[mOved] Les meilleures quotes - Page 89

[19:18:30] <Zerosquare> bon allez zou, je vais passer un coup de balai
[19:19:13] <Folco> tu vas pas me dire que t'as pas un script pour ça ?
[19:19:22] <Zerosquare> non
[...]
[19:21:07] <Folco> Je parlais d'un script "CoupDeBalai.sh" :D
[19:21:14] <Zerosquare> ^^
[19:21:33] <Zerosquare> ça serait bien, mais j'arrive déjà pas à avoir une fonction "aligner le bordel sur la grille" dans ma chambre :(

bearbecue :

	while (sCount--)
	{
		__m128	xmm2 = _mm_load_ps(data + 0x0);
		__m128	xmm3 = _mm_load_ps(data + 0x4);
		__m128	xmm4 = _mm_load_ps(data + 0x8);
		__m128	xmm5 = _mm_load_ps(data + 0xC);
		xmin = _mm_min_ps(xmin, xmm2);
		xmax = _mm_max_ps(xmax, xmm2);
		xmin = _mm_min_ps(xmin, xmm3);
		xmax = _mm_max_ps(xmax, xmm3);
		xmin = _mm_min_ps(xmin, xmm4);
		xmax = _mm_max_ps(xmax, xmm4);
		xmin = _mm_min_ps(xmin, xmm5);
		xmax = _mm_max_ps(xmax, xmm5);
		data += 0x10;
	}

Visual Studio 2010 :

01794BA0 0F 5D 00             minps       xmm0,xmmword ptr [eax]  
01794BA3 0F 5F 08             maxps       xmm1,xmmword ptr [eax]  
01794BA6 0F 5D 40 10          minps       xmm0,xmmword ptr [eax+10h]  
01794BAA 0F 5F 48 10          maxps       xmm1,xmmword ptr [eax+10h]  
01794BAE 0F 5D 40 20          minps       xmm0,xmmword ptr [eax+20h]  
01794BB2 0F 5F 48 20          maxps       xmm1,xmmword ptr [eax+20h]  
01794BB6 0F 5D 40 30          minps       xmm0,xmmword ptr [eax+30h]  
01794BBA 0F 5F 48 30          maxps       xmm1,xmmword ptr [eax+30h]  
01794BBE 83 C0 40             add         eax,40h  
01794BC1 49                   dec         ecx  
01794BC2 75 DC                jne         HellHeaven:: SIMD_SSE::FindMinMaxFloat4Stream_Aligned16+20h (1794BA0h)

bearbecue :

	__asm
	{
		[...]

		cmp	eax, ecx
		jg	_SkipUnrolled
_LoopUnrolled:
		movaps	xmm0, [eax]
		movaps	xmm1, [eax + 0x10]
		movaps	xmm2, [eax + 0x20]
		movaps	xmm3, [eax + 0x30]

		minps	xmm6, xmm0
		maxps	xmm7, xmm0
		add	eax, 64
		minps	xmm6, xmm1
		maxps	xmm7, xmm1
		cmp	eax, ecx
		minps	xmm6, xmm2
		maxps	xmm7, xmm2
		minps	xmm6, xmm3
		maxps	xmm7, xmm3

		jbe	_LoopUnrolled
_SkipUnrolled:
		[...]
	}

	__asm
	{
		mov		eax, streamAligned16
		mov		ebx, count

		movaps	xmm4, [eax]
		add		ebx, ebx
		movaps	xmm5, xmm4
		movaps	xmm6, xmm4
		movaps	xmm7, xmm4
		lea		ebx, [eax + ebx * 8]	// end = stream + count*16
		add		eax, 16
		lea		ecx, [ebx - 64]

		cmp		eax, ecx
		jg		_SkipUnrolled
_LoopUnrolled:
		movaps	xmm0, [eax]
		movaps	xmm1, [eax + 0x10]
		movaps	xmm2, [eax + 0x20]
		movaps	xmm3, [eax + 0x30]

		minps	xmm4, xmm0		// <--
		maxps	xmm5, xmm0		// <--
		add		eax, 64
		minps	xmm6, xmm1
		maxps	xmm7, xmm1
		cmp		eax, ecx
		minps	xmm4, xmm2		// <--
		maxps	xmm5, xmm2		// <--
		minps	xmm6, xmm3
		maxps	xmm7, xmm3

		jbe		_LoopUnrolled

./2660 Zerosquare :Le compilateur optimisera mieux que toi, qu'ils disaient

À votre avis ?
The best optimizer is between your ears