format ELF64 executable 3 entry _start ;------------------------------------------------------------------------------- ; NAME: DOT3 ; IN: xmm0 | ? z0 y0 x0 | ; IN: xmm1 | ? z1 y1 x1 | ; OUT: xmm0 | s s s s | s = x0*x1+y0*y1+z0*z1 ;------------------------------------------------------------------------------- macro DOT3 { mulps xmm0,xmm1 movaps xmm1,xmm0 movaps xmm2,xmm0 shufps xmm0,xmm0,0x00 shufps xmm1,xmm1,0x55 shufps xmm2,xmm2,0xaa addps xmm0,xmm1 addps xmm0,xmm2 } ;------------------------------------------------------------------------------- ; NAME: DOT4 ; IN: xmm0 | w0 z0 y0 x0 | ; IN: xmm1 | w1 z1 y1 x1 | ; OUT: xmm0 | s s s s | s = x0*x1+y0*y1+z0*z1+w0*w1 ;------------------------------------------------------------------------------- macro DOT4 { mulps xmm0,xmm1 movaps xmm1,xmm0 movaps xmm2,xmm0 movaps xmm3,xmm0 shufps xmm0,xmm0,0x00 shufps xmm1,xmm1,0x55 shufps xmm2,xmm2,0xaa shufps xmm3,xmm3,0xff addps xmm0,xmm1 addps xmm2,xmm3 addps xmm0,xmm2 } ;------------------------------------------------------------------------------- ; NAME: FLOOR ; IN: xmm0 | w z y x | ; OUT: xmm0 | floor(w) floor(z) floor(y) floor(x) | ;------------------------------------------------------------------------------- macro FLOOR { cvttps2dq xmm1,xmm0 psrld xmm0,31 psubd xmm1,xmm0 cvtdq2ps xmm0,xmm1 } ;------------------------------------------------------------------------------- ; NAME: STEP ; IN: xmm0 | ew ez ey ex | edge vector ; IN: xmm1 | w z y x | value vector ; OUT: xmm0 | step(ew,w) step(ez,z) step(ey,y) step(ex,x) | ;------------------------------------------------------------------------------- macro STEP { cmpltps xmm1,xmm0 andnps xmm1,dqword [g_1_0] movaps xmm0,xmm1 } ;------------------------------------------------------------------------------- ; NAME: MOD289 ; IN: xmm0 | w z y x | ; OUT: xmm0 | mod289(w) mod289(z) mod289(y) mod289(x) | ; mod289(s) = s - floor(s * (1.0/289.0)) * 289.0 ;------------------------------------------------------------------------------- macro MOD289 { movaps xmm2,xmm0 mulps xmm0,dqword [g_1_div_289] FLOOR mulps xmm0,dqword [g_289_0] subps xmm2,xmm0 movaps xmm0,xmm2 } ;------------------------------------------------------------------------------- ; NAME: PERMUTE ; IN: xmm0 | w z y x | ; OUT: xmm0 | perm(w) perm(z) perm(y) perm(x) | ; perm(s) = mod289(((s*34.0)+1.0)*s) ;------------------------------------------------------------------------------- macro PERMUTE { movaps xmm1,xmm0 mulps xmm0,dqword [g_34_0] addps xmm0,dqword [g_1_0] mulps xmm0,xmm1 MOD289 } ;------------------------------------------------------------------------------- ; NAME: ABS ; IN: xmm0 | w z y x | ; OUT: xmm0 | abs(w) abs(z) abs(y) abs(x) | ;------------------------------------------------------------------------------- macro ABS { xorps xmm1,xmm1 ; xmm1 = | 0 0 0 0 | subps xmm1,xmm0 ; xmm1 = neg(x) maxps xmm0,xmm1 ; xmm0 = abs(x) } ;------------------------------------------------------------------------------- segment readable executable ;------------------------------------------------------------------------------- ; NAME: snoise3 ; DESC: 3D Simplex noise (https://github.com/ashima/webgl-noise) ; IN: xmm0 | ? z y x | ; OUT: xmm0 | s s s s | s is noise value [-1.0,1.0] ;------------------------------------------------------------------------------- align 16 snoise3: v equ rbp-16 i equ rbp-32 x0 equ rbp-48 x1 equ rbp-64 x2 equ rbp-80 x3 equ rbp-96 i1 equ rbp-112 i2 equ rbp-128 push rbp mov rbp,rsp sub rsp,256 movaps [v],xmm0 ; save input on the stack ; ; Compute corners (x0, x1, x2, x3) ; ; i = floor(v + dot(v, C.yyy)) movaps xmm1,dqword [g_snoise_C] shufps xmm1,xmm1,0x55 ; xmm1 = C.yyy DOT3 ; xmm0 = dot(xmm0,xmm1) addps xmm0,[v] FLOOR movaps [i],xmm0 ; x0 = v - i + dot(i, C.xxx) movaps xmm3,[v] subps xmm3,xmm0 movaps xmm1,dqword [g_snoise_C] shufps xmm1,xmm1,0x00 ; xmm1 = C.xxx DOT3 ; xmm0 = dot(xmm0,xmm1) addps xmm3,xmm0 movaps [x0],xmm3 ; compute i1 and i2 movaps xmm0,xmm3 shufps xmm0,xmm0,11001001b ; xmm0 = | w x z y | movaps xmm1,xmm3 STEP movaps xmm7,xmm0 ; xmm7 = g movaps xmm6,dqword [g_1_0] subps xmm6,xmm7 ; xmm6 = 1.0 - g = l shufps xmm6,xmm6,11010010b ; xmm6 = | w y x z | movaps xmm0,xmm7 movaps xmm1,xmm7 minps xmm0,xmm6 ; xmm0 = min(g.xyz, l.zxy) maxps xmm1,xmm6 ; xmm1 = max(g.xyz, l.zxy) movaps [i1],xmm0 ; xmm0 = i1 movaps [i2],xmm1 ; xmm1 = i2 ; compute x1, x2 and x3 movaps xmm7,[x0] ; xmm7 = x0 movaps xmm6,xmm7 ; xmm6 = x0 movaps xmm5,xmm7 ; xmm5 = x0 movaps xmm4,dqword [g_snoise_C] movaps xmm3,xmm4 movaps xmm2,dqword [g_snoise_D] shufps xmm4,xmm4,0x00 ; xmm4 = C.xxx shufps xmm3,xmm3,0x55 ; xmm3 = C.yyy shufps xmm2,xmm2,0x55 ; xmm2 = D.yyy subps xmm5,xmm0 ; xmm5 = x0 - i1 subps xmm6,xmm1 ; xmm6 = x0 - i2 subps xmm7,xmm2 ; xmm7 = x0 - D.yyy addps xmm5,xmm4 ; xmm5 = x0 - i1 + C.xxx addps xmm6,xmm3 ; xmm6 = x0 - i2 + C.yyy movaps [x1],xmm5 movaps [x2],xmm6 movaps [x3],xmm7 ; ; Compute permutations (p) ; movaps xmm0,[i] MOD289 movaps xmm7,xmm0 movaps xmm6,xmm0 movaps xmm5,xmm0 shufps xmm7,xmm7,10101010b ; xmm7 = i.zzzz shufps xmm6,xmm6,01010101b ; xmm6 = i.yyyy shufps xmm5,xmm5,00000000b ; xmm5 = i.xxxx movaps xmm4,[i1] ; xmm4 = i1 movaps xmm3,[i2] ; xmm3 = i2 ; movaps xmm0,xmm4 ; xmm0 = i1 movaps xmm1,xmm3 ; xmm1 = i2 shufps xmm0,xmm0,10101010b ; xmm0 = | i1.z i1.z i1.z i1.z | shufps xmm1,xmm1,10101010b ; xmm1 = | i2.z i2.z i2.z i2.z | andps xmm0,dqword [g_mask_0010] ; xmm0 = | 0 0 i1.z 0 | andps xmm1,dqword [g_mask_0100] ; xmm1 = | 0 i2.z 0 0 | orps xmm0,xmm1 ; xmm0 = | 0 i2.z i1.z 0 | orps xmm0,dqword [g_1_0_w] ; xmm0 = | 1 i2.z i1.z 0 | addps xmm0,xmm7 ; xmm0 = i.zzzz + | 1 i2.z i1.z 0 | PERMUTE movaps xmm8,xmm0 ; xmm8 = p movaps xmm0,xmm4 ; xmm0 = i1 movaps xmm1,xmm3 ; xmm1 = i2 shufps xmm0,xmm0,01010101b ; xmm0 = | i1.y i1.y i1.y i1.y | shufps xmm1,xmm1,01010101b ; xmm1 = | i2.y i2.y i2.y i2.y | andps xmm0,dqword [g_mask_0010] ; xmm0 = | 0 0 i1.y 0 | andps xmm1,dqword [g_mask_0100] ; xmm1 = | 0 i2.y 0 0 | orps xmm0,xmm1 ; xmm0 = | 0 i2.y i1.y 0 | orps xmm0,dqword [g_1_0_w] ; xmm0 = | 1 i2.y i1.y 0 | addps xmm0,xmm6 ; xmm0 = i.yyyy + | 1 i2.y i1.y 0 | addps xmm0,xmm8 PERMUTE movaps xmm8,xmm0 ; xmm8 = p movaps xmm0,xmm4 ; xmm0 = i1 movaps xmm1,xmm3 ; xmm1 = i2 shufps xmm0,xmm0,00000000b ; xmm0 = | i1.x i1.x i1.x i1.x | shufps xmm1,xmm1,00000000b ; xmm1 = | i2.x i2.x i2.x i2.x | andps xmm0,dqword [g_mask_0010] ; xmm0 = | 0 0 i1.x 0 | andps xmm1,dqword [g_mask_0100] ; xmm1 = | 0 i2.x 0 0 | orps xmm0,xmm1 ; xmm0 = | 0 i2.x i1.x 0 | orps xmm0,dqword [g_1_0_w] ; xmm0 = | 1 i2.x i1.x 0 | addps xmm0,xmm5 ; xmm0 = i.xxxx + | 1 i2.x i1.x 0 | addps xmm0,xmm8 PERMUTE movaps xmm8,xmm0 ; xmm8 = p ; ; Compute gradients ; movaps xmm0,dqword [g_snoise_D] movaps xmm1,xmm0 shufps xmm0,xmm0,11100111b ; xmm0 = | D.w D.z D.y D.w | shufps xmm1,xmm1,11001000b ; xmm1 = | D.w D.x D.z D.x | mulps xmm0,dqword [g_1_div_7] subps xmm0,xmm1 movaps xmm7,xmm0 ; xmm7 = ns ; xmm8 = j = p - 49.0 * floor(p * ns.z * ns.z) shufps xmm0,xmm0,10101010b ; xmm0 = ns.zzzz mulps xmm0,xmm0 mulps xmm0,xmm8 FLOOR mulps xmm0,dqword [g_49_0] subps xmm8,xmm0 ; xmm8 = j ; x_ = floor(j * ns.zzzz) movaps xmm0,xmm7 shufps xmm0,xmm0,10101010b ; xmm0 = ns.zzzz mulps xmm0,xmm8 FLOOR ; xmm0 = x_ movaps xmm6,xmm0 ; xmm6 = x_ ; y_ = floor(j - 7.0 * x_) mulps xmm0,dqword [g_7_0] movaps xmm1,xmm8 subps xmm1,xmm0 movaps xmm0,xmm1 FLOOR movaps xmm5,xmm0 ; xmm5 = y_ ; x = x_ * ns.xxxx + ns.yyyy ; y = y_ * ns.xxxx + ns.yyyy movaps xmm0,xmm7 ; xmm0 = ns movaps xmm1,xmm7 ; xmm1 = ns shufps xmm0,xmm0,00000000b ; xmm0 = ns.xxxx shufps xmm1,xmm1,01010101b ; xmm1 = ns.yyyy mulps xmm6,xmm0 ; xmm6 = x_ * ns.xxxx mulps xmm5,xmm0 ; xmm5 = y_ * ns.xxxx addps xmm6,xmm1 ; xmm6 = x = x_ * ns.xxxx + ns.yyyy addps xmm5,xmm1 ; xmm5 = y = y_ * ns.xxxx + ns.yyyy ; h = 1.0 - abs(x) - abs(y) movaps xmm4,dqword [g_1_0] ; xmm4 = h = | 1 1 1 1 | movaps xmm0,xmm6 ; xmm0 = x ABS ; xmm0 = abs(x) movaps xmm3,xmm0 ; xmm3 = abs(x) movaps xmm0,xmm5 ; xmm0 = y ABS ; xmm0 = abs(y) subps xmm4,xmm3 ; xmm4 = h = 1.0 - abs(x) subps xmm4,xmm0 ; xmm4 = h = 1.0 - abs(x) - abs(y) ; b0 = vec4(x.xy, y.xy) movaps xmm0,xmm6 ; xmm0 = x movaps xmm1,xmm5 ; xmm1 = y unpcklps xmm0,xmm1 ; xmm0 = | y.y x.y y.x x.x | shufps xmm0,xmm0,11011000b ; xmm0 = | y.y y.x x.y x.x | movaps xmm7,xmm0 ; xmm7 = b0 ; b1 = vec4(x.zw, y.zw) movaps xmm0,xmm6 ; xmm0 = x movaps xmm1,xmm5 ; xmm1 = y unpckhps xmm0,xmm1 ; xmm0 = | y.w x.w y.z x.z | shufps xmm0,xmm0,11011000b ; xmm0 = | y.w y.z x.w x.z | movaps xmm3,xmm0 ; xmm3 = b1 ; s0 = floor(b0) * 2.0 + 1.0 movaps xmm0,xmm7 FLOOR addps xmm0,xmm0 addps xmm0,dqword [g_1_0] movaps xmm15,xmm0 ; xmm15 = s0 ; s1 = floor(b1) * 2.0 + 1.0 movaps xmm0,xmm3 FLOOR addps xmm0,xmm0 addps xmm0,dqword [g_1_0] movaps xmm14,xmm0 ; xmm14 = s1 ; sh = -step(h, vec4(0.0)) movaps xmm0,xmm4 ; xmm0 = h xorps xmm1,xmm1 ; xmm1 = | 0 0 0 0 | STEP xorps xmm1,xmm1 subps xmm1,xmm0 movaps xmm13,xmm1 ; xmm13 = sh ; a0 = b0.xzyw + s0.xzyw * sh.xxyy shufps xmm7,xmm7,11011000b ; xmm7 = b0 = | w y z x | shufps xmm15,xmm15,11011000b ; xmm15 = s0 = | w y z x | movaps xmm0,xmm13 shufps xmm0,xmm0,01010000b ; xmm0 = | y y x x | mulps xmm0,xmm15 addps xmm0,xmm7 movaps xmm7,xmm0 ; xmm7 = a0 ; a1 = b1.xzyw + s1.xzyw * sh.zzww shufps xmm3,xmm3,11011000b ; xmm3 = b1 = | w y z x | shufps xmm14,xmm14,11011000b ; xmm14 = s1 = | w y z x | shufps xmm13,xmm13,11111010b ; xmm13 = sh = | w w z z | mulps xmm14,xmm13 addps xmm14,xmm3 movaps xmm6,xmm14 ; xmm6 = a1 ; p0 = vec3(a0.xy, h.x) movaps xmm0,xmm7 ; xmm0 = a0 shufps xmm0,xmm4,00000100b ; | h.x h.x a0.y a0.x | movaps xmm5,xmm0 ; xmm5 = p0 ; p1 = vec3(a0.zw, h.y) shufps xmm7,xmm4,01011110b ; xmm7 = p1 = | h.y h.y a0.w a0.z | ; p2 = vec3(a1.xy, h.z) movaps xmm0,xmm6 ; xmm0 = a1 shufps xmm0,xmm4,10100100b ; | h.z h.z a1.y a1.x | movaps xmm3,xmm0 ; xmm3 = p2 ; p3 = vec3(a1.zw, h.w) shufps xmm6,xmm4,11111110b ; xmm6 = p3 = | h.w h.w a1.w a1.z | ; movaps xmm4,xmm3 ; xmm4 = p2 ; ; Normalize gradients ; ; xmm5 = p0, xmm7 = p1, xmm4 = p2, xmm6 = p3 ; ; xmm15 = dot(p0, p0) movaps xmm0,xmm5 ; xmm0 = p0 movaps xmm1,xmm5 ; xmm1 = p0 DOT3 movaps xmm15,xmm0 ; xmm15 = dot(p0, p0) ; xmm14 = dot(p1, p1) movaps xmm0,xmm7 ; xmm0 = p1 movaps xmm1,xmm7 ; xmm1 = p1 DOT3 movaps xmm14,xmm0 ; xmm14 = dot(p1, p1) ; xmm13 = dot(p2, p2) movaps xmm0,xmm4 ; xmm0 = p2 movaps xmm1,xmm4 ; xmm1 = p2 DOT3 movaps xmm13,xmm0 ; xmm13 = dot(p2, p2) ; xmm12 = dot(p3, p3) movaps xmm0,xmm6 ; xmm0 = p3 movaps xmm1,xmm6 ; xmm1 = p3 DOT3 movaps xmm12,xmm0 ; xmm12 = dot(p3, p3) ; movaps xmm0,dqword [g_taylor_scale] movaps xmm1,dqword [g_taylor_bias] mulps xmm15,xmm0 mulps xmm14,xmm0 mulps xmm13,xmm0 mulps xmm12,xmm0 addps xmm15,xmm1 addps xmm14,xmm1 addps xmm13,xmm1 addps xmm12,xmm1 ; normalize mulps xmm5,xmm15 ; xmm5 = p0 mulps xmm7,xmm14 ; xmm7 = p1 mulps xmm4,xmm13 ; xmm4 = p2 mulps xmm6,xmm12 ; xmm6 = p3 ; ; Mix final noise value ; ; xmm15 = dot(x0, x0) movaps xmm0,[x0] ; xmm0 = x0 movaps xmm1,xmm0 ; xmm1 = x0 DOT3 movaps xmm15,xmm0 ; xmm15 = dot(x0, x0) ; xmm14 = dot(x1, x1) movaps xmm0,[x1] ; xmm0 = x1 movaps xmm1,xmm0 ; xmm1 = x1 DOT3 movaps xmm14,xmm0 ; xmm14 = dot(x1, x1) ; xmm13 = dot(x2, x2) movaps xmm0,[x2] ; xmm0 = x2 movaps xmm1,xmm0 ; xmm1 = x2 DOT3 movaps xmm13,xmm0 ; xmm13 = dot(x2, x2) ; xmm12 = dot(x3, x3) movaps xmm0,[x3] ; xmm0 = x3 movaps xmm1,xmm0 ; xmm1 = x3 DOT3 movaps xmm12,xmm0 ; xmm12 = dot(x3, x3) ; andps xmm15,dqword [g_mask_0001] andps xmm14,dqword [g_mask_0010] andps xmm13,dqword [g_mask_0100] andps xmm12,dqword [g_mask_1000] orps xmm15,xmm14 orps xmm13,xmm12 orps xmm15,xmm13 movaps xmm0,dqword [g_0_6] subps xmm0,xmm15 maxps xmm0,dqword [g_0_0] ; xmm0 = m mulps xmm0,xmm0 mulps xmm0,xmm0 movaps xmm10,xmm0 ; xmm10 = m^4 ; ; xmm15 = dot(x0, p0) movaps xmm0,[x0] ; xmm0 = x0 movaps xmm1,xmm5 ; xmm1 = p0 DOT3 movaps xmm15,xmm0 ; xmm15 = dot(x0, p0) ; xmm14 = dot(x1, p1) movaps xmm0,[x1] ; xmm0 = x1 movaps xmm1,xmm7 ; xmm1 = p1 DOT3 movaps xmm14,xmm0 ; xmm14 = dot(x1, p1) ; xmm13 = dot(x2, p2) movaps xmm0,[x2] ; xmm0 = x2 movaps xmm1,xmm4 ; xmm1 = p2 DOT3 movaps xmm13,xmm0 ; xmm13 = dot(x2, p2) ; xmm12 = dot(x3, p3) movaps xmm0,[x3] ; xmm0 = x3 movaps xmm1,xmm6 ; xmm1 = p3 DOT3 movaps xmm12,xmm0 ; xmm12 = dot(x3, p3) ; put all above dots into xmm15 andps xmm15,dqword [g_mask_0001] andps xmm14,dqword [g_mask_0010] andps xmm13,dqword [g_mask_0100] andps xmm12,dqword [g_mask_1000] orps xmm15,xmm14 orps xmm13,xmm12 orps xmm15,xmm13 ; movaps xmm0,xmm10 ; xmm0 = m^4 movaps xmm1,xmm15 DOT4 mulps xmm0,dqword [g_42_0] mov rsp,rbp pop rbp ret restore v,i,x0,x1,x2,x3,i1,i2 ;------------------------------------------------------------------------------- ; NAME: main ; DESC: Program main function. ;------------------------------------------------------------------------------- align 16 main: imgptr equ rbp-8 push rbp mov rbp,rsp sub rsp,128 ; alloc memory for the image mov eax,9 ; sys_mmap xor edi,edi ; addr mov esi,SIZE*SIZE*4 ; length 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 mov [imgptr],rax mov rbx,rax ; begin loops xor r13d,r13d ; .LoopY index .LoopY: xor r12d,r12d ; .LoopX index .LoopX: ; compute xorps xmm0,xmm0 xorps xmm1,xmm1 cvtsi2ss xmm0,r12d ; xmm0 = | 0 0 0 x | cvtsi2ss xmm1,r13d ; xmm1 = | 0 0 0 y | unpcklps xmm0,xmm1 ; xmm0 = | 0 0 y x | divps xmm0,dqword [g_size] addps xmm0,xmm0 addps xmm0,xmm0 call snoise3 mulps xmm0,dqword [g_0_5] addps xmm0,dqword [g_0_5] ; clamp to [0.0,1.0] minps xmm0,dqword [g_1_0] maxps xmm0,dqword [g_0_0] ; convert from [0.0,1.0] to [0,255] mulps xmm0,dqword [g_255_0] cvttps2dq xmm0,xmm0 movd eax,xmm0 mov [rbx+2],al ; red pshufd xmm1,xmm0,00000001b movd eax,xmm1 mov [rbx+1],al ; green pshufd xmm1,xmm0,00000010b movd eax,xmm1 mov [rbx+0],al ; blue mov byte [rbx+3],255 ; alpha ; advance pixel pointer add rbx,4 ; continue .LoopX add r12d,1 cmp r12d,SIZE jne .LoopX ; continue .LoopY add r13d,1 cmp r13d,SIZE jne .LoopY ; create TGA file mov eax,85 mov rdi,g_tga_name mov esi,110000000b syscall mov rbx,rax ; write header mov eax,1 mov rdi,rbx mov rsi,g_tga_head mov edx,18 syscall ; write pixel data mov eax,1 mov rdi,rbx mov rsi,[imgptr] mov edx,SIZE*SIZE*4 syscall mov rsp,rbp pop rbp ret restore imgptr ;------------------------------------------------------------------------------- ; NAME: Debug ;------------------------------------------------------------------------------- align 16 Debug: v equ rbp-16 push rbp mov rbp,rsp sub rsp,128 mov dword [v+0],1.2 mov dword [v+4],2.4 mov dword [v+8],3.5 mov dword [v+12],0.0 movaps xmm0,[v] call snoise3 mov rsp,rbp pop rbp ret restore v ;------------------------------------------------------------------------------- ; NAME: _start ; DESC: Program entry point. ;------------------------------------------------------------------------------- _start: call main ; terminate process mov eax,60 xor edi,edi syscall ;------------------------------------------------------------------------------- segment readable ;------------------------------------------------------------------------------- align 1 g_tga_name db 'snoise.tga',0 g_tga_head db 0,0,2,9 dup 0 db (SIZE and 0x00ff),(SIZE and 0xff00) shr 8 db (SIZE and 0x00ff),(SIZE and 0xff00) shr 8,32,0 align 16 SIZE=800 g_size dd 4 dup 800.0 g_snoise_C dd 0.166666667,0.333333333,0.0,0.0 g_snoise_D dd 0.0,0.5,1.0,2.0 g_0_0 dd 4 dup 0.0 g_0_5 dd 4 dup 0.5 g_0_6 dd 4 dup 0.6 g_1_0 dd 4 dup 1.0 g_7_0 dd 4 dup 7.0 g_34_0 dd 4 dup 34.0 g_42_0 dd 4 dup 42.0 g_49_0 dd 4 dup 49.0 g_255_0 dd 4 dup 255.0 g_289_0 dd 4 dup 289.0 g_1_div_7 dd 4 dup 0.142857142857 g_1_div_289 dd 4 dup 0.003460208 g_mask_0001 dd 0xffffffff,0x00000000,0x00000000,0x00000000 g_mask_0010 dd 0x00000000,0xffffffff,0x00000000,0x00000000 g_mask_0100 dd 0x00000000,0x00000000,0xffffffff,0x00000000 g_mask_1000 dd 0x00000000,0x00000000,0x00000000,0xffffffff g_1_0_w dd 0.0,0.0,0.0,1.0 g_taylor_bias dd 4 dup 1.79284291400159 g_taylor_scale dd 4 dup -0.85373472095314 ;-------------------------------------------------------------------------------