JPH04216114A - Method and device for optical parallel arithmetic processing - Google Patents

Abstract

PURPOSE:To enable the title device to execute complicated arithmetic operations even when the device has a simple constitution and, in addition, to make the device usable as a semi-general purpose processor by utilizing its fast processing speed and its ability of carrying out various kinds of operations. CONSTITUTION:Input digital information is encoded to a light intensity at a spatial position by means of an encoding section 6-2. An optical mask 6-3 for executing logical operations performs a prescribed logical operation on encoded light and supplies the encoded light to a optically writing type space light modulator 6-4. The modulator 6-4 performs a modulating or non-modulating process on the input light in accordance with a prefixed threshold. When the modulator 6-4 performs the process, operations are performed plural times by means of the encoding section 6-2 and optical mask 6-3 during the writing time to the modulator 6-4 and operated results are inputted to the modulator 6-4 for arithmetic processing. Then the modulator 6-4 is irradiated with the light of a light source 6-5 through a polarization beam splitter 6-6 and an arithmetic output is read out.

Description

[Detailed description of the invention]

0001

[Field of Industrial Application] The present invention relates to an optical parallel processing method and apparatus suitable for optically performing logical operations between multiple pieces of digital pattern information, such as image processing and logical operations between various symbols. .

0002

2. Description of the Related Art FIG. 6 shows a schematic configuration of a conventional optical parallel arithmetic processing device. In the figure, 1-1 is a light source of incident light for processing, 1-2 is an encoding unit that encodes input digital information with light intensity, and 1-3 is an optical mask (computation kernel and 1-4 is a calculation output, and 1-5 is a data path including a latch memory (or delay means) for output. This configuration is commonly used in optical digital computing systems;
For example, OPTICAL PARALLEL ARRA
Y LOGIC SYSTEM (15 MAY 198
6/Vol. 25, No. 10/APPLIED OP
TICS) etc.

According to the above configuration, the light from the light source 1-1 is encoded by the encoder 1-2 according to the input. The input is usually parallel information such as a two-dimensional pattern,
The encoding unit 1-2 simultaneously encodes as many inputs as the operation targets. For example, if a logical operation is to be performed between a pattern A and a pattern B, the encoding unit 1-
2 encodes these A and B patterns simultaneously. Therefore, the encoding unit 1-2 is configured to be able to simultaneously encode the number of inputs to be calculated. Regarding encoding, we will take as an example the case where light hits spatially different positions depending on the logical value of an input (having a logical value made up of a plurality of pixels).

[0004] The optical mask 1-3 is designed to extract from the encoded light an optical path at a position corresponding to a designated logical value, and a logical operation is executed using this mask. In order to program the type of logical operation, this mask may be changed, or the encoding section may be made variable so that light reaches the transparent portion within the mask depending on the logic desired to be executed. In the above-mentioned components 1-1 to 1-4, logical operations between inputs are executed.

The data path 1-5 is a feedback path that returns the output to the input side when necessary, and enables sequential calculation. The data paths 1-5 are provided, for example, to realize 4-input processing in a 2-input processing system. To give an example, when executing A, B + C, D (logical product AND + logical sum OR) expressed in product-sum format, the number of inputs is 4 (A ~
Instead of creating four input systems corresponding to D), operations are performed sequentially by a two-input processing system.

[0006]

[Problems to be Solved by the Invention] However, the above-mentioned conventional optical arithmetic processing device has a problem in that the configuration becomes complicated because it requires data paths 1-5 that include optical memory, latches, etc. . Furthermore, since the calculation process is a feedback system, data is repeatedly fed back, which has the disadvantage that calculation time is long and errors tend to accumulate.

Furthermore, in conventional processing systems, even if the information is spatially continuous (for example, pattern information), it is spatially digitized into pixel units, and then encoded to represent logical operations. I've done it. This is because encoding using space (usually a plane) is extremely useful as encoding, but the processing content is insufficient from the viewpoint of spatial parallelism and amount of information. This problem arises from the fact that space is used redundantly as a result. However, in the conventional logical operation method in which spatial selection is made according to the logical value of each pixel, since the logical values are expanded spatially in the first place, it is difficult to avoid this problem.

The present invention was made in view of the above-mentioned circumstances, and it simplifies the configuration without requiring a data return path.
Another object of the present invention is to provide an optical parallel arithmetic processing method and apparatus that can avoid accumulation of errors, and can also achieve spatial parallelism and enrich the amount of information.

[0009]

[Means for Solving the Problems] In order to solve the above problems, in the invention according to claim 1, coded light is created by coding one-dimensional or two-dimensional digital information using light intensity. A threshold value (Pth) determined by the encoding process, the light intensity and the irradiation time is determined in advance, and the constant determined from the light intensity, irradiation time, wavelength, etc. of one or more light inputs incident within a finite time is determined in advance. A modulation process in which the presence or absence of modulation is determined depending on whether or not the threshold value (Pth) is exceeded, and modulated or unmodulated output light is sent out according to this determination, and an optical mask or the like is applied to the coded light or the coded light. and a supply step of supplying the light after the logical operation has been performed to the modulation process as an optical input.

[0010] Furthermore, in the invention according to claim 2,
It has an encoding means that creates encoded light by encoding one-dimensional or two-dimensional digital information using light intensity, and a threshold value (Pth) determined by the light intensity and irradiation time, and has a threshold value (Pth) determined by the light intensity and irradiation time. The presence or absence of modulation is selected depending on whether a constant determined from the light intensity, irradiation time, wavelength, etc. of one or more optical inputs exceeds the threshold value (Pth), and thereby modulated or unmodulated output light is selected. and an optically written spatial light modulator for transmitting data, and the encoded light is supplied to the optically written spatial light modulator as an optical input.

[0011] The invention according to claim 3 is provided with a spatial light modulator or an optical mask for extracting a predetermined logic operation output from the coded light, and the extracted logic operation output is transferred to the optical writing type space. It is characterized in that it is supplied to an optical modulator. In the invention set forth in claim 4, a first
and a second control means that sets the timing of extraction by the first control means as the light input timing of the optical writing type spatial light modulator.

In the invention as set forth in claim 5, the input surface of the optical writing type spatial light modulator is made of GaAs PI.
It is characterized by having an N structure and using a ferroelectric liquid crystal as the light modulation part.

[0013]

[Operation] The presence or absence of modulation is determined depending on whether or not the incident light exceeds the threshold of the optical writing type spatial light modulator, so the relationship between the threshold and the incident light should be set appropriately. By doing so, the desired logical operation is executed.

[0014]

Embodiments Next, embodiments of the present invention will be described with reference to the drawings. FIG. 1 is a block diagram showing the configuration of a first embodiment of the present invention, and FIGS. 2 and 3 are diagrams for explaining its operation.

In FIG. 1, light from a light source 3-1 is incident on an encoding section 3-2. The input encoding section 3-2 encodes input data from the input section 3-3. Here, we assume something like a spatial light modulator that simply modulates input data (for example, a two-dimensional digital pattern) two-dimensionally depending on the brightness of light. The encoding unit 3-2 and the input unit 3-3 have appropriate control units, and encode (or modulate) the light with specific pattern data at the necessary timing.
do. The light modulated brightly and darkly by the pattern data is
The light is incident on the writing surface of an optical writing type spatial light modulator 3-4 having storage and threshold functions. Spatial light modulator 3-4
has a dedicated drive section 3-5, and writing and reading are performed at predetermined timing. The light source 3-6 is a light source for reading out the spatial light modulator 3-4. In this embodiment, since a reflective readout type spatial light modulator 3-4 is adopted, the light enters from the readout surface opposite to the writing surface via the polarization beam splitter. Since the spatial light modulator 3-4 modulates the polarization of the readout light according to the state of the input light, the output 3-4 of the polarization beam splitter
8 is a bright and dark light pattern output modulated by the spatial light modulator 3-4.

Next, the calculation operation of the embodiment shown in FIG. 1 will be explained. FIG. 2 is a diagram showing an example of the operation, and shows a case where an AND operation and an OR operation are performed between two inputs.

FIG. 2A shows the timing of input light and output light. As shown in this figure, the spatial light modulator 3
The -4 state has write timing and read timing. Normally, this exists periodically, and if necessary, further includes an erasing timing for erasing the entire spatial light modulator 3-4. At this writing timing, the light from the light source 3-1 encoded by the signal from the input section 3-3 enters the encoding section 3-2. This light is in the form of a pulse in terms of time, and the input optical pulse 1 is inputted at the beginning of the writing timing of the spatial light modulator 3-4, and the input optical pulse 2 is inputted next. Then, an output is obtained at the read timing of the spatial light modulator 3-4. Figure 2 (b) - (
D) shows how logical product AND is performed between two inputs. That is, if the input 1 is a circular pattern, the light from the light source 3-1 passes through the encoding unit 3-2 and then changes the brightness of the circle (the shaded part is This pattern is bright and the white part is dark), and this pattern is input to the spatial light modulator 3-4 which is in the writing timing state. Similarly, the right half pattern of input 2 is encoded into brightness and darkness by the encoding unit 3-2 (see (c) in the same figure), and is input to the spatial light modulator 3-4. The spatial light modulator 3-4 has a predetermined threshold characteristic with respect to the total amount of incident light within its write timing time, and when both input lights are incident, the threshold value is exceeded. The intensity and pulse width of the input optical pulse are set so that only one of them exceeds the threshold. As a result, in the spatial light modulator 3-4, only the overlapping portion of the input optical pulses 1 and 2 is modulated, so the logical product AND of the inputs 1 and 2 is performed at the readout timing of the spatial light modulator 3-4. Output ((d) in the same figure)
reference).

[0018] Figures (e) to (g) show cases in which logical OR can be performed between two inputs. In this case, if either one of the two input lights is incident, the spatial light modulator 3
The intensity and pulse width of the input optical pulse are set so as to exceed the threshold value of −4. By setting like this, the logical sum of input 1 and input 2 is output (see figure (G)).
.

FIG. 3 is a diagram illustrating another example of the operation of the embodiment shown in FIG. Here, AND operation between three inputs and O
This shows a case where an R operation is performed.

FIG. 2(A) explains the timing, and is similar to FIG. 2(A) described above. At the illustrated write timing, light from the light source 3-1 encoded by the signal from the input section 3-3 is input to the spatial light modulator 3-4. This light is in the form of a pulse in terms of time, and the input optical pulse 1 is input at the beginning of the writing timing of the spatial light modulator 3-4, then the input optical pulse 2 is input, and then the input optical pulse 3 is input. . Then, an output is obtained at the read timing of the spatial light modulator 3-4.

[0021] Figures (B) to (E) show how logical product AND is performed between three inputs. If the input 1 is a circular pattern, the light from the light source 3-1 is transmitted to the encoder 3-2.
After passing through, the light becomes a circular light and dark pattern as shown in FIG. Similarly, the right half pattern of input 2 is encoded brightly and darkly by the encoding unit 3-2 (see (c) in the same figure), inputted to the spatial light modulator 3-4, and then input to the spatial light modulator 3-4. Half of the pattern is encoded part 3-2
The light is encoded brightly and darkly (see (d) in the same figure) and input to the spatial light modulator 3-4. In this case, the intensity and pulse width of the input optical pulses are set so that the threshold value of the spatial light modulator 3-4 is exceeded when all three input optical pulses are input, but the threshold value of the spatial light modulator 3-4 is not exceeded when only two of them are input. I'll keep it. As a result, in the spatial light modulator 3-4, the input optical pulses 1, 2, 3
Since only the overlapped portion is modulated, the AND of inputs 1, 2, and 3 is output at the read timing.

FIGS. 3(F) to 3(R) show cases in which logical OR is performed. In this case, the intensity and pulse width of the input light pulse are set so that the threshold value of the spatial light modulator 3-4 is exceeded when any one of the three incident lights is incident. As a result, the logical sum of the bright part of input 1, the bright part of input 2, and the bright part of input 3 is output.

Next, a second embodiment will be explained. FIG. 4 is a block diagram showing the configuration of the second embodiment, and FIG. 5 is a diagram explaining the operation of FIG. 4. The configuration shown in FIG. 4 is basically similar to the embodiment shown in FIG. 1, except that an optical mask 6-3 for executing logical operations is provided. Note that this embodiment is an embodiment in which the system includes a two-input logical operation system.

In FIG. 4, light from a light source 6-1 is incident on an encoding section 6-2. In this case, the encoding unit 6-2 encodes two input data simultaneously. Furthermore, we assume encoding in which the position of light is varied depending on the logical value of each pixel of input data (for example, a two-dimensional digital pattern). The encoding unit 6-2 has a suitable control unit and modulates the light with a specific input pattern at a necessary timing.

The light encoded into the two input pattern data enters an optical mask (usually called a kernel) 6-3 for performing logical operations. Note that when a logical operation is programmed by changing the illumination pattern of the light source, the light source pattern is called a kernel. Appropriate spatial filtering is performed in the logical operation execution optical mask 6-3, and a logical operation output is obtained. Up to this point, data memory 1-5 in FIG.
is the same as the system without . The light expressed as bright and dark by executing the logical operation is incident on the writing surface of the optical writing type spatial light modulator 6-4. Spatial light modulator 6-4
has a dedicated drive unit, and writing and reading are performed at predetermined timings. The light source 6-5 is a light source for reading out the spatial light modulator 6-4. Here, since the spatial light modulator 6-4 is of a reflective readout type, the light enters from the readout surface opposite to the writing surface via the polarization beam splitter. Since the spatial light modulator 6-4 modulates the readout light according to the state of the input light, the output 6-7 of the polarizing beam splitter is transmitted to the spatial light modulator 6-4.
A bright and dark light pattern output is modulated by 4.

Next, the operation will be explained. FIG. 5 shows an example in which a product-sum operation and a sum-product operation are performed between four inputs.

FIG. 3A shows input/output timing. The spatial light modulator 6-4 has a write timing and a read timing, which are the same as in the embodiment described above. At the illustrated write timing, the light after passing through the logical operation optical mask 6-3 is incident. This light is temporally pulsed, and the spatial light modulator 6
At the beginning of the write timing of −4, input optical pulse 1 is
Next, an input optical pulse 2 is input, and an output is obtained at the subsequent read timing. Figures (B) to (E) show the sum of products (A AND B) and OR (C
AND D) is shown (A, B,
C and D indicate input).

First, when input 1 (pulse 1) of the circle and input 2 (pulse 1) of the upper half are input (see FIG. 5 (b)), AND is performed by setting the optical mask 6-3 for logical operation. is executed, and pulse 1 (input 1 AND input 2) is obtained (see the upper part of the same figure (d)). This is (A AND B)
corresponds to The same applies to pulse 2, for example,
The logical product of a triangle and a square can be obtained (see the bottom of the same figure (c) and the same figure (d)). This corresponds to (C AND D). Since the result of these logical products is a pattern of light and dark, the rest can be logically summed by the spatial light modulator 6-4 in the same manner as in the previous embodiment, and the final output (see (e) in the same figure) is obtained. .

Next, if the kernel is set to logical sum and the spatial light modulator 6-4 is set to perform logical product, sum product (A OR B) AND (C OR D) will be performed. . Figures (F) and (G) show this process.

Note that as an optical writing type spatial light modulator,
For example, it is preferable that the input surface has a GaAs PIN structure and the light modulation section is made of ferroelectric liquid crystal.

[0031]

As described above, according to the present invention, the configuration can be simplified without requiring a data return path, and accumulation of errors can be avoided. Therefore, even complex processing contents can be executed by a processing device with a simple configuration, and the processing speed can be made sufficient. Moreover, it is possible to achieve spatial parallelism and enrich the amount of information.

[Brief explanation of the drawing]

FIG. 1 is a block diagram showing the configuration of a first embodiment of the present invention.

FIG. 2 is a diagram for explaining the operation of the first embodiment.

FIG. 3 is a diagram for explaining another operation of the first embodiment.

FIG. 4 is a block diagram showing the configuration of a second embodiment of the invention.

FIG. 5 is a diagram for explaining the operation of the second embodiment.

FIG. 6 is a block diagram showing a schematic configuration of a conventional optical parallel processing device.

[Explanation of symbols]

3-2 Encoding section 3-4 Spatial light modulator 6-2 Encoding section 6-3 Optical mask for logical operations 6-4 Spatial light modulator

Claims (5)

[Claims] Claim 1: An encoding process in which one-dimensional or two-dimensional digital information is encoded using light intensity to create coded light, and a threshold value (Pth) determined by the light intensity and irradiation time is predetermined, and a finite The presence or absence of modulation is determined depending on whether or not a constant determined from the light intensity, irradiation time, wavelength, etc. of one or more light inputs incident within the time exceeds the threshold value (Pth), and modulation or non-modulation is determined according to this determination. a modulation process for transmitting modulated output light; and a supply process for supplying the coded light or light after performing a logical operation on the coded light using an optical mask or the like to the modulation process as an optical input. An optical parallel processing method comprising: 2. Encoding means for creating encoded light by encoding one-dimensional or two-dimensional digital information using light intensity, and a threshold value (Pth) determined by the light intensity and irradiation time, and a finite amount of light. The presence or absence of modulation is selected depending on whether a constant determined from the light intensity, irradiation time, wavelength, etc. of one or more light inputs incident within a time exceeds the threshold value (Pth), and this determines whether or not modulation is required. An optical parallel arithmetic processing device comprising: an optical writing type spatial light modulator that sends out modulated output light, and supplying the encoded light to the optical writing type spatial light modulator as an optical input. 3. The optical system comprises a spatial light modulator or an optical mask for extracting a predetermined logic operation output from the encoded light, and supplies the extracted logic operation output to the optical writing type spatial light modulator. 3. The optical parallel processing device according to claim 2. 4. A first control means for controlling the timing of the encoding and the output of the logical operation;
3. The optical parallel arithmetic processing device according to claim 2, further comprising a second control means that sets the take-out timing by the control means as the optical input timing of the optical writing type spatial light modulator. 5. The optical device according to claim 2, wherein the optical writing type spatial light modulator has an input surface having a GaAs PIN structure and a light modulating portion made of ferroelectric liquid crystal. Parallel processing unit.