Entry point mapping and skipping method and apparatus.

BACKGROUND

The present invention relates to methods and apparatus for mapping a macro-instruction to a micro­code instruction memory.

Many computers today execute instructions by storing a number of microcode instructions for each macro-instruction in a microcode memory. The microcode instructions are then executed for the particular macro-instruction to implement the macro-instruction. A typical method is shown in Fig. 1, where a macro-­instruction is placed into an operation code instruc­tion register 10. The instruction in register 10 is then decoded by a mapping circuit 12 to produce an address which is provided to a microcode memory 14. This address will be the address of the first micro-­instruction in a series of micro-instructions to implement the particular macro-instruction. The remaining micro-instructions are obtained by increment­ing the first micro-instruction address 16 by one after that micro-instruction is executed to produce succes­sive micro-instructions 18, 20 and 22 for the macro-­instruction.

In one implementation, each macro-instruction is allocated a fixed number of micro-instructions. If this fixed number is four, for instance, then any macro-instruction which requires more than four micro-­instructions must include a jump in the last micro-­instruction to a separate area of the microcode memo­ry 14. A jump is undesirable because it places addi­tional requirements on the microcode in terms of time or space. To do a jump, a jump address must be spec­ified, and this requires space. In addition, another clock cycle may be required for a jump. To avoid the requirement of a jump, enough space can be allocated to accommodate the largest macro-instruction, but this results in a large number of wasted memory spaces for macro-instructions which require less than the maximum number of micro-instructions. Alternately, the micro­code memory could be divided into groups of different fixed sizes with the sizes varying by a power of two (i.e., groups of four, eight, sixteen, etc.). Thus, a short macro-instruction could be allocated for micro­code addresses 16-22 as shown in Fig. 1, while a longer macro-instruction can be allocated eight microcode memory addresses starting at location 23 as shown in Fig. 1. Numbers of microcode addresses which are not a power of two cannot be accommodated without requiring additional circuitry for mapping circuit 12. A power of two implementation can be done simply by blocking out unwanted bits to result in a higher number.

Map circuit 12 could be implemented to accommodate a particular instruction set so that the number of micro-instruction addresses allocated to a particular macro-instruction varies. This would require that either the instruction set be fixed at the time the map is fixed or that the map circuit itself be programmable.

SUMMARY OF THE INVENTION

The present invention uses a fixed entry-­point map to produce an entry point address of a first micro-instruction for a particular macro-instruction. That address is then incremented by a fixed number, such as 512, to produce the second, third, etc. micro-­instructions for that macro-instruction. After a fixed number of these 512 address skips, such as 6, the addresses are incremented by 1 so that successive micro-instructions are in adjacent address locations.

The determination of whether to increment the address by 512 or by 1 is done by a decode circuit which simply looks at the current address to determine whether it is in the skipping region of the microcode memory or the successive incrementing region. The present invention thus provides flexibility to the micro-instruction addressing with the minimal addition of circuitry. In addition, the method of the present invention allows the entry point map to be fixed, and thus there is no need for it to be programmable.

The macro-instructions are arranged so that any macro-instruction requiring more than six micro-­instructions is followed by one or more macro-­instructions which require six or less micro-­instructions. Thus, when the first macro-instruction starts executing the micro-instructions in the incre­ment by one region of memory, these memory addresses will be available since the next succeeding macro-­instructions will not extend that far.

Alternately, a variable skipping region can be used to give more flexibility. This flexibility enables the skipping regions to be determined when the code is written, thus enabling the optimization of memory space. A variable skipping region implementa­tion requires a slightly more complex method of de­termining whether the address should be an increment of one or a skip increment upon returns from subroutine calls, since it is not simply dependent upon the fixed address.

For a fuller understanding of the nature and advantages of the invention, reference should be made to the ensuing detailed description taken in conjunc­tion with the accompanying drawings.

BRIEF DESCRIPTION OF THE DRAWINGS

<ul id="ul0001" list-style="none"><li>Fig. 1 is a diagram of a microcode memory mapping scheme of the prior art;</li><li>Fig. 2 is a block diagram of a prior art microcode addressing circuit;</li><li>Fig. 3 is a diagram of a microcode mapping scheme according to the present invention; and</li><li>Fig. 4 is a block diagram of a microcode addressing scheme according to the present invention.</li></ul>

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

Fig. 2 shows one type of implementation of microcode address generation according to the prior art which will aid in understanding the present invention. An operation code register 24 stores a macro-­instruction which is decoded by a map circuit 26 to provide a microcode memory address to target regis­ter 28. Map circuit 26 is shown as being a program­mable read-only memory (PROM), but could take other forms. Target register 28 receives the initial address from map circuit 26 through multiplexer 27 and provides it through a multiplexer 30 to the microcode memory 32. The micro-instruction at that address is provided to microcode register 38. The address is then incremented by a +1 incrementer 34 which is then provided to a micro program counter register 36. This incremented address is provided through multiplexer 30 back to memory 32 to provide the next micro-instruction to microcode register 38. This type of incrementing continues until there is a jump or other instruction which is detected by decode logic 40. Typically, a jump is executed by providing a jump address from microcode register 38 to multiplexer 27. A select signal 37 is provided from microcode register 38 to select the microcode jump address from microcode register 38 which is then provided to target regis­ter 28. A control bit or bits are latched into control register 39 which are then decoded by decode logic 40. Decode logic 40 then provides a select signal to multiplexer 30 to select a jump address in target register 28 as opposed to the incrementing address from micro program counter register 36. At the end of a micro-instruction, an instruction dispatch jump is done in which select signal 37 from multiplexer 27 selects the next instruction address from map circuit 26, and a new instruction is then executed.

Fig. 3 is a diagram of a microcode memory mapping scheme according to the present invention. An operation code register 42 is provided to hold a macro-instruction. A fixed map 44 decodes the macro-­instruction to produce an address to a microcode memory 46. The first address provided is an entry point address 48 in an entry block 50. The next micro-instruction is found by incrementing address 48 by a fixed number to produce a second address 52. This address is in turn incremented by the same number to produce a third address 54. These addresses are located in a skipping region 56, which, in one embodi­ment, includes six skips for each macro-instruction (only three are shown in Fig. 3). After the six skips, an address 58 is incremented by one to produce address­es 60 and 62 in succession in an overflow region 64.

This mapping scheme provides flexibility for mapping macro-instructions having different numbers of micro-instructions. For instance, if entry point address 48 is address 212 and the macro-instructions having entry point addresses of 213 and 214 are less than six micro-instructions long, then address lo­cations 60 and 62 will be blank, allowing the macro-­instructions starting at address 48 to have eight micro-instructions. By appropriate placement of the starting address of each macro-instruction, the in­structions can be mapped to make optimum use of the memory by appropriate programming of the microcode while at the same time allowing for a simple fixed map 44 because of the fixed entry point address for each macro-instruction. The use of a fixed entry point allows smaller, faster logic to be used. The logic is high speed because less logic is required than a RAM or PROM. In addition, because the logic is special purpose, it takes up less space than a RAM or PROM, which is especially important when the circuit is put on an integrated circuit.

A block diagram of a circuit to implement the mapping scheme of Fig. 3 is shown in Fig. 4. The block diagram of Fig. 4 follows the basic structure of that of Fig. 2. However, instead of using a +1 incrementer 34 as in Fig. 2, a variable incrementer 66 is used with a memory region decode ct 68. Variable incrementer 66 includes a +1 incrementer 70 and a +512 incrementer 72 as well as a multiplexer 74. In opera­tion, the address provided to microcode RAM 32 is incremented by 512 by incrementer 72 with the new address being passed through multiplexer 74 to micro­code program counter register 36. This process contin­ues until memory region decode circuit 68 detects that the address corresponds to overflow region 64 of Fig. 3. At this point, decode circuit 68 provides a select signal to multiplexer 74 to select the output of +1 incrementer 70. Accordingly, thereafter the address to microcode RAM 32 is incremented by 1 as shown in Fig. 3.

Although variable incrementer 66 is shown as a pair of incrementers and a multiplexer for sake of understanding, it can be implemented with a single counter which allows for multiple carry bit inputs. Region decode circuit 68 can simply select a position of the carry bit to vary the increment. Decode cir­cuit 68 only needs to look at the upper bits of the address to determine whether it is in overflow re­gion 64.

Although a 512 increment which is repeated six times is shown, any other numbers could be used. Fig. 3 actually shows three skips in the skipping region, although any other number could be used.

In an alternate embodiment, the skipping region of Fig. 3 can be made of variable length. This provides additional flexibility when the microcode is programmed. For instance, if a number of instructions requiring only a few address locations each, such as three each, are grouped together, then a block of memory in the section which would normally correspond to the fourth address location for each instruction is freed for other uses.

In one implementation of a variable skipping region, skipping is done until there is a jump or branch in the program. In this case, upon a return from the jump or branch, the skipping incrementer is no longer used and the +1 incrementer is used. As shown in Fig. 4, this extra bit can be stored in a posi­tion 41 shown in phantom attached to micro PC regis­ter 36. A control line from bit position 41 is provid­ed to multiplexer 74. This control line replaces the select line from region decode circuit 68, which would not be used in this implementation. Upon a branch, a bit is stored in position 41 and upon return from branch, this bit selects incrementer 70 by the appro­priate control signal to multiplexer 74. Typically, the branch address is stored in a return stack (not shown). The extra bit is positioned in the return address stack so that when the address is provided to micro PC register 36, this bit is in position 41. The input to bit position 41 is provided by control signal line 37, shown in phantom. The loading of this bit into micro PC register 36 is done by a load control signal from control register 39 (shown in phantom).

The variable skipping embodiment thus allows variation in the number of address locations per instruction while still maintaining the fixed entry point mapping system. This flexibility allows changes to be made to the microcode as it becomes more com­plicated or simpler by adjusting the skipping region. In addition, regions which are not a power of two with respect to each other can be used, since the determina­tion of the region length is done by the positioning of the jump instruction, and not by the addition of circuitry to the mapping circuit.

As will be understood by those familiar with the art, the present invention may be embodied in other specific forms without departing from the spirit or essential characteristics thereof. For example, instead of using a decoding circuit to determine whether the address is in the skipping region or the overflow region, a counter could be used to count the number of skipping increments by incrementer 72. The variable skipping region could also be implemented in different ways, such as by using a bit in the microcode instruction so that the skipping region is not depen­dent upon jumps. Accordingly, the disclosure of the preferred embodiments of the invention is intended to be illustrative, but not limiting, of the scope of the invention which is set forth in the following claims.