🚀 go-pugleaf

W dniu niedziela, 29 grudnia 2019 03:06:34 UTC+1 użytkownik fir napisał:
> i got so terrible problems last half of year i was like unable to concentrate on coding, still not feeling well to be honest but i think maybe i can back for a while 
> 
> the idea is to maybe upbuild (build it up) my assembler (so called organic assembler) so maybe i will show that project and maybe some more specific questions/topics will appear 
> 
> vurrent version to download: 
> 
> http://minddetonator.htw.pl/org-asm.zip 
> 
> (for windows32, i mean it runs as win32 console tool and assembles win32 executables) 
> 
> i added some example pice of asm code in it 
> as sorta example how to use it 
> 
> 
> 
> main: 
> 
>  call test_1M_loop 
>  call test_1M_loop 
>  call test_1M_loop 
>  call test_1M_loop 
>  call test_1M_loop 
>  call test_1M_loop 
>  call test_1M_loop 
>  call test_1M_loop 
>  call test_1M_loop 
>  call test_1M_loop 
>  call test_1M_loop 
>  call test_1M_loop 
>  call test_1M_loop 
>  call test_1M_loop 
> 
> 
>  ret 
> 
> //////////////////////// 
> 
> 
> 
> test_1M_loop: 
>   nop 
>   nop 
>   mov esi 1000000 
>   rdtsc 
>   push edx 
>   push eax 
>  loop: 
> //================== 
> 
>   nop 
>   nop 
>   nop 
> //  mov ebx esi 
> //  add ebx 19 
> //  mov edx 1000 
> //  mov eax 1000 
> //  div edax ebx 
> 
> //================= 
>   dec esi 
>   jne loop 
>   rdtsc 
>   pop ebx 
>   pop ecx 
>   sub eax ebx 
>   sbb edx ecx 
>   call print_eax 
>   ret 
> 
> //////////////////////////// 
> 
> 
> inner_test: 
> 
>   ret 
> 
> ////////////////////////////// 
> 
>   @ note_eax: "\x0d\x0a rdtsc eax = %d \x00" 
> 
> print_eax: 
>   push eax 
>   push note_eax 
>   call msvcrt.printf 
>   pop eax 
>   pop eax 
>   ret 
> 
> 
> 
> this pice of asm makes loop of 1M rolls and 
> measures how many cycles it takes by rtdsc command 
> 
> this itself may be interesting for someone who would like to toy a bit and measure what amount of cycles given assembly comands and routines take on your cpu 
> 
> for example gien loop with 3 nops gives that result (here on my old cpu) 
> 
> 
> 
>  rdtsc eax = 2000103 
> 
>  rdtsc eax = 2000117 
> 
>  rdtsc eax = 2000082 
> 
>  rdtsc eax = 2000082 
> 
>  rdtsc eax = 2000082 
> 
>  rdtsc eax = 2000082 
> 
>  rdtsc eax = 2000082 
> 
>  rdtsc eax = 2000082 
> 
>  rdtsc eax = 2000082 
> 
>  rdtsc eax = 2000082 
> 
>  rdtsc eax = 2000082 
> 
>  rdtsc eax = 2042404 
> 
>  rdtsc eax = 2415266 
> 
>  rdtsc eax = 2266271 
> 
> if i delete that 3 nops it gives 
> 
> 
> 
>  rdtsc eax = 1000125 
> 
>  rdtsc eax = 1000111 
> 
>  rdtsc eax = 1000083 
> 
>  rdtsc eax = 1000076 
> 
>  rdtsc eax = 1000076 
> 
>  rdtsc eax = 1000083 
> 
>  rdtsc eax = 1000083 
> 
>  rdtsc eax = 1000076 
> 
>  rdtsc eax = 1000076 
> 
>  rdtsc eax = 1000076 
> 
>  rdtsc eax = 1000076 
> 
>  rdtsc eax = 1000076 
> 
>  rdtsc eax = 1000076 
> 
>  rdtsc eax = 1000076 
> 
> so it shows that 1 nop takes only like 0.33 cycle, and that dec esi and jne jump itself take only one cycle 
> 
> if someone is interested how fast things execute that kind of tests may be usefull 
> (it could be done in more popular assemblers but the entry level of my own may be easier, for me its easier for sure) 
> 
> things can be tested easily, for example if i want to test rand() i just put a line 
> 
>   call msvcrt.rand 
> 
> instead of that 3 nops and i get the result 
> 
> 
> 
>  rdtsc eax = 28016989 
> 
>  rdtsc eax = 28000112 
> 
>  rdtsc eax = 28043897 
> 
>  rdtsc eax = 28001589 
> 
>  rdtsc eax = 28001498 
> 
>  rdtsc eax = 28105077 
> 
>  rdtsc eax = 28000161 
> 
>  rdtsc eax = 28001904 
> 
>  rdtsc eax = 28016415 
> 
>  rdtsc eax = 28024360 
> 
>  rdtsc eax = 28000084 
> 
>  rdtsc eax = 28012999 
> 
>  rdtsc eax = 28026061 
> 
>  rdtsc eax = 28002849 
> 
> so its 27 cycles..and so on 
> 
> 
> if someone would like to toy with my assembler i could improve it (i not typed all possiblemnemonic register combinations yet as it seems to be posiibly thousands of them) 
> 
> i need yet also make some improve,ments in syntax of that asm... im quite satisfied as to cyntax of code (which i a bit refreshed if compared to intel by rejecting periods and turning square brackets into round brackets but i quess i still need to work on  data format and also repair some holes here

http://minddetonator.htw.pl/org-asm-4.zip 

i added some extremally lame asm code for drawing mandelbrot (temporarely becouse of laziness) 

it may illustrate how you do float assembler today (you simply add ss to mnemonics and have 8 registers) 

though honestly it is visible i need to rework some things in that assembly, like turning xmm1 into x1 to be shorter etc 

but im tired and will do it later 

// 
// unsigned background_color = 0x0; 
// 
// void RunFrame(int advance) 
// { 
//     ClearFrameData(background_color); 
// } 
// 
// 
// void ProcessMouseMove(int x, int y){} 
// void ProcessKeyDown(int key){} 
// void OnResize(){} 
// 
// 
// int main(void) 
// { 
// 
//         RegisterMouseMove( ProcessMouseMove ); 
//         RegisterKeyDown( ProcessKeyDown ); 
//         RegisterOnResize( OnResize ); 
//         RegisterRunFrame( RunFrame ); 
//         SetSleepValue(10); 
// 
//         SetupWindow2(" Example Green Fire App ",  20, 20,  560, 400 ); 
// 
//         return 0; 
// } 


jmp start 

ProcessMouseMove:   ret 
ProcessKeyDown:     ret 
OnResize:           ret 

////////////////////////////////////// 



calc_mandelbtot_for_esi_edi: 


           >> minus2: -2.0f four: 4.0 dx: 0.01 dy: 0.01 
           >>     c_re: 0.0 c_im: 0.0 res: 0 

 push eax 


  cvt si2ss xmm0 esi 

  mov ss xmm1 (dx) 
  mul ss xmm0 xmm1 
  add ss xmm0 (minus2) 
  mov ss (c_re) xmm0 

  cvt si2ss xmm0 edi 

  mov ss xmm1 (dy) 
  mul ss xmm0 xmm1 
  add ss xmm0 (minus2) 
  mov ss (c_im) xmm0 


  ////////////////////// 
   mov ecx 0 


   mov ss xmm3 (c_re) 
   mov ss xmm4 (c_im) 

   mov ss xmm0 xmm3 
   mov ss xmm1 xmm4 

   mov ss xmm2 (four) 

 iterations: 
                     //re 
   mov ss xmm5 xmm0 
                    //im 
   mov ss xmm6 xmm1 

                     //re * re 
   mul ss xmm0 xmm0 
                    //im * im 
   mul ss xmm1 xmm1 
                   // re*re + im*im 
   add ss xmm0 xmm1 

   mov ss xmm1 xmm0 

   cmple ss xmm1 xmm2 

   cvt ss2si eax xmm1 

   cmp eax 0 
   je out 

   mov ss xmm0 xmm5 
   mov ss xmm1 xmm6 
                     //re * re 
   mul ss xmm0 xmm0 
                     //im * im 
   mul ss xmm1 xmm1 
   sub ss xmm0 xmm1 
   add ss xmm0 xmm3 
                    // re_n = re * re - im * im + cRe; 
   mov ss xmm7 xmm0 

   mov ss  xmm0 xmm5 
   mul ss  xmm0 xmm6 
   add ss  xmm0 xmm0 
   add ss  xmm0 xmm4 

   mov ss  xmm1 xmm0 
   mov ss  xmm0 xmm7 


// 
    inc ecx 
    cmp ecx 100 
    jl iterations 

    mov ecx 0xff00ff 
    pop eax 
    ret 
out: 

   mov ecx 0xffff00 
   pop eax 
   ret 

//    double re = cRe; 
//   double im = cIm; 
// 
//    for(int n=1; n<man_maxiter; n++) 
//    { 
//      points_list[n].re = re; 
//      points_list[n].im = im; 
// 
//      if( re * re + im * im > 4.0 ) return n; 
// 
//      double re_n = re * re - im * im + cRe; 
//      double im_n = 2 * re * im       + cIm; 
// 
//      re = re_n; 
//      im = im_n; 
// 
//    } 






RunFrame: 

   push 0x007000,    call green_fire.ClearFrameData,        pop eax ;


/////////////////// loop on pixels 
   mov eax green_fire.frame_bitmap 
   mov ebp (eax) 
   mov eax green_fire.frame_size_x 
   mov ebx (eax) 

   mov edi 0 
loop_y: 
   mov esi 0 
loop_x: 
   mov eax edi,  mul eax ebx, add eax esi, shl eax 2, add eax ebp 
   call calc_mandelbtot_for_esi_edi 
   mov (eax) ecx 
   inc esi ,   cmp esi 400,   jl loop_x 
   inc edi ,   cmp edi 400,   jl loop_y 
////////// 


   ret 


start: 

    push ProcessMouseMove,  call green_fire.RegisterMouseMove,   pop eax 
    push ProcessKeyDown,    call green_fire.RegisterKeyDown,     pop eax 
    push OnResize,          call green_fire.RegisterOnResize,    pop eax 
    push RunFrame,          call green_fire.RegisterRunFrame,    pop eax 
    push 10,                call green_fire.SetSleepValue,       pop eax 

               >> title: "exampla app" 

    push 420, push 500, 
       push 20, push 20, 
          push title, call green_fire.SetupWindow2 
           add esp 20 

    mov eax 0 
    ret