Dissertation Performance Portable Short Vector Transforms

ing from Special Machine Features In the context of this thesis all short vector SIMD extensions feature the functionality required in intermediate level building blocks. However, the implementation of such building blocks depends on special features of the target architecture. For instance, a complex reordering operation like a permutation has to be implemented using register-register permutation instructions provided by the target architecture. In addition, restrictions like aligned memory access have to be handled. Thus, a set of intermediate building blocks has to be defined which (i) can be implemented on all current short vector SIMD architectures and (ii) enables all discrete linear transforms to be built on top of these building blocks. This set is called the portable SIMD API. Appendix B describes the relevant parts of the instruction sets provided by current short vector SIMD extensions. Abstracting from Special Compiler Featuresing from Special Compiler Features All compilers featuring a short vector SIMD C language extension provide the required functionality to implement the portable SIMD API. But syntax and semantics differ from platform to platform and from compiler to compiler. These specifics have to be hidden in the portable SIMD API. Table 3.1 (on page 40) shows that for any current short vector SIMD extension compilers with short vector SIMD language extensions exist. 6.1 Definition of the Portable SIMD API The portable SIMD API includes macros of four types: (i) data types, (ii) constant handling, (iii) arithmetic operations, and (iv) extended memory operations. An overview of the provided macros is given below. Appendix C contains examples of actual implementations of the portable SIMD API on various platforms. All examples of such macros displayed in this section suppose a two-way or fourway short vector SIMD extension. The portable SIMD API can be extended to arbitrary vector length ν. Thus, optimization techniques like loop interleaving (Gatlin and Carter [37]) can be implemented on top of the portable SIMD API.