@programminginc - страница 1086

Ludovic 'Archivist'

I think it does shit when it comes to cache management

Anonymous

Can you provide a example code?

Ludovic 'Archivist'

and that those 60ns are some shitty cache L3 lookups

Ludovic 'Archivist'

Can you provide a example code?

That would be quite hard as the code is not that small and kinda bundled in lots more shit

Ludovic 'Archivist'

and also under NDA

Ludovic 'Archivist'

but well

Ludovic 'Archivist'

it makes me rage

Ludovic 'Archivist'

out of the hotpath, performance is not horrible

Anonymous

Oh ok, I'll try it myself. I already heard of it

Ludovic 'Archivist'

but the hotpath is littered with stuff that should have been optimized away

Ludovic 'Archivist'

It reminds me of that time i could not compile in -O3 in G++ because it changed some initialization orders and borked the hell out of my memory

Ludovic 'Archivist'

had to add some pragmas to put some 3 lines in -O2 when i was compiling in O3 or more

Ludovic 'Archivist'

Conclusion, sometimes compilers do unexpected shit

Ludovic 'Archivist'

Deterministic unexpected shit that is

Anonymous

Yeah, the problem with g++ still exists. If your code is written in a manner which the compiler don't like it gets a memory and performance hell on O3

Ludovic 'Archivist'

Yeah, the problem with g++ still exists. If your code is written in a manner which the compiler don't like it gets a memory and performance hell on O3

Well, in my case it omitted initialization of something and borked the hell out of it

Ludovic 'Archivist'

It was a very weird bug

Anonymous

Lel

Ludovic 'Archivist'

in O2 valgrind was like: "perfect dude" then on 03 it was like "what the fuck dude? don't you ever initialize stuff?"

Ludovic 'Archivist'

Short long story two instruction were reversed to optimize a potential cache issue (which existed, but i prefer a very consistent and regular cache issue to a crash)

Ludovic 'Archivist'

in short i have at that place a single instruction that L1 cache misses every time and pumps 5ns in

Ludovic 'Archivist'

kinda infuriating but I don't want to add some assembly shit nor an intel specific _mm_prefetch for now

#39424E

Why here assert - "list iterator not dereferencable": int main() { list<string> users; list<string>::iterator iter; users.push_front("one"); users.push_front("two"); string temp; for (iter = users.begin(); iter != users.end(); iter++); { temp = *iter; } return 0; }

olli

Why here assert - "list iterator not dereferencable": int main() { list<string> users; list<string>::iterator iter; users.push_front("one"); users.push_front("two"); string temp; for (iter = users.begin(); iter != users.end(); iter++); { temp = *iter; } return 0; }

you have a semicolon ; after the for loop parenthesis, so the for loop has an empty body. After the loop, inside your {} block you dereference iter, which point past the last element Tip: - variables should use the smallest scope, since iter is only used inside the loop, delcare it as part of the loop. Mistakes like this one would result in a compile error instead of "strange" runtime behavior

#39424E

you have a semicolon ; after the for loop parenthesis, so the for loop has an empty body. After the loop, inside your {} block you dereference iter, which point past the last element Tip: - variables should use the smallest scope, since iter is only used inside the loop, delcare it as part of the loop. Mistakes like this one would result in a compile error instead of "strange" runtime behavior

Thanks!

Anonymous

#include<iostream> using namespace std; class person{ public : void introduce (){ cout <<"hi from person"<<endl; } }; class student : public person{ public : void introduce (){ cout <<"hi from student"<<endl; } }; class Farmer : public person{ public : void introduce (){ cout <<"hi from farmer"<<endl; } }; void whosThis(person &p){ p.introduce(); } int main() { Farmer anil; student alex; whosThis(alex); return 0; }

Anonymous

Only the main class is calling

Anonymous

Base class is calling but derived class is not calling

Anonymous

Why