auroraustc / opt_matmul_step_by_step

Optimize matrix multiplication step-by-step

Pre-requests

• A Processor which supports AVX2 instruction set
• GCC
• Intel MKL

How to build and run

# Check whether the processor supports AVX2 instruction set
# You should see some outputs. No output means AVX2 is not supported by the processor you are using.
cat /proc/cpuinfo | grep -i avx2
# Build using g++
g++ *.cpp  -m64 -I${MKLROOT}/include -O3 -mavx2 -fopenmp -L${MKLROOT}/lib/intel64 -Wl,--no-as-needed -lmkl_intel_lp64 -lmkl_gnu_thread -lmkl_core -lgomp -lpthread -lm -ldl
# Run the code
# ./a.out [N]
# N is the size of the matrices. Could be nothing or an integer.
# For 1024 by 1024 matrices:
./a.out
# For 2048 by 2048 matrices:
./a.out 2048