JPS63115229A - Element shift system - Google Patents

Abstract

PURPOSE:To reduce the amount of hardware by using data, which is overflowed by data shift, as data of the shift result if the shift direction is different from the shift direction indicated with a code. CONSTITUTION:'1' is set to a code 8 indicating the shift direction, and '111111' is set to the shift quantity with 2' complement expression as shown by 9. If '111111' is left shift data, data 7 is left shifted by 63 bits because '111111' indicates that data should be left shifted by 63 bits, and the result is set to a register 11 and overflowed data is set to a register 10. As the result, right shifted data is held in the register 10, and left shifted data is held in the register 11. Consequently, it is unnecessary to convert the shift quantity to an absolute value through it is expressed with 2' complement in accordance with the shift direction, and hardware for this conversion is unnecessary.

Description

[Detailed Description of the Invention] [Summary] A code indicating whether to shift leftward or rightward (for example, "0" to the left, "1" to the right), and 2.
An arithmetic processing unit capable of executing an element shift instruction that shifts vector data element by element based on a shift amount indicated by a base number or a two's complement number (for example, a binary number for a left shift and a two's complement number for a right shift). Conventionally, when the shift amount is indicated by a two's complement number, this is converted to the absolute value of the shift amount before the shift is performed, which has the problem of increasing the amount of hardware and slowing down the processing speed. was there. In order to solve these conventional problems, the present invention eliminates the need to convert the shift amount represented by two's complement into an absolute value, and can always obtain the required shift data by shifting in a constant direction. Discloses the deckle element shift method.

[Industrial Application Field] The present invention relates to shifting vector data, and in particular, shifts vector data element by element to the left or right by instructing the shift direction and shift amount using the same command. Related to control.

[Prior Art] There is an arithmetic processing device that can shift vector data element by element to the left or to the right with a single instruction.

In the shift command in such a case, the shift direction is indicated by the code '0' or 'i'.

For example, normally a code "0" is used for a left shift, and a code II I I+ is used for a right shift. When the sign is “0”, the shift amount is indicated by a positive binary number, but when the sign is “1”, the shift amount is a negative number, that is, 2
It is shown as the complement of .

FIG. 3 is a diagram showing an example of the configuration of an execution system for a conventional element shift instruction for vector data. 51 is an absolute value conversion circuit (abbreviated in the figure) that converts a shift amount expressed as a two's complement number into an absolute value. 52 is a preprocessing circuit (indicated as PRESHiFTER in the figure), 53 is a logic operation circuit (indicated as PRESHiFTER),
In the figure, it is abbreviated as ALU>, and 54 is a shift circuit that shifts left or right in units of 1 bit (B i in the figure).
T S Hi F T E R), 55 is 1
A shift circuit that shifts left in digit units and a post-processing circuit (in the figure, POST
) and 56 each represent a register.

[Problems to be solved by the invention] In the conventional shift instruction execution system as described above,
An absolute value conversion circuit 51 was required to convert the right shift amount expressed as a two's complement number into an absolute value. The right shift is performed by a preprocessing circuit 52 and a shift circuit 54, while the left shift is performed by a shift circuit 54 and a shift circuit 55, which requires hardware for an absolute value conversion circuit and has a slow processing speed. There was a problem.

In view of these conventional problems, the present invention eliminates the need for a circuit to convert the shift amount into an absolute value even when the shift amount is expressed as a two's complement number, and therefore requires less hardware than the conventional method. It is an object of the present invention to provide an arithmetic processing device that can reduce the amount of data and has a high processing speed.

[Means for Solving the Problems] According to the present invention, the above-mentioned object is achieved by using a code indicating a shift direction and a shift indicated by a binary number or a two's complement number depending on the shift direction. In a vector arithmetic processing device having a shift function capable of executing an element shift instruction that shifts vector data element by element depending on the amount, the data is shifted in a constant direction irrespective of the shift direction indicated by the symbol. This is achieved by an element shift method characterized in that when the data is shifted in a direction different from the shift direction indicated by the sign, the overflowed data is used as the shift result data.

[Embodiment] FIG. 1 is a block diagram of an embodiment of the present invention, showing an example of the configuration of a system related to execution of an element shift instruction of vector data. In Figure 1, 1 is a register, 2 is a preprocessing circuit (in the figure, PRE 5HiFT
3 is a logic operation circuit (abbreviated as AL in the figure).
4 is a shift circuit that shifts left or right in 1-bit units (described as BiT 5HiFTER in the figure), 5 is a shift circuit that shifts left in 1-digit units, and data is selected by a code indicating the shift direction. A shift circuit with a selector (in the figure, POST
6 represents a 68-bit latch.

Figure 2 is a diagram showing the principle of element shift, where 7 is the data to be shifted, 8 is a code indicating the direction of shift, 9 is data indicating the amount of shift, 10 and 11 are registers, and the contents of register 10 are shown on the right. Shifted data, the contents of register 11 represent left-shifted data.

The operation will be explained below using FIGS. 1 and 2.

In the present invention, as is clear from FIG. 1, there is no need for a circuit for converting a shift amount expressed as a two's complement into an absolute value, which was indispensable in the prior art. Instead, a shift circuit 5 is constructed by adding a 68-bit latch 6 and a circuit for selecting data according to a code indicating the shift direction to the shift circuit.
has been established.

FIG. 2 shows a case where data 7 is shifted to the right by one bit. That is, the code 8 indicating the shift direction in this case is set to "1" (if it is a left shift, it is set to "0").
), and the shift amount (1 bit) is set to "111111°" as shown at 9 in two's complement representation. ”11
1111'' is considered as left shift data and shifts to the left by 6.
Since it is a 3-bit shift, data 7 is shifted 63 to the left.
The bits are shifted, the result is set in register 11, and the overflow data is set in register 10. As a result, the right-shifted data will be held in register 10 and the left-shifted data will be held in register 11, so it is good to select one of them using the code 8 indicating the shift direction. .

[Effects of the Invention] As explained above, according to the element shift method of the present invention, even if the shift amount is indicated by a two's complement number depending on the shift direction, there is no need to convert it into an absolute value. There is no need to provide any hardware for this purpose, and therefore there are advantages in that the amount of hardware in the system can be reduced and the processing speed can be improved.

[Brief explanation of the drawing]

FIG. 1 is a block diagram of an embodiment of the present invention, FIG. 2 is a diagram showing the principle of element shift, and FIG. 3 is a diagram showing an example of the configuration of an execution system for a conventional vector data element shift instruction. . 1.10.11...Register, 2...
Preprocessing circuit, 3...Logic operation circuit, 4.5...
...Shift circuit, 6...Latch, 7...
...Data, 8... Code indicating shift direction,
9... Data showing the amount of shift Agent Patent attorney Sada Igeta − Results

Claims (1)

[Claims] A vector operation having a shift function capable of executing an element shift instruction that shifts vector data element by element using a sign indicating a shift direction and a shift amount indicated by a binary number or a two's complement depending on the shift direction. In the processing device, the data is shifted in a constant direction regardless of the shift direction indicated by the symbol, and when the shift direction is different from the shift direction indicated by the symbol, the data is shifted to remove overflow data. An element shift method characterized by being used as shift result data.