U12: New Features in TamStat summary | slides (750 KB) | video (23 mins)
#Tasky tutorial how to
All that remained was to learn how to use it and how to seamlessly integrate it in a traditional Windows-based APL application. The first step was the choice of the tools and so we went fishing and caught in our net D3.js, a very versatile open source JavaScript library for manipulating web documents based on data. We identified a few areas where the introduction of charts would improve significantly the user experience and set off on their implementation. Instead of trying to compete with dedicated reporting tools, we thought we would play to our strengths if we concentrated on a handful of tailored charts. A short while ago we decided that the time was ripe for it to start drawing pictures to better show the beauty of those numbers. Sofia, our portfolio management system for institutional investors, is capable of producing large quantities of numbers. Certainly, most human beings find an interactive chart more intuitive than a table with a thousand numeric rows. They say that a picture is worth a thousand words. Robert Bernecky, Snake Island Research (Canada) U08: A Compendium of SIMD Boolean Array Algorithms in APL summary | slides (3.4 MB) | handouts (3.7 MB) | video (45 mins) The Futhark compiler performs a number of GPU-critical optimisations before outputting GPU-optimised OpenCL code, which will be shown to integrate well with host-executed implementations of high-level languages, such as Python or even an APL interpreter. The compiler infrastructure compiles APL programs into programs written in a typed array intermediate language (called TAIL), which are then recompiled into programs written in the functional array programming language, Futhark.
He will reflect on the programmer's experience with programming in the supported, statically typeable, subset of APL and the fact that the compiler infrastructure does not necessarily embrace the use of idioms before inviting APL programmers to reproduce the obtained results he will also invite experienced APL programmers to improve on the implementation of the Dyalog versions of the benchmarks and to provide feedback about which additional APL features should be considered indispensable.
#Tasky tutorial code
He demonstrates that, for a number of benchmarks, the generated GPU programs runs in the order of 100x faster than compiled generated C code running on similarly-priced CPU hardware. Martin presents a compiler infrastructure for compiling a subset of APL, written in the style of co-dfns, into performance efficient GPU programs.
0 kommentar(er)
