I’m trying to use cublasGemmEx or cublasHgemm to perform C = A*B + C. Probably I’m missing something stupid, but the output is always a matrix filled with zeros. I tried all the combination of float and half for the matrices’ pointers.
Here’s the code: cublasEx.cpp (4.3 KB) I use the code with file extension .cu
I compile with:
nvcc -ccbin g++ -I./Common -m64 -maxrregcount=255 -gencode arch=compute_75,code=sm_75 -gencode arch=compute_75,code=compute_75 -o cublasEx.o -c cublasEx.cu
nvcc -ccbin g++ -lcublas -m64 -gencode arch=compute_75,code=sm_75 -gencode arch=compute_75,code=compute_75 -o cublasEx cublasEx.o