Pipelined dual port integrated circuit memory

Abstract

A pipelined dual port integrated circuit memory (20) includes an array (30) of static random access memory (SRAM) cells, where each of the memory cells (80) is connected to a single word line (72) and to a single bit line pair (74, 76). A control circuit (32) controls access to the memory cells, where substantially simultaneous requests for access are serviced sequentially within a single cycle of a clock signal of a data processor that is accessing the memory (20). An address collision detector (110) compares addresses provided to the two ports, and generates a match signal that is used for determining which of the two ports are serviced first, independent of which port is read from, or written to. Because dual port functionality is obtained using a single port SRAM array (30), the memory (20) may be manufactured using relatively less integrated circuit surface area, and therefore at a lower cost.

Description

FIELD OF THE INVENTION

This invention relates in general to memories, and more specifically to dual port memories.

BACKGROUND OF THE INVENTION

Dual port memories are useful for a wide variety of applications. They have special usefulness in the areas of communications and multiprocessor systems. In multiprocessor systems, one processor may write data into the array and the other processor may read data out. In particular, dual port RAMs are especially well suited for a communications application known as Asynchronous Transfer Mode (ATM). In an ATM switch, large amounts of data must be transferred between two processing devices. Another communications application is a standard IEEE 802.3 (commonly known under the trademark "Ethernet" available from Digital Equipment Corporation) communications router. These types of applications have a need for a dual port memory which is inexpensive but includes a large array.

Conventionally, dual port random access memories (RAMs) were constructed using one of two techniques. In the first technique, each memory cell was truly dual port and thus required eight transistors. Because the large dual port memory cells make the array itself quite large, integrated circuit memories based on this technique are expensive. A second technique utilizes standard single port static RAM cells with a partitioned array. If both ports simultaneously attempt to access the same partition, then one of the accesses must be delayed. As the number of partitions increases, the likelihood that a collision will occur decreases, but the cost increases due to the extra decoding and collision detection circuitry. Thus, what is needed is a large dual port RAM which uses conventional single port SRAM cells but which is also inexpensive and fast. These needs are met by the present invention whose features and advantages will be further described with reference to the drawings and the accompanying description.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG. 1 illustrates in block diagram form a pipelined dual port integrated circuit memory according to the present invention.

FIG. 2 illustrates in partial schematic and partial block diagram form a portion of the array of FIG. 1 including a single port static RAM cell.

FIG. 3 illustrates in logic diagram form the address path of FIG. 1.

FIG. 4 illustrates in logic diagram form the I/O data path of FIG. 1.

FIG. 5 illustrates in logic diagram form the clock and control path of FIG. 1.

FIG. 6 illustrates in timing diagram form signals which are relevant to understanding the operation of the memory of FIG. 1 during non-match cycles.

FIG. 7 illustrates in timing diagram form signals which are relevant to understanding the operation of the memory of FIG. 1 during match cycles.

DETAILED DESCRIPTION OF A PREFERRED EMBODIMENT

FIG. 1 illustrates in block diagram form a pipelined dual port integrated circuit memory 20 according to the present invention. Memory 20 includes generally a single port SRAM array 30, a clock and control path 32, an input/output (I/O) data path 34, an address path 36, and a set of input and/or output bonding pads 40. The input and/or output pads are as follows: a clock input pad labelled "CLOCK" 42, a read/write input pad for PORT 1 labelled "R/W.sub.-- PORT 1" 44, a read/write input pad for port 2 labelled "R/W.sub.-- PORT 2" 46, a data input/output pad for PORT 1 labelled "DATA.sub.-- PORT 1" 48, a data input/output pad for port 2 labelled "DATA .sub.-- PORT 2" 50, an output enable input pad for PORT 1 labelled "OE.sub.-- PORT 1" 52, an output enable input pad for port 2 labelled "OE.sub.-- PORT 2" 54, a pass through input pad labelled "PT.sub.-- PORT 1" 56, a pass through input pad for port 2 labelled "PT.sub.-- PORT 2" 58, an address input pad for PORT 1 labelled "ADDRESS.sub.-- PORT 1" 60, and an address input pad for port 2 labelled "ADDRESS.sub.-- PORT 2" 62. As should be apparent, the data and address terminals represent multi-bit terminals, but are just shown as single terminals to facilitate understanding of the present invention.

Clock and control path 32 has a CLOCK input terminal connected to pad 42, a R/W.sub.-- PORT1 input terminal connected to pad 44, a R/W.sub.-- PORT2 input terminal connected to pad 46, an address match input terminal labelled "MATCH", a read/write output terminal labelled "R/W SIGNALS", and a clock control output terminal labelled "CLOCKS". I/O data path 34 has terminals corresponding to bonding pads 48, 50, 52, 54, 56, and 58, an input terminal for receiving signal MATCH, an input terminal for receiving signal CLOCKS, an output terminal labelled "DATA.sub.-- IN", and an input terminal labelled "DATA.sub.-- OUT". Address path 36 has address input terminals corresponding to ADDRESS.sub.-- PORT1 pad 60 and ADDRESS.sub.-- PORT2 pad 62, an output terminal for providing the MATCH signal, an input terminal for receiving the CLOCKS, and an output terminal for providing a signal labelled "ADDRESS.sub.-- RAM". Note that the CLOCKS output by clock and control path 32 are not all received by each of I/O data path 34 and address path 36. The specific CLOCKS that are received by each of these blocks will be described further with reference to FIGS. 3-5 below.

Array 30 is a standard single port static random access memory array having control input terminals for receiving the R/W SIGNALS, a data input terminal connected to the DATA.sub.-- IN output terminal of I/O data path 34, a data output terminal connected to the DATA.sub.-- OUT input terminal of I/O data path 34, and an address input terminal for receiving the ADDRESS.sub.-- RAM output of address path 36. Note that the constitution of the R/W SIGNALS will vary from embodiment to embodiment depending on the specific design of array 30 and thus these signals are designated generically. However, specific signals which are used in memory 20 will be described further with reference to FIG. 5 below.

In operation, memory 20 functions as a full dual port static random access memory (SRAM). Unlike some dual port designs which use the array partitioning technique, however, there are no circumstances in which two accesses cannot occur within a single CLOCK cycle. In addition, memory 20 uses a standard six-transistor SRAM cell and thus avoids the need for special dual ported cells associated with the other dual ported technique.

Memory 20 accomplishes these advantages by generating two accesses to access array 30 for every period of the CLOCK signal. Array 30 is a single port memory core capable of being accessed at twice the speed of the external CLOCK signal. Generally, memory 20 generates requests for access of array 30 during the first portion of a CLOCK cycle based on PORT 1 requests for access and during the second portion of a CLOCK cycle based on PORT 2 accesses. Memory 20 reverses this order in the special case of an attempted write cycle on PORT 1 during a CLOCK cycle in which there is an attempted read to the same address on PORT 2.

The detailed operation of memory 20 will be better understood with reference to FIGS. 2-7. FIG. 2 illustrates in partial schematic and partial block diagram form a portion 70 of array 30 of FIG. 1 including a single port memory cell 80. Memory cell 80 is a static RAM cell accessed by the activation of a signal labelled "WL" conducted on a word line 72 and conducts a differential data signal labelled "BL" and "BL" on a complementary pair of bit lines 74 and 76 , respectively. Memory cell 80 includes N-channel metal oxide semiconductor (MOS) transistors 82 and 84, and inverters 86 and 88. Transistor 82 has a first current electrode connected to bit line 74, a gate connected to word line 72, and a second current electrode. Transistor 84 has a first current electrode connected to bit line 76, a gate connected to word line 72, and a second current electrode. Inverter 86 has an input terminal connected to the second current electrode of transistor 82, and an output terminal connected to the second current electrode of transistor 84. Inverter 88 has an input terminal connected to the output terminal of inverter 86, and an output terminal connected to the input terminal of inverter 86.

In operation, memory cell 80 is a standard single port six-transistor memory cell whose logic state is stored due to the operation of back-to-back inverters 86 and 88. Note that as described herein, the number of transistors in a memory cell includes the access transistors in addition to the transistors that perform the storage. Memory cell 80 is conventionally accessed by the activation of word line 72. When word line 72 is active, transistors 82 and 84 become conductive, coupling the contents of the memory cell as a relatively small differential voltage onto bit lines 74 and 76. This voltage is subsequently sensed and output. During a write cycle, external circuitry provides a relatively large differential voltage between BL and BL to over-write the contents stored in memory cell 80.

Memory cell 80 differs from a standard eight transistor dual ported memory cell. First, it only includes six instead of eight transistors. Second, it only is connected to a single complementary pair of bit lines instead of two separate pairs of bit lines, saving two additional access transistors. Furthermore, there is only a single word line which is able to access memory cell 80, as compared with two word lines for the dual ported memory cell. In addition to saving two transistors, the connection to only a single word line and a single bit line pair also reduces the amount of metal wiring into and out of memory cell 80. These effects allow array 30 to be constructed using relatively cheap conventional SRAM cells. Note that an array constructed using single port memory cells like memory cell 80 will be about 25% of the size of an array based on a corresponding eight-transistor true dual port memory cell.

FIG. 3 illustrates in logic diagram form address path 36 of FIG. 1. Address path 36 includes generally an address port input stage 90, a predecoder portion 100, and an address collision detector 110. Input stage 90 includes generally an exclusive OR gate 92, a D-type flip-flop labelled "DFF" 94, and a multiplexer labelled "MUX" 96. Exclusive OR gate 92 has a first input terminal connected to pad 60 for receiving the ADDRESS.sub.-- PORT1 signal, a second input terminal connected to pad 62 for receiving the ADDRESS.sub.-- PORT2 signal, and an output terminal. Note that input stage 90 is a representative one of a number of input stages that correspond to the number of address bits, and corresponding to each address bit is an input stage like input stage 90 shown in FIG. 3. D flip-flop 94 has a D input terminal connected to pad 62 for receiving the ADDRESS.sub.-- PORT2 signal, a clock input terminal for receiving a signal labelled "P2CLK", and a Q output terminal. MUX 96 has a first input terminal connected to pad 60 for receiving the ADDRESS.sub.-- PORT1 signal, a second input terminal connected to the Q output terminal of D flip-flop 94, a control input terminal for receiving a signal labelled "PIP2MUX", and an output terminal.

Predecoder portion 100 includes an AND gate 102 and a D flip-flop 104. AND gate 102 has a first input terminal connected to the output terminal of MUX 96, a second input terminal, and an output terminal. D flip-flop 104 has a D input terminal connected to the output terminal of AND gate 102, a clock input terminal for receiving a signal labelled "DECCLK", and a Q output terminal for providing a signal labelled "ADDRESS.sub.-- RAM". Note that predecoder portion 100 is representative of a logic function which may be performed during predecoding. Thus, the second input of AND gate 102 is not labelled and will vary based on what predecoding function is actually being performed, which will itself vary from embodiment to embodiment. Note further that AND gate 102 may have two, three, or more inputs or may be a different logic function based on the particular embodiment, and the output of MUX 96 is provided to several other logic gates like AND gate 102.

Address collision detector 110 includes NOR gate 112 and a D flip-flop 114. NOR gate 112 has a first input terminal connected to the output terminal of exclusive OR gate 92, and additional input terminals. Each of these additional input terminals is connected to a corresponding exclusive OR output terminal of other input stages corresponding to each of the other ADDRESS pins. NOR gate 112 also has an output terminal. D flip-flop 114 has a D input terminal connected to the output terminal of NOR gate 112, a clock input terminal for receiving a signal labelled "P3CLK", and a Q output terminal for providing the MATCH signal.

In operation, the overall function of address path 36 will be described with reference to each of the three portions illustrated in FIG. 3. Input stage 90 serves to select a given one of the first or second port's addresses for use in driving the input of array 30 as shown in FIG. 1. During one-half cycle of the CLOCK signal, one of ADDRESS.sub.-- PORT1 and ADDRESS.sub.-- PORT2 will be selected to be input to array 30 in a manner which will be described further below. An additional function provided by input stage 90 is to recognize when corresponding address bits are equal. The output of exclusive OR gate 92 is active when the two address bits are identical. Thus, an active low output represents a coincidence of two corresponding address bits.

As mentioned above, predecoder portion 100 performs a predecoding function for providing inputs into array 30. In other embodiments, the predecoding function may be considered to be part of array 30, but in memory 20 it is considered to be a function of address path 36. Note that during the first half of the CLOCK cycle, the address input on ADDRESS.sub.-- PORT1 will be input to array 30, and D flip-flop 104 is used to register this decoded address. Also note that during the first part of a CLOCK cycle, the first input of MUX 96 is selected and thus the address on PORT 1 flows through and is registered in the register formed by D flip-flop 104. Similarly, during the first portion of the clock cycle, the address conducted on ADDRESS.sub.-- PORT2 is registered, but in this case is registered in the register formed by D flip-flop 94. During the second portion of the CLOCK cycle, the address conducted on PORT 2 and stored in D flip-flop 94 is selected through the second input of MUX 96 and then is registered in D flip-flop 104 in order to drive the predecoded address into array 30.

Unlike input stage 90 and predecoder portion 100, address collision detector 110 is shared between all address input bits. NOR gate 112 detects the condition in which each address bit of both the PORT 1 and PORT 2 addresses coincide and provides an active high signal to indicate this MATCH condition. D flip-flop 114 is used to store this control signal coincident with clock signal P3CLK.

FIG. 4 illustrates in logic diagram form I/O data path 34 of FIG. 1. I/O data path 34 includes generally a self-timed amplifier 120, a D flip-flop 130, a MUX 132, a D flip-flop 134, a MUX 136, D flip-flops 138, 140, 142, 143, and 144, a MUX 146, a D flip-flop 148, and an AND gate 150. Self-timed amplifier 120 includes a sense amplifier 122 and inverters 124 and 126. Sense amplifier 122 has a positive input terminal for receiving a signal labelled "GDL", a negative input terminal for receiving a signal labelled "GDLB", an active-low control input terminal for receiving signal labelled "FAMPB", and an output terminal. Note that here and in the following text a signal having a "B" appended to the end of it represents an active low signal. Signals GDL and GDLB are signals conducted on complementary global data lines in array 30 which represent the contents of a memory cell selected by row and column decoding, and correspond to the DATA.sub.-- OUT signals of FIG. 1. Inverter 124 has an input terminal connected to the output terminal of sense amplifier 122, and an output terminal. Inverter 126 has an input terminal connected to the output terminal of inverter 124, and an output terminal connected to the input terminal of inverter 124.

D flip-flop 130 has a D input terminal connected to the output terminal of sense amplifier 122, a clock input terminal for receiving a signal labelled "DLCLK1", and a Q output terminal. MUX 132 has a first input terminal labelled "0", a second input terminal labelled "1" connected to the Q output terminal of D flip-flop 130, a control input terminal, and an output terminal. D flip-flop 134 has a D input terminal connected to the output terminal of MUX 132, a clock input terminal for receiving a signal labelled "QVCLK", and a Q output terminal connected to pad 48 for providing signal DATA.sub.-- PORT1 thereto.

MUX 136 has a first input terminal labelled "0", a second input terminal labelled "1", a third input terminal labelled "2" connected to the Q output terminal of D flip-flop 130, a fourth input terminal labelled "3" connected to the output terminal of sense amplifier 122, a first control input terminal labelled "4" for receiving a signal labelled "RATFB", a second control input terminal labelled "5", and an output terminal. D flip-flop 138 has a D input terminal connected to the output terminal of MUX 136, a clock input terminal for receiving signal QVCLK, and a Q output terminal connected to pad 50 for providing signal DATA.sub.-- PORT2 thereto.

D flip-flop 140 has a D input terminal connected to pad 58 for receiving signal PT-PORT2, a clock input terminal for receiving signal P3CLK, and a Q output terminal connected to the control input terminal of MUX 132. D filp-flop 142 has a D input terminal connected to pad 56 for receiving signal PT.sub.-- PORT1, a clock input terminal for receiving signal P3CLK, and a Q output terminal connected to the second control input terminal of MUX 136. D filp-flop 143 has a D input terminal connected to the Q output terminal of D filp-flop 134, a clock input terminal for receiving a signal labelled "P2CLK", and a Q output terminal connected to both the first and second input terminals of MUX 136. D flip-flop 144 has a D input terminal connected to pad 50 for receiving signal DATA.sub.-- PORT 2, a clock input terminal for receiving signal P2CLK, and a Q output terminal connected to the first input terminal of MUX 132. MUX 146 has a first input terminal connected to the Q output terminal of D flip-flop 134, a second input terminal connected to the Q output terminal of D flip-flop 144, a control input terminal, and an output terminal. D filp-flop 148 has a D input terminal connected to the output terminal of MUX 146, a clock input terminal for receiving a signal labelled "DINK", and a Q output terminal for providing a signal labelled "EQD". AND gate 150 has a first input terminal for receiving signal P1P2MUX, a second input terminal for receiving signal RAFTB, and an output terminal connected to the control input terminal of MUX 146. Note that signal EQD corresponds to DATA.sub.-- IN of FIG. 1. Signals FAMPB, DLCLK1, QVCLK, P3CLK, PIP2MUX, RATFB, P2CLK, and DINK correspond to CLOCKS from FIG. 1.

In operation, I/O data path 34 is connected to four different signal bonding pads, one each for data from DATA.sub.-- PORT1 and DATA.sub.-- PORT2, and two control signals, PT.sub.-- PORT1 and PT.sub.-- PORT2. These control signals are pass-through signals which control the pass through of data from one port to the other. Thus, when signal PT.sub.-- PORT2 is active at a logic low, then the output of D flip-flop 140 causes MUX 132 to select the first input thereof which represents the DATA.sub.-- PORT2 signal from pad 50 which will be provided through D flip-flop 134 to pad 48 as the DATA.sub.-- PORT1 signal. Conversely, when signal PT.sub.-- PORTI is active at a logic low at the input of D flip-flop 142, the Q output thereof controls MUX 136 to select the first and second inputs of MUX 136 which, through D flip-flop 143, represents the DATA.sub.-- PORTI signal from pad 48, and outputs that bit through D flip-flop 138 to pad 50 as the DATA.sub.-- PORT2 signal.

Now the operation of I/O data path 34 when not operating in passthrough mode will be described. Self-timed amplifier 120 is used to sense output data from array 30 during a read cycle. The output of self-timed amplifier 120 is input to both the D input terminal of D flip-flop 130 and the D input terminal of D flip-flop 138, through the fourth input terminal of MUX 136. Thus, data from memory array 30 is selectively provided to either pad 48 as signal DATA.sub.-- PORT1 or to pad 50 as signal DATA.sub.-- PORT2 during a read cycle depending upon which port is currently accessing the array.

Additional circuitry is used during a write cycle. MUX 146 selects either signal DATA.sub.-- PORT1 from pad 48 or signal DATA.sub.-- PORT2 from pad 50 (which is registered in D flip-flop 144). The selection is based on signal P1P2MUX logically ANDed with signal RATFB. Signal RATFB indicates the match condition and will be described further below. The selected input data then is registered by signal DINK in D flip-flop 148 and driven into array 30 as signal EQD.

FIG. 5 illustrates in logic diagram form clock and control path 32 of FIG. 1. In general, clock and control path 32 includes an I/O clock portion 160, a self-synchronizing port select circuit 170, a master clock circuit 180, a data and address clock circuit 200, a self-timed read clock and control logic circuit 210, and a read/write control circuit 230. I/O clock portion 160 includes delay elements 162, 164, and 166, each of which has an input terminal connected to pad 42 for receiving the CLOCK signal, and an output terminal. Delay element 162 has an output terminal for providing signal P2CLK. Delay element 164 has an output terminal for providing signal P3CLK. Delay element 166 has an output terminal for providing signal QVCLK.

Self-synchronizing port select circuit 170 includes an OR gate 172, an SR flip-flop (SRFF) 174, and a delay element 176. OR gate 172 has a first input terminal, a second input terminal, and an output terminal. SR flip-flop 174 has a set terminal labelled "S", a reset input terminal labelled "R" connected to the output terminal of OR gate 172, and an output terminal labelled "Q" for providing signal P1P2MUX. Delay element 176 has an input terminal connected to the Q output terminal of SR flip-flop 174, and an output terminal connected to the second input terminal of OR gate 172.

Master clock circuit 180 includes a delay element 182, an AND gate 184, an SR flip-flop 186, delay elements 188 and 190, and AND gates 192 and 194. Delay element 182 has an input terminal connected to pad 42 for receiving the CLOCK signal, and a complementary output terminal. AND gate 184 has a first input terminal connected to pad 42 for receiving the CLOCK signal, a second input terminal connected to the complementary output terminal of delay element 182, and an output terminal. SR flip-flop 186 has an S input terminal for receiving signal DLCLK1, an R input terminal connected to the output terminal of AND gate 184, and a Q output terminal. Delay element 188 has an input terminal connected to pad 42 for receiving the CLOCK signal, and a complementary output terminal. Delay element 190 has an input terminal connected to the Q output terminal of SR flip-flop 186, and a complementary output terminal. AND gate 192 has a first input terminal connected to pad 42 for receiving the CLOCK signal, a second input terminal connected to the complementary output terminal of delay element 188, and an output terminal connected to the S input terminal of SR flip-flop 174. AND gate 194 has a first input terminal connected to the Q output terminal of SR flip-flop 186, a second input terminal connected to the complementary output terminal of delay element 190, and an output terminal connected to the first input terminal of OR gate 172.

Data and address clock circuit 200 includes an OR gate 202 and an OR gate 204. OR gate 202 has a first input terminal connected to the output terminal of AND gate 192, a second input terminal connected to the output terminal of AND gate 194, and an output terminal for providing signal DINK. OR gate 204 has a first input terminal connected to the output terminal of AND gate 192, a second input terminal connected to the output terminal of AND gate 194, and an output terminal for providing signal DECCLK.

Self-timed read clocks and control logic circuit 210 includes delay elements 212 and 214, an SR flip-flop 216, an AND gate 218, delay elements 220, 222, and 224, a NOR gate 226, and an OR gate 228. Delay element 212 has an input terminal connected to the output terminal of OR gate 204, and an output terminal. Delay element 214 has an input terminal connected to the output terminal of OR gate 204, and an output terminal. SR flip-flop 216 has an R input terminal connected to the output terminal of delay element 212, an S input terminal, and a Q output terminal. AND gate 218 has a first input terminal connected to the output terminal of delay element 214, a second input terminal connected to the Q output terminal of SR flip-flop 174, and an output terminal. Delay element 220 has an input terminal connected to the output terminal of OR gate 204, and an output terminal. Delay element 222 has an input terminal connected to the Q output terminal of SR flip-flop 216, and a complementary output terminal connected to the S input terminal of SR flip-flop 216. Delay element 224 has an input terminal connected to the output terminal of AND gate 218, and an output terminal for providing signal DLCLK1. NOR gate 226 has a first input terminal connected to the output terminal of OR gate 204, a second input terminal connected to the output terminal of delay element 220, and an output terminal for providing a signal labelled "DIC". OR gate 228 has a first input terminal connected to the Q output terminal of SR flip-flop 216, a second input terminal, and an output terminal for providing a signal labelled "FAMPB".

Read/write control circuit 230 includes D flip-flops 232 and 234, an SR flip-flop 236, an inverter 238, NAND gates 240 and 242, an OR gate 246, a NAND gate 248, NOR gates 250, 252, and 254, an inverter 256, NAND gates 258 and 260, a NOR gate 262, a MUX 264, and a self-timed write clocks circuit 270. D flip-flop 232 has a D input terminal connected to pad 46 for receiving signal R/W.sub.-- PORT2, a clock input terminal for receiving signal P3CLK, and a Q output terminal. D flip-flop 234 has a D input terminal connected to pad 44 for receiving signal R/W.sub.-- PORT1, a clock input terminal for receiving P3CLK, and a Q output terminal. SR flip-flop 236 has an S input terminal connected to the output terminal of AND gate 192, an R input terminal connected to the output terminal of AND gate 194, and a Q output terminal.

Inverter 238 has an input terminal for receiving signal MATCH, and an output terminal. NAND gate 240 has a first input terminal connected to the output terminal of inverter 238, a second input terminal connected to the Q output terminal of D flip-flop 232, and an output terminal. NAND gate 242 has a first input terminal connected to the Q output terminal of D flip-flop 232, a second input terminal connected to the Q output terminal of D flip-flop 234, and an output terminal. OR gate 246 has a first input terminal connected to the output terminal of inverter 238, a second input terminal connected to the Q output terminal D flip-flop 234, and an output terminal for providing signal RATFB. NAND gate 248 has a first input terminal connected to the output terminal of NAND gate 240, a second input terminal connected to the output terminal of NAND gate 242, and an output terminal. NOR gate 250 has a first input terminal, a second input terminal for receiving signal MATCH, a third input terminal connected to the Q output terminal of D filp-flop 234, and an output terminal. NOR gate 252 has a first input connected to the output terminal of NAND gate 248, a second input terminal, and an output terminal. NOR gate 254 has a first input terminal connected to the output terminal of NOR gate 250, a second input terminal connected to the output terminal of NOR gate 252, and an output terminal for providing a signal labelled "CSWEB".

Inverter 256 has an input terminal connected to the Q output terminal of D flip-flop 234, and an output terminal. NAND gate 258 has a first input terminal for receiving signal MATCH, a second input terminal connected to the output terminal of inverter 256, and an output terminal. NAND gate 260 has a first input terminal connected to the output terminal of NAND gate 258, a second input terminal connected to the Q output terminal of D flip-flop 232, and an output terminal. NOR gate 262 has a first input terminal for receiving signal MATCH, a second input terminal connected to the Q output terminal of D flip-flop 232, and an output terminal. MUX 264 has a first input terminal connected to the output terminal of NAND gate 260, a second input terminal connected to the output terminal of NOR gate 262, a control input terminal connected to the Q output terminal of SR flip-flop 236, and an output terminal connected to the second input terminal of OR gate 228 for providing a signal labelled "WEBCSB".

Self timed write clocks circuit 270 includes SR flip-flops 272 and 274, delay elements 276 and 278, and inverters 280 and 282. SR flip-flop 272 has an S input terminal connected to the output terminal of AND gate 192, an R input terminal, and a Q output terminal. SR flip-flop 274 has an S input terminal connected to the output terminal of AND gate 194, an R input terminal, and a Q output terminal. Delay element 276 has an input terminal connected to the Q output terminal of SR flip-flop 272, and an output terminal connected to the R input terminal of SR flip-flop 272. Delay element 278 has an input terminal connected to the Q output terminal of SR flip-flop 274, and an output terminal connected to the R input terminal of SR flip-flop 274. Inverter 280 has an input terminal connected to the Q output terminal of SR flip-flop 272, and an output terminal connected to the first input terminal of NOR gate 250. Inverter 282 has an input terminal connected to the Q output terminal of SR flip-flop 274, and an output terminal connected to the second input terminal of NOR gate 252.

Now the operation of clock and control path 32 will be described briefly with reference to the circuit structure shown in FIG. 5. However additional details of the operation will be explained further with reference to FIGS. 6 and 7 below. I/O clock portion 160 generates three clock signals which are all based on the input CLOCK. Delay elements 162, 164, and 166 provide different amounts of delay in generating signals P2CLK, P3CLK, and QVCLK, respectively. The relative timing relationship between these signals and the clock signal will be explained further below.

Self-synchronizing port select circuit 170 functions essentially as a onebit counter to alternate accesses between PORT 1 and PORT 2. In general, the first part of a cycle defined by signal CLOCK will be generated by PORT 1 and the second cycle by PORT 2. Note that delay element 176 is used to insure that signal P1P2MUX starts in a logic low condition when power is first applied. After power-up, the state of the Q output signal of SR flip-flop 174 is indeterminate. However, delay element 176 and the additional logic input to OR gate 172 ensure that SR flip-flop 174 is reset to a logic 0 during this first cycle, and thus that signal PFP2MUX is registered as a logic low before the first access occurs.

Master clock circuit 180 is used to generate two clocks, at the output of AND gates 192 and 194, respectively, in which the output of AND gate 192 represents a trigger signal for a cycle 1 access and the output of AND gate 194 represents a trigger for a cycle 2 access. Note that the combination of an AND gate and corresponding delay elements forms a one-shot and therefore on a low-to-high transition of the CLOCK, the one-shot formed by delay element 188 and AND gate 192 provides an output to signify a trigger for the first cycle. At the end of the first cycle, self-timed read clocks and control logic circuit 210 activates signal DLCLK1 which causes SR flip-flop 186 to be set. When the Q output thereof transitions to a logic high, AND gate 194 outputs the trigger signal associated with the second cycle. When the CLOCK signal again transitions to logic high, the output of AND gate 184 will become a logic high, thereby resetting SR flip-flop 186. Note that throughout clock and control path 32, one shots are used to ensure that the circuitry is edge triggered. Thus, if the CLOCK signal is static by being held at a given logic state, no timing signals will be generated.

Data and address clock circuit 200 provides data input clock signal DINK to data path 34 and signal predecoder clock signal DECCLK to address path 36. Note that signals DINK and DECCLK are logically identical, but are output through separate OR gates in order to allow signal DECCLK to propagate as quickly as possible without the additional loading on the signal line driven by OR gate 202. Self-timed read clocks and control logic circuit 210 is used to generate three clocks which are important in various other operations including clocks DIC, FAMPB, and DLCLK1. Signal DIC is a decoder inhibit clock which is used to disable all decoders at the end of cycle 2 or cycle 1. Signal FAMPB is used to enable back-end sensing and signal DLCLK1 is used to register cycle 1 data as well as signaling the beginning of cycle 2.

Read/write control circuit 230 is used to provide control signals to drive array 30. The first signal is WEBCSB which indicates a pending read cycle. The second signal is CSWEB which indicates a pending write cycle. A "CYCLE 1" access to array 30 is an access which occurs during a first portion of the CLOCK cycle, and a "CYCLE 2" access is an access which follows the CYCLE 1 access during the CLOCK cycle. Normally a PORT 1 access occurs during CYCLE 1 and a PORT 2 access occurs during CYCLE 2. Signal RATFB is activated at a logic low if there is a match between PORT 1 and PORT 2 addresses and the access on PORT 1 is a write cycle. Signal RATFB is used to reverse the order of the PORT 1 and PORT 2 accesses such that during cycle 1 the PORT 2 access takes place and during cycle 2 the PORT 1 access takes place. Thus, memory 20 follows a read-before-write protocol.

The operation of memory 20 during various cycle combinations can be understood with reference to TABLE I below:

TABLE I______________________________________Case PORT 1 PORT 2 MATCH CYCLE 1 CYCLE 2______________________________________1 R R N R1 R22 R W N R1 W23 W R N W1 R24 W W N W1 W25 R R Y R1 R26 R W Y R1 W27 W R Y R2 W18 W W Y W1 W1______________________________________ in which "R" represents a read access, "W" represents a write access, "R1" represents a read access from PORT 1, "N" represents the no-match condition, and "Y" represents the match condition. So for example case 1 represents simultaneous read accesses on PORT 1 and PORT 2, but to different addresses. In this case, the PORT 1 access to array 30 occurs during CYCLE 1 and the port 2 read access occurs during CYCLE 2. In general, the access to PORT 1 occurs during CYCLE 1, and the access to PORT 2 occurs during CYCLE 2. However since memory 20 follows a read-before-write protocol, when PORT 1 attempts to write the same address that PORT 2 attempts to read, the PORT 2 access occurs during CYCLE 1 and the PORT 1 access occurs during CYCLE 2. Furthermore when PORT 1 and PORT 2 attempt to write the same address, the PORT 1 write takes precedence.

TABLE 1 is better understood with reference to FIGS. 6 and 7, which illustrate in timing diagram form signals which are relevant to understanding the operation of memory 20 of FIG. 1 during non-match (cases 1-4) and match (cases 5-8) cycles, respectively. Now considering FIGS. 6 and 7 together, note that since memory 20 is a synchronous memory, the CLOCK signal functions as a master clock signal which initiates all other clock signals. These dependencies are illustrated during the first case only, but hold for all other cases as well. A low-to-high transition of the CLOCK signal initiates the CYCLE 1-CYCLE 2 sequence. The high-to-low transition does not affect the start of CYCLE 2. Signal P1P2MUX is initially at a logic low for CYCLE 1 and at a logic high for CYCLE 2. Each CLOCK transition causes the activation of the pulsed DECCLK signal, which causes the predecoded address signals to be latched in the address path. Note that the timing signals which are illustrated in TABLE I and FIGS. 6 and 7 are a sufficient set of timing signals to control the major operation of memory 20, but will vary from embodiment to embodiment.

Although the present invention has been described with reference to a specific embodiment, further modifications and improvements will occur to those skilled in the art. For example, a dual-port memory according to the present invention may be constructed using different single port arrays of different densities. The timing signals used will also vary. Therefore it is to be understood that the invention encompasses all such modifications that do not depart from the scope of the invention as defined in the appended claims.

Claims

1. An integrated circuit memory, comprising:

a plurality of memory cells, each of the plurality of memory cells being coupled to a single word line and to a single bit line pair;

an address decoder, coupled to the plurality of memory cells, for selecting a memory cell of the plurality of memory cells in response to receiving an address;

a first address port, coupled to the address decoder, for providing a first address to the address decoder for accessing the plurality of memory cells;

a second address port, coupled to the address decoder, for providing a second address to the address decoder for accessing the plurality of memory cells;

a read data port, coupled to the plurality of memory cells, for reading data from the plurality of memory cells in response to either the first or the second address;

a write data port, coupled to the plurality of memory cells, for writing data to the plurality of memory cells in response to either the first or the second address; and

a control circuit, coupled to the address decoder, to the first and second address ports, and to the read and write data ports, the control circuit for controlling access to the plurality of memory cells, wherein substantially simultaneous requests for access to the plurality of memory cells are serviced sequentially within a single clock cycle of a clock signal of a data processor accessing the integrated circuit memory.

2. The integrated circuit memory of claim 1, wherein the integrated circuit memory is pipelined.

3. The integrated circuit memory of claim 1, further comprising an address collision detector for receiving the first and second addresses, and in response, providing a match signal of a first logic state when the first address is the same as the second address, and for providing the match signal of a second logic state when the first address is not the same as the second address.

4. The integrated circuit memory of claim 3, wherein the match signal is provided to the control circuit to insure that a read access occurs before a write access when the match signal is of the first logic state.

5. The integrated circuit memory of claim 1, wherein the read data port is characterized as being a first read/write data port and the write data port is characterized as being a second read/write data port.

6. The integrated circuit memory of claim 5, further comprising a selfsynchronizing port select circuit, coupled to the first and second read/write data ports, for insuring, on power-up of the integrated circuit memory, that the first address port is synchronized to the first read/write data port and the second address port is synchronized to the second read/write data port.

7. The integrated circuit memory of claim 5, further comprising a passthrough circuit for allowing data that is provided to the first read/write data port to be passed-through to the second read/write data port.

8. The integrated circuit memory of claim 5, wherein the first and second read/write data ports are enabled by first and second externally generated output enable signals, respectively, wherein the first and second externally generated output enable signals are generated independently of any clock signal.

9. A pipelined dual port static random access memory, comprising:

a plurality of static random access memory cells, each of the plurality of static random access memory cells being coupled to a single word line and to a single bit line pair;

an address decoder, coupled to the plurality of static random access memory cells, for selecting a memory cell of the plurality of static random access memory cells in response to receiving an address;

a first address port, coupled to the address decoder, for providing a first address to the address decoder for accessing the plurality of static random access memory cells;

a second address port, coupled to the address decoder, for providing a second address to the address decoder for accessing the plurality of static random access memory cells;

a first read/write data port, coupled to the plurality of static random access memory cells, for reading data from the plurality of static random access memory cells in response to either the first address or the second address;

a second read/write data port, coupled to the plurality of memory cells, for writing data to the plurality of static random access memory cells in response to either the first or the second address; and

a control circuit, coupled to the address decoder, to the first and second address ports, and to the first and second read/write data ports, the control circuit for controlling access to the plurality of static random access memory cells, wherein substantially simultaneous requests for access to the plurality of static random access memory are serviced sequentially within a single clock cycle of a clock signal of a data processor accessing the pipelined dual port static random access memory.

10. The pipelined dual port static random access memory of claim 9, further comprising an address collision detector for receiving the first and second addresses, and in response, providing a match signal of a first logic state when the first address is the same as the second address, and for providing the match signal of a second logic state when the first address is not the same as the second address.

11. The pipelined dual port static random access memory of claim 10, wherein the match signal is provided to the control circuit to insure that a read access occurs before a write access when the match signal is of the first logic state.

12. The pipelined dual port static random access memory of claim 9, further comprising a self-synchronizing port select circuit, coupled to the first and second read/write data ports, for insuring, on power-up of the pipelined dual-port static random access memory, that the first address port is synchronized to the first read/write data port and the second address port is synchronized to the second read/write data port.

13. The pipelined dual port static random access memory of claim 9, further comprising a pass-through circuit for allowing data that is provided to the first read/write data port to be passed-through to the second read/write data port.

14. The pipelined dual port static random access memory of claim 9, wherein the first and second read/write data ports are enabled by first and second externally generated output enable signals, respectively, wherein the first and second externally generated output enable signals are generated independently of any clock signal.

15. A pipelined dual port static random access memory, comprising:

a plurality of static random access memory cells, each of the plurality of static random access memory cells being coupled to a single word line and to a single bit line pair;

an address decoder, coupled to the plurality of static random access memory cells, for selecting a memory cell of the plurality of static random access memory cells in response to receiving an address;

a first address port, coupled to the address decoder, for providing a first address to the address decoder for accessing the plurality of static random access memory cells;

a second address port, coupled to the address decoder, for providing a second address to the address decoder for accessing the plurality of static random access memory cells;

an address collision detector, coupled to the first and second address ports, for receiving the first and second addresses, and in response, providing a match signal of a first logic state when the first address is the same as the second address, and for providing the match signal of a second logic state when the first address is not the same as the second address;

a first read/write data port, coupled to the plurality of static random access memory cells, for reading data from the plurality of static random access memory cells in response to either the first address or the second address;

a second read/write data port, coupled to the plurality of static random access memory cells, for writing data to the plurality of static random access memory cells in response to either the first address or the second address; and

a control circuit, coupled to the address decoder, to the first and second address ports, to the address collision detector, and to the first and second read/write data ports, the control circuit for controlling access to the plurality of memory cells wherein substantially simultaneous requests for access to the plurality of static random access memory cells are serviced sequentially within a single clock cycle of a clock signal of a data processor that is accessing the pipelined dual-port static random access memory, and wherein the match signal determines which of the first and second read/write data ports are serviced first.

16. The pipelined dual port static random access memory of claim 15, wherein the match signal is provided to the control circuit to insure that a read access occurs before a write access when the match signal is of the first logic state.

17. The pipelined dual port static random access memory of claim 15, further comprising a self-synchronizing port select circuit, coupled to the first and second read/write data ports, for insuring, on power-up of the pipelined dual-port static random access memory, that the first address port is synchronized to the first read/write data port and the second address port is synchronized to the second read/write data port.

18. The pipelined dual port static random access memory of claim 15, further comprising a pass-through circuit, coupled to the first and second read/write data ports, for allowing data that is provided to the first read/write data port to be passed-through to the second read/write data port.

19. The pipelined dual port static random access memory of claim 15, wherein the first and second read/write data ports are enabled by first and second externally generated output enable signals, respectively, wherein the first and second externally generated output enable signals are generated independently of any clock signal.