Re: C23 thoughts and opinions

On 26/05/2024 17:35, Michael S wrote:

#include <stdio.h>
#include <stdlib.h>
#include <ctype.h>
 int main(int argz, char** argv)
{
   if (argz > 1) {
     FILE* fp = fopen(argv[1], "wb");
     if (fp) {
       char buf[2048];
       _Bool look_for_comma = 0;
       for (;;) {
         if (fgets(buf, sizeof(buf), stdin) != buf)
           break;
          char* p = buf;
         for (;;) {
           char c = *p;
           if (isgraph(c)) {
             if (look_for_comma) {
               if (c == ',') {
                 look_for_comma = 0;
                 ++p;
               } else {
                 goto done;
               }
             } else {
               char* endp;
               long val = strtol(p, &endp, 0);
               if (endp==p) // not a number
                 goto done;
               fputc((unsigned char)val, fp);
               p = endp;
               look_for_comma = 1;
             }
           } else {
             if (c == 0)
               break; // end of line
             ++p; // skip space or control character
           }
         }
       }
       done:
       fclose(fp);
     } else {
       perror(argv[1]);
       return 1;
     }
   }
   return 0;
}

I tried this on my 600MB data like this:
   C:\c>c fred.exe <data
   C:\c>fred --version
   clang version 18.1.0rc
   Target: x86_64-pc-windows-msvc
   Thread model: posix
   InstalledDir: C:\c
Since those bytes represent the contents of the clang compiler, I was able to run it afterwards.
All versions across compilers/optimise levels seemed to give a constant time of 17-18 seconds. This is good compared with my initial 144 seconds (most compilers failed; you reported a similar test took several minutes).
However, what's involved with a compiler is much elaborate than such a program. There's syntax, type-checking, code-generation...
Still, I reported earlier an experimental change to my non-C compiler, which translated this same input to a program with that embedded binary (not just the binary itself) in under 6 seconds.
That's three times as fast as the above result:
   C:\mapps>tm \mx2\mm -ext test2          # tm is  timing tool
   Compiling test2.m to test2.exe
   TM: 5.86                                # (timings vary)
   C:\mapps>test2
   data is 119571969 bytes
   C:\mapps>type test2.m
    []byte data = (
        include "data"
    0)
    proc main=
        fprintln "data is # bytes", data.len
    end