ATTN: Lojban
Re: Credit Summary (Score Standings Q4)
Date: 10-22-2014 08:41:20

And all vegetables don't go in the pan at the same time. Onions and carrots take longer to cook and are first. Celery goes in before mushrooms and tomatoes. Then greens, if any. Garlic goes in very late. It doesn't stand overcooking. Happy Birthday.

This notification is meant to inform your of possible changes to your score.

Your Score Summary (Viewable Now)

--
Please, continue to check your reports periodically, especially for the first year after you discover any issues, to make sure no negative activity has occurred.

Periodic Changes for Lojban - 4388132

to modify your status with Hill Media
write to 22 Bond St. Reading MA 01867-2432 or continue to here.

No it is not. The compiler should expect functions to have some side effects. Some compilers have language extensions for "pure constant" functions. You could ask your compiler to show the assembler code (e.g.g++ -O2 -S -fverbose-asm your-code.ccwithGCC...)Basile StarynkevitchOct 4 at 6:43

"which guarantees different result than original code (zero vs. greater than zero time reported)"-- It doesn't guarantee that, as far as the standard is concerned, since the standard makes no mention of how long any particular operation should take (aside from calls tosleepfunctions). The call toveryLongComputation()could very well be instantaneous.Benjamin LindleyOct 4 at 7:25

@BenjaminLindleyThe call to veryLongComputation() could very well be instantaneous- I disagree. There is great number of algorithms, which given sufficiently large data are guaranteed not to complete before the end of solar system using any of computing hardware known at the time C++ standard was written.Paul JurczakOct 4 at 7:32

@Paul what qualifies as I/O is the sending-to-cout part, not the computing-r part. What you are seeing is perfectly compliant. Whatever non-negative difference you see between the clock calls, possibly small enough to be rounded to 0, is fine. Imagine a program where the difference between the 2 calls to clock is such that the you get 0 or 1 with roughly 50% probability, depending on when you start and what else is running. Should optimizations preserve this exact probability?Marc GlisseOct 4 at 17:16

How could you tell the difference between the two versions from the standpoint of the abstract machine? There is no guarantee whatsoever how long a computation will take. (Practical example: the compiler could fully evaluate your computation.). This reordering would be invalid if your computation performed a side-effect with a guaranteed minimum duration such as a sleep. An IO is not enough because, again, you cannot tell the difference between the reordering and a really fast IO. The as-if rule applies.usrOct 4 at 19:55