cublasGemmEx() doesn't support 8-bit unsigned integer multiplications!!!


cublasGemmEx() with Atype = CUDA_R_8I, Btype = CUDA_R_8I and computetype CUDA_32_8I, only does multiplication of signed 8-bit numbers(-128 to 127) .

Does it also support UNSIGNED 8-bit multiplication(0-255), something like CUDA_R_8U for Atype, Btype and CUDA_R_32I for computetype. The documentation doesn’t mention anything for this.

How can I do multiplication of unsigned 8-bit multiplications using cublasGemmEx() ?