Closed WLFJ closed 1 year ago
by running:
./build/bin/gemm-benchmark | python3 scripts/gflops.py
./build/bin/conv2d-benchmark | python3 scripts/gflops.py
Benchmark also have correctness checking, Conv2d is enabled by default and MatMul not. If you want to make benchmark running faster, just remove #define OP_TEST in benchmarks/OpOptimization/Conv2dNchwFchw/MLIROptBenchmark.cpp.
#define OP_TEST
benchmarks/OpOptimization/Conv2dNchwFchw/MLIROptBenchmark.cpp
MatMul benchmark result
by running:
Conv2dNchwFchw benchmark result
by running:
Note
Benchmark also have correctness checking, Conv2d is enabled by default and MatMul not. If you want to make benchmark running faster, just remove
#define OP_TEST
inbenchmarks/OpOptimization/Conv2dNchwFchw/MLIROptBenchmark.cpp
.