first two bytes of the word fetched from RAM to be written in that array and bus 122 supplying the ~ array 102 with IO the second two bytes of the same word to be written in that array. Register 14 temporarily stores the wri~e pointer W from the IPB control 20 which points to the 32-bit }ocation where a new word is to be written. The 32-bit location is obtained by combining both 16-bit registers A and B from the array 10 and addressing them as one 32-bit register. Thus, write address W0 addresses the A array address A0 and the B array address B
and so forth. Therefore, the W pointer is only four bits long.
Register 15 temporarily stores the read pointer ~ from the IPB control 20 which points to the 32-bit location from where the ins~ruction is to be read. The 5-bit read pointer R is supplied to the address generation and control logic 16.
This logic generates the 4-bit addresses for each of the arrays 101 and 102. Specifically, the
4-bit A address is supplied to the A decoder 103 and the 4-bit B address is supplied to the B
decoder 104. The output of the A array 101 is supplied via a first 16-bit output bus 105 to port A, and the output of the B array 102 is supplied via a second 16-bit output bus 106 to port B.
~Z23~
In order to have the instructions read out of the array 10 aligned, the beginning of the instruction has always to be read from the multiplexer 123 whether the instruction is a 16-bit instruction or a 32-bit instruction. Also, to access a 32-bit instruction from the array, both sides of the array, odd and even, have to be addressed and data brought to ports A and B. But as already described, a 16-bit portion containing the beginning of the instruction may be either in the A array 101 or the B array 102; i.e., 32-bit instructions can start in either array.
Therefore, multiplexers 123 and 12~ are provided to read the instructions out of the array 10 in the proper order. The two 16-bit inputs of each multiplexer 123 and 124 are each connected to the 16-bit F~cts A and B of the array 10 but in the opposite order from each other. The 16-bit outputs of the multiplexers 123 and 124 are 2~ combined to provide a 32-bit instruction output.
In the case of a 16-bit instruction, the high order 16 bits comprise a null field and are simply disregarded.
The subject invention relates to the IPB
control 20 which supplies the write pointer W and the read pointer ~ to the IPB array 10. An implementation of the invention is illustrated in Figure ~. The instruction fetch pointer (IFP) is loaded in a register 2~1, and the adder 203 is used to increment the IFP. The IFP is incremented every time a new instruction i5 written into the IPB array 10. Since the IFP points to 32-bit boundaries, it is incremented by two every time.
The ~our most significant bits or the IFP are used as a write pointer W for the IPB array 10. The Y0984-05~
instruction execution pointer (IEP) is loaded in a register 205. The five bits of the IEP are used as a read pointer R for the IPB array 10. The load signals FL and EL are used to load ne~ values of IFP and IEP into registers 201 and 205, respectively. The clear signals FC and EC are used to clear the registers. These clear signals occur when a branch or jump is out of range requiring the loading and reading of instructions to start from the beginning. The pointers IFP and IEP are diagramatically illustrated in Figure 6 which illustrates the relative positions of the pointers with reference to IPB address space.
Figure 6 is re~erred to in the discussion which follows to illustrate the function of the invention~
Referring now back to Figure 4, in addition to incrementing the IFP register 201, the adder 203 is used for branches to calculate the difference F
between IFP and IEP, which difference represents the number of instructions that have been written into the IPB but have yet to be encountered by the IEP as shown in Figure 6. However, this number is actually for one smaller than the difference because the position pointed to by the IFP does not yet contain the new instructionO This fact facilitates the hardware implementation because l's complement addition ~subtraction) gives the expected result. The implementation requires only that IEP be inverted before it is fed into adder 203. Thus, the IEP contained in register- 205 i5 inverted by inverter 207 and supplied to the other input of adder 203 via the multiplexer 209.
Performing l's complement (inversion) in complementing logic 211 of the output F of adder Y09~4-056 203 gives a value P which represents the number of instructions between the pointers IEP and IFP
which are the instructions accumulated in the IPB
array 10 but not yet overwritten by new instruction fetches. However, as illustrated in Figure 6, the proper interpretation of the value P
requires two separate cases to be distinguished:
lo A branch out of the region occurred recently, or 2. The IFP has rolled over zero at least once since the last branch out of the reglon.
When a branch out of the region is encountered, both IEP and IFP are set to location 0, the target instruction is fetched and decoded, and the zero cross (ZC) latch 213 is reset to zero. Note that IEP and IFP are trying to catch each other as illustrated in Figure 6. In the si`tuation illustrated in Figure 6, it is possible that IFP
reached its current position starting from 0 (after a successful branch out of the region was encountered), case l; or that it has already made a revolution, case 2. In case 1, P=Pl, while in case 2, P=P2.
Referring again to Figure 4, the carry signal Cl from adder 32 indicates two possibilities:
Cl=l; IFP>IEP both case 1 and case 2 are possible, or Cla0; IFP<IEP rollover, or case 2.
Case 1 or case 2 possibilities are resolved by the ZC latch 213. If the latch is set~ then case 2 is encountered; otherwise, it ls case 1. The ZC
latch 213 is set with the carry signal C2 from adder 215 during the increment of the IEP cycle (C~=l during the rollover). More specifically, Y~984-056 the C2 output of adder 215 is supplied to one input of AND gate 214, and the output of AND gate 214 is connected to the set input of latch 213.
AND gate 214 is enabled by the increment select (MC=0,1) so that the latch 213 can only be set during normal instruction sequencing.
Adder 215 is used to calculate the new IEP.
When no branch is encountered, IEP i5 simply incremented ~y the instruction length (1,2). In order to increment the IEP for 1 or 2, the output from the IEP register 205 is fed to the right input of the adder 215, and a 1 or a 2 is fed to the left input of the adder via a multiplexer 217.
The increment, 1 or 2, is selected by the multiplexer select code tMC) being 0 or 1; i.e., binary 00 or 01. When a successful branch is encountered and the target is within the IPB array 10, the target location is ~enerated. In this case, the inputs to adder 215 are the IEP from register 205 and the branch offset sR from register 219, the latter being supplied to the adder 21S via the multiplexer 217 when MC=2 (i.e., binary 10). BR is represented as a 2's complement number and consists of a 5-bit number derived from the five least significant bits (Lss) of the instruction read into instruction register (IRG) 222 from the array 10 plus one sign bit. A third adder 221 is used to indicate whether a branch instruction is contained within the IPB array 1 or not. One input to the adder 221 is BR from register 219 and the other is P from complementing logic 211. Since P = F, the result is as follows:
P - BR + 2 if BR < 0 P ~ BR - F - 1 if BR > 0 Z2~
The indication of the range is given by the carry-out signal C from adder by the following table:
~ . _ .
C3 Br > 0 Br < 0 ._ __ 0 In the range Out of the range _ _ ~ ___ -. 1 Out of the range In the range , _ _- I - ~ I
Carry out C is exclusive ORed with the BR sign bit in exclusive NOR gate 223 so that~the outside range signal IIR equals 1 when the target of the branch instruction is outside of the IPB array 10.
15 To summarize, i~ a successful branch instruction is encountered, the target address is used in conjunction with the output of adder 221 to determine whether the target instruction is contained within the IPB array 10 or not, and the Zo new IEP is calculated (IEP') which points to the target instruction in the IPB. If it is detemined that the instruction is in the IPB array 10, IEP' is loaded into register 205. If not, registers 201 and 205 are set to zero and the ZC latch 213 is also set to 0. These operations are triggered on the branch successful signal being true.
The instruction sequencing logic is better shown in the simplified block diagram of Figure 5.
The IFP in register 201 is incremented by adding 2 to the contents of the register in adder 203. The 2 is supplied to the adder 203 via multiplexer 209~ It will be recalled that the IFP is incremented by 2 because a full word (32-bits) is ~ ~A ~3a2 ~0984-056 written into the array 10. The IEP in register 205 is incremented by adding a 1 or a 2 to the contents of the register in adder 215. The constants 1 and 2 are supplied to adder 215 via multiplexer 217. Incrementing is by 1 if the instruction is a 16-bit instruction and by 2 if the instruction is a 32-bit instruction. On the other hand, for a branch, the offset BR in register 219 is added to the contents of register 205 to produce IEP'. The multiplexer 217 responds to an input that indicates the instruction length as determined by the length determining logic 303 shown in Figure 2 or a branch condition. The IEP
load signal is generated under the following conditions:
1. IEP increment (branch not present or branch not successful), or 2. Branch successful and branch target is in the calculted IPB range.
The function of the ~ogic shown in Figure 7 is to detect whether the Branch Target Instruction is contained in the IPB. There are two types of branches, short and long, which define the locality or range of the target. The offset fields of the Branch Instructions read into instruction register 222 in Figure 4 are shown in Figure 7 as being temporarily stored in registers 223 and 225. Register 223 is a one byte register containing the field JI of the target instruction, while register 22S is a three byte register containing the field BI of the branch instruction.
These fields are sign-extended; e.g., the signs are duplicated to pad the fields to full 32-bit words~ These sign-extended fields are then supplied to a multiplexer 127. The appropriate ., one of these two fields is selected wit:h the multiplexer 127 by the BC input (depencling on whether the instruction is of the short: or long type as determined by the logic 303 in Figure 2), S and this offset is added to the contents of the IAR 305 (see.Figure 2) in adder 129 in order to generate the address of the target instruction.
The 31 most significant bits (MSB), that is all but the least sig~ificant bit, of the multiplexer lO 127 are checked to determine whether this offset is smaller than 64 bytes, in which case there is a possibility that the target instruction might be contained in the IPB. This is done by comparing the sign field with Exclusive OR gates 131 to dete.rmine how long.this field is. If the sign field is at least 26 bits long (leaving six bits including ~he least significant 32nd bit for the value), then the offset of six bits value can not exceed 64 bytes and, therefore, it is possible 2~ that this offset is in the range of -48 to ~16 bytes. Since the least significant bit of the branch offset is always zero (instructions always begin on even byte or half word boundaries), the least significant bit ls not taken into account.
The outputs of the Exclusive OR gates 131 are connected to the inputs of NOR gate 133, and the output of NOR gate 133 indicates whether the looping logic should be disabled. That is, IR = 0 indicates that the Branch Target is certainly out of the range; conversely, I~ = 1 enables the looping logio. This check is performed in parallel with the Branch Address Generation, and if a match is found indicating that the instruction exists in the IPB, the request for the fetch from this memory location is not activatedO
Y098~-056 Instead, the instruction is taken from the IPB.
The implementation shown in Figure 7 is only one of the possible implementations, and those skilled in the art should understand that Figure 7 is illustrative of the principles of operation~
However, actual implementation depends on the .
particular case; i.e., machine structure and orginization.
The flow chart of Figure 8 illustrates the function of the IPB control according to the invention. The process is entered at block 40, and the ~irst operation, as indicated in block 42 is to initialize the IEP, IEP and ZC to zero.
That is, registers 201 and 205 and the latch 213 are all set to zero. Then, in decision block 44, a test is made to determine if the IPB is empty.
This te~t i5 performed by the comparator 208 which compares the IFP and the IEP in registers 201 and 205 to determine if they are equal. Equality is ~ the indication that the IPB is empty. Initially, the IPB is empty because the registers were initialized to zero in block 42. Therefore, a determination is made in decision block 44 that the IPB is empty, and an instruction fetch from ~5 the bus is madeO This is accomplished in block 46 by reading the instruction into the IPB array 10 at address IFP. Then, in block 48, the select signal MF for multiplexer 209 is set to zero, and in block 50, IFP is incremented by 2. The incrementing of IFP by 2 is performed by adder 203. Now, when the test is made in decision block 48 to determine if the IPB is empty, it will be assumed that the test is negative.
~29i~28;~
Y09~4-056 In this case, the instruction in the IP3 array 10 identified by the IEP pointer in register 205 is read to the instruction register IRG 222 as indicated by block 52. The instruction is decoded S in block 54, and then in decision block 56, it is tested to determine if it is a branch or jump instruction. Assuming first that it is not a branch or jump instruction, a determination is made in decision block 58 as to the length of the instruction; i.e., whether it is a 16-bi~ or a 32-bit instruction. This is the function of the instruction length logic 303 in ~igure 2. If the instruction is a 16-bit instruction, then in block 60 the IEP pointer in register 205 is incremented by 1. On the other hand, if the instruction is a 32-bit instruction, the IEP pointer in register 205 is incremented by 2 as indicate3 by block 62.
These functions are accomplished by the adder 215 and the multiplexer 217. In either case, the carry output of adder 215 is checked in decision block 64 to determine if there is a carry. If so, the zero cro~s latch 213 is set as indicated in block 66; otherwise, it is not. The process then loops back, and the steps described are repeated.
Returning now to decision block 56, assume now that the result of the test is positive; that is, the decoded instruction is a branch or jump instruction. In this case, a test is made in decision block 68 to determine if the target instruction is in range. As will be recalled, this is the function of the logic shown in Figure 7. A logic 1 output of the NOR gate 133 indicates a positive result of the test, and this will result in a test being made in decision block 70 to determine if the target is in the IPB array 10. This process is described in more detail with reference to the flow chart in Figure 9, but for now, suffice it to say that this determination is based on the output of the exclusive NOR gate 223 S in Figure 4 which receives as inputs the carry output of adder 221 and the sign from BR register 219. As will become apparentr if the test in decision block 70 is positive, the process loops back to block 52 to read the branched to instruction out of the IPB array. If the target instruction is not in the IPB array, that is the test in decision block 70 is negative, then the process loops back to decision block 44 to test for the condition of the IPB array being empty.
However, prior to the test in decision block 70, if the test in decision block 78 was negative, then the process loops back to block 42 to re-initialize the registers 201 and 205 and the latch 213.
Turning now to the flow chart of Figure 9, the process of the test made in decision block 70 is shown in more detail. The process is entered at block 72 and immediately branches. On the one hand, the MC input to multiplexer 217 is set to 2 (i.e., binary 10) to select the BR register 219 as an input to the adder 215. Adder 215 then computes the target address by adding the contents of the registers 205 and 219. Whether the result is stored in register 205 is determined by the parallel processing on the left side of the flow chart. In this parallel processing, the l's complement of the contents of register 205 is added to the contents of register 201 in adder 203 to produce an output as indicated in block 78.
The carry output is then tested in decision block ~2;~
80 to determine if there was a rollover. If not, the zero cross latch Z13 is tested in decision block 82 to determine if there was a recent branch. If there was, then the contents of the IEP register 205 are taken as P, as indicated in block 86. On the other hand, if the results of either of the tests in decision blocks 80 or 82 are negative or the result of the test in decision block 84 is positive, then in block 88, the sum output of adder 203 is complemented in complementing logic 211 and this is taken as P.
It will be recalled that P is intrepreted in this way to distinguish between the two cases of a branch the first being that a branch out of the region occurred recently and the second being that the IFP has rolled over zero at least once since the last branch out of the region. In either case, the most significant bit (MSB) of P is set to the sign in the BR register 219, as indicated in block 90 and as shown in Figure 4. Adder 221 then sums P and BR in block 92. The sign of sR is checked in decision block 94 to determine if it is zero. If it is not, the carry output of adder 221 is tested to determine if it is a one in block 95, but if it is, the carry output of adder 221 is tested to determine if it is a zero in block 96.
A negative result from either of decision blocks 95 and 96 is the negative output of decision block 70 of Figure 8. Both of these tests are performed by exclusive NOR gate 223 in Figure 4. On the other hand, a positive result from either of the tests made in decision blocks 95 or 96 results in the sum output of adder 215 being stored in IEP
register 205. This is the positive output of the decision block 70 of Figure 8.
28;~
While the invention has been particularly shown and described with respect to a preferred embodiment thereof, it will be understood by those skilled in the art that changes may be made in form and detail without departing from the spirit and scope of the invention.
decoder 104. The output of the A array 101 is supplied via a first 16-bit output bus 105 to port A, and the output of the B array 102 is supplied via a second 16-bit output bus 106 to port B.
~Z23~
In order to have the instructions read out of the array 10 aligned, the beginning of the instruction has always to be read from the multiplexer 123 whether the instruction is a 16-bit instruction or a 32-bit instruction. Also, to access a 32-bit instruction from the array, both sides of the array, odd and even, have to be addressed and data brought to ports A and B. But as already described, a 16-bit portion containing the beginning of the instruction may be either in the A array 101 or the B array 102; i.e., 32-bit instructions can start in either array.
Therefore, multiplexers 123 and 12~ are provided to read the instructions out of the array 10 in the proper order. The two 16-bit inputs of each multiplexer 123 and 124 are each connected to the 16-bit F~cts A and B of the array 10 but in the opposite order from each other. The 16-bit outputs of the multiplexers 123 and 124 are 2~ combined to provide a 32-bit instruction output.
In the case of a 16-bit instruction, the high order 16 bits comprise a null field and are simply disregarded.
The subject invention relates to the IPB
control 20 which supplies the write pointer W and the read pointer ~ to the IPB array 10. An implementation of the invention is illustrated in Figure ~. The instruction fetch pointer (IFP) is loaded in a register 2~1, and the adder 203 is used to increment the IFP. The IFP is incremented every time a new instruction i5 written into the IPB array 10. Since the IFP points to 32-bit boundaries, it is incremented by two every time.
The ~our most significant bits or the IFP are used as a write pointer W for the IPB array 10. The Y0984-05~
instruction execution pointer (IEP) is loaded in a register 205. The five bits of the IEP are used as a read pointer R for the IPB array 10. The load signals FL and EL are used to load ne~ values of IFP and IEP into registers 201 and 205, respectively. The clear signals FC and EC are used to clear the registers. These clear signals occur when a branch or jump is out of range requiring the loading and reading of instructions to start from the beginning. The pointers IFP and IEP are diagramatically illustrated in Figure 6 which illustrates the relative positions of the pointers with reference to IPB address space.
Figure 6 is re~erred to in the discussion which follows to illustrate the function of the invention~
Referring now back to Figure 4, in addition to incrementing the IFP register 201, the adder 203 is used for branches to calculate the difference F
between IFP and IEP, which difference represents the number of instructions that have been written into the IPB but have yet to be encountered by the IEP as shown in Figure 6. However, this number is actually for one smaller than the difference because the position pointed to by the IFP does not yet contain the new instructionO This fact facilitates the hardware implementation because l's complement addition ~subtraction) gives the expected result. The implementation requires only that IEP be inverted before it is fed into adder 203. Thus, the IEP contained in register- 205 i5 inverted by inverter 207 and supplied to the other input of adder 203 via the multiplexer 209.
Performing l's complement (inversion) in complementing logic 211 of the output F of adder Y09~4-056 203 gives a value P which represents the number of instructions between the pointers IEP and IFP
which are the instructions accumulated in the IPB
array 10 but not yet overwritten by new instruction fetches. However, as illustrated in Figure 6, the proper interpretation of the value P
requires two separate cases to be distinguished:
lo A branch out of the region occurred recently, or 2. The IFP has rolled over zero at least once since the last branch out of the reglon.
When a branch out of the region is encountered, both IEP and IFP are set to location 0, the target instruction is fetched and decoded, and the zero cross (ZC) latch 213 is reset to zero. Note that IEP and IFP are trying to catch each other as illustrated in Figure 6. In the si`tuation illustrated in Figure 6, it is possible that IFP
reached its current position starting from 0 (after a successful branch out of the region was encountered), case l; or that it has already made a revolution, case 2. In case 1, P=Pl, while in case 2, P=P2.
Referring again to Figure 4, the carry signal Cl from adder 32 indicates two possibilities:
Cl=l; IFP>IEP both case 1 and case 2 are possible, or Cla0; IFP<IEP rollover, or case 2.
Case 1 or case 2 possibilities are resolved by the ZC latch 213. If the latch is set~ then case 2 is encountered; otherwise, it ls case 1. The ZC
latch 213 is set with the carry signal C2 from adder 215 during the increment of the IEP cycle (C~=l during the rollover). More specifically, Y~984-056 the C2 output of adder 215 is supplied to one input of AND gate 214, and the output of AND gate 214 is connected to the set input of latch 213.
AND gate 214 is enabled by the increment select (MC=0,1) so that the latch 213 can only be set during normal instruction sequencing.
Adder 215 is used to calculate the new IEP.
When no branch is encountered, IEP i5 simply incremented ~y the instruction length (1,2). In order to increment the IEP for 1 or 2, the output from the IEP register 205 is fed to the right input of the adder 215, and a 1 or a 2 is fed to the left input of the adder via a multiplexer 217.
The increment, 1 or 2, is selected by the multiplexer select code tMC) being 0 or 1; i.e., binary 00 or 01. When a successful branch is encountered and the target is within the IPB array 10, the target location is ~enerated. In this case, the inputs to adder 215 are the IEP from register 205 and the branch offset sR from register 219, the latter being supplied to the adder 21S via the multiplexer 217 when MC=2 (i.e., binary 10). BR is represented as a 2's complement number and consists of a 5-bit number derived from the five least significant bits (Lss) of the instruction read into instruction register (IRG) 222 from the array 10 plus one sign bit. A third adder 221 is used to indicate whether a branch instruction is contained within the IPB array 1 or not. One input to the adder 221 is BR from register 219 and the other is P from complementing logic 211. Since P = F, the result is as follows:
P - BR + 2 if BR < 0 P ~ BR - F - 1 if BR > 0 Z2~
The indication of the range is given by the carry-out signal C from adder by the following table:
~ . _ .
C3 Br > 0 Br < 0 ._ __ 0 In the range Out of the range _ _ ~ ___ -. 1 Out of the range In the range , _ _- I - ~ I
Carry out C is exclusive ORed with the BR sign bit in exclusive NOR gate 223 so that~the outside range signal IIR equals 1 when the target of the branch instruction is outside of the IPB array 10.
15 To summarize, i~ a successful branch instruction is encountered, the target address is used in conjunction with the output of adder 221 to determine whether the target instruction is contained within the IPB array 10 or not, and the Zo new IEP is calculated (IEP') which points to the target instruction in the IPB. If it is detemined that the instruction is in the IPB array 10, IEP' is loaded into register 205. If not, registers 201 and 205 are set to zero and the ZC latch 213 is also set to 0. These operations are triggered on the branch successful signal being true.
The instruction sequencing logic is better shown in the simplified block diagram of Figure 5.
The IFP in register 201 is incremented by adding 2 to the contents of the register in adder 203. The 2 is supplied to the adder 203 via multiplexer 209~ It will be recalled that the IFP is incremented by 2 because a full word (32-bits) is ~ ~A ~3a2 ~0984-056 written into the array 10. The IEP in register 205 is incremented by adding a 1 or a 2 to the contents of the register in adder 215. The constants 1 and 2 are supplied to adder 215 via multiplexer 217. Incrementing is by 1 if the instruction is a 16-bit instruction and by 2 if the instruction is a 32-bit instruction. On the other hand, for a branch, the offset BR in register 219 is added to the contents of register 205 to produce IEP'. The multiplexer 217 responds to an input that indicates the instruction length as determined by the length determining logic 303 shown in Figure 2 or a branch condition. The IEP
load signal is generated under the following conditions:
1. IEP increment (branch not present or branch not successful), or 2. Branch successful and branch target is in the calculted IPB range.
The function of the ~ogic shown in Figure 7 is to detect whether the Branch Target Instruction is contained in the IPB. There are two types of branches, short and long, which define the locality or range of the target. The offset fields of the Branch Instructions read into instruction register 222 in Figure 4 are shown in Figure 7 as being temporarily stored in registers 223 and 225. Register 223 is a one byte register containing the field JI of the target instruction, while register 22S is a three byte register containing the field BI of the branch instruction.
These fields are sign-extended; e.g., the signs are duplicated to pad the fields to full 32-bit words~ These sign-extended fields are then supplied to a multiplexer 127. The appropriate ., one of these two fields is selected wit:h the multiplexer 127 by the BC input (depencling on whether the instruction is of the short: or long type as determined by the logic 303 in Figure 2), S and this offset is added to the contents of the IAR 305 (see.Figure 2) in adder 129 in order to generate the address of the target instruction.
The 31 most significant bits (MSB), that is all but the least sig~ificant bit, of the multiplexer lO 127 are checked to determine whether this offset is smaller than 64 bytes, in which case there is a possibility that the target instruction might be contained in the IPB. This is done by comparing the sign field with Exclusive OR gates 131 to dete.rmine how long.this field is. If the sign field is at least 26 bits long (leaving six bits including ~he least significant 32nd bit for the value), then the offset of six bits value can not exceed 64 bytes and, therefore, it is possible 2~ that this offset is in the range of -48 to ~16 bytes. Since the least significant bit of the branch offset is always zero (instructions always begin on even byte or half word boundaries), the least significant bit ls not taken into account.
The outputs of the Exclusive OR gates 131 are connected to the inputs of NOR gate 133, and the output of NOR gate 133 indicates whether the looping logic should be disabled. That is, IR = 0 indicates that the Branch Target is certainly out of the range; conversely, I~ = 1 enables the looping logio. This check is performed in parallel with the Branch Address Generation, and if a match is found indicating that the instruction exists in the IPB, the request for the fetch from this memory location is not activatedO
Y098~-056 Instead, the instruction is taken from the IPB.
The implementation shown in Figure 7 is only one of the possible implementations, and those skilled in the art should understand that Figure 7 is illustrative of the principles of operation~
However, actual implementation depends on the .
particular case; i.e., machine structure and orginization.
The flow chart of Figure 8 illustrates the function of the IPB control according to the invention. The process is entered at block 40, and the ~irst operation, as indicated in block 42 is to initialize the IEP, IEP and ZC to zero.
That is, registers 201 and 205 and the latch 213 are all set to zero. Then, in decision block 44, a test is made to determine if the IPB is empty.
This te~t i5 performed by the comparator 208 which compares the IFP and the IEP in registers 201 and 205 to determine if they are equal. Equality is ~ the indication that the IPB is empty. Initially, the IPB is empty because the registers were initialized to zero in block 42. Therefore, a determination is made in decision block 44 that the IPB is empty, and an instruction fetch from ~5 the bus is madeO This is accomplished in block 46 by reading the instruction into the IPB array 10 at address IFP. Then, in block 48, the select signal MF for multiplexer 209 is set to zero, and in block 50, IFP is incremented by 2. The incrementing of IFP by 2 is performed by adder 203. Now, when the test is made in decision block 48 to determine if the IPB is empty, it will be assumed that the test is negative.
~29i~28;~
Y09~4-056 In this case, the instruction in the IP3 array 10 identified by the IEP pointer in register 205 is read to the instruction register IRG 222 as indicated by block 52. The instruction is decoded S in block 54, and then in decision block 56, it is tested to determine if it is a branch or jump instruction. Assuming first that it is not a branch or jump instruction, a determination is made in decision block 58 as to the length of the instruction; i.e., whether it is a 16-bi~ or a 32-bit instruction. This is the function of the instruction length logic 303 in ~igure 2. If the instruction is a 16-bit instruction, then in block 60 the IEP pointer in register 205 is incremented by 1. On the other hand, if the instruction is a 32-bit instruction, the IEP pointer in register 205 is incremented by 2 as indicate3 by block 62.
These functions are accomplished by the adder 215 and the multiplexer 217. In either case, the carry output of adder 215 is checked in decision block 64 to determine if there is a carry. If so, the zero cro~s latch 213 is set as indicated in block 66; otherwise, it is not. The process then loops back, and the steps described are repeated.
Returning now to decision block 56, assume now that the result of the test is positive; that is, the decoded instruction is a branch or jump instruction. In this case, a test is made in decision block 68 to determine if the target instruction is in range. As will be recalled, this is the function of the logic shown in Figure 7. A logic 1 output of the NOR gate 133 indicates a positive result of the test, and this will result in a test being made in decision block 70 to determine if the target is in the IPB array 10. This process is described in more detail with reference to the flow chart in Figure 9, but for now, suffice it to say that this determination is based on the output of the exclusive NOR gate 223 S in Figure 4 which receives as inputs the carry output of adder 221 and the sign from BR register 219. As will become apparentr if the test in decision block 70 is positive, the process loops back to block 52 to read the branched to instruction out of the IPB array. If the target instruction is not in the IPB array, that is the test in decision block 70 is negative, then the process loops back to decision block 44 to test for the condition of the IPB array being empty.
However, prior to the test in decision block 70, if the test in decision block 78 was negative, then the process loops back to block 42 to re-initialize the registers 201 and 205 and the latch 213.
Turning now to the flow chart of Figure 9, the process of the test made in decision block 70 is shown in more detail. The process is entered at block 72 and immediately branches. On the one hand, the MC input to multiplexer 217 is set to 2 (i.e., binary 10) to select the BR register 219 as an input to the adder 215. Adder 215 then computes the target address by adding the contents of the registers 205 and 219. Whether the result is stored in register 205 is determined by the parallel processing on the left side of the flow chart. In this parallel processing, the l's complement of the contents of register 205 is added to the contents of register 201 in adder 203 to produce an output as indicated in block 78.
The carry output is then tested in decision block ~2;~
80 to determine if there was a rollover. If not, the zero cross latch Z13 is tested in decision block 82 to determine if there was a recent branch. If there was, then the contents of the IEP register 205 are taken as P, as indicated in block 86. On the other hand, if the results of either of the tests in decision blocks 80 or 82 are negative or the result of the test in decision block 84 is positive, then in block 88, the sum output of adder 203 is complemented in complementing logic 211 and this is taken as P.
It will be recalled that P is intrepreted in this way to distinguish between the two cases of a branch the first being that a branch out of the region occurred recently and the second being that the IFP has rolled over zero at least once since the last branch out of the region. In either case, the most significant bit (MSB) of P is set to the sign in the BR register 219, as indicated in block 90 and as shown in Figure 4. Adder 221 then sums P and BR in block 92. The sign of sR is checked in decision block 94 to determine if it is zero. If it is not, the carry output of adder 221 is tested to determine if it is a one in block 95, but if it is, the carry output of adder 221 is tested to determine if it is a zero in block 96.
A negative result from either of decision blocks 95 and 96 is the negative output of decision block 70 of Figure 8. Both of these tests are performed by exclusive NOR gate 223 in Figure 4. On the other hand, a positive result from either of the tests made in decision blocks 95 or 96 results in the sum output of adder 215 being stored in IEP
register 205. This is the positive output of the decision block 70 of Figure 8.
28;~
While the invention has been particularly shown and described with respect to a preferred embodiment thereof, it will be understood by those skilled in the art that changes may be made in form and detail without departing from the spirit and scope of the invention.
Claims (9)
1. A method of controlling an instruction prefetch buffer that stores the code for a number of instructions that have already been executed as well as the code for a number of instructions yet to be executed, said method comprising the steps of:
storing an instruction fetch pointer which is supplied to said buffer as a write pointer that points to the location where a new word is to be written into the buffer;
writting instructions into said buffer at the locations pointed to by said instruction fetch pointer;
storing an instruction execution pointer which is supplied to said buffer as a read pointer that points to the location from which an instruction is to be read from the buffer;
reading instructions from locations in the buffer pointed to by said instruction execution buffer;
decoding each instruction read from the buffer to determine if it is a branch or jump instruction;
determining if a target instruction of a branch or jump instruction is within a range of instructions which may be stored in the buffer;
determining for those target instructions found to be within said range if the target instruction is in the buffer; and setting the instruction execution pointer to the location of the target instruction found to be in the buffer.
storing an instruction fetch pointer which is supplied to said buffer as a write pointer that points to the location where a new word is to be written into the buffer;
writting instructions into said buffer at the locations pointed to by said instruction fetch pointer;
storing an instruction execution pointer which is supplied to said buffer as a read pointer that points to the location from which an instruction is to be read from the buffer;
reading instructions from locations in the buffer pointed to by said instruction execution buffer;
decoding each instruction read from the buffer to determine if it is a branch or jump instruction;
determining if a target instruction of a branch or jump instruction is within a range of instructions which may be stored in the buffer;
determining for those target instructions found to be within said range if the target instruction is in the buffer; and setting the instruction execution pointer to the location of the target instruction found to be in the buffer.
2. The method recited in claim 1 wherein the instruction prefetch buffer stores instructions of variable lengths, said method further comprising the step of determining the length of an instruction read out of the buffer and incrementing the instruction execution buffer according to the length of the instruction if the instruction is not a branch or jump instruction.
3. The method recited in claim 2 further comprising the steps of:
setting the instruction fetch pointer and the instruction execution pointer to zero if the instruction read out of the buffer is a branch or jump instruction and the target instruction is not within said range; and reading the next instruction from a main memory storing said instructions.
setting the instruction fetch pointer and the instruction execution pointer to zero if the instruction read out of the buffer is a branch or jump instruction and the target instruction is not within said range; and reading the next instruction from a main memory storing said instructions.
4. The method recited in claim 3 further comprising the steps of:
testing for the condition of equality of the instruction fetch pointer and the instruction execution pointer as an indication that the buffer is empty; and reading the next instruction from said main memory if the buffer is empty.
testing for the condition of equality of the instruction fetch pointer and the instruction execution pointer as an indication that the buffer is empty; and reading the next instruction from said main memory if the buffer is empty.
5. An instruction prefetch buffer control for buffer means that stores the code for a number of instructions that have already been executed as well as the code for a number of instructions yet to be executed, said instruction prefetch buffer control comprising:
first register means for storing an instruction fetch pointer, said instruction fetch pointer being supplied to said buffer means by said first register means as a write pointer which points to the location where a new word is to be written into said buffer means;
second register means for storing an instruction execution pointer, said instruction execution pointer being supplied to said buffer means by said second register means as a read pointer which points to the location in said buffer means where a word is to be read from said buffer means;
first adder means for incrementing said first register means to increment said instruction fetch pointer for sequential instructions, said first adder means calculating a new instruction fetch pointer for branch instructions;
second adder means for incrementing said second register means to increment said instruction execution pointer for sequential instructions, said second adder means calculating a new instruction execution pointer for branch instructions; and third adder means responsive to the output of said first adder means and a branch target address to calculate whether the target instruction is contained in said buffer means and, if it is, causing the new instruction execution pointer calculated by said second adder means to be loaded into said second register means.
first register means for storing an instruction fetch pointer, said instruction fetch pointer being supplied to said buffer means by said first register means as a write pointer which points to the location where a new word is to be written into said buffer means;
second register means for storing an instruction execution pointer, said instruction execution pointer being supplied to said buffer means by said second register means as a read pointer which points to the location in said buffer means where a word is to be read from said buffer means;
first adder means for incrementing said first register means to increment said instruction fetch pointer for sequential instructions, said first adder means calculating a new instruction fetch pointer for branch instructions;
second adder means for incrementing said second register means to increment said instruction execution pointer for sequential instructions, said second adder means calculating a new instruction execution pointer for branch instructions; and third adder means responsive to the output of said first adder means and a branch target address to calculate whether the target instruction is contained in said buffer means and, if it is, causing the new instruction execution pointer calculated by said second adder means to be loaded into said second register means.
6. The instruction prefetch buffer control recited in claim 5 wherein said buffer means stores instructions of variable length, further comprising length means for determining the length of an instruction read out of said buffer means, said second adder means being responsive to said length determining means for incrementing said instruction execution pointer according to the length of the instruction read out of said buffer means.
7. The instruction prefetch buffer control recited in claim 6 further comprising:
decoding means for decoding an instruction read from said buffer means to determine if that instruction is a branch or jump instruction;
first logic means for determining if a target instruction of a branch or jump instruction read from said buffer means is within a range of instructions which may be in said buffer means;
and second logic means for determining if a target instruction found to be within said range is in said buffer means, said second register means being responsive to said second logic means for storing the calculated new instruction execution pointer.
decoding means for decoding an instruction read from said buffer means to determine if that instruction is a branch or jump instruction;
first logic means for determining if a target instruction of a branch or jump instruction read from said buffer means is within a range of instructions which may be in said buffer means;
and second logic means for determining if a target instruction found to be within said range is in said buffer means, said second register means being responsive to said second logic means for storing the calculated new instruction execution pointer.
8. The instruction prefetch buffer control recited in claim 7 further comprising means responsive to said first and second logic means for setting said first and second registers to zero if the target instruction is not within said range or, if within said range, not in said buffer means.
9. The instruction prefetch buffer control recited in claim 8 further comprising comparing means connected to said first and second register means for determining the condition of equality of the instruction fetch pointer and the instruction execution pointer as an indication that said buffer means is empty, said comparing means causing an instruction to be read from a main memory storing instructions.
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
US06/728,724 US4714994A (en) | 1985-04-30 | 1985-04-30 | Instruction prefetch buffer control |
US06/728,724 | 1985-04-30 |
Publications (1)
Publication Number | Publication Date |
---|---|
CA1242282A true CA1242282A (en) | 1988-09-20 |
Family
ID=24928044
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CA000502803A Expired CA1242282A (en) | 1985-04-30 | 1986-02-26 | Instruction prefetch buffer control |
Country Status (4)
Country | Link |
---|---|
US (1) | US4714994A (en) |
EP (1) | EP0199946A3 (en) |
JP (1) | JPS61256446A (en) |
CA (1) | CA1242282A (en) |
Families Citing this family (76)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US4833651A (en) * | 1986-07-24 | 1989-05-23 | National Semiconductor Corporation | High-speed, asynchronous, No-Fall-Through, first-in-first out memory with high data integrity |
JPS6393038A (en) * | 1986-10-07 | 1988-04-23 | Mitsubishi Electric Corp | Computer |
JPS6393041A (en) * | 1986-10-07 | 1988-04-23 | Mitsubishi Electric Corp | Computer |
JPS6398737A (en) * | 1986-10-15 | 1988-04-30 | Mitsubishi Electric Corp | Data processor |
DE3802025C1 (en) * | 1988-01-25 | 1989-07-20 | Otto 7750 Konstanz De Mueller | |
US5522053A (en) * | 1988-02-23 | 1996-05-28 | Mitsubishi Denki Kabushiki Kaisha | Branch target and next instruction address calculation in a pipeline processor |
JP2752076B2 (en) * | 1988-02-23 | 1998-05-18 | 株式会社東芝 | Programmable controller |
EP0389175A3 (en) * | 1989-03-15 | 1992-11-19 | Fujitsu Limited | Data prefetch system |
US5150468A (en) * | 1989-06-30 | 1992-09-22 | Bull Hn Information Systems Inc. | State controlled instruction logic management apparatus included in a pipelined processing unit |
US5440749A (en) * | 1989-08-03 | 1995-08-08 | Nanotronics Corporation | High performance, low cost microprocessor architecture |
DE69030648T2 (en) * | 1990-01-02 | 1997-11-13 | Motorola Inc | Method for sequential prefetching of 1-word, 2-word or 3-word instructions |
US5185868A (en) * | 1990-01-16 | 1993-02-09 | Advanced Micro Devices, Inc. | Apparatus having hierarchically arranged decoders concurrently decoding instructions and shifting instructions not ready for execution to vacant decoders higher in the hierarchy |
AU7305491A (en) * | 1990-01-29 | 1991-08-21 | Teraplex, Inc. | Architecture for minimal instruction set computing system |
US5230068A (en) * | 1990-02-26 | 1993-07-20 | Nexgen Microsystems | Cache memory system for dynamically altering single cache memory line as either branch target entry or pre-fetch instruction queue based upon instruction sequence |
US5454089A (en) * | 1991-04-17 | 1995-09-26 | Intel Corporation | Branch look ahead adder for use in an instruction pipeline sequencer with multiple instruction decoding |
US5539911A (en) * | 1991-07-08 | 1996-07-23 | Seiko Epson Corporation | High-performance, superscalar-based computer system with out-of-order instruction execution |
US5493687A (en) | 1991-07-08 | 1996-02-20 | Seiko Epson Corporation | RISC microprocessor architecture implementing multiple typed register sets |
US5961629A (en) * | 1991-07-08 | 1999-10-05 | Seiko Epson Corporation | High performance, superscalar-based computer system with out-of-order instruction execution |
JP2881049B2 (en) * | 1991-07-30 | 1999-04-12 | 株式会社日立製作所 | Prefetch buffer |
US5345569A (en) * | 1991-09-20 | 1994-09-06 | Advanced Micro Devices, Inc. | Apparatus and method for resolving dependencies among a plurality of instructions within a storage device |
US5434986A (en) * | 1992-01-09 | 1995-07-18 | Unisys Corporation | Interdependency control of pipelined instruction processor using comparing result of two index registers of skip instruction and next sequential instruction |
GB2263565B (en) * | 1992-01-23 | 1995-08-30 | Intel Corp | Microprocessor with apparatus for parallel execution of instructions |
SG45269A1 (en) * | 1992-02-06 | 1998-01-16 | Intel Corp | End bit markers for instruction decode |
EP0636256B1 (en) | 1992-03-31 | 1997-06-04 | Seiko Epson Corporation | Superscalar risc processor instruction scheduling |
US5438668A (en) | 1992-03-31 | 1995-08-01 | Seiko Epson Corporation | System and method for extraction, alignment and decoding of CISC instructions into a nano-instruction bucket for execution by a RISC computer |
US5371864A (en) * | 1992-04-09 | 1994-12-06 | International Business Machines Corporation | Apparatus for concurrent multiple instruction decode in variable length instruction set computer |
EP0638183B1 (en) | 1992-05-01 | 1997-03-05 | Seiko Epson Corporation | A system and method for retiring instructions in a superscalar microprocessor |
KR100248903B1 (en) * | 1992-09-29 | 2000-03-15 | 야스카와 히데아키 | System and method for handling load and/or store operating in a superscalar microprocessor |
US6735685B1 (en) * | 1992-09-29 | 2004-05-11 | Seiko Epson Corporation | System and method for handling load and/or store operations in a superscalar microprocessor |
US5628021A (en) | 1992-12-31 | 1997-05-06 | Seiko Epson Corporation | System and method for assigning tags to control instruction processing in a superscalar processor |
EP0682789B1 (en) | 1992-12-31 | 1998-09-09 | Seiko Epson Corporation | System and method for register renaming |
EP0612013A1 (en) * | 1993-01-21 | 1994-08-24 | Advanced Micro Devices, Inc. | Combination prefetch buffer and instruction cache cross references to related applications |
US5588128A (en) * | 1993-04-02 | 1996-12-24 | Vlsi Technology, Inc. | Dynamic direction look ahead read buffer |
US5544342A (en) * | 1993-06-30 | 1996-08-06 | International Business Machines Corporation | System and method for prefetching information in a processing system |
JP2596712B2 (en) * | 1993-07-01 | 1997-04-02 | インターナショナル・ビジネス・マシーンズ・コーポレイション | System and method for managing execution of instructions, including adjacent branch instructions |
US5574887A (en) * | 1993-09-20 | 1996-11-12 | Apple Computer, Inc. | Apparatus and method for emulation routine pointer prefetch |
US5408622A (en) * | 1993-09-23 | 1995-04-18 | Apple Computer, Inc. | Apparatus and method for emulation routine control transfer via host jump instruction creation and insertion |
JP3290280B2 (en) * | 1994-01-13 | 2002-06-10 | 株式会社東芝 | Information processing device |
US5590351A (en) * | 1994-01-21 | 1996-12-31 | Advanced Micro Devices, Inc. | Superscalar execution unit for sequential instruction pointer updates and segment limit checks |
US5537559A (en) * | 1994-02-08 | 1996-07-16 | Meridian Semiconductor, Inc. | Exception handling circuit and method |
JP3590427B2 (en) * | 1994-08-30 | 2004-11-17 | 株式会社ルネサステクノロジ | Instruction cache memory with read-ahead function |
US5758116A (en) * | 1994-09-30 | 1998-05-26 | Intel Corporation | Instruction length decoder for generating output length indicia to identity boundaries between variable length instructions |
US5875315A (en) * | 1995-06-07 | 1999-02-23 | Advanced Micro Devices, Inc. | Parallel and scalable instruction scanning unit |
US5905881A (en) * | 1995-11-30 | 1999-05-18 | Unisys Corporation | Delayed state writes for an instruction processor |
US5734881A (en) * | 1995-12-15 | 1998-03-31 | Cyrix Corporation | Detecting short branches in a prefetch buffer using target location information in a branch target cache |
US5887174A (en) * | 1996-06-18 | 1999-03-23 | International Business Machines Corporation | System, method, and program product for instruction scheduling in the presence of hardware lookahead accomplished by the rescheduling of idle slots |
US5867699A (en) * | 1996-07-25 | 1999-02-02 | Unisys Corporation | Instruction flow control for an instruction processor |
US5796972A (en) * | 1997-01-14 | 1998-08-18 | Unisys Corporation | Method and apparatus for performing microcode paging during instruction execution in an instruction processor |
US5852727A (en) * | 1997-03-10 | 1998-12-22 | Advanced Micro Devices, Inc. | Instruction scanning unit for locating instructions via parallel scanning of start and end byte information |
US6122729A (en) * | 1997-05-13 | 2000-09-19 | Advanced Micro Devices, Inc. | Prefetch buffer which stores a pointer indicating an initial predecode position |
US6044222A (en) * | 1997-06-23 | 2000-03-28 | International Business Machines Corporation | System, method, and program product for loop instruction scheduling hardware lookahead |
US6189092B1 (en) * | 1997-06-30 | 2001-02-13 | Matsushita Electric Industrial Co., Ltd. | Pipeline processor capable of reducing branch hazards with small-scale circuit |
DE69841256D1 (en) * | 1997-12-17 | 2009-12-10 | Panasonic Corp | Command masking for routing command streams to a processor |
US6237078B1 (en) * | 1998-08-21 | 2001-05-22 | Advanced Micro Devices, Inc. | Code segment default operation determination |
EP0992894A1 (en) * | 1998-10-06 | 2000-04-12 | Texas Instruments Inc. | Apparatus and method for loop execution |
EP1050804A1 (en) | 1999-05-03 | 2000-11-08 | STMicroelectronics SA | Execution of instruction loops |
EP1050802B1 (en) * | 1999-05-03 | 2007-02-07 | STMicroelectronics S.A. | Instruction prefetch unit |
US7085915B1 (en) | 2000-02-29 | 2006-08-01 | International Business Machines Corporation | Programmable prefetching of instructions for a processor executing a non-procedural program |
US6643755B2 (en) * | 2001-02-20 | 2003-11-04 | Koninklijke Philips Electronics N.V. | Cyclically sequential memory prefetch |
CA2467404A1 (en) * | 2001-11-15 | 2003-05-30 | Visto Corporation | System and methods for asychronous synchronization |
US6772179B2 (en) * | 2001-12-28 | 2004-08-03 | Lucent Technologies Inc. | System and method for improving index performance through prefetching |
US7493607B2 (en) | 2002-07-09 | 2009-02-17 | Bluerisc Inc. | Statically speculative compilation and execution |
JP3811140B2 (en) * | 2003-05-12 | 2006-08-16 | 株式会社日立製作所 | Information processing device |
US7680990B2 (en) * | 2003-05-30 | 2010-03-16 | Hewlett-Packard Development Company, L.P. | Superword memory-access instructions for data processor |
US20050114850A1 (en) | 2003-10-29 | 2005-05-26 | Saurabh Chheda | Energy-focused re-compilation of executables and hardware mechanisms based on compiler-architecture interaction and compiler-inserted control |
US7996671B2 (en) | 2003-11-17 | 2011-08-09 | Bluerisc Inc. | Security of program executables and microprocessors based on compiler-architecture interaction |
JP4247132B2 (en) | 2004-01-29 | 2009-04-02 | 株式会社ルネサステクノロジ | Information processing device |
US8607209B2 (en) | 2004-02-04 | 2013-12-10 | Bluerisc Inc. | Energy-focused compiler-assisted branch prediction |
US9304773B2 (en) * | 2006-03-21 | 2016-04-05 | Freescale Semiconductor, Inc. | Data processor having dynamic control of instruction prefetch buffer depth and method therefor |
US20080022080A1 (en) * | 2006-07-20 | 2008-01-24 | Arm Limited | Data access handling in a data processing system |
US20080126766A1 (en) | 2006-11-03 | 2008-05-29 | Saurabh Chheda | Securing microprocessors against information leakage and physical tampering |
US20100115239A1 (en) * | 2008-10-29 | 2010-05-06 | Adapteva Incorporated | Variable instruction width digital signal processor |
JP2013131036A (en) * | 2011-12-21 | 2013-07-04 | Fujitsu Ltd | Arithmetic processing unit and method for controlling arithmetic processing unit |
GB2501299A (en) * | 2012-04-19 | 2013-10-23 | Ibm | Analysing computer program instructions to determine if an instruction can be replaced with a trap or break point. |
US10628163B2 (en) * | 2014-04-17 | 2020-04-21 | Texas Instruments Incorporated | Processor with variable pre-fetch threshold |
GB2572954B (en) * | 2018-04-16 | 2020-12-30 | Advanced Risc Mach Ltd | An apparatus and method for prefetching data items |
Family Cites Families (12)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US3401376A (en) * | 1965-11-26 | 1968-09-10 | Burroughs Corp | Central processor |
US3551895A (en) * | 1968-01-15 | 1970-12-29 | Ibm | Look-ahead branch detection system |
US3764988A (en) * | 1971-03-01 | 1973-10-09 | Hitachi Ltd | Instruction processing device using advanced control system |
US3928857A (en) * | 1973-08-30 | 1975-12-23 | Ibm | Instruction fetch apparatus with combined look-ahead and look-behind capability |
US3967247A (en) * | 1974-11-11 | 1976-06-29 | Sperry Rand Corporation | Storage interface unit |
US4099229A (en) * | 1977-02-14 | 1978-07-04 | The United States Of America As Represented By The Secretary Of The Navy | Variable architecture digital computer |
JPS54100635A (en) * | 1978-01-25 | 1979-08-08 | Nec Corp | Information processor |
US4403284A (en) * | 1980-11-24 | 1983-09-06 | Texas Instruments Incorporated | Microprocessor which detects leading 1 bit of instruction to obtain microcode entry point address |
JPS5835627A (en) * | 1981-08-26 | 1983-03-02 | Toshiba Corp | Controlling system for memory data prefetch |
JPH0743647B2 (en) * | 1982-03-18 | 1995-05-15 | 横河電機株式会社 | Instruction loop catching mechanism in data processor |
DE3305693A1 (en) * | 1983-02-18 | 1984-08-30 | Nixdorf Computer Ag | CIRCUIT FOR THE INTERMEDIATE STORAGE OF COMMAND WORDS |
JPS59223850A (en) * | 1983-06-03 | 1984-12-15 | Fuji Electric Co Ltd | Pre-read control system of instruction |
-
1985
- 1985-04-30 US US06/728,724 patent/US4714994A/en not_active Expired - Fee Related
-
1986
- 1986-02-26 CA CA000502803A patent/CA1242282A/en not_active Expired
- 1986-03-14 EP EP86103437A patent/EP0199946A3/en not_active Ceased
- 1986-03-18 JP JP61058468A patent/JPS61256446A/en active Pending
Also Published As
Publication number | Publication date |
---|---|
EP0199946A2 (en) | 1986-11-05 |
EP0199946A3 (en) | 1989-10-25 |
US4714994A (en) | 1987-12-22 |
JPS61256446A (en) | 1986-11-14 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CA1242282A (en) | Instruction prefetch buffer control | |
US5136696A (en) | High-performance pipelined central processor for predicting the occurrence of executing single-cycle instructions and multicycle instructions | |
US5167026A (en) | Simultaneously or sequentially decoding multiple specifiers of a variable length pipeline instruction based on detection of modified value of specifier registers | |
US4476525A (en) | Pipeline-controlled data processing system capable of performing a plurality of instructions simultaneously | |
US5193167A (en) | Ensuring data integrity by locked-load and conditional-store operations in a multiprocessor system | |
US4827402A (en) | Branch advanced control apparatus for advanced control of a branch instruction in a data processing system | |
US5142631A (en) | System for queuing individual read or write mask and generating respective composite mask for controlling access to general purpose register | |
US5142633A (en) | Preprocessing implied specifiers in a pipelined processor | |
US5560036A (en) | Data processing having incircuit emulation function | |
US4399507A (en) | Instruction address stack in the data memory of an instruction-pipelined processor | |
US5586276A (en) | End bit markers for indicating the end of a variable length instruction to facilitate parallel processing of sequential instructions | |
US4860197A (en) | Branch cache system with instruction boundary determination independent of parcel boundary | |
AU635016B2 (en) | Decoding multiple specifiers in a variable length instruction architecture | |
CA1186801A (en) | Central processing unit for executing instructions of variable length | |
US6463520B1 (en) | Processor for executing instruction codes of two different lengths and device for inputting the instruction codes | |
US4912635A (en) | System for reexecuting branch instruction without fetching by storing target instruction control information | |
KR100404257B1 (en) | Method and apparatus for verifying that instructions are pipelined in correct architectural sequence | |
EP0094535B1 (en) | Pipe-line data processing system | |
US6615339B1 (en) | VLIW processor accepting branching to any instruction in an instruction word set to be executed consecutively | |
US4739470A (en) | Data processing system | |
JP3486690B2 (en) | Pipeline processor | |
US5034880A (en) | Apparatus and method for executing a conditional branch instruction | |
EP0471888B1 (en) | Microprocessor for enhancing initiation of instruction execution after the execution of conditional branch instruction | |
US6757809B1 (en) | Data processor having 2n bits width data bus for context switching functions | |
US4309753A (en) | Apparatus and method for next address generation in a data processing system |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
MKEX | Expiry |