format ELF64 executable 3 entry _start ;------------------------------------------------------------------------------- ; NAME: XORWOW ; DESC: Pseudo random number generator. ; OUT: eax [0;2^32-1] ;------------------------------------------------------------------------------- macro XORWOW { mov edx,[g_xorwow_x] ; edx = x shr edx,2 ; edx = x >> 2 xor edx,[g_xorwow_x] ; t = x ^ (x >> 2) mov eax,[g_xorwow_y] ; eax = y mov [g_xorwow_x],eax ; x = y mov eax,[g_xorwow_z] ; eax = z mov [g_xorwow_y],eax ; y = z mov eax,[g_xorwow_w] ; eax = w mov [g_xorwow_z],eax ; z = w mov eax,[g_xorwow_v] ; eax = v mov [g_xorwow_w],eax ; w = v mov edi,eax ; edi = v shl edi,4 ; edi = v << 4 xor edi,eax ; edi = (v ^ (v << 4)) mov eax,edx ; eax = t shl eax,1 ; eax = t << 1 xor eax,edx ; eax = (t ^ (t << 1)) xor eax,edi ; eax = (v ^ (v << 4)) ^ (t ^ (t << 1)) mov [g_xorwow_v],eax ; v = eax add [g_xorwow_d],362437 ; d += 362437 mov eax,[g_xorwow_d] ; eax = d add eax,[g_xorwow_v] ; eax = d + v } ;------------------------------------------------------------------------------- ; NAME: RANDOM ; DESC: Returns pseudo random number in the range [-0.5;0.5). ; OUT: xmm0.x [-0.5;0.5) ;------------------------------------------------------------------------------- macro RANDOM { XORWOW cvtsi2ss xmm0,eax mulss xmm0,[g_rand_scale] } ;------------------------------------------------------------------------------- segment readable executable ;------------------------------------------------------------------------------- ; NAME: GenerateSequence ; IN: xmm0.x re (c0.x) ; IN: xmm1.x im (c0.y) ; IN: rdi array size ; IN/OUT: rsi pointer to the allocated array ; OUT: rax generated sequence size ;------------------------------------------------------------------------------- align 16 GenerateSequence: xor eax,eax ; eax is index loop xorps xmm4,xmm4 ; xmm4 is c.x xorps xmm5,xmm5 ; xmm5 is c.y .Loop: ; cn.x = c.x * c.x - c.y * c.y + c0.x movaps xmm2,xmm4 movaps xmm3,xmm5 mulss xmm2,xmm4 mulss xmm3,xmm5 subss xmm2,xmm3 addss xmm2,xmm0 movaps xmm6,xmm2 ; xmm6 is cn.x ; cn.y = 2.0 * c.x * c.y + c0.y movaps xmm7,xmm4 mulss xmm7,xmm5 addss xmm7,xmm7 addss xmm7,xmm1 ; xmm7 is cn.y ; store cn movd dword [rsi+rax*8],xmm6 movd dword [rsi+rax*8+4],xmm7 ; if (cn.x * cn.x + cn.y * cn.y > 10.0) return eax; movaps xmm2,xmm6 movaps xmm3,xmm7 mulss xmm2,xmm6 mulss xmm3,xmm7 addss xmm2,xmm3 ucomiss xmm2,[g_max_dist] ja .EndLoop movaps xmm4,xmm6 ; c.x = cn.x movaps xmm5,xmm7 ; c.y = cn.y ; continue loop add eax,1 cmp eax,edi jb .Loop ; return 0 xor eax,eax .EndLoop: ret ;------------------------------------------------------------------------------- ; NAME: AllocateMemory ; IN: rdi size in bytes ; OUT: rax memory address ;------------------------------------------------------------------------------- align 16 AllocateMemory: mov eax,9 ; sys_mmap xor esi,esi ; addr xchg esi,edi mov edx,0x1+0x2 ; PROT_READ | PROT_WRITE mov r10d,0x02+0x20 ; MAP_PRIVATE | MAP_ANONYMOUS mov r8,-1 ; fd xor r9d,r9d ; offset syscall ret ;------------------------------------------------------------------------------- ; NAME: main ; DESC: Program main function. ;------------------------------------------------------------------------------- align 16 main: img_ptr equ rbp-8 seq_ptr equ rbp-16 pixel equ rbp-24 push rbp mov rbp,rsp sub rsp,128 ; alloc mem for the sequence mov edi,SEQ_SIZE*8 call AllocateMemory mov [seq_ptr],rax ; alloc mem for the image mov edi,IMG_SIZE*IMG_SIZE*4 call AllocateMemory mov [img_ptr],rax ; begin loops xor r13d,r13d ; .LoopIterations counter .LoopIterations: xor r12d,r12d ; .LoopOneMillion counter .LoopOneMillion: RANDOM mulss xmm0,[g_range] movaps xmm1,xmm0 RANDOM mulss xmm0,[g_range] mov edi,SEQ_SIZE mov rsi,[seq_ptr] call GenerateSequence ; eax = n sequence size test eax,eax jz .LoopSequenceEnd xor ecx,ecx ; ecx = i = 0 loop counter mov r9,[seq_ptr] ; r9 = sequence base address mov r8,[img_ptr] ; r8 = image base address movss xmm2,[g_img_size] movaps xmm3,xmm2 mulss xmm3,[g_0_5] ; xmm3 = (g_img_size)/2 movss xmm4,[g_zoom] mulss xmm4,xmm2 ; xmm4 = g_zoom * g_img_size movss xmm5,[g_offsetx] ; xmm5 = g_offsetx movss xmm6,[g_offsety] ; xmm6 = g_offsety .LoopSequence: cmp ecx,eax ; i < n je .LoopSequenceEnd movd xmm0,[r9+rcx*8] ; load re movd xmm1,[r9+rcx*8+4] ; load im addss xmm0,xmm5 ; xmm0 = re+g_offsetx addss xmm1,xmm6 ; xmm1 = im+g_offsety mulss xmm0,xmm4 ; xmm0 = (re+g_offsetx)*g_img_size*g_zoom mulss xmm1,xmm4 ; xmm1 = (im+g_offsety)*g_img_size*g_zoom addss xmm0,xmm3 ; xmm0 = (re+g_offsetx)*g_img_size*g_zoom+g_img_size/2 addss xmm1,xmm3 ; xmm1 = (im+g_offsety)*g_img_size*g_zoom+g_img_size/2 cvtss2si edi,xmm0 ; edi = x = int(xmm0.x) cvtss2si esi,xmm1 ; esi = y = int(xmm1.x) cmp edi,0 jl @f cmp edi,IMG_SIZE jge @f cmp esi,0 jl @f cmp esi,IMG_SIZE jge @f imul esi,esi,IMG_SIZE add esi,edi add dword [r8+rsi*4],1 @@: add ecx,1 jmp .LoopSequence .LoopSequenceEnd: ; continue .LoopOneMillion add r12d,1 cmp r12d,1000000 jb .LoopOneMillion ; continue .LoopIterations add r13d,1 cmp r13d,ITERATIONS jb .LoopIterations ; find max value mov r8,[img_ptr] ; r8 = image base address xor r12d,r12d ; r12d = max_val = 0 xor eax,eax ; eax = i = loop counter .LoopMax: cmp dword [r8+rax*4],r12d cmova r12d,dword [r8+rax*4] add eax,1 cmp eax,IMG_SIZE*IMG_SIZE jb .LoopMax ; find min value mov r13d,r12d ; r13d = min_val = max_val xor eax,eax ; eax = i = loop counter .LoopMin: cmp dword [r8+rax*4],r13d cmovb r13d,dword [r8+rax*4] add eax,1 cmp eax,IMG_SIZE*IMG_SIZE jb .LoopMin ; create TGA file mov eax,85 mov rdi,g_tga_name mov esi,110000000b syscall mov rbx,rax ; write TGA header mov eax,1 mov rdi,rbx mov rsi,g_tga_head mov edx,18 syscall ; write image pixels mov byte [pixel+3],255 mov r14,[img_ptr] ; r14 = image base address xor r15d,r15d ; r15d = i = loop counter cvtsi2ss xmm0,r12d ; load max_value cvtsi2ss xmm1,r13d ; load min_value movaps xmm2,xmm0 subss xmm2,xmm1 ; xmm2 = r = max_value - min_value .LoopWrite: mov eax,[r14+r15*4] ; eax = image_value sub eax,r13d ; eax = image_value - min_value cvtsi2ss xmm0,eax ; xmm0 = float(image_value - min_value) addss xmm0,xmm0 ; xmm0 = 2.0f * float(image_value - min_value) divss xmm0,xmm2 ; xmm0 = 2.0f * float(image_value - min_value) / r minss xmm0,[g_1_0] ; clamp to 1.0 maxss xmm0,[g_0_0] ; clamp to 0.0 mulss xmm0,[g_255_0] ; convert to 0 - 255 cvtss2si eax,xmm0 mov [pixel],al ; store B component mov [pixel+1],al ; store G component mov [pixel+2],al ; store R component ; write pixel data mov eax,1 mov rdi,rbx lea rsi,[pixel] mov edx,4 syscall ; continue .LoopWrite add r15d,1 cmp r15d,IMG_SIZE*IMG_SIZE jb .LoopWrite mov rsp,rbp pop rbp ret restore img_ptr,seq_ptr,pixel ;------------------------------------------------------------------------------- ; NAME: _start ; DESC: Program entry point. ;------------------------------------------------------------------------------- _start: call main mov eax,60 ; syscall number xor edi,edi ; exit code syscall ; exit ;------------------------------------------------------------------------------- segment readable writeable ;------------------------------------------------------------------------------- align 1 g_tga_name db 'buddhabrot.tga',0 g_tga_head db 0,0,2,9 dup 0 db (IMG_SIZE and 0x00ff),(IMG_SIZE and 0xff00) shr 8 db (IMG_SIZE and 0x00ff),(IMG_SIZE and 0xff00) shr 8,32,0 align 4 g_xorwow_x dd 123456789 g_xorwow_y dd 362436069 g_xorwow_z dd 521288629 g_xorwow_w dd 88675123 g_xorwow_v dd 5783321 g_xorwow_d dd 6615241 g_rand_scale dd 2.3283064e-10 ; 1.0 / 2^32 IMG_SIZE=800 SEQ_SIZE=50 ITERATIONS=1000 g_img_size dd 800.0 g_offsetx dd 0.5 g_offsety dd 0.0 g_zoom dd 0.4 g_max_dist dd 10.0 g_range dd 4.2 g_0_5 dd 0.5 g_0_0 dd 0.0 g_1_0 dd 1.0 g_255_0 dd 255.0 ;-------------------------------------------------------------------------------