Numerical floating-point code: precision-independence, base-independence, and other programming considerations

The knowledge and skills of the floating-point-literate mathematician are seldom lauded, though central to the art of translating the world of mathematics into the language of computers. This thesis highlights the importance of floatingpoint literacy through various high-utility techniques relevant to the beginning numerical programmer. Particular attention is paid to the pursuit of writing precision-independent and base-independent code. Consideration is also given to special values like NaN , error-calculation techniques, and the fused multiply-add function. Also included is a brief discussion about the future utility of numerical techniques in light of computer hardware trends.