High-performance CPU SpMM (N:M sparsity) algorithm accelerated using AVX.

TODOs:

• More performance optimization?
• Parameterized N and M (currently only works for 2:4 sparsity)

Metrics are in GFLOPs; speedups are compared to either OpenBLAS or MKL, whichever is faster. Types are fp32.

Single-threaded, n=4096 average of 10 runs

| Kernel | CPU | This algorithm | BLAS | Speedup | |:-------|:----|--------:|---------------:|---------:|:------------------| AVX-512 | Tiger Lake i5-1135G7 | 146 | 122 | 1.20 | AVX-512 | Granite Rapids Xeon 6972P | 158 | --- | --- |

Multithreaded, n=4096 average of 10 runs

A modern x86 processor supporting AVX2 or AVX-512 must be used.

For the library:

• pthreads support (for multithreading)
• A version of make that supports the shell directive

For the bench program:

• OpenBLAS (for verifying correctness and providing a baseline to compare against)

To build, use the given makefile, specifying your march. For example:

$ make TARGET=skylake

When compiling for a target that supports AVX-512, AVX-512 operations will automatically be used.

By default, multithreading is enabled using as many threads as logical processors on your system. To change the number of threads, pass the NUM_THREADS=<num> option:

$ make TARGET=skylake NUM_THREADS=4

You'll need to make clean first before changing configurations.