nb_kernel212_x86_64_sse2.s
来自「最著名最快的分子模拟软件」· S 代码 · 共 2,263 行 · 第 1/5 页
S
2,263 行
addsd nb212_fizH2(%rsp),%xmm15 movsd %xmm7,nb212_fixO(%rsp) movsd %xmm8,nb212_fiyO(%rsp) movsd %xmm9,nb212_fizO(%rsp) movsd %xmm10,nb212_fixH1(%rsp) movsd %xmm11,nb212_fiyH1(%rsp) movsd %xmm12,nb212_fizH1(%rsp) movsd %xmm13,nb212_fixH2(%rsp) movsd %xmm14,nb212_fiyH2(%rsp) movsd %xmm15,nb212_fizH2(%rsp) ## store back j O forces from xmm0-xmm2 movsd %xmm0,(%rdi,%rax,8) movsd %xmm1,8(%rdi,%rax,8) movsd %xmm2,16(%rdi,%rax,8) ## move j H1 coordinates to local temp variables movsd 24(%rsi,%rax,8),%xmm0 movsd 32(%rsi,%rax,8),%xmm1 movsd 40(%rsi,%rax,8),%xmm2 ## xmm0 = H1x ## xmm1 = H1y ## xmm2 = H1z movsd %xmm0,%xmm3 movsd %xmm1,%xmm4 movsd %xmm2,%xmm5 movsd %xmm0,%xmm6 movsd %xmm1,%xmm7 movsd %xmm2,%xmm8 subsd nb212_ixO(%rsp),%xmm0 subsd nb212_iyO(%rsp),%xmm1 subsd nb212_izO(%rsp),%xmm2 subsd nb212_ixH1(%rsp),%xmm3 subsd nb212_iyH1(%rsp),%xmm4 subsd nb212_izH1(%rsp),%xmm5 subsd nb212_ixH2(%rsp),%xmm6 subsd nb212_iyH2(%rsp),%xmm7 subsd nb212_izH2(%rsp),%xmm8 movsd %xmm0,nb212_dxOH1(%rsp) movsd %xmm1,nb212_dyOH1(%rsp) movsd %xmm2,nb212_dzOH1(%rsp) mulsd %xmm0,%xmm0 mulsd %xmm1,%xmm1 mulsd %xmm2,%xmm2 movsd %xmm3,nb212_dxH1H1(%rsp) movsd %xmm4,nb212_dyH1H1(%rsp) movsd %xmm5,nb212_dzH1H1(%rsp) mulsd %xmm3,%xmm3 mulsd %xmm4,%xmm4 mulsd %xmm5,%xmm5 movsd %xmm6,nb212_dxH2H1(%rsp) movsd %xmm7,nb212_dyH2H1(%rsp) movsd %xmm8,nb212_dzH2H1(%rsp) mulsd %xmm6,%xmm6 mulsd %xmm7,%xmm7 mulsd %xmm8,%xmm8 addsd %xmm1,%xmm0 addsd %xmm2,%xmm0 addsd %xmm4,%xmm3 addsd %xmm5,%xmm3 addsd %xmm7,%xmm6 addsd %xmm8,%xmm6 ## start doing invsqrt for jH1 atoms cvtsd2ss %xmm0,%xmm1 cvtsd2ss %xmm3,%xmm4 cvtsd2ss %xmm6,%xmm7 rsqrtss %xmm1,%xmm1 rsqrtss %xmm4,%xmm4 rsqrtss %xmm7,%xmm7 cvtss2sd %xmm1,%xmm1 cvtss2sd %xmm4,%xmm4 cvtss2sd %xmm7,%xmm7 movsd %xmm1,%xmm2 movsd %xmm4,%xmm5 movsd %xmm7,%xmm8 mulsd %xmm1,%xmm1 ## lu*lu mulsd %xmm4,%xmm4 ## lu*lu mulsd %xmm7,%xmm7 ## lu*lu movsd nb212_three(%rsp),%xmm9 movsd %xmm9,%xmm10 movsd %xmm9,%xmm11 mulsd %xmm0,%xmm1 ## rsq*lu*lu mulsd %xmm3,%xmm4 ## rsq*lu*lu mulsd %xmm6,%xmm7 ## rsq*lu*lu subsd %xmm1,%xmm9 subsd %xmm4,%xmm10 subsd %xmm7,%xmm11 ## 3-rsq*lu*lu mulsd %xmm2,%xmm9 mulsd %xmm5,%xmm10 mulsd %xmm8,%xmm11 ## lu*(3-rsq*lu*lu) movsd nb212_half(%rsp),%xmm15 mulsd %xmm15,%xmm9 ## first iteration for rinvOH1 mulsd %xmm15,%xmm10 ## first iteration for rinvH1H1 mulsd %xmm15,%xmm11 ## first iteration for rinvH2OH1 ## second iteration step movsd %xmm9,%xmm2 movsd %xmm10,%xmm5 movsd %xmm11,%xmm8 mulsd %xmm2,%xmm2 ## lu*lu mulsd %xmm5,%xmm5 ## lu*lu mulsd %xmm8,%xmm8 ## lu*lu movsd nb212_three(%rsp),%xmm1 movsd %xmm1,%xmm4 movsd %xmm1,%xmm7 mulsd %xmm0,%xmm2 ## rsq*lu*lu mulsd %xmm3,%xmm5 ## rsq*lu*lu mulsd %xmm6,%xmm8 ## rsq*lu*lu subsd %xmm2,%xmm1 subsd %xmm5,%xmm4 subsd %xmm8,%xmm7 ## 3-rsq*lu*lu mulsd %xmm1,%xmm9 mulsd %xmm4,%xmm10 mulsd %xmm7,%xmm11 ## lu*(3-rsq*lu*lu) movsd nb212_half(%rsp),%xmm15 mulsd %xmm15,%xmm9 ## rinvOH1 mulsd %xmm15,%xmm10 ## rinvH1H1 mulsd %xmm15,%xmm11 ## rinvH2H1 ## H1 interactions ## rsq in xmm0,xmm3,xmm6 ## rinv in xmm9, xmm10, xmm11 movsd %xmm9,%xmm1 ## copy of rinv movsd %xmm10,%xmm4 movsd %xmm11,%xmm7 movsd nb212_krf(%rsp),%xmm2 mulsd %xmm9,%xmm9 ## rinvsq mulsd %xmm10,%xmm10 mulsd %xmm11,%xmm11 mulsd %xmm2,%xmm0 ## k*rsq mulsd %xmm2,%xmm3 mulsd %xmm2,%xmm6 movsd %xmm0,%xmm2 ## copy of k*rsq movsd %xmm3,%xmm5 movsd %xmm6,%xmm8 addsd %xmm1,%xmm2 ## rinv+krsq addsd %xmm4,%xmm5 addsd %xmm7,%xmm8 movsd nb212_crf(%rsp),%xmm14 subsd %xmm14,%xmm2 ## rinv+krsq-crf subsd %xmm14,%xmm5 subsd %xmm14,%xmm8 movsd nb212_qqOH(%rsp),%xmm12 movsd nb212_qqHH(%rsp),%xmm13 mulsd %xmm12,%xmm2 ## xmm6=voul=qq*(rinv+ krsq-crf) mulsd %xmm13,%xmm5 ## xmm6=voul=qq*(rinv+ krsq-crf) mulsd %xmm13,%xmm8 ## xmm6=voul=qq*(rinv+ krsq-crf) addsd %xmm0,%xmm0 ## 2*krsq addsd %xmm3,%xmm3 addsd %xmm6,%xmm6 subsd %xmm0,%xmm1 ## rinv-2*krsq subsd %xmm3,%xmm4 subsd %xmm6,%xmm7 mulsd %xmm12,%xmm1 ## (rinv-2*krsq)*qq mulsd %xmm13,%xmm4 mulsd %xmm13,%xmm7 addsd nb212_vctot(%rsp),%xmm2 addsd %xmm8,%xmm5 addsd %xmm5,%xmm2 movsd %xmm2,nb212_vctot(%rsp) mulsd %xmm1,%xmm9 ## fscal mulsd %xmm4,%xmm10 mulsd %xmm7,%xmm11 ## move j H1 forces to xmm0-xmm2 movsd 24(%rdi,%rax,8),%xmm0 movsd 32(%rdi,%rax,8),%xmm1 movsd 40(%rdi,%rax,8),%xmm2 movsd %xmm9,%xmm7 movsd %xmm9,%xmm8 movsd %xmm11,%xmm13 movsd %xmm11,%xmm14 movsd %xmm11,%xmm15 movsd %xmm10,%xmm11 movsd %xmm10,%xmm12 mulsd nb212_dxOH1(%rsp),%xmm7 mulsd nb212_dyOH1(%rsp),%xmm8 mulsd nb212_dzOH1(%rsp),%xmm9 mulsd nb212_dxH1H1(%rsp),%xmm10 mulsd nb212_dyH1H1(%rsp),%xmm11 mulsd nb212_dzH1H1(%rsp),%xmm12 mulsd nb212_dxH2H1(%rsp),%xmm13 mulsd nb212_dyH2H1(%rsp),%xmm14 mulsd nb212_dzH2H1(%rsp),%xmm15 addsd %xmm7,%xmm0 addsd %xmm8,%xmm1 addsd %xmm9,%xmm2 addsd nb212_fixO(%rsp),%xmm7 addsd nb212_fiyO(%rsp),%xmm8 addsd nb212_fizO(%rsp),%xmm9 addsd %xmm10,%xmm0 addsd %xmm11,%xmm1 addsd %xmm12,%xmm2 addsd nb212_fixH1(%rsp),%xmm10 addsd nb212_fiyH1(%rsp),%xmm11 addsd nb212_fizH1(%rsp),%xmm12 addsd %xmm13,%xmm0 addsd %xmm14,%xmm1 addsd %xmm15,%xmm2 addsd nb212_fixH2(%rsp),%xmm13 addsd nb212_fiyH2(%rsp),%xmm14 addsd nb212_fizH2(%rsp),%xmm15 movsd %xmm7,nb212_fixO(%rsp) movsd %xmm8,nb212_fiyO(%rsp) movsd %xmm9,nb212_fizO(%rsp) movsd %xmm10,nb212_fixH1(%rsp) movsd %xmm11,nb212_fiyH1(%rsp) movsd %xmm12,nb212_fizH1(%rsp) movsd %xmm13,nb212_fixH2(%rsp) movsd %xmm14,nb212_fiyH2(%rsp) movsd %xmm15,nb212_fizH2(%rsp) ## store back j H1 forces from xmm0-xmm2 movsd %xmm0,24(%rdi,%rax,8) movsd %xmm1,32(%rdi,%rax,8) movsd %xmm2,40(%rdi,%rax,8) ## move j H2 coordinates to local temp variables movsd 48(%rsi,%rax,8),%xmm0 movsd 56(%rsi,%rax,8),%xmm1 movsd 64(%rsi,%rax,8),%xmm2 ## xmm0 = H2x ## xmm1 = H2y ## xmm2 = H2z movsd %xmm0,%xmm3 movsd %xmm1,%xmm4 movsd %xmm2,%xmm5 movsd %xmm0,%xmm6 movsd %xmm1,%xmm7 movsd %xmm2,%xmm8 subsd nb212_ixO(%rsp),%xmm0 subsd nb212_iyO(%rsp),%xmm1 subsd nb212_izO(%rsp),%xmm2 subsd nb212_ixH1(%rsp),%xmm3 subsd nb212_iyH1(%rsp),%xmm4 subsd nb212_izH1(%rsp),%xmm5 subsd nb212_ixH2(%rsp),%xmm6 subsd nb212_iyH2(%rsp),%xmm7 subsd nb212_izH2(%rsp),%xmm8 movsd %xmm0,nb212_dxOH2(%rsp) movsd %xmm1,nb212_dyOH2(%rsp) movsd %xmm2,nb212_dzOH2(%rsp) mulsd %xmm0,%xmm0 mulsd %xmm1,%xmm1 mulsd %xmm2,%xmm2 movsd %xmm3,nb212_dxH1H2(%rsp) movsd %xmm4,nb212_dyH1H2(%rsp) movsd %xmm5,nb212_dzH1H2(%rsp) mulsd %xmm3,%xmm3 mulsd %xmm4,%xmm4 mulsd %xmm5,%xmm5 movsd %xmm6,nb212_dxH2H2(%rsp) movsd %xmm7,nb212_dyH2H2(%rsp) movsd %xmm8,nb212_dzH2H2(%rsp) mulsd %xmm6,%xmm6 mulsd %xmm7,%xmm7 mulsd %xmm8,%xmm8 addsd %xmm1,%xmm0 addsd %xmm2,%xmm0 addsd %xmm4,%xmm3 addsd %xmm5,%xmm3 addsd %xmm7,%xmm6 addsd %xmm8,%xmm6 ## start doing invsqrt for jH2 atoms cvtsd2ss %xmm0,%xmm1 cvtsd2ss %xmm3,%xmm4 cvtsd2ss %xmm6,%xmm7 rsqrtss %xmm1,%xmm1 rsqrtss %xmm4,%xmm4 rsqrtss %xmm7,%xmm7 cvtss2sd %xmm1,%xmm1 cvtss2sd %xmm4,%xmm4 cvtss2sd %xmm7,%xmm7 movsd %xmm1,%xmm2 movsd %xmm4,%xmm5 movsd %xmm7,%xmm8 mulsd %xmm1,%xmm1 ## lu*lu mulsd %xmm4,%xmm4 ## lu*lu mulsd %xmm7,%xmm7 ## lu*lu movsd nb212_three(%rsp),%xmm9 movsd %xmm9,%xmm10 movsd %xmm9,%xmm11 mulsd %xmm0,%xmm1 ## rsq*lu*lu mulsd %xmm3,%xmm4 ## rsq*lu*lu mulsd %xmm6,%xmm7 ## rsq*lu*lu subsd %xmm1,%xmm9 subsd %xmm4,%xmm10 subsd %xmm7,%xmm11 ## 3-rsq*lu*lu mulsd %xmm2,%xmm9 mulsd %xmm5,%xmm10 mulsd %xmm8,%xmm11 ## lu*(3-rsq*lu*lu) movsd nb212_half(%rsp),%xmm15 mulsd %xmm15,%xmm9 ## first iteration for rinvOH2 mulsd %xmm15,%xmm10 ## first iteration for rinvH1H2 mulsd %xmm15,%xmm11 ## first iteration for rinvH2H2 ## second iteration step movsd %xmm9,%xmm2 movsd %xmm10,%xmm5 movsd %xmm11,%xmm8 mulsd %xmm2,%xmm2 ## lu*lu mulsd %xmm5,%xmm5 ## lu*lu mulsd %xmm8,%xmm8 ## lu*lu movsd nb212_three(%rsp),%xmm1 movsd %xmm1,%xmm4 movsd %xmm1,%xmm7 mulsd %xmm0,%xmm2 ## rsq*lu*lu mulsd %xmm3,%xmm5 ## rsq*lu*lu mulsd %xmm6,%xmm8 ## rsq*lu*lu subsd %xmm2,%xmm1 subsd %xmm5,%xmm4 subsd %xmm8,%xmm7 ## 3-rsq*lu*lu mulsd %xmm1,%xmm9 mulsd %xmm4,%xmm10 mulsd %xmm7,%xmm11 ## lu*(3-rsq*lu*lu) movsd nb212_half(%rsp),%xmm15 mulsd %xmm15,%xmm9 ## rinvOH2 mulsd %xmm15,%xmm10 ## rinvH1H2 mulsd %xmm15,%xmm11 ## rinvH2H2 ## H2 interactions ## rsq in xmm0,xmm3,xmm6 ## rinv in xmm9, xmm10, xmm11 movsd %xmm9,%xmm1 ## copy of rinv movsd %xmm10,%xmm4 movsd %xmm11,%xmm7 movsd nb212_krf(%rsp),%xmm2 mulsd %xmm9,%xmm9 ## rinvsq mulsd %xmm10,%xmm10 mulsd %xmm11,%xmm11 mulsd %xmm2,%xmm0 ## k*rsq mulsd %xmm2,%xmm3 mulsd %xmm2,%xmm6 movsd %xmm0,%xmm2 ## copy of k*rsq movsd %xmm3,%xmm5 movsd %xmm6,%xmm8 addsd %xmm1,%xmm2 ## rinv+krsq addsd %xmm4,%xmm5 addsd %xmm7,%xmm8 movsd nb212_crf(%rsp),%xmm14 subsd %xmm14,%xmm2 ## rinv+krsq-crf subsd %xmm14,%xmm5 subsd %xmm14,%xmm8 movsd nb212_qqOH(%rsp),%xmm12 movsd nb212_qqHH(%rsp),%xmm13 mulsd %xmm12,%xmm2 ## xmm6=voul=qq*(rinv+ krsq-crf) mulsd %xmm13,%xmm5 ## xmm6=voul=qq*(rinv+ krsq-crf) mulsd %xmm13,%xmm8 ## xmm6=voul=qq*(rinv+ krsq-crf) addsd %xmm0,%xmm0 ## 2*krsq addsd %xmm3,%xmm3 addsd %xmm6,%xmm6 subsd %xmm0,%xmm1 ## rinv-2*krsq subsd %xmm3,%xmm4 subsd %xmm6,%xmm7 mulsd %xmm12,%xmm1 ## (rinv-2*krsq)*qq mulsd %xmm13,%xmm4 mulsd %xmm13,%xmm7 addsd nb212_vctot(%rsp),%xmm2 addsd %xmm8,%xmm5 addsd %xmm5,%xmm2 movsd %xmm2,nb212_vctot(%rsp) mulsd %xmm1,%xmm9 ## fscal mulsd %xmm4,%xmm10 mulsd %xmm7,%xmm11 ## move j H2 forces to xmm0-xmm2 movsd 48(%rdi,%rax,8),%xmm0 movsd 56(%rdi,%rax,8),%xmm1 movsd 64(%rdi,%rax,8),%xmm2 movsd %xmm9,%xmm7 movsd %xmm9,%xmm8 movsd %xmm11,%xmm13 movsd %xmm11,%xmm14 movsd %xmm11,%xmm15 movsd %xmm10,%xmm11 movsd %xmm10,%xmm12 mulsd nb212_dxOH2(%rsp),%xmm7 mulsd nb212_dyOH2(%rsp),%xmm8 mulsd nb212_dzOH2(%rsp),%xmm9 mulsd nb212_dxH1H2(%rsp),%xmm10 mulsd nb212_dyH1H2(%rsp),%xmm11 mulsd nb212_dzH1H2(%rsp),%xmm12 mulsd nb212_dxH2H2(%rsp),%xmm13 mulsd nb212_dyH2H2(%rsp),%xmm14 mulsd nb212_dzH2H2(%rsp),%xmm15 addsd %xmm7,%xmm0 addsd %xmm8,%xmm1 addsd %xmm9,%xmm2 addsd nb212_fixO(%rsp),%xmm7 addsd nb212_fiyO(%rsp),%xmm8 addsd nb212_fizO(%rsp),%xmm9 addsd %xmm10,%xmm0 addsd %xmm11,%xmm1 addsd %xmm12,%xmm2 addsd nb212_fixH1(%rsp),%xmm10 addsd nb212_fiyH1(%rsp),%xmm11 addsd nb212_fizH1(%rsp),%xmm12 addsd %xmm13,%xmm0 addsd %xmm14,%xmm1 addsd %xmm15,%xmm2 addsd nb212_fixH2(%rsp),%xmm13⌨️ 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?