56m
comment The Archimedean place of $\mathbb{Q}$
It's worth noting that the nonnegative rational numbers are precisely the sums of squares, so in this case the ordering can be recovered from algebraic properties. (also, the nonnegative elements of $\mathbb{R}$ are precisely the squares)
8h
comment Has there ever been an application of dividing by zero?
@jpmc26: Sure; in this setting, $1/x$ is continuous at $x=0$ so you can always replace it with a limit (and vice versa), but again, why go through all the trouble with limits when you can just do arithmetic?
1d
comment Has there ever been an application of dividing by zero?
@jpmc26: Why would you go through all the trouble with limits when you can just do arithmetic?
Apr
29
comment Why does the enhanced GCC 6 optimizer break practical C++ code?
@Ben: Do you know if there has ever been discussion about changing the standard to make dereferencing a null pointer to call a non-virtual non-static member function a well-defined thing?
Apr
29
comment Why does the enhanced GCC 6 optimizer break practical C++ code?
@Ben: That's right -- you can continue using legacy compilers (or even modern compilers with the right flags!) to compile your ancient nonconforming code. Thus, you offer no good reason why things should be ruined for those of us who are writing standard-conforming code.
Apr
29
comment Why does the enhanced GCC 6 optimizer break practical C++ code?
@Ben: If you get to a point where this is null, it's too late: you've already done something wrong -- in the examples people have given, it's because you've already dereferenced a null pointer, and for whatever reason, the compiler decided to turn that into a method call with a null this rather than have some other effect.
Apr
28
comment Why does the enhanced GCC 6 optimizer break practical C++ code?
@jcast: I'd really prefer that the C++ standard wasn't written by google. To prevent that, we have to hold google accountable for writing bad code, rather than shift the responsibility onto compiler vendors and the standards committee to accommodate google's errors.
Apr
26
comment "Spatial (geometrical)" realization of Elementary topos?
My instincts are that doing so wouldn't lead towards getting 'more geometry', but instead doing geometry in a more restrictive logical setting.
Apr
15
awarded Necromancer
Apr
15
comment Were Michelson-Morley results a surprise?
@Marco: Historical accuracy aside, they had this line of thought along these lines: "Maxwell's equations require Lorentz invariance. Lorentz invariance is wrong. Thus we expect to see violations of Maxwell's equations".
Apr
11
awarded Yearling
Apr
11
awarded Yearling
Apr
9
awarded Yearling
Apr
9
awarded Yearling
Apr
5
comment Why do C++ optimizers have problems with these temporary variables or rather why `v[]` should be avoided in tight loops?
@Mooing: I don't believe the standard guarantees correct behavior -- any enterprising programmer could do horribly unsafe but legal things to modify the internals so that the class invariants aren't maintained. To correctly assume safety, the compiler would actually have to know that every non-inline function (e.g. all of the startup code in libstdc++ and the iostream implementation) that got called prior to the loop running is well-behaved.
Feb
20
awarded Yearling
Feb
20
awarded Yearling
Feb
17
awarded Yearling
Feb
17
awarded Yearling
Feb
11
awarded Notable Question
1 2 3 4 5