v0.0.1

First Release of float8_matmul

This is the first release of the float8_matmul project, featuring basic matrix multiplication for 8-bit floating-point matrices.

Key Features:

• Basic 8-bit float matrix multiplication (matmul).
• About 25~50% of native FP32 performance for older CUDA device

Passed test:

When:

• Matrix A: Row x0 filled with 1.0.
• Matrix B: Column y0 filled with 1.0.

Expected Output:

• Result at (x0, y0) will be the matrix dimension.
• All other positions will be 0.0.