nb_kernel113_ia32_sse2.intel_syntax.s
来自「最著名最快的分子模拟软件」· S 代码 · 共 2,098 行 · 第 1/4 页
S
2,098 行
cvtpd2ps xmm2, xmm5 rsqrtps xmm2, xmm2 cvtps2pd xmm2, xmm2 movapd xmm3, xmm2 mulpd xmm2, xmm2 movapd xmm1, [esp + nb113_three] mulpd xmm2, xmm5 ;# rsq*lu*lu subpd xmm1, xmm2 ;# 30-rsq*lu*lu mulpd xmm1, xmm3 ;# lu*(3-rsq*lu*lu) mulpd xmm1, [esp + nb113_half] ;# iter1 ( new lu) movapd xmm3, xmm1 mulpd xmm1, xmm1 ;# lu*lu mulpd xmm5, xmm1 ;# rsq*lu*lu movapd xmm1, [esp + nb113_three] subpd xmm1, xmm5 ;# 3-rsq*lu*lu mulpd xmm1, xmm3 ;# lu*( 3-rsq*lu*lu) mulpd xmm1, [esp + nb113_half] ;# rinv movapd [esp + nb113_rinvH2], xmm1 ;# rsqM - seed in xmm2 cvtpd2ps xmm2, xmm4 rsqrtps xmm2, xmm2 cvtps2pd xmm2, xmm2 movapd xmm3, xmm2 mulpd xmm2, xmm2 movapd xmm1, [esp + nb113_three] mulpd xmm2, xmm4 ;# rsq*lu*lu subpd xmm1, xmm2 ;# 30-rsq*lu*lu mulpd xmm1, xmm3 ;# lu*(3-rsq*lu*lu) mulpd xmm1, [esp + nb113_half] ;# iter1 ( new lu) movapd xmm3, xmm1 mulpd xmm1, xmm1 ;# lu*lu mulpd xmm4, xmm1 ;# rsq*lu*lu movapd xmm1, [esp + nb113_three] subpd xmm1, xmm4 ;# 3-rsq*lu*lu mulpd xmm1, xmm3 ;# lu*( 3-rsq*lu*lu) mulpd xmm1, [esp + nb113_half] ;# rinv movapd [esp + nb113_rinvM], xmm1 ;# do O interactions directly - rsqO is in xmm7 cvtpd2ps xmm2, xmm7 movapd xmm6, xmm7 rcpps xmm2, xmm2 cvtps2pd xmm2, xmm2 movapd xmm1, [esp + nb113_two] movapd xmm0, xmm1 mulpd xmm7, xmm2 subpd xmm1, xmm7 mulpd xmm2, xmm1 ;# iter1 mulpd xmm6, xmm2 subpd xmm0, xmm6 mulpd xmm0, xmm2 ;# xmm0=rinvsq movapd xmm1, xmm0 mulpd xmm1, xmm1 ;# rinv4 mulpd xmm1, xmm0 ;#rinvsix movapd xmm2, xmm1 mulpd xmm2, xmm2 ;# rinvtwelve mulpd xmm1, [esp + nb113_c6] mulpd xmm2, [esp + nb113_c12] movapd xmm3, xmm2 subpd xmm3, xmm1 ;# Vvdw=Vvdw12-Vvdw6 addpd xmm3, [esp + nb113_Vvdwtot] mulpd xmm1, [esp + nb113_six] mulpd xmm2, [esp + nb113_twelve] subpd xmm2, xmm1 mulpd xmm2, xmm0 movapd xmm4, xmm2 ;# total fsO movapd [esp + nb113_Vvdwtot], xmm3 movapd xmm0, [esp + nb113_dxO] movapd xmm1, [esp + nb113_dyO] movapd xmm2, [esp + nb113_dzO] mulpd xmm0, xmm4 mulpd xmm1, xmm4 mulpd xmm2, xmm4 ;# update O forces movapd xmm3, [esp + nb113_fixO] movapd xmm4, [esp + nb113_fiyO] movapd xmm7, [esp + nb113_fizO] addpd xmm3, xmm0 addpd xmm4, xmm1 addpd xmm7, xmm2 movapd [esp + nb113_fixO], xmm3 movapd [esp + nb113_fiyO], xmm4 movapd [esp + nb113_fizO], xmm7 ;# update j forces with water O movapd [esp + nb113_fjx], xmm0 movapd [esp + nb113_fjy], xmm1 movapd [esp + nb113_fjz], xmm2 ;# H1 interactions movapd xmm6, [esp + nb113_rinvH1] movapd xmm4, xmm6 mulpd xmm4, xmm4 ;# xmm6=rinv, xmm4=rinvsq mulpd xmm6, [esp + nb113_qqH] ;# xmm6=vcoul mulpd xmm4, xmm6 ;# total fsH1 in xmm4 addpd xmm6, [esp + nb113_vctot] movapd xmm0, [esp + nb113_dxH1] movapd xmm1, [esp + nb113_dyH1] movapd xmm2, [esp + nb113_dzH1] movapd [esp + nb113_vctot], xmm6 mulpd xmm0, xmm4 mulpd xmm1, xmm4 mulpd xmm2, xmm4 ;# update H1 forces movapd xmm3, [esp + nb113_fixH1] movapd xmm4, [esp + nb113_fiyH1] movapd xmm7, [esp + nb113_fizH1] addpd xmm3, xmm0 addpd xmm4, xmm1 addpd xmm7, xmm2 movapd [esp + nb113_fixH1], xmm3 movapd [esp + nb113_fiyH1], xmm4 movapd [esp + nb113_fizH1], xmm7 ;# update j forces with water H1 addpd xmm0, [esp + nb113_fjx] addpd xmm1, [esp + nb113_fjy] addpd xmm2, [esp + nb113_fjz] movapd [esp + nb113_fjx], xmm0 movapd [esp + nb113_fjy], xmm1 movapd [esp + nb113_fjz], xmm2 ;# H2 interactions movapd xmm5, [esp + nb113_rinvH2] movapd xmm4, xmm5 mulpd xmm4, xmm4 ;# xmm5=rinv, xmm4=rinvsq mulpd xmm5, [esp + nb113_qqH] ;# xmm5=vcoul mulpd xmm4, xmm5 ;# total fsH1 in xmm4 addpd xmm5, [esp + nb113_vctot] movapd xmm0, [esp + nb113_dxH2] movapd xmm1, [esp + nb113_dyH2] movapd xmm2, [esp + nb113_dzH2] movapd [esp + nb113_vctot], xmm5 mulpd xmm0, xmm4 mulpd xmm1, xmm4 mulpd xmm2, xmm4 ;# update H2 forces movapd xmm3, [esp + nb113_fixH2] movapd xmm4, [esp + nb113_fiyH2] movapd xmm7, [esp + nb113_fizH2] addpd xmm3, xmm0 addpd xmm4, xmm1 addpd xmm7, xmm2 movapd [esp + nb113_fixH2], xmm3 movapd [esp + nb113_fiyH2], xmm4 movapd [esp + nb113_fizH2], xmm7 ;# update j forces with water H2 addpd xmm0, [esp + nb113_fjx] addpd xmm1, [esp + nb113_fjy] addpd xmm2, [esp + nb113_fjz] movapd [esp + nb113_fjx], xmm0 movapd [esp + nb113_fjy], xmm1 movapd [esp + nb113_fjz], xmm2 ;# M interactions movapd xmm5, [esp + nb113_rinvM] movapd xmm4, xmm5 mulpd xmm4, xmm4 ;# xmm5=rinv, xmm4=rinvsq mulpd xmm5, [esp + nb113_qqM] ;# xmm5=vcoul mulpd xmm4, xmm5 ;# total fsM in xmm4 addpd xmm5, [esp + nb113_vctot] movapd xmm0, [esp + nb113_dxM] movapd xmm1, [esp + nb113_dyM] movapd xmm2, [esp + nb113_dzM] movapd [esp + nb113_vctot], xmm5 mulpd xmm0, xmm4 mulpd xmm1, xmm4 mulpd xmm2, xmm4 ;# update M forces movapd xmm3, [esp + nb113_fixM] movapd xmm4, [esp + nb113_fiyM] movapd xmm7, [esp + nb113_fizM] addpd xmm3, xmm0 addpd xmm4, xmm1 addpd xmm7, xmm2 movapd [esp + nb113_fixM], xmm3 movapd [esp + nb113_fiyM], xmm4 movapd [esp + nb113_fizM], xmm7 mov edi, [ebp + nb113_faction] ;# update j forces addpd xmm0, [esp + nb113_fjx] addpd xmm1, [esp + nb113_fjy] addpd xmm2, [esp + nb113_fjz] movlpd xmm3, [edi + eax*8] movlpd xmm4, [edi + eax*8 + 8] movlpd xmm5, [edi + eax*8 + 16] movhpd xmm3, [edi + ebx*8] movhpd xmm4, [edi + ebx*8 + 8] movhpd xmm5, [edi + ebx*8 + 16] subpd xmm3, xmm0 subpd xmm4, xmm1 subpd xmm5, xmm2 movlpd [edi + eax*8], xmm3 movlpd [edi + eax*8 + 8], xmm4 movlpd [edi + eax*8 + 16], xmm5 movhpd [edi + ebx*8], xmm3 movhpd [edi + ebx*8 + 8], xmm4 movhpd [edi + ebx*8 + 16], xmm5 ;# should we do one more iteration? sub dword ptr [esp + nb113_innerk], 2 jl .nb113_checksingle jmp .nb113_unroll_loop.nb113_checksingle: mov edx, [esp + nb113_innerk] and edx, 1 jnz .nb113_dosingle jmp .nb113_updateouterdata.nb113_dosingle: mov edx, [esp + nb113_innerjjnr] ;# pointer to jjnr[k] mov eax, [edx] add dword ptr [esp + nb113_innerjjnr], 4 mov esi, [ebp + nb113_charge] ;# base of charge[] xorpd xmm3, xmm3 movlpd xmm3, [esi + eax*8] movapd xmm4, xmm3 mulpd xmm3, [esp + nb113_iqM] mulpd xmm4, [esp + nb113_iqH] movd mm0, eax ;# use mmx registers as temp storage movapd [esp + nb113_qqM], xmm3 movapd [esp + nb113_qqH], xmm4 mov esi, [ebp + nb113_type] mov eax, [esi + eax*4] mov esi, [ebp + nb113_vdwparam] shl eax, 1 mov edi, [esp + nb113_ntia] add eax, edi movlpd xmm6, [esi + eax*8] ;# c6a movhpd xmm6, [esi + eax*8 + 8] ;# c6a c12a xorpd xmm7, xmm7 movapd xmm4, xmm6 unpcklpd xmm4, xmm7 unpckhpd xmm6, xmm7 movd eax, mm0 movd ebx, mm1 movapd [esp + nb113_c6], xmm4 movapd [esp + nb113_c12], xmm6 mov esi, [ebp + nb113_pos] ;# base of pos[] lea eax, [eax + eax*2] ;# replace jnr with j3 ;# move coordinates to xmm0-xmm2 movlpd xmm0, [esi + eax*8] movlpd xmm1, [esi + eax*8 + 8] movlpd xmm2, [esi + eax*8 + 16] ;# move ixO-izO to xmm4-xmm6 movapd xmm4, [esp + nb113_ixO] movapd xmm5, [esp + nb113_iyO] movapd xmm6, [esp + nb113_izO] ;# calc dr subsd xmm4, xmm0 subsd xmm5, xmm1 subsd xmm6, xmm2 ;# store dr movapd [esp + nb113_dxO], xmm4 movapd [esp + nb113_dyO], xmm5 movapd [esp + nb113_dzO], xmm6 ;# square it mulsd xmm4,xmm4 mulsd xmm5,xmm5 mulsd xmm6,xmm6 addsd xmm4, xmm5 addsd xmm4, xmm6 movapd xmm7, xmm4 ;# rsqO in xmm7 ;# move ixH1-izH1 to xmm4-xmm6 movapd xmm4, [esp + nb113_ixH1] movapd xmm5, [esp + nb113_iyH1] movapd xmm6, [esp + nb113_izH1] ;# calc dr subsd xmm4, xmm0 subsd xmm5, xmm1 subsd xmm6, xmm2 ;# store dr movapd [esp + nb113_dxH1], xmm4 movapd [esp + nb113_dyH1], xmm5 movapd [esp + nb113_dzH1], xmm6 ;# square it mulsd xmm4,xmm4 mulsd xmm5,xmm5 mulsd xmm6,xmm6 addsd xmm6, xmm5 addsd xmm6, xmm4 ;# rsqH1 in xmm6 ;# move ixH2-izH2 to xmm3-xmm5 movapd xmm3, [esp + nb113_ixH2] movapd xmm4, [esp + nb113_iyH2] movapd xmm5, [esp + nb113_izH2] ;# calc dr subsd xmm3, xmm0 subsd xmm4, xmm1 subsd xmm5, xmm2 ;# store dr movapd [esp + nb113_dxH2], xmm3 movapd [esp + nb113_dyH2], xmm4 movapd [esp + nb113_dzH2], xmm5 ;# square it mulsd xmm3,xmm3 mulsd xmm4,xmm4 mulsd xmm5,xmm5 addsd xmm5, xmm4 addsd xmm5, xmm3 ;# move ixM-izM to xmm2-xmm4 movapd xmm3, [esp + nb113_iyM] movapd xmm4, [esp + nb113_izM] subpd xmm3, xmm1 subpd xmm4, xmm2 movapd xmm2, [esp + nb113_ixM] subpd xmm2, xmm0 ;# store dr movapd [esp + nb113_dxM], xmm2 movapd [esp + nb113_dyM], xmm3 movapd [esp + nb113_dzM], xmm4 ;# square it mulpd xmm2,xmm2 mulpd xmm3,xmm3 mulpd xmm4,xmm4 addpd xmm4, xmm3 addpd xmm4, xmm2 ;# rsqM in xmm4, rsqH2 in xmm5, rsqH1 in xmm6, rsqO in xmm7 ;# start with rsqH1 - put seed in xmm2 cvtsd2ss xmm2, xmm6 rsqrtss xmm2, xmm2 cvtss2sd xmm2, xmm2 movapd xmm3, xmm2 mulsd xmm2, xmm2 movapd xmm1, [esp + nb113_three] mulsd xmm2, xmm6 ;# rsq*lu*lu subsd xmm1, xmm2 ;# 30-rsq*lu*lu mulsd xmm1, xmm3 ;# lu*(3-rsq*lu*lu) mulsd xmm1, [esp + nb113_half] ;# iter1 ( new lu) movapd xmm3, xmm1 mulsd xmm1, xmm1 ;# lu*lu mulsd xmm6, xmm1 ;# rsq*lu*lu movapd xmm1, [esp + nb113_three] subsd xmm1, xmm6 ;# 3-rsq*lu*lu mulsd xmm1, xmm3 ;# lu*( 3-rsq*lu*lu) mulsd xmm1, [esp + nb113_half] ;# rinv movapd [esp + nb113_rinvH1], xmm1 ;# rsqH2 - seed in xmm2 cvtsd2ss xmm2, xmm5 rsqrtss xmm2, xmm2 cvtss2sd xmm2, xmm2 movapd xmm3, xmm2 mulsd xmm2, xmm2 movapd xmm1, [esp + nb113_three] mulsd xmm2, xmm5 ;# rsq*lu*lu subsd xmm1, xmm2 ;# 30-rsq*lu*lu mulsd xmm1, xmm3 ;# lu*(3-rsq*lu*lu) mulsd xmm1, [esp + nb113_half] ;# iter1 ( new lu) movapd xmm3, xmm1 mulsd xmm1, xmm1 ;# lu*lu mulsd xmm5, xmm1 ;# rsq*lu*lu movapd xmm1, [esp + nb113_three] subsd xmm1, xmm5 ;# 3-rsq*lu*lu mulsd xmm1, xmm3 ;# lu*( 3-rsq*lu*lu) mulsd xmm1, [esp + nb113_half] ;# rinv movapd [esp + nb113_rinvH2], xmm1 ;# rsqM - seed in xmm2 cvtsd2ss xmm2, xmm4 rsqrtss xmm2, xmm2 cvtss2sd xmm2, xmm2 movapd xmm3, xmm2 mulsd xmm2, xmm2 movapd xmm1, [esp + nb113_three] mulsd xmm2, xmm4 ;# rsq*lu*lu subsd xmm1, xmm2 ;# 30-rsq*lu*lu mulsd xmm1, xmm3 ;# lu*(3-rsq*lu*lu) mulsd xmm1, [esp + nb113_half] ;# iter1 ( new lu) movapd xmm3, xmm1 mulsd xmm1, xmm1 ;# lu*lu mulsd xmm4, xmm1 ;# rsq*lu*lu movapd xmm1, [esp + nb113_three] subsd xmm1, xmm4 ;# 3-rsq*lu*lu mulsd xmm1, xmm3 ;# lu*( 3-rsq*lu*lu) mulsd xmm1, [esp + nb113_half] ;# rinv movapd [esp + nb113_rinvM], xmm1 ;# do O interactions directly. xmm7=rsq cvtsd2ss xmm2, xmm7 movapd xmm6, xmm7 rcpps xmm2, xmm2 cvtss2sd xmm2, xmm2 movapd xmm1, [esp + nb113_two] movapd xmm0, xmm1 mulsd xmm7, xmm2 subsd xmm1, xmm7 mulsd xmm2, xmm1 ;# iter1 mulsd xmm6, xmm2 subsd xmm0, xmm6 mulsd xmm0, xmm2 ;# xmm0=rinvsq movapd xmm1, xmm0 mulsd xmm1, xmm1 ;# rinv4 mulsd xmm1, xmm0 ;#rinvsix movapd xmm2, xmm1 mulsd xmm2, xmm2 ;# rinvtwelve mulsd xmm1, [esp + nb113_c6] mulsd xmm2, [esp + nb113_c12] movapd xmm3, xmm2 subsd xmm3, xmm1 ;# Vvdw=Vvdw12-Vvdw6 addsd xmm3, [esp + nb113_Vvdwtot] mulsd xmm1, [esp + nb113_six] mulsd xmm2, [esp + nb113_twelve] subsd xmm2, xmm1 mulsd xmm2, xmm0 movapd xmm4, xmm2 ;# total fsO movsd [esp + nb113_Vvdwtot], xmm3 movapd xmm0, [esp + nb113_dxO] movapd xmm1, [esp + nb113_dyO] movapd xmm2, [esp + nb113_dzO] mulsd xmm0, xmm4 mulsd xmm1, xmm4 mulsd xmm2, xmm4 ;# update O forces movapd xmm3, [esp + nb113_fixO] movapd xmm4, [esp + nb113_fiyO] movapd xmm7, [esp + nb113_fizO] addsd xmm3, xmm0 addsd xmm4, xmm1 addsd xmm7, xmm2 movsd [esp + nb113_fixO], xmm3 movsd [esp + nb113_fiyO], xmm4 movsd [esp + nb113_fizO], xmm7 ;# update j forces with water O movsd [esp + nb113_fjx], xmm0 movsd [esp + nb113_fjy], xmm1 movsd [esp + nb113_fjz], xmm2 ;# H1 interactions movapd xmm6, [esp + nb113_rinvH1] movapd xmm4, xmm6 mulsd xmm4, xmm4 ;# xmm6=rinv, xmm4=rinvsq mulsd xmm6, [esp + nb113_qqH] ;# xmm6=vcoul mulsd xmm4, xmm6 ;# total fsH1 in xmm4 addsd xmm6, [esp + nb113_vctot] movapd xmm0, [esp + nb113_dxH1] movapd xmm1, [esp + nb113_dyH1] movapd xmm2, [esp + nb113_dzH1] movsd [esp + nb113_vctot], xmm6 mulsd xmm0, xmm4 mulsd xmm1, xmm4 mulsd xmm2, xmm4 ;# update H1 forces movapd xmm3, [esp + nb113_fixH1] movapd xmm4, [esp + nb113_fiyH1] movapd xmm7, [esp + nb113_fizH1] addsd xmm3, xmm0 addsd xmm4, xmm1 addsd xmm7, xmm2 movsd [esp + nb113_fixH1], xmm3 movsd [esp + nb113_fiyH1], xmm4 movsd [esp + nb113_fizH1], xmm7 ;# update j forces with water H1 addsd xmm0, [esp + nb113_fjx] addsd xmm1, [esp + nb113_fjy] addsd xmm2, [esp + nb113_fjz] movsd [esp + nb113_fjx], xmm0 movsd [esp + nb113_fjy], xmm1 movsd [esp + nb113_fjz], xmm2 ;# H2 interactions movapd xmm5, [esp + nb113_rinvH2] movapd xmm4, xmm5 mulsd xmm4, xmm4 ;# xmm5=rinv, xmm4=rinvsq mulsd xmm5, [esp + nb113_qqH] ;# xmm5=vcoul mulsd xmm4, xmm5 ;# total fsH1 in xmm4 addsd xmm5, [esp + nb113_vctot] movapd xmm0, [esp + nb113_dxH2] movapd xmm1, [esp + nb113_dyH2] movapd xmm2, [esp + nb113_dzH2] movsd [esp + nb113_vctot], xmm5 mulsd xmm0, xmm4 mulsd xmm1, xmm4 mulsd xmm2, xmm4⌨️ 快捷键说明
复制代码Ctrl + C
搜索代码Ctrl + F
全屏模式F11
增大字号Ctrl + =
减小字号Ctrl + -
显示快捷键?