The most significant place (and probably only one where it makes sense) where assembly use in C/C++ is used is in embedded systems development, where it is the only way to access some peripherals or where it is required to do so because we need this exact order of instructions no matter what the compilation flags are. This last case is most often seen in bootloaders or critical OS parts like the context switch.
The example that comes to mind is accessing the special registers in a MicroBlaze processor: even Xilinx recommends accessing those registers using assembly (there is also no other way that I know of) because there is no other way.
By embedding it in C, it would be used like this:
unsigned long get_msr(void) {
unsigned long msr;
asm ("mfs %0, rmsr" : "=d" (msr)); // note: this is a gnu extention
return msr;
}
Another example is the implementation for a context switch provided by ARM also uses assembly (although not embedded in C code in this case).
That's GNU C inline asm, where you can write input / output constraints to tell the compiler which registers are inputs and outputs. The OP is asking about MSVC inline asm, which is so bad that MSVC dropped it for x64 (brittle implementation in MSVC, and inefficient syntax). The could do this because they have intrinsics for most things that you'd use GNU C inline asm for, even stuff like x86
cli/stito disable / enable interrupts in kernel code.