Can someone suggest me any ideas to start with this project?

Hi Do you have any idea how to show the binary form of a double?

You can use an union: // in C you have to put typedef, in C++ is not necesary typedef union { double d; uint8_t bytes[sizeof(double)]; } double_bytes_t; // ... double_bytes_t x; x.d = 2.3; // supposing little endian: for(int i = sizeof(double)-1; i >= 0; i--) { printf("%02X", x.bytes[i]); // HEX format of binary for IEEE754 64-bits }

You can use an union: // in C you have to put typedef, in C++ is not necesary typedef union { double d; uint8_t bytes[sizeof(double)]; } double_bytes_t; // ... double_bytes_t x; x.d = 2.3; // supposing little endian: for(int i = sizeof(double)-1; i >= 0; i--) { printf("%02X", x.bytes[i]); // HEX format of binary for IEEE754 64-bits }

I think that the correct way is to use std::bit_set. I can not give you an example right now, but you can google it

You probably can, but it is an undefined behavior, you can read about it here: https://en.cppreference.com/w/cpp/language/union

You probably can, but it is an undefined behavior, you can read about it here: https://en.cppreference.com/w/cpp/language/union

You meant to say std::bit_cast.

Aha, i see, but in C with most of the compilers i assure you that this works just fine... The only thing is knowing if the hardware system is little endian or big endian... but most are little endian.

No, I meant to say std::bitset, not std::bit_set :) But yes, probably std::bit_cast will be needed

It is not Undefined Behavior in C. In C++ it is UB.

No, I meant to say std::bitset, not std::bit_set :) But yes, probably std::bit_cast will be needed

Aha, i see, but in C with most of the compilers i assure you that this works just fine... The only thing is knowing if the hardware system is little endian or big endian... but most are little endian.

You must ensure C code will be compiled. Elsewise the compiler you choose will make asumptions for you

This union example i showed will work always in every C compiler... try it yourself hehe

You can use an union: // in C you have to put typedef, in C++ is not necesary typedef union { double d; uint8_t bytes[sizeof(double)]; } double_bytes_t; // ... double_bytes_t x; x.d = 2.3; // supposing little endian: for(int i = sizeof(double)-1; i >= 0; i--) { printf("%02X", x.bytes[i]); // HEX format of binary for IEEE754 64-bits }

C++ support std::optional Instead of union

std::optional is not a replacement for union. std::variant is.

std::optional is not a replacement for union. std::variant is.

I thought std:: optional is for handling not critical exceptions to "substitute" try catch

I thought std:: optional is for handling not critical exceptions to "substitute" try catch

Or maybe suitable for case of returning pointer which can be nullptr?

I've recently stumbled upon the article with a somewhat similiar idea, and actually I wanted to share godbolt code, taken from the same article, where some form of vectorization is achieved: shorturl.at/xFGT7 https://siboehm.com/articles/22/Fast-MMM-on-CPU You mentioned spatial locality: can I increase it with a contiguous pool of memory in order to make the algorithm more cache friendly in this regard? I can't use stack for the pool because there will be enormous amounts of data.

No. You can't effectively increase spatial locality in matrix multiplication. The way to increase spatial locality is not to have the 2nd matrix in row major form. It must be converted to a column major form and then the matrix multiplication needs to be done in an inverted sense. This will be easily vectorized by the compiler. The reason why spatial locality is not there in the first case is because you are picking elements off a column in the 2nd matrix in regular matrix multiplication and these elements are far away in memory and not more than 1 or 2 elements can fit in the same cache line depending on the size of the row in the 2nd matrix and also on the cache line size.Replacing the 2nd matrix by its transpose addresses this issue(computing transpose is a n^2 operation by itself and order n^2 space complexity). Now you will be able to compute dot products across rows of both the matrices which is a cache friendly operation. And this can be auto vectorized by the compiler. If you want to avoid computing the transpose (which you would if your matrix is huge and you want to avoid the space overheads), you must vectorize the code yourself as the compiler won't be of much help like I said earlier.

I made a linked list and now I want to pop() all of its nodes but something is keeping my mind busy!!! When my linked list has got one node I must delete (Head) which is an attribute in my class , but it does not delete like others normal pointers !!! why??

What is the difference between iostream and studio.h?

And do we use all of them?

Who can help me with <cmath> ?

This is sooo vague question, can you get directly to the point? dontasktoask.com

There are many and yes, all og them have a valid use case

vector calls deallocation that would call delete on unique_ptr, so you should be safe :)

Hello! I have a question about unique_ptr. I want to create an array of pointers of a superclass, so then with the magic of polymorphism, I insert objects of type of the subclass (obviously dynamically allocated). What I wanted to ask is: do these two lines of code ALL handle the new and delete issue by THEM? (so I don't have a problem, I know that if an element is removed from the vector, the space in heap memory is deleted by itself). vector<std::unique_ptr<BaseClass>> my_vector; my_vector.emplace_back(std::make_unique<SubClass>(parameters, ...));

I got also questions about unique ptr. Is it possible to pass list of unique ptrs to constructor without implementing your own list derived from std::list? I mean if you pass list of unique ptrs to constructor and then assign it to some variable it will failed but maybe I did somethings wrong (I used std::move)

I am not sure about this one, it seems like it should be possible but i cannot give certain answer (and i don't have time to research it rn)

I got also questions about unique ptr. Is it possible to pass list of unique ptrs to constructor without implementing your own list derived from std::list? I mean if you pass list of unique ptrs to constructor and then assign it to some variable it will failed but maybe I did somethings wrong (I used std::move)

This will work only if the destructor in BaseClass is declared virtual btw.

Your question is not clear. Showing the code you wrote would help us understand the issue

https://godbolt.org/z/7KaodE1hr

https://godbolt.org/z/7KaodE1hr

Why?

#include <stdio.h> int main() { int i=10,j; j= ++i + ++i; printf("\n%d",j); return 0; }

++i return reference to i in this case we got this: 12 + 12

Logically op=23 I got=24

Sir output is 23

24 is the crt output i am looking for the reason

Re run

Sir output is 23

24 is the crt output i am looking for the reason

i = 10 ++i i = 11 ++i i = 12 i+i = 12+12 = 24

11+12?

you have j = (++i) + (++i)

That what i said

i + i

11+12

i = 10 ++i i = 11 ++i i = 12 i+i = 12+12 = 24

please share you test code

#include <stdio.h> int main() { int i=1,j; j= ++i + ++i + ++i; printf("\n%d",j); return 0; }

23 [Program finished]

9?