JP4391554B2 - Associative memory - Google Patents

Description

The present invention relates to the associative memory.

  In order to respond to the increase in the number of Internet users and the increase in the construction of corporate LANs in recent years, a relay node router in a network is a TCAM that is a kind of content addressable memory (CAM) in order to improve the processing capability. (Ternary CAM) is frequently used as an address search device. Below, TCAM is demonstrated with reference to FIGS. FIG. 15 is a diagram illustrating a configuration example of a network system in which TCAM as an address search device is used. FIG. 16 is a diagram for explaining a packet classification algorithm by hardware processing. FIG. 17 is a conceptual diagram illustrating the basic configuration and search operation of TCAM. FIG. 18 is a block diagram showing a detailed configuration of TCAM. FIG. 19 is a conceptual diagram showing a configuration of the conventional TCAM cell shown in FIG.

  FIG. 15 shows a system example shown in Non-Patent Document 1. In FIG. 15, the network 1601 is managed by the Internet service provider (ISP1). An internal network (E1) 1603 is connected to the network 1601 via a relay point router 1602, and an internal network (E2) 1605 is connected via a relay point router 1604.

  Further, a network access point (NAP) 1607 is connected to the network 1601 via a relay point router 1606. A network 1608 managed by the Internet service provider (ISP2) and a network 1609 managed by the Internet service provider (ISP3) are connected to the NAP 1607.

  In the above configuration, the relay point routers 1602, 1604, and 1606 perform packet classification by looking up the rules described in the header of the received packet. For example, in the relay router 1606, the packet received from the NAP 1607 may be “This packet should go to the corporate network (E1)” or “This packet is not the corporate network (E1), but the corporate network (E2). Classify them as “hop” or “this packet is nowhere”.

  Such a variety of information is described in the form of rules in the header of the packet, and serves as a search key for classification within the relay router. In the relay point router, it is necessary to look up what action each of the various rules of various packets sent has and look up to which router (hop) or the like.

  There are various methods for packet classification (see Non-Patent Document 1). The packet classification algorithm by hardware processing as shown in FIG. 16 is most frequently used today. For this, special hardware called TCAM (Ternary Content Addressable Memory) is used.

  In FIG. 16, a packet classification algorithm by hardware processing uses a TCAM 1701, a priority control unit (priority encoder) 1702, and an action memory 1703.

  The TCAM 1701 holds a rule for each entry. Today's TCAMs are so expensive that each entry typically holds only a rule name, ie a label. The action to be performed next is stored in an action memory 1703 constituted by a relatively inexpensive DRAM or SRAM.

  In the TCAM 1701, when a rule is given as a destination address from a packet to an input pin, coincidence comparison is started simultaneously in parallel with a rule that internally holds the rule given to the input pin. Then, the entry number (single or plural) that matches or is closest is output. For example, when a rule {rule: R4, R5, R6} is given, entry number # 3 is output.

  When there are a plurality of TCAM 1701 search results, the priority control unit 1702 gives the action memory 1703 the entry with the lowest number as a high priority entry. The action memory 1703 searches for an action to be performed next for the entry input from the priority control unit 1702. As a result, the next router is designated, the packet is forwarded, and so on.

  Here, for example, when the rule is an IP address, an “X” value (don't care) is often used. This means that it is excluded from the search target for matching comparison. In the TCAM 1701, if the stored value is “X”, it is determined that the condition is unconditional. Therefore, the information to be stored in the internal cell of the TCAM 1701 is also three values including “X” in addition to “0” and “1”. This is the origin of the name Ternary type CAM.

  Next, a basic hardware configuration of TCAM and a basic operation of search will be described with reference to FIG. In FIG. 17, TCAM is provided with a search port 1801 in which a data string of search commands (search commands) is arranged. Each entry (entries # 0 to #n) is provided with a plurality of TCAM cells 1802. The plurality of TCAM cells 1802 are connected to the data string of the search port 1801 by a search operation bit line (search line) 1803 in a one-to-one correspondence. Further, a plurality of TCAM cells 1802 provided for each entry (entries # 0 to #n) are connected to a search output line (match line) 1804 in common. As described above, the TCAM cell 1802 stores three values of “0”, “1”, or “X”.

  During a search operation, when a search command is applied to an input pin (not shown), the search command is transferred to the search port 1801, and a data string of the search command is arranged on the search operation bit line 1803 and transmitted to all TCAM cells 1802 of each entry. Is done. As a result, in all entries, comparison of coincidence / mismatch between the data string and the data held in the TCAM cell 1802 is started simultaneously.

  The search output line 1804 is precharged to “H” level before the search operation is started. Normally, the search output line 1804 to which the matched TCAM cell 1802 is connected maintains the “H” level as it is. On the other hand, the search output line 1804 to which the mismatching TCAM cell 1802 is connected is discharged through the transistor in the cell and pulled down to the “L” level. That is, in the TCAM 1701 as the address search device shown in FIG. 16, the entry number of the search output line 1804 holding the “H” level is checked and the entry number is output to the priority control unit 1702. The above is the role of the TCAM 1701.

  Next, the detailed configuration of the TCAM and the outline of the search operation and the like will be described with reference to FIG. As shown in FIG. 18, data input / output pins 1901, instruction control pins 1902, address input pins 1903, and search result output pins 1904 are provided as connection pins to the outside.

  An instruction decoder 1921 is connected to the instruction control pin 1902. The instruction given to the instruction control pin 1902 is decoded by the instruction decoder 1921 and an operation instruction based on the decoding result is notified to each unit.

  A decoder 1931 is connected to the address input pin 1903. A row decoder 1932 is connected to the output terminal of the decoder 1931. The search result output pin 1904 is connected to the output terminal of the search encoder 1941.

  A plurality of TCAM cells 1910 are arranged in a horizontal row between each row output end of the row decoder 1932 and each row input end of the search encoder 1941. Each row output end of the row decoder 1932 is provided with a mask data word line 1933 and an effective data word line 1943. A plurality of TCAM cells 1910 arranged in a horizontal row are connected in parallel to a mask data word line 1933 and a valid data word line 1934, respectively.

  A search output line (match line) 1942 is provided at each row input end of the search encoder 1941, and a plurality of TCAM cells 1910 arranged in a horizontal row are connected in parallel to each search output line 1942. Yes.

  A plurality of sense amplifiers 1911 and a plurality of search drivers 1912 and 1913 are connected to the data input / output pin 1901. The plurality of sense amplifiers 1911 and the plurality of search drivers 1912 and 1913 are provided in one-to-one correspondence on one end side of each column of the plurality of TCAM cells 1910 arranged in a horizontal row.

  Each of the plurality of sense amplifiers 1911 is provided with bit lines 1914 and 1915 for performing write, read, and refresh operations. The two bit lines 1914 and 1915 are connected in parallel to a plurality of TCAM cells 1910 in each column of the plurality of TCAM cells 1910 arranged in a horizontal row.

  Each of the search drivers 1912 and 1913 is provided with search operation bit lines (search lines) 1916 and 1917. A plurality of TCAM cells 1910 in each column of the plurality of TCAM cells 1910 arranged in a horizontal row are connected in parallel to the two search operation bit lines 1916 and 1917.

  In the above configuration, first, when data is written to the TCAM cell 1910, a command to write is given to the command control pin 1902. Then, the data to be written is given to the data input / output pin 1901, and the address and entry number to be written are given to the address input pin 1903.

  Note that it is not necessary to input an address for the write command. What is ultimately needed is the next action corresponding to each rule (input data), so even if the data is entered into the TCAM entry, even a lookup into the TCAM entry and the next action memory This is because it only has to be managed. Therefore, the data is sequentially input to the vacant entries, and to which entry the data is written may be output to the search result output pin 1904 at the time of data writing.

  Next, write data is transmitted to the two bit lines 1914 and 1915 by a write buffer circuit built in the sense amplifier 1911. At the same time, the write address is input to the row decoder 1932 via the decoder 1931, and writing is performed to the TCAM cell 1910 in the desired entry.

  Next, the search operation will also be described. A search command (search) is also applied to the instruction control pin 1902, and a data string to be searched is also applied to the data input / output pin 1901. During a search, the data string is transmitted to search operation bit lines 1916 and 1917 through search drivers 1912 and 1913. At this time, the search output line 1942 is precharged to “H” level in advance before the search operation is started.

  In the TCAM cell 1910, the comparison between the retained data and the search operation bit lines 1916 and 1917 is performed. Then, the search output line 1942 connected only to the entry in which all the data strings match holds the “H” level. If even one bit does not match, in the entry to which the search output line 1942 is connected, the level of the search output line 1942 is discharged through the transistor in the cell and lowered to the “L” level. This is done by wired OR processing.

  In this manner, the search encoder 1941 finds the search output line 1942 holding all the data strings and holding the “H” level, and belongs to the search output line 1942 holding the “H” level. The entry number is output from the search result output pin 1904.

  Next, the conventional TCAM cell 1910 shown in FIG. 18 is configured as shown in FIG. 19, for example. In FIG. 19, a conventional TCAM cell 1910 has three values (ternary values) “1”, “0”, and “X” by physically combining two cells of an effective data storage cell 2001 and a mask data storage cell 2002. Is configured to hold.

  Each of the valid data storage cell 2001 and the mask data storage cell 2002 is configured by two inverters connected in reverse parallel so as to form a flip-flop.

  One storage end of the effective data storage cell 2001 is connected to the source electrode of the MOS transistor 2011 and the gate electrode of the MOS transistor 2012, and the other storage end is connected to the source electrode of the MOS transistor 2013 and the gate electrode of the MOS transistor 2014. It is connected to the. The drain electrode of the MOS transistor 2011 is connected to the bit line 2015. The drain electrode of the MOS transistor 2013 is connected to the bit line 2016. The gate electrodes of the MOS transistors 2011 and 2013 are commonly connected to the effective data word line 2017. The MOS transistors 2012 and 2014 are connected in series, one end is connected to the search operation bit line 2018 and the other end is connected to the search operation bit line 2019.

  One storage end of the mask data storage cell 2002 is connected to the source electrode of the MOS transistor 2021, and the other storage end is connected to the source electrode of the MOS transistor 2022 and the gate electrode of the MOS transistor 2023. The drain electrode of the MOS transistor 2021 is connected to the bit line 2024, and the drain electrode of the MOS transistor 2023 is connected to the bit line 2025. The gate electrodes of the MOS transistors 2021 and 2022 are connected to the mask data word line 2026 in common. The source electrode of the MOS transistor 2023 is grounded, and the drain electrode is connected to the source electrode of the MOS transistor 2027. The gate electrode of the MOS transistor 2027 is connected to the connection end of the MOS transistors 2012 and 2014 connected in series, and the drain electrode is connected to the search output line 2030.

  In this way, the TCAM cell used in the conventional network address search device (LSI) has, for example, 16 MOS transistors to express the three values “1”, “0”, and “X” as shown in FIG. Are used to physically configure two SRAM (Static Random Access Memory) cells and a cell to be searched (matched comparison).

The search operation in this TCAM cell is expressed by the following equation.
If ([val = 1'b0] && [mas = 1'b1]) then WORD <= 1'b0;
If ([val = 1'b1] && [mas = 1'b1]) then WORD <= 1'b1;
If ([val = 1'bx] && [mas = 1'b0]) then WORD <= 1'bx;
That is, if the mask bit (Mask bit) held in the mask data storage cell 2002 is “1”, the valid bit (valid bit) held in the valid data storage cell 2001 and the search operation bit lines 2018 and 2019 Are compared, and the search result is output to the search output line 2030. When the mask bit (Mask bit) held in the mask data storage cell 2002 is “0”, the MOS transistors 2023 and 2027 are not turned on, so that the search output line 2030 is absolutely connected to the ground (GND) level. Never happen.

IEEE Network Mar / Apr2001 Algorithms for Packet Classification

  However, the conventional TCAM cell has the following problems. First, as shown in FIG. 19, the conventional TCAM cell uses, for example, 16 MOS transistors to physically configure two SRAM cells and a cell to be searched (matched comparison). There is a problem that the occupied area per cell is large and a large capacity cannot be mounted on the LSI silicon. The reality is that only 9 Mbit capacity can be secured. This is apparent when compared to the fact that today's SRAM is composed of six MOS transistors, and DRAM (dynamic random access memory) is composed of one MOS transistor and one capacitor element.

  FIG. 20 shows the above-described search operation for easy understanding. The TCAM cell array 2101 shown in FIG. 20 is configured by the total number of rows of a plurality of TCAM cells 1910 arranged in one horizontal row in each row shown in FIG. A search driver 2102 collectively represents the search drivers 1912 and 1913 shown in FIG. The search driver 2102 is provided with a search line SL serving as a search operation bit line for each of the plurality of TCAM cells arranged in each column. In addition, a match line ML serving as a search output line is provided for each of a plurality of TCAM cells arranged in a row in each row.

  In FIG. 20, in the conventional search operation, search data is given to all TCAM cells simultaneously in parallel by a search line SL. As a result, all TCAM cells drive all match lines ML. That is, the basic operations of the memory such as reading and writing are usually performed for a limited part of the entire memory, but unlike the search operation, the vertical direction and the horizontal direction are the object of operation for all TCAM cells. It becomes. For this reason, the power consumption becomes extremely large, for example, it may reach 6 W for a 4 Mbit capacity and 10 W for a 9 Mbit capacity. This point also makes it difficult to install a large capacity. In addition, the large power consumption has a problem that it is difficult to realize a system LSI as a network address search device that speeds up the search operation.

  Further, the conventional TCAM generally includes the TCAM 1701 and the priority control unit (priority encoder) 1702 shown in FIG. 16, and within the TCAM, the higher the priority, the smaller the entry number can be arranged. ing. Therefore, in the conventional TCAM, when a hit occurs in a plurality of entries in the search operation, the one having a high priority, that is, the one having a small entry number is output. Also, for example, IP addresses that no longer accessed by anyone are erased at any time (referred to as aging), and are rearranged in descending order of priority.

  This re-addressing is actually a very high overhead task. This is because data rewriting must be executed for all entries. During this period, the search operation is not accepted. In order to solve these problems, for example, a configuration as shown in FIG. 21 can be considered. FIG. 21 is a block diagram showing a configuration example in the case of solving a problem at the time of address relocation using a conventional TCAM.

  21, TCAMs 2201 and 2202 include the TCAM 1701 and the priority control unit (priority encoder) 1702 shown in FIG. 16 as described above. The TCAM 2201 is connected to a write line (Write) 2204 and a search line (Search) 2205 via an address decoder 2203. Similarly, the TCAM 2202 is connected to a write line (Write) 2204 and a search line (Search) 2205 via an address decoder 2206.

  According to this configuration, the same data is written to each of the TCAMs 2201 and 2202. If priority order rewriting occurs, for example, rewriting is performed only on the TCAM 2201. During this period, the search can be permitted for the TCAM 2202 and the search result is output. Of course, the search is performed only on old data before rewriting. Next, when rewriting to the TCAM 2201 is completed, the search port is moved from the TCAM 2202 to the TCAM 2201 and executed. When rewriting is necessary, rewriting can be performed on the TCAM 2202. The search command can be executed continuously regardless of the rewrite command.

  However, the configuration shown in FIG. 21 is more expensive because it requires a plurality of TCAMs even though they are very expensive hardware. Further, although the search speed has not kept up with the demands of recent network systems, there is a problem that improvement is very difficult because of the large power consumption.

  That is, the search speed that can be realized with today's TCAM hardware is at most about 100 Mpps (100 megapackets per second). However, as a market requirement, for example, in 40GLAN, (40 Gbit / 20 bytes) = 250 MHz, that is, 250 Mpps (250 megabits). Packet per second), and 25 times the current performance is required.

  In addition, the conventional TCAM is generally configured as an LSI including the TCAM 1701 and the priority control unit (priority encoder) 1702 shown in FIG. 16, and is not a system LSI including the action memory 1703. Absent. Therefore, conventionally, a TCMMA LSI including a TCAM 1701 and a priority control unit (priority encoder) 1702 and an action memory LSI including only an action memory 1703 are arranged on a system board, and a search result in the TCMA LSI is encoded. Since it is configured to input it to the action memory LSI, there is also a problem that it is inconvenient in terms of application.

The present invention has been made in view of the above, an object of the present invention to provide an associative memory which take measures that can perform a search operation under low power to permit faster.

To achieve the above object, an associative memory according to the present invention is connected to the data input and output lines Ubi Tsu preparative lines, a memory cell having that Symbol憶node to hold the written data, service over Ji for matching comparison with data held in the search data and the previous Kiki憶node supplied to the line, the content addressable memory cell and a coincidence comparison circuit for changing the logic level of the output line in accordance with the comparison result, a plurality of An associative memory cell array configured by the associative memory cell array , wherein the associative memory cell array includes a first sub-array and a second sub-array that performs a coincidence comparison according to a coincidence comparison result of the first sub-array; the third sub-array to perform coincidence comparison according to the matching comparison result of the second sub-array, the detection result the third according to the output line in said first subarray Characterized in that all or part of the search line belonging to Buarei and a search control means which are not subjected to the search data.

According to the present invention, an effect that an associative memory which allows the search operation to be executed under a low power consumption, and high speed is obtained.

  Exemplary embodiments of an associative memory cell, an associative memory cell array, an address search memory, and a network address search device according to the present invention will be described below in detail with reference to the drawings.

Embodiment 1 FIG.
FIG. 1 is a circuit diagram showing a configuration of a TCAM cell as an associative memory cell according to Embodiment 1 of the present invention. In FIG. 1, a TCAM cell that is an associative memory cell includes four MOS transistors 101, 102, 103, and 104 and two capacitor elements 105 and 106. Among these, the MOS transistor 101 and the capacitor element 105, and the MOS transistor 102 and the capacitor element 106 constitute a dynamic memory cell, and the MOS transistors 103 and 104 constitute a coincidence comparison circuit. The bit line (BL) 111 and the bit line (ZBL) 112 are bit lines that perform a write operation, a read operation, and a refresh operation.

  One bit line (BL) 111 is connected to the drain electrode of the MOS transistor 101 and the source electrode of the MOS transistor 101 is connected to the storage node of the capacitor element 105. Further, the drain electrode of the MOS transistor 102 is connected to the other bit line (ZBL) 112, and the storage node of the capacitor element 106 is connected to the source electrode of the MOS transistor 102. The gate electrodes of the MOS transistors 101 and 102 are connected to a word line (WL) 121 that performs a write operation, a read operation, and a refresh operation, respectively.

  The search line (SL) 131 and the search line (ZSL) 132 whose potential levels change complementarily are bit lines for search operation. A match line (ML) 141 is an output line for outputting a search result. The gate electrode of the MOS transistor 103 is connected to the storage node of the capacitor element 105, and the drain electrode is connected to the search line (SL) 131. The gate electrode of the MOS transistor 104 is connected to the storage node of the capacitor element 106, and the drain electrode is connected to the search line (ZSL) 132. The source electrodes of the MOS transistors 102 and 103 are connected to a match line (ML) 141, respectively.

According to this configuration, ternary values “1”, “0”, and “X” can be held in two dynamic cell memories. That is,
If ([1a = 1'b0] && [1b = 1'b1]) then WORD <= 1'b0;
If ([1a = 1'b1] && [1b = 1'b0]) then WORD <= 1'b1;
If ([1a = 1'b0] && [1b = 1'b0]) then WORD <= 1'bx;
It can be expressed as

  In writing of ternary data, the word line (WL) 121 is set to a high level (hereinafter referred to as “H” level), and the bit line (BL) 111 and the bit line (ZBL) 112 are set to “H” level and low. This is performed by changing to a level (hereinafter referred to as “L” level). For example, “1” is written by setting the bit line (BL) 111 to the “H” level and the bit line (ZBL) 112 to the “L” level. Writing “0” is performed by setting the bit line (BL) 111 to the “L” level and the bit line (ZBL) 112 to the “H” level. “X” is written by setting the bit line (BL) 111 to the “L” level and the bit line (ZBL) 112 to the “L” level.

  The search data consists of “H” level and “L”, and is input from the search line (SL) 131 and the search line (ZSL) 132. The coincidence comparison circuit using the MOS transistors 103 and 104 performs coincidence comparison with the data held in the cell. The comparison result (search result) is output to the match line (ML) 141.

  Here, search line (SL) 131, search line (ZSL) 132, and match line (ML) 141 are each precharged before the start of the search operation. In this embodiment, the precharge level is Vcc. It is not a level but a level of Vcc−Vth that is lowered by the threshold value Vth.

  As a result of the match comparison (search result), if a match is found, the match comparison circuit holds the level of the match line (ML) 141 at the precharge level Vcc−Vth. On the other hand, when there is a mismatch, the match comparator circuit performs an operation of extracting the charge of the match line (ML) 141 and reducing it to the ground (GND) level. At this time, in this embodiment, the match line (ML) 141 is provided with a clamp circuit 151 so that the mismatch match line (ML) 141 does not completely drop to the ground (GND) level.

  The clamp circuit 151 is configured by a MOS transistor Q. That is, the source electrode and the gate electrode of the MOS transistor Q are respectively connected to the power source, and the drain electrode is connected to the match line (ML) 141. Thereby, for example, it stops at the level of (1/2) Vcc.

  Next, the operation of the TCAM cell which is the content addressable memory cell according to the first embodiment will be described with reference to FIGS. FIG. 2 is an explanatory diagram of a read operation and a write operation performed in the TCAM cell shown in FIG. FIG. 3 is an explanatory diagram of a search operation performed in the TCAM cell shown in FIG. FIG. 4 is an operation time chart of the TCAM cell shown in FIG.

  First, a read operation and a write operation will be described with reference to FIG. In FIG. 2, for a M × N-bit TCAM cell array 201, a row decoder 202 that controls the word line WL, a sense amplifier 203 that controls the bit line BL, and a column decoder that gives the sense amplifier 203 an indication of the column to be selected. 204 is provided. An I / O bus 205 is connected to the sense amplifier 203. The bit line BL collectively represents the pair of bit lines (BL, ZBL) shown in FIG.

  In the above configuration, in the read operation, first, row selection is performed by operating one of the M word lines WL by the row decoder 202. By activation of the word line WL, in the TCAM cell connected to the word line WL, the bit line BL is slightly driven to the “H” level or the “L” level by the held data. The sense amplifier 203 amplifies the potential level of the bit line BL in the column selected by the column decoder 204 at high speed. As a result, the data of the pair of bit lines BL in the N pairs of bit lines BL is sent from the sense amplifier 203 to the I / O bus 205. In the write operation, data sent from the I / O bus 205 is written into the TCAM cell array 201 in the same procedure as described above.

  Next, the search operation will be described with reference to FIG. In FIG. 3, four TCAM cells 301, 302, 303, and 304 for two entries having a 2-bit length are shown. The TCAM cells 301 and 302 in the row direction are connected to the word line WL_0 and the match line ML_0 of the first entry. The TCAM cells 303 and 304 in the row direction are connected to the word line WL_1 and the match line ML_1 of the second entry. The column-direction TCAM cells 301 and 303 are connected to the bit lines BL_0 and ZBL_0 and the search lines SL_0 and ZSL_0. Further, the TCAM cells 302 and 304 in the column direction are connected to the bit lines BL_1 and ZBL_1 and the search lines SL_1 and ZSL_1. Although not shown, the match lines ML_0 and ML_1 are provided with the clamp circuit 151 shown in FIG. Here, it is assumed that the upper bits are held in the TCAM cells 301 and 303 and the lower bits are held in the TCAM cells 302 and 304.

  In the search operation, as a preparation, both the search line and the match line are precharged to the “H” level. As described above, the precharge level is the level of Vcc−Vth for both the search line and the match line. This is a measure for reducing power consumption when charging and discharging at “H” level and “L” level are repeated.

  Search is started in this state. That is, search data is given to SL_0, ZSL_0, and SL_1, ZSL_1, respectively. Now, the data SL_0 = “H”, ZSL_0 = “L”, SL_1 = “L”, and ZSL_1 = “H” are given. In the first entry, assuming that “H” and “L” are held in the TCAM cell 301 and “L” and “H” are held in the TCAM cell 302, the comparison results match (in FIG. 3). , Pass and pass). Therefore, match line ML_0 holds precharge level Vcc-Vth.

  On the other hand, in the first entry, assuming that “H” and “L” are held in the TCAM cell 303 and “H” and “L” are held in the TCAM cell 304, the comparison results in the TCAM cell 303. However, in the TCAM cell 304 of the lower bit, they do not match (displayed as “fail”). As a result, the charge precharged in the match line ML_1 is started to be discharged to the search lines SL_1 and ZSL_1 via the match comparison circuit of the TCAM cell 304, and the potential of the match line ML_1 is set to the precharge level Vcc. Start descent from −Vth to ground (GND) level.

  Here, when the charge of the match line ML_1 is completely discharged and the potential becomes the ground (GND) level, the cell determination is the same in the upper bits connected to the match line ML_1, that is, the TCAM cell 303. ), The MOS transistor of the coincidence comparison circuit performing the OFF operation performs the ON operation, and a current path is generated between the match line ML_1 and the search line SL_0 or the search line ZSL_0 via the MOS transistor.

  However, in this embodiment, the match line ML_1 is not brought to the ground (GND) level by the clamp circuit 151 (see FIG. 1) provided on the match line ML_1, and drops to (1/2) Vcc, for example. The level is to be maintained. As a result, in the present example, in the TCAM cell 303, the source potential of the MOS transistor of the coincidence comparison circuit is (1/2) Vcc. Since the operation threshold value of the MOS transistor increases due to the substrate effect of the transistor, the ON operation is not performed and no malfunction occurs.

  FIG. 4 is a time chart collectively showing the above-described write operation, read operation, and search operation. In FIG. 4, (1): each operation is sequentially executed according to the clock CK. (2): As the instruction OP, there are shown a case where four write instructions W are continued and four search (search) instructions S are subsequently continued. (3): The word line WL becomes “H” level and “L” which are MOS levels in response to the write command W. That is, it is the amplitude of the MOS level. (4): At the same time, the bit line BL (ZBL) changes complementarily between the “H” level and the “L” level. This is also the amplitude of the MOS level. (5) The search line SL (ZSL) has a precharge level of Vcc-Vth, and is complementary between the Vcc-Vth level and the ground (GND) level in response to the search (search) instruction S. Change. That is, it changes with an amplitude much smaller than the amplitude of the MOS level. (6): Similarly, the match line ML has a precharge level of Vcc−Vth, and changes between the Vcc−Vth level and the ground (GND) level according to the comparison result. That is, it changes with an amplitude much smaller than the amplitude of the MOS level. (7): The signal is transmitted to the next stage after the level of the match line ML is amplified to the original MOS level.

  As described above, according to the first embodiment, since the TCAM cell is configured by the four MOS transistors and the two capacitor elements, the area occupied by the cell can be reduced. A large capacity cell can be installed. Specifically, with the same design technology as the conventional example (FIG. 18), about four times the capacity can be mounted. For example, in the case of 0.13 μm technology, the conventional example has a capacity of up to about 9 Mbit, but according to the first embodiment, a capacity can be mounted up to about 27 Mbit.

  And since the clamp circuit 151 which is a level maintenance means is provided, the following operations and effects can be obtained. First, since the search line and the match line operate with a low voltage amplitude by limiting the operating voltage value range, it is possible to reduce power consumption during the search operation. In addition, since the match line does not reach the ground potential level when the comparison result does not match, the operation threshold value of the matching MOS transistor can be increased as a result of the comparison, and malfunction can be prevented. .

Embodiment 2. FIG.
FIG. 5 is a circuit diagram showing a configuration of a TCAM cell as an associative memory cell according to the second embodiment of the present invention. In FIG. 5, the same reference numerals are given to components that are the same as or equivalent to those of the TCAM cell shown in FIG. 1. Here, the description will focus on the parts related to the second embodiment.

  As shown in FIG. 5, a TCAM cell, which is an associative memory cell according to the second embodiment, is replaced with a match comparison circuit (MOS transistors 103 and 104) and a match line (ML) in place of the clamp circuit 151 shown in FIG. 141, a current limiting circuit 501 is provided. The rest is the same as the configuration shown in FIG.

  The current limiting circuit 501 includes a MOS transistor 511 provided between the MOS transistor 103 and the match line (ML) 141, and a MOS transistor 512 provided between the MOS transistor 104 and the match line (ML) 141. ing. The gate electrodes of the MOS transistors 511 and 512 are commonly connected to the match line (ML) 141.

  According to this configuration, at the time of the search operation, the match comparison circuit (MOS transistors 103 and 104) determines that there is a mismatch, the charge extraction of the match line (ML) 141 is started, and the potential of the match line (ML) 141 decreases. Since the current flowing through the coincidence comparison circuit (MOS transistors 103 and 104) is limited by the current limiting circuit 501, the potential of the match line (ML) 141 is set to the ground (GND) level as described in the first embodiment. Instead, it can be held at a level of (1/2) Vcc, for example. Therefore, as in the first embodiment, it is possible to obtain the action and effect of reducing the power consumption and preventing the malfunction during the search operation.

Embodiment 3 FIG.
FIG. 6 is a block diagram showing a configuration of a TCAM cell array as an associative memory cell array that realizes a low power consumption type search operation according to Embodiment 3 of the present invention. In the third embodiment, the entire TCAM cell array by the TCAM cell shown in the first embodiment (FIG. 1) or the second embodiment (FIG. 5) is divided into, for example, four subarrays, and the search operation is executed for each subarray. However, a configuration example is shown in which the low power consumption type search operation is realized by thinning out the match lines one after another, that is, by thinning out hierarchically.

  In FIG. 6, when the entire TCAM cell array is composed of, for example, 288 bits, it is divided into four subarrays of [287: 216] [215: 144] [143: 72] [71: 0]. In FIG. 6, subarray [287: 216] is referred to as stage 1, sublay [215: 144] as stage 2, subarray [143: 72] as stage 3, and subarray [71: 0] as stage 4. Each stage is connected by a so-called pipeline system.

  In stage 1, the input terminal of the AND gate 611 is connected to the precharge line 601 and the power supply line, and the output terminal of the AND gate 611 is connected to the gate electrode of the MOS transistor 612. The source electrode of the MOS transistor 612 is connected to the power supply, and the drain electrode is connected to the match line ML. As a result, the match line ML is precharged to the level of Vcc-Vth before the search operation starts. The match line ML is connected to the positive phase input terminal (+) of the amplifier 613, and the reference voltage ref is applied to the negative phase input terminal (−) of the amplifier 613.

  In this stage 1, a search operation is performed on the entries of all bit groups [287: 216]. Since the matched match line ML holds the level of Vcc−Vth, the amplifier 613 connected to the match line ML sets the output to the “H” level of the MOS level. On the other hand, a level drop occurs in the mismatched match line ML. When the level drops from the level of Vcc-Vth to, for example, (1/2) Vcc, the amplifier 613 to which the match line ML is connected outputs its output to the MOS level. Set to “L” level.

  In the precharge stage that connects the stage 1 and the next stage 2, a pipeline register 622 that receives the output of the amplifier 613, and an AND gate 623 to which the output terminal of the pipeline register 622 and the precharge line 621 are connected are provided. Is provided. The pipeline register 622 sets the output to “H” level when the output of the corresponding amplifier 613 is “H” level, and sets the output to “L” level when the output of the corresponding amplifier 613 is “L” level. To. Accordingly, only the AND gate 623 corresponding to the pipeline register 622 outputting the “H” level among the plurality of AND gates 623 sets the output to the “H” level.

  In stage 2, a MOS transistor 624 and an amplifier 625 are provided. The gate electrode of the MOS transistor 624 is connected to the output terminal of the AND gate 623, the source electrode is connected to the power supply, and the drain electrode is connected to the match line ML. The match line ML is connected to the positive phase input terminal (+) of the amplifier 625, and the reference voltage ref is applied to the negative phase input terminal (−) of the amplifier 625. The output terminal of the pipeline register 622 is connected to the control signal input terminal of the amplifier 625.

  That is, in this stage 2, only the match line ML related to the AND gate 623 that has received the output of the pipeline register 622 whose output is set to the “H” level among the plurality of pipeline registers 622 is precharged. Only the amplifier 625 to which the match line ML is connected performs an amplification operation and transmits the search result to the next stage 3. Therefore, in the stage 2, the search operation is not executed for the entries of all the bit groups [215: 144], but the search operation is executed only for the match line ML for which the match determination is performed in the preceding stage 1. Will be. At this time, the match line ML that has been subjected to the mismatch determination in the preceding stage 1 maintains the (1/2) Vcc level and is not affected by the potential change of the search lines (SL, ZSL), so that a malfunction may occur. Absent.

  In the precharge stage connected to the stage 2 and the next stage 3, a pipeline register 632 that receives the output of the amplifier 625, and an AND gate 633 that connects the output terminal of the pipeline register 632 and the precharge line 631 are provided. It has been. Although not shown in the drawings after stage 3, the configuration is the same.

  Next, a low power consumption type search operation according to the third embodiment will be described with reference to FIG. In FIG. 7, as in FIG. 6, when the entire TCAM cell array is composed of, for example, 288 bits, four subarrays of [287: 216] [215: 144] [143: 72] [71: 0] The subarray [287: 216] is referred to as stage 1, the subarray [215: 144] is referred to as stage 2, the subarray [143: 72] is referred to as stage 3, and the subarray [71: 0] is referred to as stage 4. Each stage is connected by a so-called pipeline system.

  In FIG. 7, in stage 1, a search operation is performed on all search lines and all match lines for entries in all bit groups [287: 216]. Then, the match line ML for which the match is determined in the stage 1 is detected, and is transmitted to the stage 2 through the precharge stage.

  In stage 2, a search operation is performed on the entries related to the match line determined to be a match in stage 1 among all the bit groups of [215: 144], and the match lines are narrowed down. The performed match line is transmitted to the stage 3.

  Thus, in stage 2, since all the bit groups [215: 144] are not targeted for search, power consumption can be reduced. In stage 3, since the target match line is further narrowed down, power consumption is similarly reduced in stages 3 and 4. In this way, power consumption can be reduced hierarchically.

  Here, the breakdown of the power consumed in the search operation is mainly composed of three components. The first is charge / discharge (a) of the search line. The second is charge / discharge (b) of the match line. The third is amplifier activation (c). The power consumption is shown in the following breakdown.

  Now, assuming that the TCAM cell array is composed of N entries of bits [287: 0], for example, in the search of stage 1, 20 entries hit (match), in stage 2 10 entries hit, in stage 3 Assume that three entries are hit and one entry finally matches the search data 288 bits in stage 4.

  The power consumption in this case is as follows. In stage 1, [N entries × 288 bits] a + [N entries × 288 bits] b + [N entries × 288 bits] c. In stage 2, [N entries × 288 bits] a + [20 entries × 288 bits] b + [20 entries × 288 bits] c. In stage 3, [N entries × 288 bits] a + [10 entries × 288 bits] b + [10 entries × 288 bits] c. In stage 4, [N entries × 288 bits] a + [3 entries × 288 bits] b + [3 entries × 288 bits] c.

  The same breakdown shows the power consumption in the TCAM cell array in the conventional example. Since all entries are targeted in all stages, in stage 1 to stage 4, they are equal to [N entries × 288 bits] a + [N entries × 288 bits] b + [N entries × 288 bits] c. From this, it can be understood that the power consumption is greatly reduced in the third embodiment.

  As described above, according to the third embodiment, in the search operation peculiar to TCAM, the search operation is not applied to the entire memory array. A hierarchical search method is adopted in which the upper N / 2 bits are searched and the next lower N / 2 bits of only the matching entries are searched.

  Therefore, power consumption can be reduced during the search operation. At this time, as the time required for the search becomes longer, the performance deteriorates. Therefore, in order to prevent the performance from being deteriorated, the operations between the bits are connected by so-called pipelines to ensure the throughput rate.

Embodiment 4 FIG.
FIG. 8 is a block diagram showing a configuration of a TCAM cell array as an associative memory cell array that realizes a low power consumption type search operation according to Embodiment 4 of the present invention. In the fourth embodiment, the entire TCAM cell array by the TCAM cell shown in the first embodiment (FIG. 1) or the second embodiment (FIG. 5) is divided into, for example, four subarrays, and the search operation is executed for each subarray. However, a configuration example is shown that realizes a low power consumption type search operation by thinning out search lines in addition to match lines.

  In FIG. 8, as in the third embodiment, when the entire TCAM cell array is composed of, for example, 288 bits, there are four [287: 216] [215: 144] [143: 72] [71: 0] It is divided into subarrays. In FIG. 8, the subarray [287: 216] is referred to as stage 1, the subarray [215: 144] is referred to as stage 2, the subarray [143: 72] is referred to as stage 3, and the subarray [71: 0] is referred to as stage 4. Each stage is connected by a so-called pipeline system.

  In stage 1, as in the third embodiment, the potential state (search result) of match line ML is amplified to the MOS amplitude level by amplifier 801. The stage 2 includes a pipeline register 802 that takes in the output of the amplifier 801, and a MOS transistor 803 in which the output of the pipeline register 802 is input to the gate electrode. The drain electrodes of the plurality of MOS transistors 803 are grounded (GND), respectively, and the source electrodes are commonly wired or connected to the drain electrode of the MOS transistor 804.

  The drain electrode of the MOS transistor 804 is connected to the input terminal of the pipeline register 805 in the stage 3, and the drive circuit 806 of the search line SL is connected in parallel to the output terminal of the pipeline register 805.

  In short, in stage 1, the potential state (search result) of match line ML is amplified, and the search result is transmitted to stage 2. In stage 2, it is determined that the search result is wired-or and there is no match for all entries. Then, the wired-or signal is supplied to the search line drive circuit 806 via the pipeline register 805, and the partial drive / all pause control of the search line SL is performed.

  Next, a low power consumption type search operation according to the fourth embodiment will be described with reference to FIG. In FIG. 9, as in FIG. 8, when the entire TCAM cell array is composed of, for example, 288 bits, four subarrays of [287: 216] [215: 144] [143: 72] [71: 0] The subarray [287: 216] is referred to as stage 1, the subarray [215: 144] is referred to as stage 2, the subarray [143: 72] is referred to as stage 3, and the subarray [71: 0] is referred to as stage 4. Each stage is connected by a so-called pipeline system. Each subarray is provided with search drivers 901, 902, 903, and 904, which are drive circuits for the search lines SL.

  In stage 1, all search lines SL and all match lines ML are subjected to a search operation. In stage 2, all search lines SL and partial match lines ML are subjected to a search operation. Here, in stage 3, it is assumed that there is no match line ML. That is, it is understood that the data string that matches the search data is not already in the subarray when it comes to stage 3.

  In this case, a disable signal (disable) is sent not only to the match line ML in the stage 4 but also to the search driver 904 that is the source of the search line SL, and the operation is stopped. Therefore, according to the fourth embodiment, the same effect as that of the third embodiment can be obtained.

Embodiment 5 FIG.
FIG. 10 is a block diagram showing a configuration of a TCAM cell array as an associative memory cell array that realizes a low power consumption type search operation according to the fifth embodiment of the present invention. In the fifth embodiment, when the entire TCAM cell array by the TCAM cells shown in the first embodiment (FIG. 1) or the second embodiment (FIG. 5) is divided into subarrays, and the search operation is executed for each subarray. A configuration example of matching the timings of the match line and the search line is shown.

  In FIG. 10, when the entire TCAM cell array is composed of, for example, 143 bits, it is divided into sub-arrays of [143: 108] [107: 72] [71:36].

  In the sub-array of [143: 108], the match line ML is precharged by receiving the outputs of the AND gate 1001 and the AND gate 1001 whose input ends are connected to the precharge line and the power supply line. A MOS transistor 1002 is provided, and the potential state (search result) of the match line ML is taken into the pipeline register 1011. On the other hand, for the search line SL, one register 1003 for timing adjustment is provided.

  In the subarray of [107: 72], with respect to the match line ML, the outputs of the AND gate 1012 and the AND gate 1012 whose input ends are connected to the precharge line and the output end of the pipeline register 1011 are received. A MOS transistor 1013 for performing precharge is provided, and the potential state (search result) of the match line ML is taken into the pipeline register 1021. On the other hand, for the search line SL, two registers 1014 and 1015 for timing adjustment are provided.

  In the sub-array of [71:36], regarding the match line ML, the outputs of the AND gate 1022 and the AND gate 1022 whose input terminals are connected to the precharge line and the output terminal of the pipeline register 1021 are received. A MOS transistor 1023 for performing precharge is provided, and the potential state (search result) of the match line ML is similarly taken into a pipeline register at the next stage (not shown). On the other hand, for the search line SL, three registers 1024, 1025 and 1026 for timing adjustment are provided.

  According to this configuration, the problem of timing when a plurality of subarrays are connected by a pipeline is solved.

Embodiment 6 FIG.
FIG. 11 is a circuit diagram showing a structure of an address search memory according to the sixth embodiment of the present invention. In FIG. 11, TCAM cell arrays 1101 and 1102 which are associative memory cell arrays arranged in parallel are configured by the TCAM cells shown in the first embodiment (FIG. 1) or the second embodiment (FIG. 5). A drive circuit 1104 and a sense amplifier 1103 are provided between the TCAM cell arrays 1101 and 1102. The drive circuit 1104 includes a write drive circuit and a search line drive circuit.

  In addition, row decoders 1105 and 1106 and search encoders 1107 and 1108 are provided independently for the TCAM cell arrays 1101 and 1102, and outputs of the search encoders 1107 and 1108 are output to the outside via the selector 1109. .

  FIG. 12 is a circuit diagram showing a relationship between each TCAM cell of two TCAM cell arrays arranged in parallel and a sense amplifier and a drive circuit arranged therebetween. 12, TCAM cells 1201 and 1201 indicate two adjacent TCAM cells in the TCAM cell array 1101 in FIG. Similarly, TCAM cells 1202 and 1202 indicate two adjacent TCAM cells in the TCAM cell array 1102 in FIG.

  Between the TCAM cell 1201 and the TCAM cell 1202, complementary bit lines 1210 and 1211 for performing write, read, and refresh operations and complementary search lines 1221 and 1222 for performing a search operation are provided. It is connected.

  The bit lines 1210 and 1211 are provided with buffer circuits 1213 and 1214 formed of MOS transistors in the middle. The MOS transistors of the buffer circuits 1213 and 1214 can be controlled ON / OFF by control lines 1215 and 1216, respectively. A sense amplifier 1203 and a write drive circuit 1204 are arranged between the buffer circuits 1213 and 1214. The sense amplifier 1203 and the write drive circuit 1204 are connected to bit lines 1210 and 1211, respectively.

  In addition, the search lines 1221 and 1222 are provided with buffer circuits 1223 and 1224 formed of MOS transistors on the way. The MOS transistors of the buffer circuits 1223 and 1224 can be ON / OFF controlled by control lines 1225 and 1226, respectively. A search line driving circuit 1205 is disposed between the buffer circuits 1223 and 1224. The search line driving circuit 1205 is connected to the search lines 1221 and 1222.

  Next, the operation of the address search memory according to the sixth embodiment will be described with reference to FIG. 11 and FIG. In the address search memory according to the sixth embodiment, the following three operation modes are possible. First, the first operation mode is an operation mode in which the TCAM cell arrays 1101 and 1102 are used as one memory space. That is, one of the write command and the search command is given to both the TCAM cell arrays 1101 and 1102. According to this, although writing and searching are limited to one, a low-speed and large-capacity operation mode that can use a large memory space can be realized.

  Referring to FIG. 12, when writing, reading, and refreshing, one of the control lines 1215 and 1216 turns one of the buffer circuits 1213 and 1214 on. At this time, the buffer circuits 1223 and 1224 are all turned off by the control lines 1225 and 1226. As a result, the TCAM cell arrays 1101 and 1102 can be used as one continuous memory area.

  On the other hand, at the time of the search operation, on the contrary, the buffer circuits 1223 and 1224 are both turned ON by the control lines 1225 and 1226. The buffer circuits 1213 and 1214 are all turned off by the control lines 1215 and 1216.

  The second operation mode is an operation mode in which the TCAM cell arrays 1101 and 1102 are used as independent memory spaces. That is, when the same data is written into the TCAM cell arrays 1101 and 1102 and rewriting becomes necessary, the write buffer in the common sense amplifier 1103 provided between the TCAM cell arrays 1101 and 1102 is used to write the array on one side. This is a mode that only executes. For the other array, the search operation can be executed without interruption using the common search line driving circuit 1104 provided between the TCAM cell arrays 1101 and 1102.

  Referring to FIG. 12, when writing, reading, and refreshing, both the buffer circuits 1213 and 1214 are turned on by the control lines 1215 and 1216. At this time, the buffer circuits 1223 and 1224 are all turned off by the control lines 1225 and 1226. As a result, the same data is stored in the TCAM cell arrays 1101 and 1102. The memory area is 1/2.

  When rewriting is necessary, for example, when rewriting is performed on the TCAM cell array 1101, only the buffer circuit 1213 is turned on by the control line 1215. The TCAM cell array 1102 can be searched by turning on the buffer circuit 1224 with the control line 1226.

  As a result, when there is a need for relocation of addresses (re-addressing), the search instruction can be executed continuously regardless of the rewrite instruction. Therefore, since the sense amplifier shared by both arrays can be configured at low cost, it is possible to provide an address search device that can meet the search speed required by the recent network system at low cost.

  The third operation mode is a mode in which the search operation for both arrays is executed at high speed while the same data is written to both arrays in the second operation mode. That is, in FIG. 12, at the “H” edge of the system clock, the buffer circuit 1223 is turned on by the control line 1225 and the buffer circuit 1224 is turned off by the control line 1226 to search the TCAM cell array 1101. At the “L” edge of the system clock, the buffer circuit 1223 is turned off by the control line 1225, the buffer circuit 1224 is turned on by the control line 1226, and the TCAM cell array 1102 is searched.

  According to this, the memory area is halved, but the search operation speed is doubled. For example, when the system clock is running at 125 MHz, a high-speed search operation of 250 MHz is possible as a whole.

  The outputs of the TCAM cell arrays 1101 and 1102 in each of the above operations are independently output as parallel data or serial data from the search encoders 1107 and 1108, and are output to the outside via the selector 1109.

Embodiment 7 FIG.
FIG. 13 is a block diagram showing a configuration of a system LSI that functions as a network address search device according to Embodiment 7 of the present invention.

  As described above, the TCAM cell, which is the associative memory cell according to the first embodiment (FIG. 1), includes four MOS transistors and two capacitor elements, and the associative memory according to the second embodiment (FIG. 5). Since the TCAM cell, which is a cell, is composed of six MOS transistors and two capacitor elements, the occupied area is remarkably reduced and a large capacity can be mounted. Further, according to Embodiments 3 to 5 (FIGS. 6 to 10), hierarchical partial search can be performed, so that power consumption can be significantly reduced.

  In other words, according to the present invention, the degree of integration of TCAM that performs a special search operation can be further improved. Therefore, as shown in FIG. 13, the conventional TCAM is an LSI in which a priority control unit (priority encoder) is included in the TCAM cell array 1401, and the action memory 1403 cannot be included. A system LSI including the action memory 1403 can be obtained.

  As described above, what is actually required from the standpoint of the application is not the entry number indicated by the TCAM but an action for next hopping. Therefore, according to the present invention, it functions as a network address search device with excellent usability. It is possible to configure a system LSI that performs this.

  FIG. 14 is a block diagram showing a specific configuration of a system LSI that functions as the network address search device shown in FIG. As shown in FIG. 14, the match line ML of the TCAM cell array 1501 is amplified by an amplifier 1502 and held in a pipeline register 1503, and the output of the pipeline register 1503 is a word of an action memory 1504 composed of dynamic RAM (DRAM). It is configured to be used directly as the line WL. In this way, the action of the hit entry is automatically amplified by the sense amplifier 1505, and a desired output can be obtained. That is, it is possible to configure a system LSI that functions as an easy-to-use network address search device in terms of application.

  In the embodiment described above, the dynamic memory cell is shown as the associative memory cell. However, the present invention is not limited to the dynamic memory cell. The same applies to the memory cell.

  As described above, the content addressable memory cell according to the present invention is useful for reducing the power consumption during the search operation.

  The content addressable memory cell array according to the present invention is useful for increasing the speed by enabling the search operation to be performed with low power consumption.

  The address search memory according to the present invention is useful for realizing an address search device that can meet the search speed required by a recent network system at low cost.

  The network address search device according to the present invention is useful for realizing a system LSI with excellent usability.

It is a circuit diagram which shows the structure of a TCAM cell as an associative memory cell by Embodiment 1 of this invention. FIG. 3 is an explanatory diagram of a read operation and a write operation performed in the TCAM cell shown in FIG. It is explanatory drawing of the search operation performed in the TCAM cell shown in FIG. 2 is an operation time chart of the TCAM cell shown in FIG. 1. It is a circuit diagram which shows the structure of a TCAM cell as an associative memory cell by Embodiment 2 of this invention. It is a block diagram which shows the structure of a TCAM cell array as an associative memory cell array which implement | achieves the low power consumption type search operation by Embodiment 3 of this invention. FIG. 7 is an explanatory diagram of a low power consumption type search operation performed in the TCAM cell array shown in FIG. 6. It is a block diagram which shows the structure of a TCAM cell array as an associative memory cell array which implement | achieves the low power consumption type search operation by Embodiment 4 of this invention. FIG. 9 is an explanatory diagram of a low power consumption type search operation performed in the TCAM cell array shown in FIG. 8. It is a block diagram which shows the structure of a TCAM cell array as an associative memory cell array which implement | achieves the low power consumption type search operation by Embodiment 5 of this invention. It is a block diagram which shows the structure of the address search memory by Embodiment 6 of this invention. FIG. 12 is a circuit diagram showing a relationship between each TCAM cell of the two TCAM cell arrays arranged in parallel shown in FIG. 11 and a sense amplifier and a drive circuit arranged therebetween. It is a block diagram which shows the structure of the system LSI which functions as a network address search device which is Embodiment 7 of this invention. It is a block diagram which shows the specific structure of the system LSI which functions as a network address search apparatus shown in FIG. It is a figure which shows the structural example of the network system in which TCAM as an address search apparatus is used. It is a figure explaining the packet classification algorithm by hardware processing. It is a conceptual diagram explaining the basic composition and search operation | movement of TCAM. It is a block diagram which shows the detailed structure of TCAM. It is a conceptual diagram which shows the structure of the conventional TCAM cell shown in FIG. It is explanatory drawing of the search operation performed by the conventional TCAM shown in FIG. It is a block diagram which shows the structural example in the case of solving the problem at the time of address rearrangement using the conventional TCAM.

Explanation of symbols

101, 102, 103, 104 MOS transistor 105, 106 Capacitor element 151 Clamp circuit 201, 1101, 1102, 1401, 1501 Content addressable memory cell array (TCAM cell array)
203, 1103, 1203 Sense amplifier 501 Current limit circuit 1104 Drive circuit (write drive circuit, search line drive circuit)
1107, 1108 Search encoder 1201, 1202 TCAM cell (associative memory cell)
1204 Write drive circuit 1205 Search line drive circuit 1402 Priority control unit (priority encoder)
1403, 1504 Action memory

Claims (4)

Is connected to data input and output lines Ubi Tsu preparative lines, a memory cell having that Symbol憶node to hold the written data,
For matching comparison with data held in the search data and the previous Kiki憶node supplied to the service over switch line, a content addressable memory cell and a coincidence comparison circuit for changing the logic level of the output line in accordance with the comparison result ,
An associative memory cell array configured by a plurality of the associative memory cell arrays ,
The associative memory cell array includes:
A first subarray;
A second subarray that performs a match comparison in response to a match comparison result of the first subarray;
A third subarray that performs a match comparison in response to a match comparison result of the second subarray;
And search control means so as not to give the search data for all or a portion of the search line belonging to the third sub-array according to the detection result of the output lines before Symbol first subarray,
Associative memory, characterized in that it comprises a. The search control unit selects and precharges an output line belonging to the second sub-array according to the match comparison result of the first sub-array, and the third control unit according to the match comparison result of the second sub-array. 2. The associative memory according to claim 1, wherein an output line belonging to the sub-array is selected and precharged. The search control means includes
A plurality of comparison circuits for comparing a potential of each output line belonging to the first sub-array and a reference potential;
Are precharged to a predetermined potential during a match comparison, a precharge line logic level changes in response to the comparison result of the plurality of comparison circuits,
Claim 1 or 2, characterized in that it comprises a control circuit to avoid giving the search data for all or a portion of the search line belonging to the third sub-array in accordance with the logic level of the precharge line associative memory as claimed in. In each of sub-arrays, associative memory according to any one of claims 1-3, characterized in that it comprises a regulating means for adjusting the timing of the output lines before and hexa over switch line.

Images