Day 5

The focus of Day 5 is cuBLAS Matrix Multiplication

Code Descriptions

cublas_matmul.cu (Matrix Multiplication using cuBLAS)

• Implements matrix multiplication using the cuBLAS cublasSgemm operation.
• Avoids explicit cudaMemcpy by utilizing cublasSetMatrix and cublasGetMatrix.
• Uses the cuBLAS API for optimized GPU-accelerated BLAS operations.
• Performs C = A × B efficiently with cublasSgemm.

Profiling and Running

To compile and run the CUDA program, use:

nvcc -o cublas_matmul cublas_matmul.cu -lcublas -lcurand
./cublas_matmul