A domain-specific architecture for elementary function evaluation
Summary: We propose a Domain-Specific Architecture for elementary function computation to improve throughput while reducing power consumption as a model for more general applications: support fine-grained parallelism by eliminating branches, and eliminate the duplication required by coprocessors by decomposing computation into instructions which fit existing pipelined execution models and standard register files. Our example instruction architecture (ISA) extension supports scalar and vector/SIMD implementations of table-based methods of calculating all common special functions, with the aim of improving throughput by (1) eliminating the need for tables in memory, (2) eliminating all branches for special cases, and (3) reducing the total number of instructions. Two new instructions are required, a table lookup instruction and an extended-precision floating-point multiply-add instruction with special treatment for exceptional inputs. To estimate the performance impact of these instructions, we implemented them in a modified Cell/B.E. SPU simulator and observed an average throughput improvement of 2.5 times for optimized loops mapping single functions over long vectors.
- Architecture-independent approximation of functions
- scientific article; zbMATH DE number 3978260
- Modular hardware architecture for somewhat homomorphic function evaluation
- Elementary Functions Hardware Implementation Using Constrained Piecewise-Polynomial Approximations
- Efficient Multiple-Precision Evaluation of Elementary Functions
- A poly-algorithmic approach to simplifying elementary functions
- scientific article; zbMATH DE number 1844677
- An accurate elementary mathematical library for the IEEE floating point standard
- An Optimized Cell BE Special Function Library Generated by Coconut
- scientific article; zbMATH DE number 3978260 (Why is no real title available?)
- Multiplication/ division/ square root module for massively parallel computers
- Starting approximations for square root calculation on IBM system /360
- Table-driven implementation of the Expm1 function in IEEE floating-point arithmetic
- Table-driven implementation of the logarithm function in IEEE floating-point arithmetic
- Unified tables for exponential and logarithm families
This page was built for publication: A domain-specific architecture for elementary function evaluation
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q1751387)
