nb_kernel313_x86_64_sse2.s
来自「最著名最快的分子模拟软件」· S 代码 · 共 2,129 行 · 第 1/5 页
S
2,129 行
movlpd 8(%rsi,%rax,8),%xmm5 movlpd 16(%rsi,%rax,8),%xmm6 movapd %xmm4,%xmm0 movapd %xmm5,%xmm1 movapd %xmm6,%xmm2 ## calc dr subsd nb313_ixO(%rsp),%xmm4 subsd nb313_iyO(%rsp),%xmm5 subsd nb313_izO(%rsp),%xmm6 ## store dr movapd %xmm4,nb313_dxO(%rsp) movapd %xmm5,nb313_dyO(%rsp) movapd %xmm6,nb313_dzO(%rsp) ## square it mulsd %xmm4,%xmm4 mulsd %xmm5,%xmm5 mulsd %xmm6,%xmm6 addsd %xmm5,%xmm4 addsd %xmm6,%xmm4 movapd %xmm4,%xmm7 ## rsqO in xmm7 ## move j coords to xmm4-xmm6 movapd %xmm0,%xmm4 movapd %xmm1,%xmm5 movapd %xmm2,%xmm6 ## calc dr subsd nb313_ixH1(%rsp),%xmm4 subsd nb313_iyH1(%rsp),%xmm5 subsd nb313_izH1(%rsp),%xmm6 ## store dr movapd %xmm4,nb313_dxH1(%rsp) movapd %xmm5,nb313_dyH1(%rsp) movapd %xmm6,nb313_dzH1(%rsp) ## square it mulsd %xmm4,%xmm4 mulsd %xmm5,%xmm5 mulsd %xmm6,%xmm6 addsd %xmm5,%xmm6 addsd %xmm4,%xmm6 ## rsqH1 in xmm6 ## move j coords to xmm3-xmm5 movapd %xmm0,%xmm3 movapd %xmm1,%xmm4 movapd %xmm2,%xmm5 ## calc dr subsd nb313_ixH2(%rsp),%xmm3 subsd nb313_iyH2(%rsp),%xmm4 subsd nb313_izH2(%rsp),%xmm5 ## store dr movapd %xmm3,nb313_dxH2(%rsp) movapd %xmm4,nb313_dyH2(%rsp) movapd %xmm5,nb313_dzH2(%rsp) ## square it mulsd %xmm3,%xmm3 mulsd %xmm4,%xmm4 mulsd %xmm5,%xmm5 addsd %xmm4,%xmm5 addsd %xmm3,%xmm5 ## move j coords to xmm4-xmm2 movapd %xmm0,%xmm4 movapd %xmm1,%xmm3 ## xmm2 already contains z ## calc dr subsd nb313_ixM(%rsp),%xmm4 subsd nb313_iyM(%rsp),%xmm3 subsd nb313_izM(%rsp),%xmm2 ## store dr movapd %xmm4,nb313_dxM(%rsp) movapd %xmm3,nb313_dyM(%rsp) movapd %xmm2,nb313_dzM(%rsp) ## square it mulpd %xmm2,%xmm2 mulpd %xmm3,%xmm3 mulpd %xmm4,%xmm4 addpd %xmm3,%xmm4 addpd %xmm2,%xmm4 ## rsqM in xmm4, rsqH2 in xmm5, rsqH1 in xmm6, rsqO in xmm7 ## 1/x for O - rsqO is in xmm7 cvtsd2ss %xmm7,%xmm2 movsd %xmm7,%xmm3 rcpps %xmm2,%xmm2 cvtss2sd %xmm2,%xmm2 movsd nb313_two(%rsp),%xmm1 movsd %xmm1,%xmm0 mulsd %xmm2,%xmm7 subsd %xmm7,%xmm1 mulsd %xmm1,%xmm2 ## iter1 mulsd %xmm2,%xmm3 subsd %xmm3,%xmm0 mulsd %xmm2,%xmm0 ## xmm0=rinvsq movsd %xmm0,nb313_rinvsqO(%rsp) ## rsqH1 - seed in xmm2 cvtsd2ss %xmm6,%xmm2 rsqrtss %xmm2,%xmm2 cvtss2sd %xmm2,%xmm2 movapd %xmm2,%xmm3 mulsd %xmm2,%xmm2 movapd nb313_three(%rsp),%xmm0 mulsd %xmm6,%xmm2 ## rsq*lu*lu subsd %xmm2,%xmm0 ## 30-rsq*lu*lu mulsd %xmm3,%xmm0 ## lu*(3-rsq*lu*lu) mulsd nb313_half(%rsp),%xmm0 ## iter1 ( new lu) movapd %xmm6,%xmm2 movapd %xmm0,%xmm3 mulsd %xmm0,%xmm0 ## lu*lu mulsd %xmm0,%xmm2 ## rsq*lu*lu movapd nb313_three(%rsp),%xmm0 subsd %xmm2,%xmm0 ## 3-rsq*lu*lu mulsd %xmm3,%xmm0 ## lu*( 3-rsq*lu*lu) mulsd nb313_half(%rsp),%xmm0 ## rinv movapd %xmm0,nb313_rinvH1(%rsp) ## rinvH1 mulsd %xmm0,%xmm6 movapd %xmm6,nb313_rH1(%rsp) ## rH1 ## rsqH2 - seed in xmm2 cvtsd2ss %xmm5,%xmm2 rsqrtss %xmm2,%xmm2 cvtss2sd %xmm2,%xmm2 movapd %xmm2,%xmm3 mulsd %xmm2,%xmm2 movapd nb313_three(%rsp),%xmm0 mulsd %xmm5,%xmm2 ## rsq*lu*lu subsd %xmm2,%xmm0 ## 30-rsq*lu*lu mulsd %xmm3,%xmm0 ## lu*(3-rsq*lu*lu) mulsd nb313_half(%rsp),%xmm0 ## iter1 ( new lu) movapd %xmm5,%xmm2 movapd %xmm0,%xmm3 mulsd %xmm0,%xmm0 ## lu*lu mulsd %xmm0,%xmm2 ## rsq*lu*lu movapd nb313_three(%rsp),%xmm0 subsd %xmm2,%xmm0 ## 3-rsq*lu*lu mulsd %xmm3,%xmm0 ## lu*( 3-rsq*lu*lu) mulsd nb313_half(%rsp),%xmm0 ## rinv movapd %xmm0,nb313_rinvH2(%rsp) ## rinv mulsd %xmm0,%xmm5 movapd %xmm5,nb313_rH2(%rsp) ## r ## rsqM - seed in xmm2 cvtsd2ss %xmm4,%xmm2 rsqrtss %xmm2,%xmm2 cvtss2sd %xmm2,%xmm2 movapd %xmm2,%xmm3 mulsd %xmm2,%xmm2 movapd nb313_three(%rsp),%xmm0 mulsd %xmm4,%xmm2 ## rsq*lu*lu subsd %xmm2,%xmm0 ## 30-rsq*lu*lu mulsd %xmm3,%xmm0 ## lu*(3-rsq*lu*lu) mulsd nb313_half(%rsp),%xmm0 ## iter1 ( new lu) movapd %xmm4,%xmm2 movapd %xmm0,%xmm3 mulsd %xmm0,%xmm0 ## lu*lu mulsd %xmm0,%xmm2 ## rsq*lu*lu movapd nb313_three(%rsp),%xmm0 subsd %xmm2,%xmm0 ## 3-rsq*lu*lu mulsd %xmm3,%xmm0 ## lu*( 3-rsq*lu*lu) mulsd nb313_half(%rsp),%xmm0 ## rinv movapd %xmm0,nb313_rinvM(%rsp) ## rinv mulsd %xmm0,%xmm4 movapd %xmm4,nb313_rM(%rsp) ## r ## do O interactions movapd nb313_rinvsqO(%rsp),%xmm0 movapd %xmm0,%xmm1 mulsd %xmm1,%xmm1 ## rinv4 mulsd %xmm0,%xmm1 ##rinvsix movapd %xmm1,%xmm2 mulsd %xmm2,%xmm2 ## rinvtwelve mulsd nb313_c6(%rsp),%xmm1 mulsd nb313_c12(%rsp),%xmm2 movapd %xmm2,%xmm3 subsd %xmm1,%xmm3 ## Vvdw=Vvdw12-Vvdw6 addsd nb313_Vvdwtot(%rsp),%xmm3 mulsd nb313_six(%rsp),%xmm1 mulsd nb313_twelve(%rsp),%xmm2 subsd %xmm1,%xmm2 mulsd %xmm0,%xmm2 movapd %xmm2,%xmm4 ## total fsO movsd %xmm3,nb313_Vvdwtot(%rsp) movapd nb313_dxO(%rsp),%xmm0 movapd nb313_dyO(%rsp),%xmm1 movapd nb313_dzO(%rsp),%xmm2 mulsd %xmm4,%xmm0 mulsd %xmm4,%xmm1 mulsd %xmm4,%xmm2 ## tx in xmm0-xmm2 ## update O forces movapd nb313_fixO(%rsp),%xmm3 movapd nb313_fiyO(%rsp),%xmm4 movapd nb313_fizO(%rsp),%xmm7 addsd %xmm0,%xmm3 addsd %xmm1,%xmm4 addsd %xmm2,%xmm7 movlpd %xmm3,nb313_fixO(%rsp) movlpd %xmm4,nb313_fiyO(%rsp) movlpd %xmm7,nb313_fizO(%rsp) ## update j forces with water O movlpd %xmm0,nb313_fjx(%rsp) movlpd %xmm1,nb313_fjy(%rsp) movlpd %xmm2,nb313_fjz(%rsp) ## Done with O interactions - now H1! movapd nb313_rH1(%rsp),%xmm7 mulpd nb313_tsc(%rsp),%xmm7 cvttsd2si %xmm7,%r8d ## mm6 = lu idx cvtsi2sd %r8d,%xmm6 subpd %xmm6,%xmm7 movapd %xmm7,%xmm1 ## xmm1=eps movapd %xmm1,%xmm2 mulpd %xmm2,%xmm2 ## xmm2=eps2 shll $2,%r8d ## idx *= 4 movq nb313_VFtab(%rbp),%rsi movapd (%rsi,%r8,8),%xmm4 ## Y1 F1 xorpd %xmm3,%xmm3 movapd %xmm4,%xmm5 unpcklpd %xmm3,%xmm4 ## Y1 unpckhpd %xmm3,%xmm5 ## F1 movapd 16(%rsi,%r8,8),%xmm6 ## G1 H1 xorpd %xmm3,%xmm3 movapd %xmm6,%xmm7 unpcklpd %xmm3,%xmm6 ## G1 unpckhpd %xmm3,%xmm7 ## H1 ## coulomb table ready, in xmm4-xmm7 mulsd %xmm1,%xmm6 ## xmm6=Geps mulsd %xmm2,%xmm7 ## xmm7=Heps2 addsd %xmm6,%xmm5 addsd %xmm7,%xmm5 ## xmm5=Fp mulsd nb313_two(%rsp),%xmm7 ## two*Heps2 movapd nb313_qqH(%rsp),%xmm3 addsd %xmm6,%xmm7 addsd %xmm5,%xmm7 ## xmm7=FF mulsd %xmm1,%xmm5 ## xmm5=eps*Fp addsd %xmm4,%xmm5 ## xmm5=VV mulsd %xmm3,%xmm5 ## vcoul=qq*VV mulsd %xmm7,%xmm3 ## fijC=FF*qq ## at this point mm5 contains vcoul and xmm3 fijC ## increment vcoul xorpd %xmm4,%xmm4 addsd nb313_vctot(%rsp),%xmm5 mulsd nb313_rinvH1(%rsp),%xmm3 movlpd %xmm5,nb313_vctot(%rsp) mulsd nb313_tsc(%rsp),%xmm3 subsd %xmm3,%xmm4 movapd nb313_dxH1(%rsp),%xmm0 movapd nb313_dyH1(%rsp),%xmm1 movapd nb313_dzH1(%rsp),%xmm2 mulsd %xmm4,%xmm0 mulsd %xmm4,%xmm1 mulsd %xmm4,%xmm2 ## update H1 forces movapd nb313_fixH1(%rsp),%xmm3 movapd nb313_fiyH1(%rsp),%xmm4 movapd nb313_fizH1(%rsp),%xmm7 addsd %xmm0,%xmm3 addsd %xmm1,%xmm4 addsd %xmm2,%xmm7 movlpd %xmm3,nb313_fixH1(%rsp) movlpd %xmm4,nb313_fiyH1(%rsp) movlpd %xmm7,nb313_fizH1(%rsp) ## update j forces with water H1 addsd nb313_fjx(%rsp),%xmm0 addsd nb313_fjy(%rsp),%xmm1 addsd nb313_fjz(%rsp),%xmm2 movlpd %xmm0,nb313_fjx(%rsp) movlpd %xmm1,nb313_fjy(%rsp) movlpd %xmm2,nb313_fjz(%rsp) ## H2 interactions movapd nb313_rH2(%rsp),%xmm7 mulsd nb313_tsc(%rsp),%xmm7 cvttsd2si %xmm7,%r8d ## mm6 = lu idx cvtsi2sd %r8d,%xmm6 subsd %xmm6,%xmm7 movapd %xmm7,%xmm1 ## xmm1=eps movapd %xmm1,%xmm2 mulsd %xmm2,%xmm2 ## xmm2=eps2 shll $2,%r8d ## idx *= 4 movq nb313_VFtab(%rbp),%rsi movapd (%rsi,%r8,8),%xmm4 ## Y1 F1 xorpd %xmm3,%xmm3 movapd %xmm4,%xmm5 unpcklpd %xmm3,%xmm4 ## Y1 unpckhpd %xmm3,%xmm5 ## F1 movapd 16(%rsi,%r8,8),%xmm6 ## G1 H1 xorpd %xmm3,%xmm3 movapd %xmm6,%xmm7 unpcklpd %xmm3,%xmm6 ## G1 unpckhpd %xmm3,%xmm7 ## H1 ## coulomb table ready, in xmm4-xmm7 mulsd %xmm1,%xmm6 ## xmm6=Geps mulsd %xmm2,%xmm7 ## xmm7=Heps2 addsd %xmm6,%xmm5 addsd %xmm7,%xmm5 ## xmm5=Fp mulsd nb313_two(%rsp),%xmm7 ## two*Heps2 movapd nb313_qqH(%rsp),%xmm3 addsd %xmm6,%xmm7 addsd %xmm5,%xmm7 ## xmm7=FF mulsd %xmm1,%xmm5 ## xmm5=eps*Fp addsd %xmm4,%xmm5 ## xmm5=VV mulsd %xmm3,%xmm5 ## vcoul=qq*VV mulsd %xmm7,%xmm3 ## fijC=FF*qq ## at this point mm5 contains vcoul and xmm3 fijC ## increment vcoul xorpd %xmm4,%xmm4 addsd nb313_vctot(%rsp),%xmm5 mulsd nb313_rinvH2(%rsp),%xmm3 movlpd %xmm5,nb313_vctot(%rsp) mulsd nb313_tsc(%rsp),%xmm3 subsd %xmm3,%xmm4 movapd nb313_dxH2(%rsp),%xmm0 movapd nb313_dyH2(%rsp),%xmm1 movapd nb313_dzH2(%rsp),%xmm2 mulsd %xmm4,%xmm0 mulsd %xmm4,%xmm1 mulsd %xmm4,%xmm2 ## update H2 forces movapd nb313_fixH2(%rsp),%xmm3 movapd nb313_fiyH2(%rsp),%xmm4 movapd nb313_fizH2(%rsp),%xmm7 addsd %xmm0,%xmm3 addsd %xmm1,%xmm4 addsd %xmm2,%xmm7 movlpd %xmm3,nb313_fixH2(%rsp) movlpd %xmm4,nb313_fiyH2(%rsp) movlpd %xmm7,nb313_fizH2(%rsp) ## update j forces with water H1 addsd nb313_fjx(%rsp),%xmm0 addsd nb313_fjy(%rsp),%xmm1 addsd nb313_fjz(%rsp),%xmm2 movlpd %xmm0,nb313_fjx(%rsp) movlpd %xmm1,nb313_fjy(%rsp) movlpd %xmm2,nb313_fjz(%rsp) ## M interactions movapd nb313_rM(%rsp),%xmm7 mulsd nb313_tsc(%rsp),%xmm7 cvttsd2si %xmm7,%r8d ## mm6 = lu idx cvtsi2sd %r8d,%xmm6 subsd %xmm6,%xmm7 movapd %xmm7,%xmm1 ## xmm1=eps movapd %xmm1,%xmm2 mulsd %xmm2,%xmm2 ## xmm2=eps2 shll $2,%r8d ## idx *= 4 movq nb313_VFtab(%rbp),%rsi movapd (%rsi,%r8,8),%xmm4 ## Y1 F1 xorpd %xmm3,%xmm3 movapd %xmm4,%xmm5 unpcklpd %xmm3,%xmm4 ## Y1 unpckhpd %xmm3,%xmm5 ## F1 movapd 16(%rsi,%r8,8),%xmm6 ## G1 H1 xorpd %xmm3,%xmm3 movapd %xmm6,%xmm7 unpcklpd %xmm3,%xmm6 ## G1 unpckhpd %xmm3,%xmm7 ## H1 ## coulomb table ready, in xmm4-xmm7 mulsd %xmm1,%xmm6 ## xmm6=Geps mulsd %xmm2,%xmm7 ## xmm7=Heps2 addsd %xmm6,%xmm5 addsd %xmm7,%xmm5 ## xmm5=Fp mulsd nb313_two(%rsp),%xmm7 ## two*Heps2 movapd nb313_qqM(%rsp),%xmm3 addsd %xmm6,%xmm7 addsd %xmm5,%xmm7 ## xmm7=FF mulsd %xmm1,%xmm5 ## xmm5=eps*Fp addsd %xmm4,%xmm5 ## xmm5=VV mulsd %xmm3,%xmm5 ## vcoul=qq*VV mulsd %xmm7,%xmm3 ## fijC=FF*qq ## at this point mm5 contains vcoul and xmm3 fijC ## increment vcoul xorpd %xmm4,%xmm4 addsd nb313_vctot(%rsp),%xmm5 mulsd nb313_rinvM(%rsp),%xmm3 movlpd %xmm5,nb313_vctot(%rsp) mulsd nb313_tsc(%rsp),%xmm3 subsd %xmm3,%xmm4 movapd nb313_dxM(%rsp),%xmm0 movapd nb313_dyM(%rsp),%xmm1 movapd nb313_dzM(%rsp),%xmm2 mulsd %xmm4,%xmm0 mulsd %xmm4,%xmm1 mulsd %xmm4,%xmm2 ## update M forces movapd nb313_fixM(%rsp),%xmm3 movapd nb313_fiyM(%rsp),%xmm4 movapd nb313_fizM(%rsp),%xmm7 addsd %xmm0,%xmm3 addsd %xmm1,%xmm4 addsd %xmm2,%xmm7 movlpd %xmm3,nb313_fixM(%rsp) movlpd %xmm4,nb313_fiyM(%rsp)⌨️ 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?