nb_kernel213_ia32_sse2.intel_syntax.s
来自「最著名最快的分子模拟软件」· S 代码 · 共 2,246 行 · 第 1/5 页
S
2,246 行
movapd xmm7, xmm4 ;# rsqO in xmm7 ;# move ixH1-izH1 to xmm4-xmm6 movapd xmm4, [esp + nb213_ixH1] movapd xmm5, [esp + nb213_iyH1] movapd xmm6, [esp + nb213_izH1] ;# calc dr subpd xmm4, xmm0 subpd xmm5, xmm1 subpd xmm6, xmm2 ;# store dr movapd [esp + nb213_dxH1], xmm4 movapd [esp + nb213_dyH1], xmm5 movapd [esp + nb213_dzH1], xmm6 ;# square it mulpd xmm4,xmm4 mulpd xmm5,xmm5 mulpd xmm6,xmm6 addpd xmm6, xmm5 addpd xmm6, xmm4 ;# rsqH1 in xmm6 ;# move ixH2-izH2 to xmm3-xmm5 movapd xmm3, [esp + nb213_ixH2] movapd xmm4, [esp + nb213_iyH2] movapd xmm5, [esp + nb213_izH2] ;# calc dr subpd xmm3, xmm0 subpd xmm4, xmm1 subpd xmm5, xmm2 ;# store dr movapd [esp + nb213_dxH2], xmm3 movapd [esp + nb213_dyH2], xmm4 movapd [esp + nb213_dzH2], xmm5 ;# square it mulpd xmm3,xmm3 mulpd xmm4,xmm4 mulpd xmm5,xmm5 addpd xmm5, xmm4 addpd xmm5, xmm3 ;# move ixM-izM to xmm2-xmm4 movapd xmm3, [esp + nb213_iyM] movapd xmm4, [esp + nb213_izM] subpd xmm3, xmm1 subpd xmm4, xmm2 movapd xmm2, [esp + nb213_ixM] subpd xmm2, xmm0 ;# store dr movapd [esp + nb213_dxM], xmm2 movapd [esp + nb213_dyM], xmm3 movapd [esp + nb213_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 ;# calculate krsq movapd xmm0, [esp + nb213_krf] movapd xmm1, xmm0 movapd xmm2, xmm0 mulpd xmm0, xmm4 mulpd xmm1, xmm5 mulpd xmm2, xmm6 movapd [esp + nb213_krsqM], xmm0 movapd [esp + nb213_krsqH2], xmm1 movapd [esp + nb213_krsqH1], xmm2 ;# start with rsqH1 - put seed in xmm2 cvtpd2ps xmm2, xmm6 rsqrtps xmm2, xmm2 cvtps2pd xmm2, xmm2 movapd xmm3, xmm2 mulpd xmm2, xmm2 movapd xmm1, [esp + nb213_three] mulpd xmm2, xmm6 ;# rsq*lu*lu subpd xmm1, xmm2 ;# 30-rsq*lu*lu mulpd xmm1, xmm3 ;# lu*(3-rsq*lu*lu) mulpd xmm1, [esp + nb213_half] ;# iter1 ( new lu) movapd xmm3, xmm1 mulpd xmm1, xmm1 ;# lu*lu mulpd xmm6, xmm1 ;# rsq*lu*lu movapd xmm1, [esp + nb213_three] subpd xmm1, xmm6 ;# 3-rsq*lu*lu mulpd xmm1, xmm3 ;# lu*( 3-rsq*lu*lu) mulpd xmm1, [esp + nb213_half] ;# rinv movapd [esp + nb213_rinvH1], xmm1 ;# rsqH2 - seed in xmm2 cvtpd2ps xmm2, xmm5 rsqrtps xmm2, xmm2 cvtps2pd xmm2, xmm2 movapd xmm3, xmm2 mulpd xmm2, xmm2 movapd xmm1, [esp + nb213_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 + nb213_half] ;# iter1 ( new lu) movapd xmm3, xmm1 mulpd xmm1, xmm1 ;# lu*lu mulpd xmm5, xmm1 ;# rsq*lu*lu movapd xmm1, [esp + nb213_three] subpd xmm1, xmm5 ;# 3-rsq*lu*lu mulpd xmm1, xmm3 ;# lu*( 3-rsq*lu*lu) mulpd xmm1, [esp + nb213_half] ;# rinv movapd [esp + nb213_rinvH2], xmm1 ;# rsqM - seed in xmm2 cvtpd2ps xmm2, xmm4 rsqrtps xmm2, xmm2 cvtps2pd xmm2, xmm2 movapd xmm3, xmm2 mulpd xmm2, xmm2 movapd xmm1, [esp + nb213_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 + nb213_half] ;# iter1 ( new lu) movapd xmm3, xmm1 mulpd xmm1, xmm1 ;# lu*lu mulpd xmm4, xmm1 ;# rsq*lu*lu movapd xmm1, [esp + nb213_three] subpd xmm1, xmm4 ;# 3-rsq*lu*lu mulpd xmm1, xmm3 ;# lu*( 3-rsq*lu*lu) mulpd xmm1, [esp + nb213_half] ;# rinv movapd [esp + nb213_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 + nb213_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 + nb213_c6] mulpd xmm2, [esp + nb213_c12] movapd xmm3, xmm2 subpd xmm3, xmm1 ;# Vvdw=Vvdw12-Vvdw6 addpd xmm3, [esp + nb213_Vvdwtot] mulpd xmm1, [esp + nb213_six] mulpd xmm2, [esp + nb213_twelve] subpd xmm2, xmm1 mulpd xmm2, xmm0 movapd xmm4, xmm2 ;# total fsO movapd [esp + nb213_Vvdwtot], xmm3 movapd xmm0, [esp + nb213_dxO] movapd xmm1, [esp + nb213_dyO] movapd xmm2, [esp + nb213_dzO] mulpd xmm0, xmm4 mulpd xmm1, xmm4 mulpd xmm2, xmm4 ;# update O forces movapd xmm3, [esp + nb213_fixO] movapd xmm4, [esp + nb213_fiyO] movapd xmm7, [esp + nb213_fizO] addpd xmm3, xmm0 addpd xmm4, xmm1 addpd xmm7, xmm2 movapd [esp + nb213_fixO], xmm3 movapd [esp + nb213_fiyO], xmm4 movapd [esp + nb213_fizO], xmm7 ;# update j forces with water O movapd [esp + nb213_fjx], xmm0 movapd [esp + nb213_fjy], xmm1 movapd [esp + nb213_fjz], xmm2 ;# H1 interactions movapd xmm6, [esp + nb213_rinvH1] movapd xmm4, xmm6 mulpd xmm4, xmm4 ;# xmm6=rinv, xmm4=rinvsq movapd xmm7, xmm6 movapd xmm0, [esp + nb213_krsqH1] addpd xmm6, xmm0 ;# xmm6=rinv+ krsq mulpd xmm0, [esp + nb213_two] subpd xmm6, [esp + nb213_crf] subpd xmm7, xmm0 ;# xmm7=rinv-2*krsq mulpd xmm6, [esp + nb213_qqH] ;# vcoul mulpd xmm7, [esp + nb213_qqH] mulpd xmm4, xmm7 ;# total fsH1 in xmm4 addpd xmm6, [esp + nb213_vctot] movapd xmm0, [esp + nb213_dxH1] movapd xmm1, [esp + nb213_dyH1] movapd xmm2, [esp + nb213_dzH1] movapd [esp + nb213_vctot], xmm6 mulpd xmm0, xmm4 mulpd xmm1, xmm4 mulpd xmm2, xmm4 ;# update H1 forces movapd xmm3, [esp + nb213_fixH1] movapd xmm4, [esp + nb213_fiyH1] movapd xmm7, [esp + nb213_fizH1] addpd xmm3, xmm0 addpd xmm4, xmm1 addpd xmm7, xmm2 movapd [esp + nb213_fixH1], xmm3 movapd [esp + nb213_fiyH1], xmm4 movapd [esp + nb213_fizH1], xmm7 ;# update j forces with water H1 addpd xmm0, [esp + nb213_fjx] addpd xmm1, [esp + nb213_fjy] addpd xmm2, [esp + nb213_fjz] movapd [esp + nb213_fjx], xmm0 movapd [esp + nb213_fjy], xmm1 movapd [esp + nb213_fjz], xmm2 ;# H2 interactions movapd xmm5, [esp + nb213_rinvH2] movapd xmm4, xmm5 mulpd xmm4, xmm4 ;# xmm5=rinv, xmm4=rinvsq movapd xmm7, xmm5 movapd xmm0, [esp + nb213_krsqH2] addpd xmm5, xmm0 ;# xmm5=rinv+ krsq mulpd xmm0, [esp + nb213_two] subpd xmm5, [esp + nb213_crf] subpd xmm7, xmm0 ;# xmm7=rinv-2*krsq mulpd xmm5, [esp + nb213_qqH] ;# vcoul mulpd xmm7, [esp + nb213_qqH] mulpd xmm4, xmm7 ;# total fsH2 in xmm4 addpd xmm5, [esp + nb213_vctot] movapd xmm0, [esp + nb213_dxH2] movapd xmm1, [esp + nb213_dyH2] movapd xmm2, [esp + nb213_dzH2] movapd [esp + nb213_vctot], xmm5 mulpd xmm0, xmm4 mulpd xmm1, xmm4 mulpd xmm2, xmm4 ;# update H2 forces movapd xmm3, [esp + nb213_fixH2] movapd xmm4, [esp + nb213_fiyH2] movapd xmm7, [esp + nb213_fizH2] addpd xmm3, xmm0 addpd xmm4, xmm1 addpd xmm7, xmm2 movapd [esp + nb213_fixH2], xmm3 movapd [esp + nb213_fiyH2], xmm4 movapd [esp + nb213_fizH2], xmm7 ;# update j forces with water H2 addpd xmm0, [esp + nb213_fjx] addpd xmm1, [esp + nb213_fjy] addpd xmm2, [esp + nb213_fjz] movapd [esp + nb213_fjx], xmm0 movapd [esp + nb213_fjy], xmm1 movapd [esp + nb213_fjz], xmm2 ;# M interactions movapd xmm5, [esp + nb213_rinvM] movapd xmm4, xmm5 mulpd xmm4, xmm4 ;# xmm5=rinv, xmm4=rinvsq movapd xmm7, xmm5 movapd xmm0, [esp + nb213_krsqM] addpd xmm5, xmm0 ;# xmm5=rinv+ krsq mulpd xmm0, [esp + nb213_two] subpd xmm5, [esp + nb213_crf] subpd xmm7, xmm0 ;# xmm7=rinv-2*krsq mulpd xmm5, [esp + nb213_qqM] ;# vcoul mulpd xmm7, [esp + nb213_qqM] mulpd xmm4, xmm7 ;# total fsH2 in xmm4 addpd xmm5, [esp + nb213_vctot] movapd xmm0, [esp + nb213_dxM] movapd xmm1, [esp + nb213_dyM] movapd xmm2, [esp + nb213_dzM] movapd [esp + nb213_vctot], xmm5 mulpd xmm0, xmm4 mulpd xmm1, xmm4 mulpd xmm2, xmm4 ;# update H2 forces movapd xmm3, [esp + nb213_fixM] movapd xmm4, [esp + nb213_fiyM] movapd xmm7, [esp + nb213_fizM] addpd xmm3, xmm0 addpd xmm4, xmm1 addpd xmm7, xmm2 movapd [esp + nb213_fixM], xmm3 movapd [esp + nb213_fiyM], xmm4 movapd [esp + nb213_fizM], xmm7 mov edi, [ebp + nb213_faction] ;# update j forces addpd xmm0, [esp + nb213_fjx] addpd xmm1, [esp + nb213_fjy] addpd xmm2, [esp + nb213_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 + nb213_innerk], 2 jl .nb213_checksingle jmp .nb213_unroll_loop.nb213_checksingle: mov edx, [esp + nb213_innerk] and edx, 1 jnz .nb213_dosingle jmp .nb213_updateouterdata.nb213_dosingle: mov edx, [esp + nb213_innerjjnr] ;# pointer to jjnr[k] mov eax, [edx] add dword ptr [esp + nb213_innerjjnr], 4 mov esi, [ebp + nb213_charge] ;# base of charge[] xorpd xmm3, xmm3 movlpd xmm3, [esi + eax*8] movapd xmm4, xmm3 mulsd xmm3, [esp + nb213_iqM] mulsd xmm4, [esp + nb213_iqH] movd mm0, eax ;# use mmx registers as temp storage movapd [esp + nb213_qqM], xmm3 movapd [esp + nb213_qqH], xmm4 mov esi, [ebp + nb213_type] mov eax, [esi + eax*4] mov esi, [ebp + nb213_vdwparam] shl eax, 1 mov edi, [esp + nb213_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 + nb213_c6], xmm4 movapd [esp + nb213_c12], xmm6 mov esi, [ebp + nb213_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 + nb213_ixO] movapd xmm5, [esp + nb213_iyO] movapd xmm6, [esp + nb213_izO] ;# calc dr subsd xmm4, xmm0 subsd xmm5, xmm1 subsd xmm6, xmm2 ;# store dr movapd [esp + nb213_dxO], xmm4 movapd [esp + nb213_dyO], xmm5 movapd [esp + nb213_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 + nb213_ixH1] movapd xmm5, [esp + nb213_iyH1] movapd xmm6, [esp + nb213_izH1] ;# calc dr subsd xmm4, xmm0 subsd xmm5, xmm1 subsd xmm6, xmm2 ;# store dr movapd [esp + nb213_dxH1], xmm4 movapd [esp + nb213_dyH1], xmm5 movapd [esp + nb213_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 + nb213_ixH2] movapd xmm4, [esp + nb213_iyH2] movapd xmm5, [esp + nb213_izH2] ;# calc dr subsd xmm3, xmm0 subsd xmm4, xmm1 subsd xmm5, xmm2 ;# store dr movapd [esp + nb213_dxH2], xmm3 movapd [esp + nb213_dyH2], xmm4 movapd [esp + nb213_dzH2], xmm5 ;# square it ⌨️ 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?