diff options
| -rw-r--r-- | README.md | 124 | ||||
| -rwxr-xr-x | bin/benchmark | bin | 267872 -> 288832 bytes | |||
| -rw-r--r-- | build/application.o | bin | 16688 -> 18904 bytes | |||
| -rw-r--r-- | build/benchmark.o | bin | 31232 -> 35080 bytes | |||
| -rw-r--r-- | build/crossbar_generator.o | bin | 17784 -> 20472 bytes | |||
| -rw-r--r-- | build/online_crossbar.o | bin | 0 -> 29384 bytes | |||
| -rw-r--r-- | build/read_crossbar.o | bin | 15584 -> 15472 bytes | |||
| -rw-r--r-- | makefile | 6 |
8 files changed, 89 insertions, 41 deletions
@@ -1,50 +1,94 @@ -# benchmark_suite_IREU -A benchmark suite for memristive and memcapacitive crossbar computing systems +# MemDevice Benchmark + +A benchmark suite for memristive and memcapacitive crossbar computing systems. ## Research question -How does a chosen device subcircuit affect the same computing system? -The memristive computing field lacks standardized benchmarks that control for -workload characteristics, expose device-type differences, and report results -comparable to conventional CMOS hardware. This project designs, implements, and -releases a benchmark suite targeting memristive and memcapacitive crossbar -architectures on CMOS substrates. +How does a chosen device subcircuit affect the same computing system? + +The memristive-computing field lacks standardized benchmarks that control for +workload characteristics, expose device-level differences, and produce results +that can be compared across implementations. MemDevice Benchmark evaluates +different SPICE device models while keeping the reservoir, crossbar topology, +training procedure, and workload fixed. + +## Architecture + +The SPICE crossbar replaces the software readout layer of a SPIRES reservoir +computer. Ridge regression first trains the readout weights in software. Those +signed weights are mapped to differential positive and negative conductances, +which become the initial resistances of the crossbar devices. Reservoir states +are scaled into crossbar row voltages, and the simulated column voltages are +decoded back into software-scale predictions. + +The benchmark supports two execution modes: + +- **Online mode (default):** SPIRES produces one reservoir state at a time and + passes it directly to a persistent shared-ngspice simulation. Ngspice runs on + a background worker thread, allowing crossbar timestep `t` to overlap with + reservoir timestep `t+1`. The pipeline has one timestep of output latency and + uses bounded backpressure instead of dropping or reordering states. +- **Offline mode:** SPIRES runs the complete input series first and stores every + reservoir state. The stored states are written as PWL voltage sources in a + batch SPICE netlist. Ngspice is launched as a separate process, and its output + data file is parsed after the simulation completes. + +Online mode uses a main SPIRES thread and a background ngspice thread. The +operating system may schedule them on separate CPU cores, but the benchmark does +not enforce CPU affinity. OpenMP or BLAS dependencies may also create additional +threads. + +## Dependencies + +- A C11 compiler with OpenMP support +- [SPIRES](https://github.com/txmastin/spires) +- ngspice, including the shared library and development headers +- PLplot +- OpenBLAS and LAPACKE +- POSIX threads + +Build SPIRES in the repository's `spires` directory before building this +benchmark. Install the remaining development packages using the package manager +for your operating system. + +## Build and run + +From the project root: + +```bash +make +``` + +Run the default online benchmark: + +```bash +make run +``` -## abstract -MemDevice Benchmark is a benchmarking framework for simulated memristor crossbar's -using reservoir computing tasks as standardized workloads with the goal of revealing -unique device model effects. The framework evaluates memristor devices under the -same circuit, network, and workload characteristics. In the proposed architecture -the SPICE crossbar is implemented as the readout layer for a reservoir computer. -Ridge regressiong training is done on a software reservoir, training the software -weights. An input series is given to the reservoir, the resulting temporal states -are converted into voltage signals applied to the crossbar rows. The trained readout -weights are converted into equivalent resistances to serve as the initial device -values. The crossbar columns are read and mapped back to human readable data that -prove task prediciton capability. +Select a mode explicitly: -Because the reservoir, crossbar, training procedure and benchmark tasks remain fixed, -different SPICE device models can be hot swapped and benchmarked in one continous -process, making fair comparisons possible and observing how device model choice -affects the same computing system +```bash +make run MODE=online +make run MODE=offline +``` -## Setup -**Download Spires:** -git clone https://github.com/txmastin/spires.git -cd spires -make +The benchmark prints the mean squared error for each device model and writes +generated netlists, simulation data, and SVG plots under `output/`. -**Download plplot:** -git clone git://git.code.sf.net/p/plplot/plplot plplot.git -cd plplot.git -mkdir build -cd build -make -sudo make install +## Embedded deployment -**Other dependencies may be required for the listed libraries** +The shared-ngspice backend targets a general-purpose computer. Typical +microcontrollers do not have the memory, numerical libraries, filesystem, or +operating-system services required to run ngspice. -**from the projects root directory (~/MemDevice_Benchmark/)** -make run +For an embedded deployment, a small SPIRES reservoir (for example, up to 16 +neurons) can run on a microcontroller and transmit each state vector to a host +computer. The host runs the persistent ngspice crossbar and returns the decoded +prediction. A fixed-size binary protocol should include a timestep number, the +state values, and integrity checking so missing or reordered messages can be +detected. -## +The producer/consumer pipeline remains useful in that configuration, but the +concurrency spans the microcontroller and host. The two-thread local +implementation is primarily useful when both the software reservoir and ngspice +readout run on the same general-purpose computer. diff --git a/bin/benchmark b/bin/benchmark Binary files differindex 5eeebe0..c7996b2 100755 --- a/bin/benchmark +++ b/bin/benchmark diff --git a/build/application.o b/build/application.o Binary files differindex 6c47157..66e9c44 100644 --- a/build/application.o +++ b/build/application.o diff --git a/build/benchmark.o b/build/benchmark.o Binary files differindex bd9f548..8d29915 100644 --- a/build/benchmark.o +++ b/build/benchmark.o diff --git a/build/crossbar_generator.o b/build/crossbar_generator.o Binary files differindex ee5613f..c654c93 100644 --- a/build/crossbar_generator.o +++ b/build/crossbar_generator.o diff --git a/build/online_crossbar.o b/build/online_crossbar.o Binary files differnew file mode 100644 index 0000000..cfcf3e1 --- /dev/null +++ b/build/online_crossbar.o diff --git a/build/read_crossbar.o b/build/read_crossbar.o Binary files differindex eb1bb7f..7d280dc 100644 --- a/build/read_crossbar.o +++ b/build/read_crossbar.o @@ -6,12 +6,14 @@ BUILD_DIR = build BIN_DIR = bin TARGET = $(BIN_DIR)/benchmark +MODE ?= online SOURCES = \ $(SRC_DIR)/application.c \ $(SRC_DIR)/spires_interface.c \ $(SRC_DIR)/crossbar_generator.c \ $(SRC_DIR)/read_crossbar.c \ + $(SRC_DIR)/online_crossbar.c \ $(SRC_DIR)/benchmark.c OBJECTS = $(SOURCES:$(SRC_DIR)/%.c=$(BUILD_DIR)/%.o) @@ -38,6 +40,8 @@ LDLIBS = \ -lopenblas \ -llapacke \ -lplplot \ + -lngspice \ + -lpthread \ -lm \ -fopenmp @@ -58,7 +62,7 @@ $(BIN_DIR): mkdir -p $(BIN_DIR) run: $(TARGET) - ./$(TARGET) + ./$(TARGET) --$(MODE) clean: rm -rf $(BUILD_DIR) $(BIN_DIR) |