Recently, I have become interested in x86 assembly, and I wanted to try out my skills on CodeForces using __asm__ extension in GCC. For example, I was very pleased when I solved problem 495B using a very naive algorithm implemented in assembly which would probably have TLEd if it was written in regular C (basically, looped all the way up to $$$\frac{n}{4}$$$ to find all divisors of $$$n$$$ for $$$n \leq 10^9$$$ rather than using $$$O(\sqrt{n})$$$ algorithm).

However, this algorithm was implemented by using only 32-bit numbers and registers, so I was curious as to how the performance would be affected by using 64-bit numbers and registers. Therefore, I tried to use the %rax, %rbx, etc. registers instead of the %eax, %ebx, etc. registers, but for some reason, this gave me a "bad register" error. I then tried running my code using Clang++17 Diagnostics and this gave me a much clearer error:

p71.cpp:24:13: error: register %rax is only available in 64-bit mode
    __asm__("push %%rax \n\t"

Of course, this implies that CodeForces is compiling our programs using 32-bit mode, so using 64-bit integers is even slower than it would normally be, because GCC/G++/Clang++ has to manipulate multiple registers in order to implement 64-bit operations in 32-bit x86. The fact that CodeForces uses 32-bit mode surprised me very much, since 64-bit has been the standard for computers and even smartphones for years. In fact, Ubuntu recently decided to drop its support for 32-bit computers in its upcoming versions, showing how old 32-bit really is.

Given all this, why does CodeForces continue to use GCC/G++/Clang++ in 32-bit version? Is this to penalize programmers from using 64-bit integers when it's not necessary, or do they use 32-bit mode for a more technical reason? I am not sure using 32-bit mode makes a huge difference since I find most problems can be solved using only 32-bit numbers anyway, but I am very curious as to why CodeForces made the decision to choose 32-bit over 64-bit.