156. WG21 October mailing (cont.), fast math

Media

Video

Podcast

WG21 October mailing (cont.), fast math

Mailing - 80 papers, 41 new!

New papers

Follow-up C++ is the next C++
- Reddit thread exploded a bit – curiously, it looks like most commenters didn’t actually read the paper but are responding to the catchy title.
  - Deep in the thread, an interesting comment about Carbon
P2632R0. A plan for better template meta programming facilities in C++26. Big plans. Separate proposals for each planned feature. Uses Circle by Sean Baxter as an implementation example.
- P2662R0. Pack indexing
- P2671R0. Syntax Choices for Generalized Pack Declaration and Usage
P2644R0. Get Fix of Broken Range-based for Loop Finally Done, by Nicolai Josuttis. A more general fix is not coming, it seems. But we have P2012. Maybe it is good enough?
P2656R0. C++ Ecosystem International Standard. Interoperable tools, what a concept.
P2658R0. Temporary storage class specifiers. More dialects?
P2666R0. Last use optimization
P2668R0. Role based parameter passing. Parameter attributes like in Cpp2, or automatic parameter handling like in Carbon.
- D0708. Parameter passing, by Herb Sutter
P2670R0. Non-transient constexpr allocation, by Barry Revzin.
P2672R0. Exploring the Design Space for a Pipeline Operator, by Barry Revzin. The pipeline operator |> lives (maybe).
P2676R0. The Val Object Model. Food for thought for a future C++. (David Sankel is at Adobe now, huh.)
P2677R0. Reconsidering concepts in-place syntax. Bring back auto{T}!
P2688R0. Pattern Matching Discussion for Kona 2022. Can we have pattern matching in C++26, pretty please? Also, match is nicer than inspect, IMHO.

Contracts!!

P2659R0. A Proposal to Publish a Technical Specification for Contracts, by Brian Bi, Bloomberg.
P2660R0. Proposed Contracts TS. P2661R0. Amendments. Uses attribute-based contract syntax.
P2680R0. Contracts for C++: Prioritizing Safety, by Gabriel Dos Reis

Fast math flag in libraries

Brendan Dolan-Gavitt wrote an article called Someone’s Been Messing With My Subnormals!, in which he describes an intrusive nature of the GCC/Clang compiler switch -ffast-math.

TL;DR: After noticing an annoying warning, I went on an absurd yak shave, and discovered that because of a tiny handful of Python packages built with an appealing-sounding but dangerous compiler option, more than 2,500 Python packages—some with more than a million downloads per month—could end up causing any program that uses them to compute incorrect numerical results.

The discovery:

It turns out (somewhat insanely) that when -ffast-math is enabled, the compiler will link in a constructor that sets the FTZ/DAZ flags whenever the library is loaded — even on shared libraries, which means that any application that loads that library will have its floating point behavior changed for the whole process. And -Ofast, which sounds appealingly like a “make my program go fast” flag, automatically enables -ffast-math, so some projects may unwittingly turn it on without realizing the implications.

Also, apparently, if you use -Ofast, -fno-fast-math does NOT disable the auto-enabled -ffast-math! WHAAAAT

The author wrote a Python script (Gist) to check if a library was compiled with -ffast-math.

Pretty crazy. Be careful with your math compiler options.