Multithreaded derivative computation with generated libraries

The computation of derivatives via automatic differentiation is a valuable technique in many science and engineering applications. While the implementation of automatic differentiation via source transformation yields the highest efficiency results, particularly for gradient computations, the implementation via operator overloading remains a viable alternative for some application contexts. Examples include the computation of higher order derivatives or cases where C++ as the language of choice still proves to be too complicated for the currently available source transformation tools. In this paper, we utilize a code generator to create libraries that overload intrinsics for derivative computation, and discuss approaches to improve the efficiency of the generated code. We first discuss the use of limited loop unrolling, but the main focus of the paper is multithreaded derivative computation, in particular an asynchronous scheme for C++ and a synchronous scheme for Fortran. We present test results obtained with a proof-of-concept implementation.     

Brillouin scattering from ferromagnetic metallic glasses

We report the first measurements of thermally excited surface acoustic phonons, and bulk and surface magnons from metallic glasses using Brillouin scattering. In particular we present data on the compositions Fe₄₀Ni₄₀P₁₄B₆ and Fe₈₀B₂₀. The velocity of the surface phonons was found to agree within 1.5% with the value calculated from elastic continuum theory using known bulk elastic constants. The observed field dependence of the bulk magnon frequency can be fit within experimental error to the classical relation for the k = 0 mode by including alldemagnetizing corrections. The values for the magnetization obtained from both the surface and bulk magnons agree with the bulk values determined previously with a magnetometer. The measured field dependences of the surface magnon frequency allow us to determine the g-factors, yielding 2.08±0.02 for FeB and 2.03±0.03 for FeNiPB.