Why is assembly code emitted by "zig cc -S" much larger than with gcc?

Question

I have some very basic C code for which I want to view the equivalent assembly instructions. I'm interested to use zig cc because it can cross compile to M1 ARM, X86, etc. Here is my code:

void variable_sort_2(int length, int *a) {
  if(length == 2){
    if(a[0] > a[1]){
      int temp = a[0];
      a[0] = a[1];
      a[1] = temp;
    }
  }
}
static int a0[] = {};
int main(void){
  variable_sort_2(0, a0);
  return 0;
}

This will successfully do what I want using gcc:

gcc -S sort2.c -o sort2.s

According to gcc --help that will "Only run preprocess and compilation steps."

How to do the equivalent with zig cc?

I tried doing the same -S operation with zig cc but it gives me a much larger output file:

$ zig cc -S sort2.c -o sort2.zig_cc.s
$ gcc -S sort2.c -o sort2.gcc.s
$ ls -l sort2*
-rw-r--r--  1 mee  wheel    236 Jun 13 12:32 sort2.c
-rw-r--r--  1 mee  wheel  27646 Jun 13 16:32 sort2.zig_cc.s
-rw-r--r--  1 mee  wheel   1455 Jun 13 16:33 sort2.gcc.s
$ wc -l *s
  63 sort2.gcc.s
 688 sort2.zig_cc.s

I don't understand assembly well enough to explain why the zig cc file is ten times longer; my assumption is it's linking in system stuff, but I don't know. I see lots of lines like this:

    .byte   17                              ## DW_TAG_compile_unit
    .byte   1                               ## DW_CHILDREN_yes
    .byte   37                              ## DW_AT_producer
    .byte   14                              ## DW_FORM_strp
    .byte   19                              ## DW_AT_language
    .byte   5                               ## DW_FORM_data2

That's lines 300 through 305 of the zig cc -S compiled output on my computer:

$ uname -a
Darwin meee 22.2.0 Darwin Kernel Version 22.2.0: Fri Nov 11 02:08:47 PST 2022; root:xnu-8792.61.2~4/RELEASE_X86_64 x86_64

I created 3 github gists to contain the C code and the two assembly files: sort2.c, sort2.zig_cc.s, sort2.gcc.s:

• https://gist.github.com/purplejacket/4f5678f985e8a1154b3e9e3116ce6a64
• https://gist.github.com/purplejacket/5a12a602e7a26fbbcbd2493ef69d0a0b
• https://gist.github.com/purplejacket/65b63cb7ef81ae1ecb8413eb5903ea45

@ShadowRanger suggested trying -g0 which makes the file smaller, but still not as small as I get with gcc.

$ zig cc -S -g0 sort2.c -o sort2.g0.s
$ ls -l sort2.g0.s
-rw-r--r--  1 mee  wheel  5042 Jun 14 17:40 sort2.g0.s
$ wc -l sort2.g0.s 
     239 sort2.g0.s

And here is a gist for that output file:

• https://gist.github.com/purplejacket/df7a119d7a067a2f64d5a1e791b52f9e

Answer 1

Looks like your zig cc assembly output has DWARF debugging information in it, which is what is bloating the file size significantly; gcc isn't producing said debug info by default.

I can't find good docs on what you can use to turn off DWARF debug generation for zig cc (other forms of using zig accept -fstrip to avoid generating it; you might try that?), but in short, nothing is wrong, zig cc is just more verbose by default.