nb_kernel202_x86_64_sse.intel_syntax.s
来自「最著名最快的分子模拟软件」· S 代码 · 共 2,206 行 · 第 1/5 页
S
2,206 行
movaps xmm4, xmm0 unpcklps xmm0, xmm1 ;# jxH2a jxH2c jyH2a jyH2c unpckhps xmm4, xmm1 ;# jxH2b jxH2d jyH2b jyH2d movaps xmm1, xmm0 unpcklps xmm0, xmm4 ;# x unpckhps xmm1, xmm4 ;# y shufps xmm2, xmm3, 136 ;# 10001000 => jzH2a jzH2b jzH2c jzH2d ;# xmm0 = H2x ;# xmm1 = H2y ;# xmm2 = H2z movaps xmm3, xmm0 movaps xmm4, xmm1 movaps xmm5, xmm2 movaps xmm6, xmm0 movaps xmm7, xmm1 movaps xmm8, xmm2 subps xmm0, [rsp + nb202_ixO] subps xmm1, [rsp + nb202_iyO] subps xmm2, [rsp + nb202_izO] subps xmm3, [rsp + nb202_ixH1] subps xmm4, [rsp + nb202_iyH1] subps xmm5, [rsp + nb202_izH1] subps xmm6, [rsp + nb202_ixH2] subps xmm7, [rsp + nb202_iyH2] subps xmm8, [rsp + nb202_izH2] movaps [rsp + nb202_dxOH2], xmm0 movaps [rsp + nb202_dyOH2], xmm1 movaps [rsp + nb202_dzOH2], xmm2 mulps xmm0, xmm0 mulps xmm1, xmm1 mulps xmm2, xmm2 movaps [rsp + nb202_dxH1H2], xmm3 movaps [rsp + nb202_dyH1H2], xmm4 movaps [rsp + nb202_dzH1H2], xmm5 mulps xmm3, xmm3 mulps xmm4, xmm4 mulps xmm5, xmm5 movaps [rsp + nb202_dxH2H2], xmm6 movaps [rsp + nb202_dyH2H2], xmm7 movaps [rsp + nb202_dzH2H2], xmm8 mulps xmm6, xmm6 mulps xmm7, xmm7 mulps xmm8, xmm8 addps xmm0, xmm1 addps xmm0, xmm2 addps xmm3, xmm4 addps xmm3, xmm5 addps xmm6, xmm7 addps xmm6, xmm8 ;# start doing invsqrt for jH2 atoms rsqrtps xmm1, xmm0 rsqrtps xmm4, xmm3 rsqrtps xmm7, xmm6 movaps xmm2, xmm1 movaps xmm5, xmm4 movaps xmm8, xmm7 mulps xmm1, xmm1 ;# lu*lu mulps xmm4, xmm4 ;# lu*lu mulps xmm7, xmm7 ;# lu*lu movaps xmm9, [rsp + nb202_three] movaps xmm10, xmm9 movaps xmm11, xmm9 mulps xmm1, xmm0 ;# rsq*lu*lu mulps xmm4, xmm3 ;# rsq*lu*lu mulps xmm7, xmm6 ;# rsq*lu*lu subps xmm9, xmm1 subps xmm10, xmm4 subps xmm11, xmm7 ;# 3-rsq*lu*lu mulps xmm9, xmm2 mulps xmm10, xmm5 mulps xmm11, xmm8 ;# lu*(3-rsq*lu*lu) movaps xmm4, [rsp + nb202_half] mulps xmm9, xmm4 ;# rinvOH2 mulps xmm10, xmm4 ;# rinvH1H2 mulps xmm11, xmm4 ;# rinvH2H2 ;# H2 interactions ;# rsq in xmm0,xmm3,xmm6 ;# rinv in xmm9, xmm10, xmm11 movaps xmm1, xmm9 ;# copy of rinv movaps xmm4, xmm10 movaps xmm7, xmm11 movaps xmm2, [rsp + nb202_krf] mulps xmm9, xmm9 ;# rinvsq mulps xmm10, xmm10 mulps xmm11, xmm11 mulps xmm0, xmm2 ;# k*rsq mulps xmm3, xmm2 mulps xmm6, xmm2 movaps xmm2, xmm0 ;# copy of k*rsq movaps xmm5, xmm3 movaps xmm8, xmm6 addps xmm2, xmm1 ;# rinv+krsq addps xmm5, xmm4 addps xmm8, xmm7 movaps xmm14, [rsp + nb202_crf] subps xmm2, xmm14 ;# rinv+krsq-crf subps xmm5, xmm14 subps xmm8, xmm14 movaps xmm12, [rsp + nb202_qqOH] movaps xmm13, [rsp + nb202_qqHH] mulps xmm2, xmm12 ;# xmm6=voul=qq*(rinv+ krsq-crf) mulps xmm5, xmm13 ;# xmm6=voul=qq*(rinv+ krsq-crf) mulps xmm8, xmm13 ;# xmm6=voul=qq*(rinv+ krsq-crf) addps xmm0, xmm0 ;# 2*krsq addps xmm3, xmm3 addps xmm6, xmm6 subps xmm1, xmm0 ;# rinv-2*krsq subps xmm4, xmm3 subps xmm7, xmm6 mulps xmm1, xmm12 ;# (rinv-2*krsq)*qq mulps xmm4, xmm13 mulps xmm7, xmm13 addps xmm5, xmm8 addps xmm2, xmm15 addps xmm2, xmm5 movaps [rsp + nb202_vctot], xmm2 mulps xmm1, xmm9 ;# fscal mulps xmm4, xmm10 mulps xmm7, xmm11 ;# move j H2 forces to local temp variables movlps xmm9, [rdi + rax*4 + 24] ;# jxH2a jyH2a - - movlps xmm10, [rdi + rcx*4 + 24] ;# jxH2c jyH2c - - movhps xmm9, [rdi + rbx*4 + 24] ;# jxH2a jyH2a jxH2b jyH2b movhps xmm10, [rdi + rdx*4 + 24] ;# jxH2c jyH2c jxH2d jyH2d movss xmm11, [rdi + rax*4 + 32] ;# jzH2a - - - movss xmm12, [rdi + rcx*4 + 32] ;# jzH2c - - - movss xmm2, [rdi + rbx*4 + 32] ;# jzH2b - - - movss xmm3, [rdi + rdx*4 + 32] ;# jzH2d - - - movlhps xmm11, xmm2 ;# jzH2a - jzH2b - movlhps xmm12, xmm3 ;# jzH2c - jzH2d - shufps xmm11, xmm12, 136 ;# 10001000 => jzH2a jzH2b jzH2c jzH2d ;# xmm9: jxH2a jyH2a jxH2b jyH2b ;# xmm10: jxH2c jyH2c jxH2d jyH2d ;# xmm11: jzH2a jzH2b jzH2c jzH2d movaps xmm0, xmm1 movaps xmm2, xmm1 movaps xmm3, xmm4 movaps xmm5, xmm4 movaps xmm6, xmm7 movaps xmm8, xmm7 mulps xmm0, [rsp + nb202_dxOH2] mulps xmm1, [rsp + nb202_dyOH2] mulps xmm2, [rsp + nb202_dzOH2] mulps xmm3, [rsp + nb202_dxH1H2] mulps xmm4, [rsp + nb202_dyH1H2] mulps xmm5, [rsp + nb202_dzH1H2] mulps xmm6, [rsp + nb202_dxH2H2] mulps xmm7, [rsp + nb202_dyH2H2] mulps xmm8, [rsp + nb202_dzH2H2] movaps xmm13, xmm0 movaps xmm14, xmm1 addps xmm11, xmm2 addps xmm0, [rsp + nb202_fixO] addps xmm1, [rsp + nb202_fiyO] addps xmm2, [rsp + nb202_fizO] addps xmm13, xmm3 addps xmm14, xmm4 addps xmm11, xmm5 addps xmm3, [rsp + nb202_fixH1] addps xmm4, [rsp + nb202_fiyH1] addps xmm5, [rsp + nb202_fizH1] addps xmm13, xmm6 addps xmm14, xmm7 addps xmm11, xmm8 addps xmm6, [rsp + nb202_fixH2] addps xmm7, [rsp + nb202_fiyH2] addps xmm8, [rsp + nb202_fizH2] movaps [rsp + nb202_fixO], xmm0 movaps [rsp + nb202_fiyO], xmm1 movaps [rsp + nb202_fizO], xmm2 movaps [rsp + nb202_fixH1], xmm3 movaps [rsp + nb202_fiyH1], xmm4 movaps [rsp + nb202_fizH1], xmm5 movaps [rsp + nb202_fixH2], xmm6 movaps [rsp + nb202_fiyH2], xmm7 movaps [rsp + nb202_fizH2], xmm8 ;# xmm0 = fH2x ;# xmm1 = fH2y ;# xmm2 = fH2z movaps xmm15, xmm13 unpcklps xmm13, xmm14 unpckhps xmm15, xmm14 addps xmm9, xmm13 addps xmm10, xmm15 movhlps xmm12, xmm11 ;# fH2zc fH2zd movlps [rdi + rax*4 + 24], xmm9 movhps [rdi + rbx*4 + 24], xmm9 movlps [rdi + rcx*4 + 24], xmm10 movhps [rdi + rdx*4 + 24], xmm10 movss [rdi + rax*4 + 32], xmm11 movss [rdi + rcx*4 + 32], xmm12 shufps xmm11, xmm11, 1 shufps xmm12, xmm12, 1 movss [rdi + rbx*4 + 32], xmm11 movss [rdi + rdx*4 + 32], xmm12 ;# should we do one more iteration? sub dword ptr [rsp + nb202_innerk], 4 jl .nb202_single_check jmp .nb202_unroll_loop.nb202_single_check: add dword ptr [rsp + nb202_innerk], 4 jnz .nb202_single_loop jmp .nb202_updateouterdata.nb202_single_loop: mov rdx, [rsp + nb202_innerjjnr] ;# pointer to jjnr[k] mov eax, [rdx] add qword ptr [rsp + nb202_innerjjnr], 4 mov rsi, [rbp + nb202_pos] lea rax, [rax + rax*2] ;# fetch j coordinates xorps xmm0, xmm0 xorps xmm1, xmm1 xorps xmm2, xmm2 movss xmm0, [rsi + rax*4] ;# jxO - - - movss xmm1, [rsi + rax*4 + 4] ;# jyO - - - movss xmm2, [rsi + rax*4 + 8] ;# jzO - - - movlps xmm6, [rsi + rax*4 + 12] ;# xmm6 = jxH1 jyH1 - - movss xmm7, [rsi + rax*4 + 20] ;# xmm7 = jzH1 - - - movhps xmm6, [rsi + rax*4 + 24] ;# xmm6 = jxH1 jyH1 jxH2 jyH2 movss xmm5, [rsi + rax*4 + 32] ;# xmm5 = jzH2 - - - ;# have all coords, time for some shuffling. shufps xmm6, xmm6, 216 ;# 11011000 ;# xmm6 = jxH1 jxH2 jyH1 jyH2 unpcklps xmm7, xmm5 ;# xmm7 = jzH1 jzH2 - - movlhps xmm0, xmm6 ;# xmm0 = jxO 0 jxH1 jxH2 shufps xmm1, xmm6, 228 ;# 11100100 ;# xmm1 = jyO 0 jyH1 jyH2 shufps xmm2, xmm7, 68 ;# 01000100 ;# xmm2 = jzO 0 jzH1 jzH2 movaps [rsp + nb202_jxO], xmm0 movaps [rsp + nb202_jyO], xmm1 movaps [rsp + nb202_jzO], xmm2 subps xmm0, [rsp + nb202_ixO] subps xmm1, [rsp + nb202_iyO] subps xmm2, [rsp + nb202_izO] movaps [rsp + nb202_dxOO], xmm0 movaps [rsp + nb202_dyOO], xmm1 movaps [rsp + nb202_dzOO], xmm2 mulps xmm0, xmm0 mulps xmm1, xmm1 mulps xmm2, xmm2 addps xmm0, xmm1 addps xmm0, xmm2 ;# have rsq in xmm0 movaps xmm6, xmm0 ;# do invsqrt rsqrtps xmm1, xmm0 mulps xmm6, [rsp + nb202_krf] ;# xmm6=krsq movaps xmm2, xmm1 movaps xmm7, xmm6 ;# xmm7=krsq mulps xmm1, xmm1 movaps xmm3, [rsp + nb202_three] mulps xmm1, xmm0 subps xmm3, xmm1 mulps xmm3, xmm2 mulps xmm3, [rsp + nb202_half] ;# rinv iO - j water addps xmm6, xmm3 ;# xmm6=rinv+ krsq mulps xmm7, [rsp + nb202_two] subps xmm6, [rsp + nb202_crf] ;# xmm6=rinv+ krsq-crf xorps xmm1, xmm1 movaps xmm0, xmm3 subps xmm3, xmm7 ;# xmm3=rinv-2*krsq xorps xmm4, xmm4 mulps xmm0, xmm0 ;# xmm0=rinvsq ;# fetch charges to xmm4 (temporary) movss xmm4, [rsp + nb202_qqOO] movhps xmm4, [rsp + nb202_qqOH] mulps xmm6, xmm4 ;# vcoul mulps xmm3, xmm4 ;# coul part of fs addps xmm6, [rsp + nb202_vctot] mulps xmm0, xmm3 ;# total fscal movaps [rsp + nb202_vctot], xmm6 movaps xmm1, xmm0 movaps xmm2, xmm0 mulps xmm0, [rsp + nb202_dxOO] mulps xmm1, [rsp + nb202_dyOO] mulps xmm2, [rsp + nb202_dzOO] ;# initial update for j forces xorps xmm3, xmm3 xorps xmm4, xmm4 xorps xmm5, xmm5 addps xmm3, xmm0 addps xmm4, xmm1 addps xmm5, xmm2 movaps [rsp + nb202_fjxO], xmm3 movaps [rsp + nb202_fjyO], xmm4 movaps [rsp + nb202_fjzO], xmm5 addps xmm0, [rsp + nb202_fixO] addps xmm1, [rsp + nb202_fiyO] addps xmm2, [rsp + nb202_fizO] movaps [rsp + nb202_fixO], xmm0 movaps [rsp + nb202_fiyO], xmm1 movaps [rsp + nb202_fizO], xmm2 ;# done with i O Now do i H1 & H2 simultaneously first get i particle coords: movaps xmm0, [rsp + nb202_jxO] movaps xmm1, [rsp + nb202_jyO] movaps xmm2, [rsp + nb202_jzO] movaps xmm3, xmm0 movaps xmm4, xmm1 movaps xmm5, xmm2 subps xmm0, [rsp + nb202_ixH1] subps xmm1, [rsp + nb202_iyH1] subps xmm2, [rsp + nb202_izH1] subps xmm3, [rsp + nb202_ixH2] subps xmm4, [rsp + nb202_iyH2] subps xmm5, [rsp + nb202_izH2] movaps [rsp + nb202_dxH1O], xmm0 movaps [rsp + nb202_dyH1O], xmm1 movaps [rsp + nb202_dzH1O], xmm2 movaps [rsp + nb202_dxH2O], xmm3 movaps [rsp + nb202_dyH2O], xmm4 movaps [rsp + nb202_dzH2O], xmm5 mulps xmm0, xmm0 mulps xmm1, xmm1 mulps xmm2, xmm2 mulps xmm3, xmm3 mulps xmm4, xmm4 mulps xmm5, xmm5 addps xmm0, xmm1 addps xmm4, xmm3 addps xmm0, xmm2 ;# have rsqH1 in xmm0 addps xmm4, xmm5 ;# have rsqH2 in xmm4 ;# do invsqrt rsqrtps xmm1, xmm0 rsqrtps xmm5, xmm4 movaps xmm2, xmm1 movaps xmm6, xmm5 mulps xmm1, xmm1 mulps xmm5, xmm5 movaps xmm3, [rsp + nb202_three] movaps xmm7, xmm3 mulps xmm1, xmm0 mulps xmm5, xmm4 subps xmm3, xmm1 subps xmm7, xmm5 mulps xmm3, xmm2 mulps xmm7, xmm6 mulps xmm3, [rsp + nb202_half] ;# rinv H1 - j water mulps xmm7, [rsp + nb202_half] ;# rinv H2 - j water mulps xmm0, [rsp + nb202_krf] ;# krsq mulps xmm4, [rsp + nb202_krf] ;# krsq ;# assemble charges in xmm6 xorps xmm6, xmm6 movss xmm6, [rsp + nb202_qqOH] movhps xmm6, [rsp + nb202_qqHH] movaps xmm1, xmm0 movaps xmm5, xmm4 addps xmm0, xmm3 ;# krsq+ rinv addps xmm4, xmm7 ;# krsq+ rinv subps xmm0, [rsp + nb202_crf] subps xmm4, [rsp + nb202_crf] mulps xmm1, [rsp + nb202_two] mulps xmm5, [rsp + nb202_two] mulps xmm0, xmm6 ;# vcoul mulps xmm4, xmm6 ;# vcoul addps xmm4, xmm0 addps xmm4, [rsp + nb202_vctot] movaps [rsp + nb202_vctot], xmm4 movaps xmm0, xmm3 movaps xmm4, xmm7 mulps xmm3, xmm3 mulps xmm7, xmm7 subps xmm0, xmm1 subps xmm4, xmm5 mulps xmm0, xmm6 mulps xmm4, xmm6 mulps xmm0, xmm3 ;# fscal mulps xmm7, xmm4 ;# fscal movaps xmm1, xmm0 movaps xmm2, xmm0 mulps xmm0, [rsp + nb202_dxH1O] mulps xmm1, [rsp + nb202_dyH1O] mulps xmm2, [rsp + nb202_dzH1O] ;# update forces H1 - j water movaps xmm3, [rsp + nb202_fjxO] movaps xmm4, [rsp + nb202_fjyO] movaps xmm5, [rsp + nb202_fjzO] addps xmm3, xmm0 addps xmm4, xmm1 addps xmm5, xmm2 movaps [rsp + nb202_fjxO], xmm3 movaps [rsp + nb202_fjyO], xmm4 movaps [rsp + nb202_fjzO], xmm5 addps xmm0, [rsp + nb202_fixH1] addps xmm1, [rsp + nb202_fiyH1] addps xmm2, [rsp + nb202_fizH1] movaps [rsp + nb202_fixH1], xmm0 movaps [rsp + nb202_fiyH1], xmm1 movaps [rsp + nb202_fizH1], xmm2⌨️ 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?