Certain aspects of the present disclosure provide a method, including: storing a depthwise convolution kernel in a first one or more columns of a CIM array; storing a fused convolution kernel in a second one or more columns of the CIM array; storing pre-activations in one or more input data buffers associated with a plurality of rows of the CIM array; processing the pre-activations with the depthwise convolution kernel in order to generate depthwise output; modifying one or more of the pre-activations based on the depthwise output to generate modified pre-activations; and processing the modified pre-activations with the fused convolution kernel to generate fused output.

An integrated circuit includes: a dual port Static Random Access Memory (SRAM) cell including a plurality of transistors; a bit line pair connected to the dual port SRAM cell, the bit line pair including a first bit line and a second bit line spaced apart from each other in a first direction and extending in a second direction perpendicular to the first direction; a power line group including a plurality of power lines spaced apart from each other in the first direction, spaced apart from the bit line pair placed in the first direction, and extending in the second direction, the power line group being configured to apply a voltage to the dual-port SRAM cell; and a first word line provided between the first bit line and the second bit line and connected to the dual port SRAM cell.

An integrated circuit includes: a dual port Static Random Access Memory (SRAM) cell including a plurality of transistors; a bit line pair connected to the dual port SRAM cell, the bit line pair including a first bit line and a second bit line spaced apart from each other in a first direction and extending in a second direction perpendicular to the first direction; a power line group including a plurality of power lines spaced apart from each other in the first direction, spaced apart from the bit line pair placed in the first direction, and extending in the second direction, the power line group being configured to apply a voltage to the dual-port SRAM cell; and a first word line provided between the first bit line and the second bit line and connected to the dual port SRAM cell.

A semiconductor device includes a semiconductor substrate including a fin of semiconductor material having a fin width and a fin length. The fin length is greater than the fin width and extends between a first fin end and a second fin end. A gate electrode extends over the fin at a first fin location between the first fin end and the second fin end. A dummy gate electrode extends over the first fin end and is floating.

A semiconductor device includes a semiconductor substrate including a fin of semiconductor material having a fin width and a fin length. The fin length is greater than the fin width and extends between a first fin end and a second fin end. A gate electrode extends over the fin at a first fin location between the first fin end and the second fin end. A dummy gate electrode extends over the first fin end and is floating.

An SRAM (static random access memory) includes a semiconductor substrate; a plurality of PD transistors, each including a first fin structure formed on the semiconductor substrate, a PD gate structure formed across the first fin structure and covering a portion of a top and sidewall surfaces of the first fin structure, and a first source/drain doped layer formed in the first fin structure on both sides of the PD gate structure; a plurality of adjacent transistors, each including a second fin structure formed on the semiconductor substrate and a second source/drain doped layer formed in the second fin structure; an isolation layer, formed on the semiconductor substrate; a fin sidewall film, formed on the isolation layer and covering sidewall surfaces of each PD gate structure; and a first PD dielectric layer, formed on the isolation layer and covering sidewall surfaces of the first source/drain doped layer.

A memory device has a plurality of bit cells, each of which includes an SRAM cell having a storage node selectively connectable to a first bit line in response to a control signal received on a first word line. Each bit cell further includes an MRAM cell selectively connectable to the storage node of the SRAM cell in response to a control signal received on a second word line.

A memory device has a plurality of bit cells, each of which includes an SRAM cell having a storage node selectively connectable to a first bit line in response to a control signal received on a first word line. Each bit cell further includes an MRAM cell selectively connectable to the storage node of the SRAM cell in response to a control signal received on a second word line.

Certain aspects provide an apparatus for signal processing in a neural network. The apparatus generally includes first computation in memory (CIM) cells configured as a first kernel for a neural network computation, the first set of CIM cells comprising one or more first columns and a first plurality of rows of a CIM array. The apparatus also include a second set of CIM cells configured as a second kernel for the neural network computation, the second set of CIM cells comprising the one or more first columns and a second plurality of rows of the CIM array. The first plurality of rows may be different than the second plurality of rows.

An apparatus and method for providing efficient floor planning, power, and performance tradeoffs of memory accesses. A dual read port and single write port memory bit cell uses two asymmetrical read access circuits for conveying stored data on two read bit lines. The two read bit lines are pre-charged to different voltage reference levels. The layout of the memory bit cell places the two read bit lines on an opposed edge from the single write bit line. The layout uses a dummy gate placed over both p-type diffusion and n-type diffusion between the edges. The layout has a same number of p-type transistors as n-type transistors despite using asymmetrical read access circuits. The layout also has a contacted gate pitch that is one more than the number of p-type transistors.