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
mulsd xmm3,xmm3 mulsd xmm4,xmm4 mulsd xmm5,xmm5 addsd xmm5, xmm4 addsd 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 movsd xmm0, [esp + nb213_krf] movsd xmm1, xmm0 movsd xmm2, xmm0 mulsd xmm0, xmm4 mulsd xmm1, xmm5 mulsd xmm2, xmm6 movsd [esp + nb213_krsqM], xmm0 movsd [esp + nb213_krsqH2], xmm1 movsd [esp + nb213_krsqH1], xmm2 ;# 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 + nb213_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 + nb213_half] ;# iter1 ( new lu) movapd xmm3, xmm1 mulsd xmm1, xmm1 ;# lu*lu mulsd xmm6, xmm1 ;# rsq*lu*lu movapd xmm1, [esp + nb213_three] subsd xmm1, xmm6 ;# 3-rsq*lu*lu mulsd xmm1, xmm3 ;# lu*( 3-rsq*lu*lu) mulsd xmm1, [esp + nb213_half] ;# rinv movapd [esp + nb213_rinvH1], xmm1 ;# rsqH2 - seed in xmm2 cvtsd2ss xmm2, xmm5 rsqrtss xmm2, xmm2 cvtss2sd xmm2, xmm2 movapd xmm3, xmm2 mulsd xmm2, xmm2 movapd xmm1, [esp + nb213_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 + nb213_half] ;# iter1 ( new lu) movapd xmm3, xmm1 mulsd xmm1, xmm1 ;# lu*lu mulsd xmm5, xmm1 ;# rsq*lu*lu movapd xmm1, [esp + nb213_three] subsd xmm1, xmm5 ;# 3-rsq*lu*lu mulsd xmm1, xmm3 ;# lu*( 3-rsq*lu*lu) mulsd xmm1, [esp + nb213_half] ;# rinv movapd [esp + nb213_rinvH2], xmm1 ;# rsqM - seed in xmm2 cvtsd2ss xmm2, xmm4 rsqrtss xmm2, xmm2 cvtss2sd xmm2, xmm2 movapd xmm3, xmm2 mulsd xmm2, xmm2 movapd xmm1, [esp + nb213_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 + nb213_half] ;# iter1 ( new lu) movapd xmm3, xmm1 mulsd xmm1, xmm1 ;# lu*lu mulsd xmm4, xmm1 ;# rsq*lu*lu movapd xmm1, [esp + nb213_three] subsd xmm1, xmm4 ;# 3-rsq*lu*lu mulsd xmm1, xmm3 ;# lu*( 3-rsq*lu*lu) mulsd xmm1, [esp + nb213_half] ;# rinv movapd [esp + nb213_rinvM], xmm1 ;# do O interactions directly. xmm7=rsq cvtsd2ss xmm2, xmm7 movapd xmm6, xmm7 rcpps xmm2, xmm2 cvtss2sd xmm2, xmm2 movapd xmm1, [esp + nb213_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 + nb213_c6] mulsd xmm2, [esp + nb213_c12] movapd xmm3, xmm2 subsd xmm3, xmm1 ;# Vvdw=Vvdw12-Vvdw6 addsd xmm3, [esp + nb213_Vvdwtot] mulsd xmm1, [esp + nb213_six] mulsd xmm2, [esp + nb213_twelve] subsd xmm2, xmm1 mulsd xmm2, xmm0 movapd xmm4, xmm2 ;# total fsO movsd [esp + nb213_Vvdwtot], xmm3 movapd xmm0, [esp + nb213_dxO] movapd xmm1, [esp + nb213_dyO] movapd xmm2, [esp + nb213_dzO] mulsd xmm0, xmm4 mulsd xmm1, xmm4 mulsd xmm2, xmm4 ;# update O forces movapd xmm3, [esp + nb213_fixO] movapd xmm4, [esp + nb213_fiyO] movapd xmm7, [esp + nb213_fizO] addsd xmm3, xmm0 addsd xmm4, xmm1 addsd xmm7, xmm2 movsd [esp + nb213_fixO], xmm3 movsd [esp + nb213_fiyO], xmm4 movsd [esp + nb213_fizO], xmm7 ;# update j forces with water O movsd [esp + nb213_fjx], xmm0 movsd [esp + nb213_fjy], xmm1 movsd [esp + nb213_fjz], xmm2 ;# H1 interactions movsd xmm6, [esp + nb213_rinvH1] movsd xmm4, xmm6 mulsd xmm4, xmm4 ;# xmm6=rinv, xmm4=rinvsq movsd xmm7, xmm6 movsd xmm0, [esp + nb213_krsqH1] addsd xmm6, xmm0 ;# xmm6=rinv+ krsq mulsd xmm0, [esp + nb213_two] subsd xmm6, [esp + nb213_crf] subsd xmm7, xmm0 ;# xmm7=rinv-2*krsq mulsd xmm6, [esp + nb213_qqH] ;# vcoul mulsd xmm7, [esp + nb213_qqH] mulsd xmm4, xmm7 ;# total fsH1 in xmm4 addsd xmm6, [esp + nb213_vctot] movapd xmm0, [esp + nb213_dxH1] movapd xmm1, [esp + nb213_dyH1] movapd xmm2, [esp + nb213_dzH1] movsd [esp + nb213_vctot], xmm6 mulsd xmm0, xmm4 mulsd xmm1, xmm4 mulsd xmm2, xmm4 ;# update H1 forces movapd xmm3, [esp + nb213_fixH1] movapd xmm4, [esp + nb213_fiyH1] movapd xmm7, [esp + nb213_fizH1] addsd xmm3, xmm0 addsd xmm4, xmm1 addsd xmm7, xmm2 movsd [esp + nb213_fixH1], xmm3 movsd [esp + nb213_fiyH1], xmm4 movsd [esp + nb213_fizH1], xmm7 ;# update j forces with water H1 addsd xmm0, [esp + nb213_fjx] addsd xmm1, [esp + nb213_fjy] addsd xmm2, [esp + nb213_fjz] movsd [esp + nb213_fjx], xmm0 movsd [esp + nb213_fjy], xmm1 movsd [esp + nb213_fjz], xmm2 ;# H2 interactions movsd xmm5, [esp + nb213_rinvH2] movsd xmm4, xmm5 mulsd xmm4, xmm4 ;# xmm5=rinv, xmm4=rinvsq movsd xmm7, xmm5 movsd xmm0, [esp + nb213_krsqH2] addsd xmm5, xmm0 ;# xmm5=rinv+ krsq mulsd xmm0, [esp + nb213_two] subsd xmm5, [esp + nb213_crf] subsd xmm7, xmm0 ;# xmm7=rinv-2*krsq mulsd xmm5, [esp + nb213_qqH] ;# vcoul mulsd xmm7, [esp + nb213_qqH] mulsd xmm4, xmm7 ;# total fsH2 in xmm4 addsd xmm5, [esp + nb213_vctot] movapd xmm0, [esp + nb213_dxH2] movapd xmm1, [esp + nb213_dyH2] movapd xmm2, [esp + nb213_dzH2] movsd [esp + nb213_vctot], xmm5 mulsd xmm0, xmm4 mulsd xmm1, xmm4 mulsd xmm2, xmm4 ;# update H2 forces movapd xmm3, [esp + nb213_fixH2] movapd xmm4, [esp + nb213_fiyH2] movapd xmm7, [esp + nb213_fizH2] addsd xmm3, xmm0 addsd xmm4, xmm1 addsd xmm7, xmm2 movsd [esp + nb213_fixH2], xmm3 movsd [esp + nb213_fiyH2], xmm4 movsd [esp + nb213_fizH2], xmm7 ;# update j forces with water H2 addsd xmm0, [esp + nb213_fjx] addsd xmm1, [esp + nb213_fjy] addsd xmm2, [esp + nb213_fjz] movsd [esp + nb213_fjx], xmm0 movsd [esp + nb213_fjy], xmm1 movsd [esp + nb213_fjz], xmm2 ;# M interactions movsd xmm5, [esp + nb213_rinvM] movsd xmm4, xmm5 mulsd xmm4, xmm4 ;# xmm5=rinv, xmm4=rinvsq movsd xmm7, xmm5 movsd xmm0, [esp + nb213_krsqM] addsd xmm5, xmm0 ;# xmm5=rinv+ krsq mulsd xmm0, [esp + nb213_two] subsd xmm5, [esp + nb213_crf] subsd xmm7, xmm0 ;# xmm7=rinv-2*krsq mulsd xmm5, [esp + nb213_qqM] ;# vcoul mulsd xmm7, [esp + nb213_qqM] mulsd xmm4, xmm7 ;# total fsH2 in xmm4 addsd xmm5, [esp + nb213_vctot] movapd xmm0, [esp + nb213_dxM] movapd xmm1, [esp + nb213_dyM] movapd xmm2, [esp + nb213_dzM] movsd [esp + nb213_vctot], xmm5 mulsd xmm0, xmm4 mulsd xmm1, xmm4 mulsd xmm2, xmm4 ;# update M forces movapd xmm3, [esp + nb213_fixM] movapd xmm4, [esp + nb213_fiyM] movapd xmm7, [esp + nb213_fizM] addsd xmm3, xmm0 addsd xmm4, xmm1 addsd xmm7, xmm2 movsd [esp + nb213_fixM], xmm3 movsd [esp + nb213_fiyM], xmm4 movsd [esp + nb213_fizM], xmm7 mov edi, [ebp + nb213_faction] ;# update j forces addsd xmm0, [esp + nb213_fjx] addsd xmm1, [esp + nb213_fjy] addsd xmm2, [esp + nb213_fjz] movlpd xmm3, [edi + eax*8] movlpd xmm4, [edi + eax*8 + 8] movlpd xmm5, [edi + eax*8 + 16] subsd xmm3, xmm0 subsd xmm4, xmm1 subsd xmm5, xmm2 movlpd [edi + eax*8], xmm3 movlpd [edi + eax*8 + 8], xmm4 movlpd [edi + eax*8 + 16], xmm5.nb213_updateouterdata: mov ecx, [esp + nb213_ii3] mov edi, [ebp + nb213_faction] mov esi, [ebp + nb213_fshift] mov edx, [esp + nb213_is3] ;# accumulate Oi forces in xmm0, xmm1, xmm2 movapd xmm0, [esp + nb213_fixO] movapd xmm1, [esp + nb213_fiyO] movapd xmm2, [esp + nb213_fizO] movhlps xmm3, xmm0 movhlps xmm4, xmm1 movhlps xmm5, xmm2 addsd xmm0, xmm3 addsd xmm1, xmm4 addsd xmm2, xmm5 ;# sum is in low xmm0-xmm2 ;# increment i force movsd xmm3, [edi + ecx*8] movsd xmm4, [edi + ecx*8 + 8] movsd xmm5, [edi + ecx*8 + 16] addsd xmm3, xmm0 addsd xmm4, xmm1 addsd xmm5, xmm2 movsd [edi + ecx*8], xmm3 movsd [edi + ecx*8 + 8], xmm4 movsd [edi + ecx*8 + 16], xmm5 ;# accumulate force in xmm6/xmm7 for fshift movapd xmm6, xmm0 movsd xmm7, xmm2 unpcklpd xmm6,xmm1 ;# accumulate H1i forces in xmm0, xmm1, xmm2 movapd xmm0, [esp + nb213_fixH1] movapd xmm1, [esp + nb213_fiyH1] movapd xmm2, [esp + nb213_fizH1] movhlps xmm3, xmm0 movhlps xmm4, xmm1 movhlps xmm5, xmm2 addsd xmm0, xmm3 addsd xmm1, xmm4 addsd xmm2, xmm5 ;# sum is in low xmm0-xmm2 ;# increment i force movsd xmm3, [edi + ecx*8 + 24] movsd xmm4, [edi + ecx*8 + 32] movsd xmm5, [edi + ecx*8 + 40] addsd xmm3, xmm0 addsd xmm4, xmm1 addsd xmm5, xmm2 movsd [edi + ecx*8 + 24], xmm3 movsd [edi + ecx*8 + 32], xmm4 movsd [edi + ecx*8 + 40], xmm5 ;# accumulate force in xmm6/xmm7 for fshift addsd xmm7, xmm2 unpcklpd xmm0, xmm1 addpd xmm6, xmm0 ;# accumulate H2i forces in xmm0, xmm1, xmm2 movapd xmm0, [esp + nb213_fixH2] movapd xmm1, [esp + nb213_fiyH2] movapd xmm2, [esp + nb213_fizH2] movhlps xmm3, xmm0 movhlps xmm4, xmm1 movhlps xmm5, xmm2 addsd xmm0, xmm3 addsd xmm1, xmm4 addsd xmm2, xmm5 ;# sum is in low xmm0-xmm2 ;# increment i force movsd xmm3, [edi + ecx*8 + 48] movsd xmm4, [edi + ecx*8 + 56] movsd xmm5, [edi + ecx*8 + 64] addsd xmm3, xmm0 addsd xmm4, xmm1 addsd xmm5, xmm2 movsd [edi + ecx*8 + 48], xmm3 movsd [edi + ecx*8 + 56], xmm4 movsd [edi + ecx*8 + 64], xmm5 ;# accumulate force in xmm6/xmm7 for fshift addsd xmm7, xmm2 unpcklpd xmm0, xmm1 addpd xmm6, xmm0 ;# accumulate Mi forces in xmm0, xmm1, xmm2 movapd xmm0, [esp + nb213_fixM] movapd xmm1, [esp + nb213_fiyM] movapd xmm2, [esp + nb213_fizM] movhlps xmm3, xmm0 movhlps xmm4, xmm1 movhlps xmm5, xmm2 addsd xmm0, xmm3 addsd xmm1, xmm4 addsd xmm2, xmm5 ;# sum is in low xmm0-xmm2 ;# increment i force movsd xmm3, [edi + ecx*8 + 72] movsd xmm4, [edi + ecx*8 + 80] movsd xmm5, [edi + ecx*8 + 88] addsd xmm3, xmm0 addsd xmm4, xmm1 addsd xmm5, xmm2 movsd [edi + ecx*8 + 72], xmm3 movsd [edi + ecx*8 + 80], xmm4 movsd [edi + ecx*8 + 88], xmm5 ;# accumulate force in xmm6/xmm7 for fshift addsd xmm7, xmm2 unpcklpd xmm0, xmm1 addpd xmm6, xmm0 ;# increment fshift force movlpd xmm3, [esi + edx*8] movhpd xmm3, [esi + edx*8 + 8] movsd xmm4, [esi + edx*8 + 16] addpd xmm3, xmm6 addsd xmm4, xmm7 movlpd [esi + edx*8], xmm3 movhpd [esi + edx*8 + 8], xmm3 movsd [esi + edx*8 + 16], xmm4 ;# get n from stack mov esi, [esp + nb213_n] ;# get group index for i particle mov edx, [ebp + nb213_gid] ;# base of gid[] mov edx, [edx + esi*4] ;# ggid=gid[n] ;# accumulate total potential energy and update it movapd xmm7, [esp + nb213_vctot] ;# accumulate movhlps xmm6, xmm7 addsd xmm7, xmm6 ;# low xmm7 has the sum now ;# add earlier value from mem mov eax, [ebp + nb213_Vc] addsd xmm7, [eax + edx*8] ;# move back to mem movsd [eax + edx*8], xmm7 ;# accumulate total lj energy and update it movapd xmm7, [esp + nb213_Vvdwtot] ;# accumulate movhlps xmm6, xmm7 addsd xmm7, xmm6 ;# low xmm7 has the sum now ;# add earlier value from mem mov eax, [ebp + nb213_Vvdw] addsd xmm7, [eax + edx*8] ;# move back to mem movsd [eax + edx*8], xmm7 ;# finish if last mov ecx, [esp + nb213_nn1] ;# esi already loaded with n inc esi⌨️ 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?