I want to practice with assembler inserts. I wanted to write code to calculate the sum of the numbers from 1 to n, implementing it in assembly language. However, he has not yet been able to compile even simple code that writes the value n to sum. Here is the code:

#include <stdio.h> int sumN(int n) { // calculate 1+2+...+n int sum; __asm__( "mov %1, %eax\n\t" "mov %eax, %0\n\t" :"=d"(sum) :"c"(n) :"%eax"); return sum; } int main() { for (int i = 1; i < 100; ++i) { printf("n: %d sum(n): %d\n",i,sumN(i) ); } return 0; }

However, the g ++ compiler produces an error.

main.cpp: 11: 11: error: invalid 'asm': operand number missing after% -letter: "% eax");
^
main.cpp: 11: 11: error: invalid 'asm': operand number missing after% -letter.

I thought that the problem is that I did not specify all the registers that I use, so I rewrote the code a bit:

 #include <stdio.h> int sumN(int n) { // calculate 1+2+...+n int sum; __asm__( "mov %1, %eax\n\t" "mov %eax, %0\n\t" :"=d"(sum) :"c"(n) :"%eax","%edx","%ecx"); return sum; } int main() { for (int i = 1; i < 100; ++i) { printf("n: %d sum(n): %d\n",i,sumN(i) ); } return 0; }

However, such code is not compiled, gcc gives an error

main.cpp: 11:25: error: 'asm' operand has impossible constraints
: "% eax", "% edx", "% ecx");

Tell me, what is the error and how to fix it? PS was guided by this article

  • The article was a switch to the Intel syntax - Evgeny Shmidt

1 answer 1

In the assembler code itself, the % sign must be escaped, i.e. write %%eax , but on the clobber list, on the contrary, the % sign is optional.

 int sumN(int n) { // calculate 1+2+...+n int sum; __asm__( "mov %1, %%eax\n\t" "mov %%eax, %0\n\t" :"=d"(sum) :"c"(n) :"eax"); return sum; } ... $ gcc test.c -m32 -o test $ ./test n: 1 sum(n): 1 ... n: 99 sum(n): 99

By the way, keep in mind that about compilations under x86-64, writing to eax clears the high-order bits of rax. This will need to be reflected in the clobber list.