17e783b241
Adds `nvvm.addf` and `nvvm.subf` Ops to the NVVM dialect. `nvvm.addf` performs a floating-point addition between two operands. `nvvm.subf` performs a floating-point subtraction between two operands and is equivalent to an `llvm.fneg` followed by an `nvvm.addf` operation. PTX ISA Reference: 1. https://docs.nvidia.com/cuda/parallel-thread-execution/#floating-point-instructions-add 2. https://docs.nvidia.com/cuda/parallel-thread-execution/#half-precision-floating-point-instructions-add