Analog neural memory system for deep learning neural network comprising multiple vector-by-matrix multiplication arrays and shared components
Numerous embodiments are disclosed for an analog neuromorphic memory system for use in a deep learning neural network. The analog neuromorphic memory system comprises a plurality of vector-by-matrix multiplication arrays and various components shared by those arrays. The shared components include high voltage generation blocks, verify blocks, and testing blocks. The analog neuromorphic memory system optionally is used within a long short term memory system or a gated recurrent unit system.
1. An analog neuromorphic memory system, comprising:
a plurality of vector-by-matrix multiplication systems, each vector-by-matrix multiplication system comprising:
an array of memory cells;
a low voltage row decoder for providing a bias voltage to one or more rows of memory cells in the array of memory cells during read and program operations and for providing a decoding signal;
a high voltage row decoder for receiving the decoding signal and providing a high voltage bias signal to one or more rows of memory cells in the array of memory cells during program and erase operations; and
a low voltage column decoder,
a plurality of output blocks, each output block providing an output in response to at least one of the plurality of vector-by-matrix multiplication systems; and
a shared high voltage generator block configured to concurrently provide high voltages to high voltage row decoders in each of the plurality of vector-by-matrix multiplication systems for one or more of erase operations and programming operations.
2. The system of claim 1 , further comprising a high voltage compensation block for each of the plurality of vector-by-matrix multiplication systems.
3. The system of claim 1 , further comprising a high voltage compensation block for all of the plurality of vector-by-matrix multiplication systems.
4. The system of claim 1 , where the system is configured to concurrently perform programming operations to two or more vector-by-matrix multiplication systems.
5. The system of claim 1 , wherein the concurrent programming operations are performed in response to a single command.
6. The system of claim 4 , where the system is configured to concurrently perform verify operations to the two or more vector-by-matrix multiplication systems after the concurrent programming operations.
7. The system of claim 1 , wherein the high voltage generator block is able to provide sufficient voltage and current to perform program and erase operations concurrently on all arrays of memory cells in all of the vector-by-matrix multiplication systems.
8. The system of claim 1 , wherein the memory cells are split-gate flash memory cells.
9. The system of claim 1 , wherein each vector-by-matrix multiplication system is a cell in a long short term memory system.
10. The system of claim 1 , wherein each vector-by-matrix multiplication system is a cell in a gated recurrent unit memory system.
11. The system of claim 1 , wherein the high voltage generator block comprises:
a charge pump for outputting a high voltage; and
a charge pump regulator for receiving the high voltage from the charge pump and generating a trimmed high voltage in response to input trim bits.
12. The system of claim 1 , wherein each of the plurality of output blocks comprises a summer and activation block.
13. The system of claim 12 , wherein each summer and activation block is configured to perform a summing and activation function for at least one of the plurality of vector-by-matrix multiplication systems.