|
From: Natalija K. <nat...@ma...> - 2026-10-23 11:17:07
|
Hi all, Dmitri Goloubentsev presented the AADC Tracer at the WBS Quant Finance Conference in Valletta this month. It records a QuantLib calculation and links each input, constant and executed branch to the source line that used it, so you can see what actually produced a number. It is built on AADC, our AAD compiler, which is in production at major banks. Here is a small question to try it on. A 25-line pricer for a European swaption, Hull-White model, Jamshidian engine: how many hard-coded constants reach its price? Make a guess from the code first, then run (Linux x86-64 or macOS on Apple silicon, Python 3.10 to 3.14; on Windows, the product page below has a browser version): python3 -m venv tracer && . tracer/bin/activate pip install "aadc==2.25.2" "aadc-quantlib-tracing==1.41.4" curl -O https://matlogica.com/downloads/swaption.py -O https://matlogica.com/downloads/swaption_trace.py python swaption_trace.py It records one pricing and prints the price, the number of constant sites and the number of branches the run fixed. f.traced_leaves() then gives each constant with its file and line. The Tracer is free for non-commercial and academic use; details are on the product page <https://matlogica.com/products/aadc-tracer/?utm_campaign=crm&utm_source=email&utm_medium=direct&tid=quantlib-users-list-2026-10-23-01> . If you try it on your own pricing code, I'd like to hear what you traced. Thanks again to Luigi for adding the Tracer to the QuantLib extensions page. Best regards, Natalija, MatLogica *About AADC: it turns object-oriented C++ and Python code into optimised machine code at run time, with an adjoint factor below 1, so all first-order Greeks cost less than one extra pricing. An AAD-enabled QuantLib is on PyPI (pip install aadc-quantlib), and AADC also accelerates ORE. Recent recognition: Celent Technology Standout 2027, Global BankTech Awards 2026 winner, and Chartis Category Leader in AAD every year since 2023.* |