Abstract: Systems and methods are provided to perform multiply-accumulate operations of reduced precision numbers in a systolic array. Each row of the systolic array can receive reduced inputs from a respective reducer. The reduced input can include a reduced input data element and/or a reduced weight. The systolic array may lack support for inputs with a first bit-length and the reducers may reduce the bit-length of a given input from the first bit-length to a second shorter bit-length and provide the reduced input to the array. In order to reduce the bit-length, the reducer may reduce the number of trailing bits of the input. Further, the systolic array can receive a reduced and rounded input. The systolic array can propagate the reduced input through the processing elements in the systolic array. Each processing element may include a multiplier and/or an adder to perform arithmetical operations based on the reduced input.

Abstract: Systems and methods are provided to perform multiply-accumulate operations of reduced precision numbers in a systolic array. Each row of the systolic array can receive reduced inputs from a respective reducer. The reduced input can include a reduced input data element and/or a reduced weight. The systolic array may lack support for inputs with a first bit-length and the reducers may reduce the bit-length of a given input from the first bit-length to a second shorter bit-length and provide the reduced input to the array. In order to reduce the bit-length, the reducer may reduce the number of trailing bits of the input. Further, the systolic array can receive a reduced and rounded input. The systolic array can propagate the reduced input through the processing elements in the systolic array. Each processing element may include a multiplier and/or an adder to perform arithmetical operations based on the reduced input.

Abstract: Systems and methods are provided to perform multiply-accumulate operations of reduced precision numbers in a systolic array. Each row of the systolic array can receive reduced inputs from a respective reducer. The reducer can receive a particular input and generate multiple reduced inputs from the input. The reduced inputs can include reduced input data elements and/or a reduced weights. The systolic array may lack support for inputs with a first bit-length and the reducers may reduce the bit-length of a given input from the first bit-length to a second shorter bit-length and provide multiple reduced inputs with second shorter bit-length to the array. The systolic array may perform multiply-accumulate operations on each unique combination of the multiple reduced input data elements and the reduced weights to generate multiple partial outputs. The systolic array may sum the partial outputs to generate the output.

Abstract: To reduce power consumption, data bits or a portion of a data register that is not expected to toggle frequently can be grouped together, and be clock-gated independently from the rest of the data register. The grouping of the data bits can be determined based on the data types of the workload being operated on. For a data register configured to store a numeric value that supports multiple data types, the portion of the data register being clock-gated may store a group of data bits that are unused for one or more data types of the multiple data types supported by the data register. The portion of the data register being clock-gated can also be a group of data bits that remain unchanged or have a constant value for numeric values within a certain numeric range that is frequently operated on.

Abstract: Clock skew may be increased along a critical path of a systolic array. Pipelined registers may be added between a bus that provides input data signals to a systolic array and between a bus that receives output data signals from the systolic array. Skew circuitry for the pipelined registers may be implemented to delay a clock signal to the pipelined registries to allow a clock skew accumulated along a critical path of the systolic array to exceed a single clock cycle.

Abstract: Systems and methods for power analysis of a hardware device design. In various examples, a target circuit can be defined within the hardware device design. The target circuit can include a plurality of digital circuit elements linking a plurality of input nodes with a plurality of output nodes. A solver can be used to search for a transition pattern that, when applied to the input nodes, causes a number of output nodes equal to a counter to transition from a first binary value to a second binary value. If a transition pattern cannot be found, the counter is decremented and a new transition pattern is searched for. Once a transition pattern is found, it is determined whether the transition pattern satisfies a constraint.

Abstract: A write driver for a memory circuit includes a control circuit configured to: operate a first push-pull driver to generate a first drive signal in a first voltage domain at a first node based on an input signal in a second domain and in response to a mode select signal being in a first mode, wherein the first drive signal is at a same logic level as the input signal; operate a second push-pull driver to generate a second drive signal in the first voltage domain at a second node based on the input signal and in response to the mode select signal being in the first mode, wherein the second drive signal is at a complement logic level with respect to the input signal; and operate the first and second push-pull drivers to float the first and second nodes in response to the mode select signal being in a second mode.

Abstract: A write-assisted memory. The write-assisted memory includes a word-line decoder that is implemented within a low VDD power domain. The write-assisted memory also includes a write-segment controller that is partially implemented within the low VDD power domain and is partially implemented within a high VDD power domain. The write-assisted memory further includes a local write word-line decoder that is implemented within the high VDD power domain.

Abstract: Various apparatuses and methods are disclosed. The system describes a pulse generator comprising a first stage configured to be powered by a first voltage; and a second stage configured to be powered by a second voltage different from the first voltage, wherein the second stage is further configured to generate a pulse in response to an input to the first stage comprising a trigger and feedback from the second stage.

Abstract: A static, ternary content addressable memory (TCAM) includes a key cell and a mask cell coupled to intermediate match lines. The key cell is coupled to a first pull-down transistor and a first pull-up transistor. The mask cell is coupled to a second pull-down transistor and a second pull-up transistor. The first pull-down transistor and second pull-down transistor are connected in parallel and the first pull-up transistor and second pull-up transistor are connected in series. A match line output is also coupled to the first pull-down transistor and second pull-down transistor and further coupled to the first pull-up transistor and second pull-up transistor.

Abstract: A write assist driver circuit is provided that assists a memory cell (e.g., volatile memory bit cell) in write operations to keep the voltage at the memory core sufficiently high for correct write operations, even when the supply voltage is lowered. The write assist driver circuit may be configured to provide a memory supply voltage VddM to a bit cell core during a standby mode of operation. In a write mode of operation, the write assist driver circuit may provide a lowered memory supply voltage VddMlower to the bit cell core as well as to at least one of the local write bitline (lwbl) and local write bitline bar (lwblb). Additionally, the write assist driver circuit may also provide a periphery supply voltage VddP to a local write wordline (lwwl), where VddP?VddM>VddMlower.

Abstract: A static, ternary content addressable memory (TCAM) includes a key cell and a mask cell coupled to intermediate match lines. The key cell is coupled to a first pull-down transistor and a first pull-up transistor. The mask cell is coupled to a second pull-down transistor and a second pull-up transistor. The first pull-down transistor and second pull-down transistor are connected in parallel and the first pull-up transistor and second pull-up transistor are connected in series. A match line output is also coupled to the first pull-down transistor and second pull-down transistor and further coupled to the first pull-up transistor and second pull-up transistor.

Abstract: Various aspects of a fast, energy efficient write driver capable of efficient operation in a dual-voltage domain memory architecture are provided herein. Specifically, various aspects of the write driver described herein combine a high speed driver with voltage level shifting capabilities that may be implemented efficiently in reducing use of silicon area while using lower power. The write driver circuit shifts or adjusts voltage levels between a first voltage domain to a second voltage domain. In one example, the write driver circuit is coupled to a global write bitline and a local write bitline that is coupled to one or more bitcells (of SRAM memory). The write driver circuit converts a first voltage level at the global write bitline to a second voltage level at the local write bitline during a write operation.

Abstract: A multiport bitcell including a pair of cross-coupled inverters is provided with increased write speed and enhanced operating voltage range by the selective isolation of a first one of the cross-coupled inverters from a power supply and ground during a write operation. The write operation occurs through a write port that includes a transmission gate configured to couple a first node driven by the first cross-coupled inverter to a write bit line. A remaining second cross-coupled inverter in the bitcell is configured to drive a second node that couples to a plurality of read ports.

Abstract: A static, ternary content addressable memory (TCAM) includes a key cell and a mask cell coupled to intermediate match lines. The key cell is coupled to a first pull-down transistor and a first pull-up transistor. The mask cell is coupled to a second pull-down transistor and a second pull-up transistor. The first pull-down transistor and second pull-down transistor are connected in parallel and the first pull-up transistor and second pull-up transistor are connected in series. A match line output is also coupled to the first pull-down transistor and second pull-down transistor and further coupled to the first pull-up transistor and second pull-up transistor.

Abstract: A multiport bitcell including a pair of cross-coupled inverters is provided with increased write speed and enhanced operating voltage range by the selective isolation of a first one of the cross-coupled inverters from a power supply and ground during a write operation. The write operation occurs through a write port that includes a transmission gate configured to couple a first node driven by the first cross-coupled inverter to a write bit line. A remaining second cross-coupled inverter in the bitcell is configured to drive a second node that couples to a plurality of read ports.

Abstract: A method within a hybrid ternary content addressable memory (TCAM) includes comparing a first portion of a search word to a first portion of a stored word in a first TCAM stage. The method further includes interfacing an output of the first TCAM stage to an input of the second TCAM stage. The method also includes comparing a second portion of the search word to a second portion of the stored word in a second TCAM stage when the first portion of the search word matches the first portion of the stored word. The first TCAM stage is different from the second TCAM stage.

Abstract: Various circuits and methods of operating circuits are disclosed. A circuit may include a pulse generator and a latch having an output configured to trigger the pulse generator, wherein the latch is configured to be set by an input signal and reset by feedback from the pulse generator. A method may include resetting a latch using feedback from a pulse generator by setting a latch using an input signal, triggering a pulse generator using an output from the latch, and resetting the latch using feedback from the pulse generator.

Abstract: A method within a ternary content addressable memory (TCAM) includes receiving a match line output from a previous TCAM stage at a gate of a pull-up transistor of a current TCAM stage and at a gate of a pull-down transistor of the current TCAM stage. The method sets a match line bar at the current TCAM stage to a low value, via the pull-down transistor, when the match line output from the previous TCAM stage indicates a mismatch. The method also sets the match line bar at the current TCAM stage to a high value, via the pull-up transistor, when the match line output from the previous TCAM stage indicates a match.

Abstract: A write driver for a memory circuit includes a control circuit configured to: operate a first push-pull driver to generate a first drive signal in a first voltage domain at a first node based on an input signal in a second domain and in response to a mode select signal being in a first mode, wherein the first drive signal is at a same logic level as the input signal; operate a second push-pull driver to generate a second drive signal in the first voltage domain at a second node based on the input signal and in response to the mode select signal being in the first mode, wherein the second drive signal is at a complement logic level with respect to the input signal; and operate the first and second push-pull drivers to float the first and second nodes in response to the mode select signal being in a second mode.