nb_kernel112_ia32_3dnow.intel_syntax.s

最著名最快的分子模拟软件

	
	pfadd mm0, mm5		;# add up total force on j particle.  
	pfadd mm1, mm7

	;# update j particle force 
	movq mm2,  [edi + eax*4]
	movd mm3,  [edi + eax*4 + 8]
	pfsub mm2, mm0
	pfsub mm3, mm1
	movq [edi + eax*4], mm2
	movd [edi + eax*4 +8], mm3

	;# interactions with j H1 
	movq  mm0, [esi + eax*4 + 12]
	movd  mm1, [esi + eax*4 + 20]
	;# copy & expand to mm2-mm4 for the H interactions 
	movq  mm2, mm0
	movq  mm3, mm0
	movq  mm4, mm1
	punpckldq mm2,mm2
	punpckhdq mm3,mm3
	punpckldq mm4,mm4
	
	movd mm6, [esp + nb112_qqOH]
	movq mm7, [esp + nb112_qqHH]
	
	pfsubr mm0, [esp + nb112_ixO]
	pfsubr mm1, [esp + nb112_izO]
		
	movq  [esp + nb112_dxO], mm0
	pfmul mm0,mm0
	movd  [esp + nb112_dzO], mm1	
	pfmul mm1,mm1
	pfacc mm0, mm1
	pfadd mm0, mm1		;# mm0=rsqO 
	
	punpckldq mm2, mm2
	punpckldq mm3, mm3
	punpckldq mm4, mm4  ;# mm2-mm4 is jx-jz 
	pfsubr mm2, [esp + nb112_ixH]
	pfsubr mm3, [esp + nb112_iyH]
	pfsubr mm4, [esp + nb112_izH] ;# mm2-mm4 is dxH-dzH 
	
	movq [esp + nb112_dxH], mm2
	movq [esp + nb112_dyH], mm3
	movq [esp + nb112_dzH], mm4
	pfmul mm2,mm2
	pfmul mm3,mm3
	pfmul mm4,mm4

	pfadd mm3,mm2
	pfadd mm3,mm4		;# mm3=rsqH 

    	pfrsqrt mm1,mm0

    	movq mm2,mm1
    	pfmul mm1,mm1
    	pfrsqit1 mm1,mm0				
    	pfrcpit2 mm1,mm2	;# mm1=invsqrt 
	movq  mm4, mm1
	pfmul mm4, mm4		;# mm4=invsq 
	;# calculate potential and scalar force 
	pfmul mm6, mm1		;# mm6=vcoul 
	pfmul mm4, mm6		;# mm4=fscalar  

	pfrsqrt mm5, mm3
	pswapd mm3,mm3
	pfrsqrt mm2, mm3
	pswapd mm3,mm3
	punpckldq mm5,mm2	;# seeds are in mm5 now, and rsq in mm3 

	movq mm2, mm5
	pfmul mm5,mm5
    	pfrsqit1 mm5,mm3				
    	pfrcpit2 mm5,mm2	;# mm5=invsqrt 
	movq mm3,mm5
	pfmul mm3,mm3		;# mm3=invsq 
	pfmul mm7, mm5		;# mm7=vcoul 
	pfmul mm3, mm7		;# mm3=fscal for the two H's. 

	;# update vctot 
	pfadd mm7, mm6
	pfadd mm7, [esp + nb112_vctot]
	movq [esp + nb112_vctot], mm7
	
	;# spread oxygen fscalar to both positions 
	punpckldq mm4,mm4
	;# calc vectorial force for O 
	movq mm0,  [esp + nb112_dxO]
	movd mm1,  [esp + nb112_dzO]
	pfmul mm0, mm4
	pfmul mm1, mm4

	;# calc vectorial force for H's 
	movq mm5, [esp + nb112_dxH]
	movq mm6, [esp + nb112_dyH]
	movq mm7, [esp + nb112_dzH]
	pfmul mm5, mm3
	pfmul mm6, mm3
	pfmul mm7, mm3
	
	;# update iO particle force 
	movq mm2,  [esp + nb112_fixO]
	movd mm3,  [esp + nb112_fizO]
	pfadd mm2, mm0
	pfadd mm3, mm1
	movq [esp + nb112_fixO], mm2
	movd [esp + nb112_fizO], mm3

	;# update iH forces 
	movq mm2, [esp + nb112_fixH]
	movq mm3, [esp + nb112_fiyH]
	movq mm4, [esp + nb112_fizH]
	pfadd mm2, mm5
	pfadd mm3, mm6
	pfadd mm4, mm7
	movq [esp + nb112_fixH], mm2
	movq [esp + nb112_fiyH], mm3
	movq [esp + nb112_fizH], mm4
	
	;# pack j forces from H in the same form as the oxygen force. 
	pfacc mm5, mm6		;# mm5(l)=fjx(H1+ h2) mm5(h)=fjy(H1+ h2) 
	pfacc mm7, mm7		;# mm7(l)=fjz(H1+ h2) 
	
	pfadd mm0, mm5		;# add up total force on j particle.  
	pfadd mm1, mm7

	;# update j particle force 
	movq mm2,  [edi + eax*4 + 12]
	movd mm3,  [edi + eax*4 + 20]
	pfsub mm2, mm0
	pfsub mm3, mm1
	movq [edi + eax*4 + 12], mm2
	movd [edi + eax*4 + 20], mm3

	;# interactions with j H2 
	movq  mm0, [esi + eax*4 + 24]
	movd  mm1, [esi + eax*4 + 32]
	;# copy & expand to mm2-mm4 for the H interactions 
	movq  mm2, mm0
	movq  mm3, mm0
	movq  mm4, mm1
	punpckldq mm2,mm2
	punpckhdq mm3,mm3
	punpckldq mm4,mm4

	movd mm6, [esp + nb112_qqOH]
	movq mm7, [esp + nb112_qqHH]

	pfsubr mm0, [esp + nb112_ixO]
	pfsubr mm1, [esp + nb112_izO]
		
	movq  [esp + nb112_dxO], mm0
	pfmul mm0,mm0
	movd  [esp + nb112_dzO], mm1	
	pfmul mm1,mm1
	pfacc mm0, mm1
	pfadd mm0, mm1		;# mm0=rsqO 
	
	punpckldq mm2, mm2
	punpckldq mm3, mm3
	punpckldq mm4, mm4  ;# mm2-mm4 is jx-jz 
	pfsubr mm2, [esp + nb112_ixH]
	pfsubr mm3, [esp + nb112_iyH]
	pfsubr mm4, [esp + nb112_izH] ;# mm2-mm4 is dxH-dzH 
	
	movq [esp + nb112_dxH], mm2
	movq [esp + nb112_dyH], mm3
	movq [esp + nb112_dzH], mm4
	pfmul mm2,mm2
	pfmul mm3,mm3
	pfmul mm4,mm4

	pfadd mm3,mm2
	pfadd mm3,mm4		;# mm3=rsqH 

    	pfrsqrt mm1,mm0

    	movq mm2,mm1
    	pfmul mm1,mm1
    	pfrsqit1 mm1,mm0				
    	pfrcpit2 mm1,mm2	;# mm1=invsqrt 
	movq  mm4, mm1
	pfmul mm4, mm4		;# mm4=invsq 
	;# calculate potential and scalar force 
	pfmul mm6, mm1		;# mm6=vcoul 
	pfmul mm4, mm6		;# mm4=fscalar  

	pfrsqrt mm5, mm3
	pswapd mm3,mm3
	pfrsqrt mm2, mm3
	pswapd mm3,mm3
	punpckldq mm5,mm2	;# seeds are in mm5 now, and rsq in mm3. 

	movq mm2, mm5
	pfmul mm5,mm5
    	pfrsqit1 mm5,mm3				
    	pfrcpit2 mm5,mm2	;# mm5=invsqrt 
	movq mm3,mm5
	pfmul mm3,mm3		;# mm3=invsq 
	pfmul mm7, mm5		;# mm7=vcoul 
	pfmul mm3, mm7		;# mm3=fscal for the two H's. 

	;# update vctot 
	pfadd mm7, mm6
	pfadd mm7, [esp + nb112_vctot]
	movq [esp + nb112_vctot], mm7
	
	;# spread oxygen fscalar to both positions 
	punpckldq mm4,mm4
	;# calc vectorial force for O 
	movq mm0,  [esp + nb112_dxO]
	movd mm1,  [esp + nb112_dzO]
	pfmul mm0, mm4
	pfmul mm1, mm4

	;# calc vectorial force for H's 
	movq mm5, [esp + nb112_dxH]
	movq mm6, [esp + nb112_dyH]
	movq mm7, [esp + nb112_dzH]
	pfmul mm5, mm3
	pfmul mm6, mm3
	pfmul mm7, mm3
	
	;# update iO particle force 
	movq mm2,  [esp + nb112_fixO]
	movd mm3,  [esp + nb112_fizO]
	pfadd mm2, mm0
	pfadd mm3, mm1
	movq [esp + nb112_fixO], mm2
	movd [esp + nb112_fizO], mm3

	;# update iH forces 
	movq mm2, [esp + nb112_fixH]
	movq mm3, [esp + nb112_fiyH]
	movq mm4, [esp + nb112_fizH]
	pfadd mm2, mm5
	pfadd mm3, mm6
	pfadd mm4, mm7
	movq [esp + nb112_fixH], mm2
	movq [esp + nb112_fiyH], mm3
	movq [esp + nb112_fizH], mm4	

	;# pack j forces from H in the same form as the oxygen force. 
	pfacc mm5, mm6		;# mm5(l)=fjx(H1+ h2) mm5(h)=fjy(H1+ h2) 
	pfacc mm7, mm7		;# mm7(l)=fjz(H1+ h2) 
	
	pfadd mm0, mm5		;# add up total force on j particle.  
	pfadd mm1, mm7

	;# update j particle force 
	movq mm2,  [edi + eax*4 + 24]
	movd mm3,  [edi + eax*4 + 32]
	pfsub mm2, mm0
	pfsub mm3, mm1
	movq [edi + eax*4 + 24], mm2
	movd [edi + eax*4 + 32], mm3
	
	;#  done  - one more? 
	dec dword ptr [esp + nb112_innerk]
	jz  .nb112_updateouterdata
	jmp .nb112_inner_loop	
.nb112_updateouterdata:	
	mov   ecx, [esp + nb112_ii3]

	movq  mm6, [edi + ecx*4]       ;# increment iO force  
	movd  mm7, [edi + ecx*4 + 8]	
	pfadd mm6, [esp + nb112_fixO]
	pfadd mm7, [esp + nb112_fizO]
	movq  [edi + ecx*4],    mm6
	movd  [edi + ecx*4 +8], mm7

	movq  mm0, [esp + nb112_fixH]
	movq  mm3, [esp + nb112_fiyH]
	movq  mm1, [esp + nb112_fizH]
	movq  mm2, mm0
	punpckldq mm0, mm3	;# mm0(l)=fxH1, mm0(h)=fyH1 
	punpckhdq mm2, mm3	;# mm2(l)=fxH2, mm2(h)=fyH2 
	movq mm3, mm1
	pswapd mm3,mm3		
	;# mm1 is fzH1 
	;# mm3 is fzH2 

	movq  mm6, [edi + ecx*4 + 12]       ;# increment iH1 force  
	movd  mm7, [edi + ecx*4 + 20] 	
	pfadd mm6, mm0
	pfadd mm7, mm1
	movq  [edi + ecx*4 + 12],  mm6
	movd  [edi + ecx*4 + 20],  mm7
	
	movq  mm6, [edi + ecx*4 + 24]       ;# increment iH2 force 
	movd  mm7, [edi + ecx*4 + 32] 	
	pfadd mm6, mm2
	pfadd mm7, mm3
	movq  [edi + ecx*4 + 24],  mm6
	movd  [edi + ecx*4 + 32],  mm7

	
	mov   ebx, [ebp + nb112_fshift]    ;# increment fshift force 
	mov   edx, [esp + nb112_is3]

	movq  mm6, [ebx + edx*4]	
	movd  mm7, [ebx + edx*4 + 8]	
	pfadd mm6, [esp + nb112_fixO]
	pfadd mm7, [esp + nb112_fizO]
	pfadd mm6, mm0
	pfadd mm7, mm1
	pfadd mm6, mm2
	pfadd mm7, mm3
	movq  [ebx + edx*4],     mm6
	movd  [ebx + edx*4 + 8], mm7
	
	;# get n from stack
	mov esi, [esp + nb112_n]
        ;# get group index for i particle 
        mov   edx, [ebp + nb112_gid]      	;# base of gid[]
        mov   edx, [edx + esi*4]		;# ggid=gid[n]

	movq  mm7, [esp + nb112_vctot]     
	pfacc mm7,mm7	          ;# get and sum the two parts of total potential 

	mov   eax, [ebp + nb112_Vc]
	movd  mm6, [eax + edx*4] 
	pfadd mm6, mm7
	movd  [eax + edx*4], mm6          ;# increment vc[gid] 

	movq  mm7, [esp + nb112_Vvdwtot]     
	pfacc mm7,mm7	          ;# get and sum the two parts of total potential 

	mov   eax, [ebp + nb112_Vvdw]
	movd  mm6, [eax + edx*4] 
	pfadd mm6, mm7
	movd  [eax + edx*4], mm6          ;# increment Vvdwtot[gid] 
       ;# finish if last 
        mov ecx, [esp + nb112_nn1]
	;# esi already loaded with n
	inc esi
        sub ecx, esi
        jecxz .nb112_outerend

        ;# not last, iterate outer loop once more!  
        mov [esp + nb112_n], esi
        jmp .nb112_outer
.nb112_outerend:
        ;# check if more outer neighborlists remain
        mov   ecx, [esp + nb112_nri]
	;# esi already loaded with n above
        sub   ecx, esi
        jecxz .nb112_end
        ;# non-zero, do one more workunit
        jmp   .nb112_threadloop
.nb112_end:
	femms

	mov eax, [esp + nb112_nouter] 	
	mov ebx, [esp + nb112_ninner]
	mov ecx, [ebp + nb112_outeriter]
	mov edx, [ebp + nb112_inneriter]
	mov [ecx], eax
	mov [edx], ebx
	
	add esp, 220
	pop edi
	pop esi
    	pop edx
    	pop ecx
    	pop ebx
    	pop eax
	leave
	ret


.globl nb_kernel112nf_ia32_3dnow
.globl _nb_kernel112nf_ia32_3dnow
nb_kernel112nf_ia32_3dnow:	
_nb_kernel112nf_ia32_3dnow:	
.equiv		nb112nf_p_nri,		8
.equiv		nb112nf_iinr,		12
.equiv		nb112nf_jindex,		16
.equiv		nb112nf_jjnr,		20
.equiv		nb112nf_shift,		24
.equiv		nb112nf_shiftvec,	28
.equiv		nb112nf_fshift,		32
.equiv		nb112nf_gid,		36
.equiv		nb112nf_pos,		40		
.equiv		nb112nf_faction,	44
.equiv		nb112nf_charge,		48
.equiv		nb112nf_p_facel,		52
.equiv		nb112nf_p_krf,		56	
.equiv		nb112nf_p_crf,		60	
.equiv		nb112nf_Vc,		64	
.equiv		nb112nf_type,		68
.equiv		nb112nf_p_ntype,	72
.equiv		nb112nf_vdwparam,	76	
.equiv		nb112nf_Vvdw,		80	
.equiv		nb112nf_p_tabscale,	84	
.equiv		nb112nf_VFtab,		88
.equiv		nb112nf_invsqrta,	92	
.equiv		nb112nf_dvda,		96
.equiv          nb112nf_p_gbtabscale,   100
.equiv          nb112nf_GBtab,          104
.equiv          nb112nf_p_nthreads,     108
.equiv          nb112nf_count,          112
.equiv          nb112nf_mtx,            116
.equiv          nb112nf_outeriter,      120
.equiv          nb112nf_inneriter,      124
.equiv          nb112nf_work,           128
			;# stack offsets for local variables 
.equiv		nb112nf_is3,		0
.equiv		nb112nf_ii3,		4
.equiv		nb112nf_ixO,		8
.equiv		nb112nf_iyO,		12
.equiv		nb112nf_izO,		16	
.equiv		nb112nf_ixH,		20  
.equiv		nb112nf_iyH,		28  
.equiv		nb112nf_izH,		36  
.equiv		nb112nf_qqOO,		44  
.equiv		nb112nf_qqOH,		52  
.equiv		nb112nf_qqHH,		60  
.equiv		nb112nf_c6,		68  
.equiv		nb112nf_c12,		76 
.equiv		nb112nf_vctot,		84 
.equiv		nb112nf_Vvdwtot,	92 
.equiv		nb112nf_innerjjnr,	100
.equiv		nb112nf_innerk,		104
.equiv          nb112nf_n,              108 ;# idx for outer loop
.equiv          nb112nf_nn1,            112 ;# number of outer iterations