makeThe current status of this entry is:
STATUS: INABIAF - please **DO NOT** fixFor more detailed information see 2015 hou bugs.
echo something | ./prog
./prog < file./try.shA good start for understanding this entry might be rfc1321.txt. How does it compute this using only a single looping construct? The original reference source code for this algorithm is about 355 lines of code, yet this entry is only 199 lines with one statement per line and it isn’t using lots of macro expansion tricks like the reference code!
Some things that make you go, “huh?”
./prog < prog.c | cut -c-5
./prog < Makefile | cut -c-4How are those values magically returned?
What happens with a large file? Something over 256K.
./prog < large_fileWhere does that message come from?
This looks like a normal C program.
Almost everything is a double.
There are tell-tale signs like pow(2,(?/12)) and exp(-?*?) (names redacted).
You can also find a bunch of cosines and polynomial evaluations.
It looks like a ????? ???????????, right?
Run and find out.
Some poor scripting folks only have access to doubles (Jav?script, L?a, BA?IC, etc.).
As a C programmer, I feel obliged to show a gesture of sympathy, to experience their pain and provide an answer.
Except when absolutely necessary (printf, int main(), array indexing), there
isn’t a single integer. Even getchar() and EOF are immediately converted.
Since bool counts as an integer type too, there is no boolean either. No
comparison, no if, no ?:, no || &&. Everything is done in a single loop
that terminates on printf(3). Straight up arithmetic.
floor(3) and ceil(3) are technically double-to-double functions, but they feel
like cheating. Yes cos(3), no floor(3).
Hard requirements:
The platform must implement the double type as IEEE754-compliant 64-bit
floating point numbers. The 80-bit intermediate format used by x87 is
considered as an violation of this. The code should print an error message on
such platforms.
The program must start with the CPU / FPU in round-to-nearest mode.
Soft requirements:
The compiler must respect volatile. The code is formatted to warn about
that, though.
The printed result is only correct on little-endian machines. The program takes care to warn about this issue after printing an incorrect big-endian result. Error messages become garbled, though.
Tested platforms: