nb_kernel202_x86_64_sse.s
来自「最著名最快的分子模拟软件」· S 代码 · 共 2,118 行 · 第 1/5 页
S
2,118 行
movaps %xmm4,nb202_dyH1O(%rsp) movaps %xmm5,nb202_dzH1O(%rsp) mulps %xmm3,%xmm3 mulps %xmm4,%xmm4 mulps %xmm5,%xmm5 movaps %xmm6,nb202_dxH2O(%rsp) movaps %xmm7,nb202_dyH2O(%rsp) movaps %xmm8,nb202_dzH2O(%rsp) mulps %xmm6,%xmm6 mulps %xmm7,%xmm7 mulps %xmm8,%xmm8 addps %xmm1,%xmm0 addps %xmm2,%xmm0 addps %xmm4,%xmm3 addps %xmm5,%xmm3 addps %xmm7,%xmm6 addps %xmm8,%xmm6 ## start doing invsqrt for jO atoms rsqrtps %xmm0,%xmm1 rsqrtps %xmm3,%xmm4 rsqrtps %xmm6,%xmm7 movaps %xmm1,%xmm2 movaps %xmm4,%xmm5 movaps %xmm7,%xmm8 mulps %xmm1,%xmm1 ## lu*lu mulps %xmm4,%xmm4 ## lu*lu mulps %xmm7,%xmm7 ## lu*lu movaps nb202_three(%rsp),%xmm9 movaps %xmm9,%xmm10 movaps %xmm9,%xmm11 mulps %xmm0,%xmm1 ## rsq*lu*lu mulps %xmm3,%xmm4 ## rsq*lu*lu mulps %xmm6,%xmm7 ## rsq*lu*lu subps %xmm1,%xmm9 subps %xmm4,%xmm10 subps %xmm7,%xmm11 ## 3-rsq*lu*lu mulps %xmm2,%xmm9 mulps %xmm5,%xmm10 mulps %xmm8,%xmm11 ## lu*(3-rsq*lu*lu) movaps nb202_half(%rsp),%xmm4 mulps %xmm4,%xmm9 ## rinvOO mulps %xmm4,%xmm10 ## rinvH1O mulps %xmm4,%xmm11 ## rinvH2O ## O interactions ## rsq in xmm0,xmm3,xmm6 ## rinv in xmm9, xmm10, xmm11 movaps %xmm9,%xmm1 ## copy of rinv movaps %xmm10,%xmm4 movaps %xmm11,%xmm7 movaps nb202_krf(%rsp),%xmm2 mulps %xmm9,%xmm9 ## rinvsq mulps %xmm10,%xmm10 mulps %xmm11,%xmm11 mulps %xmm2,%xmm0 ## k*rsq mulps %xmm2,%xmm3 mulps %xmm2,%xmm6 movaps %xmm0,%xmm2 ## copy of k*rsq movaps %xmm3,%xmm5 movaps %xmm6,%xmm8 addps %xmm1,%xmm2 ## rinv+krsq addps %xmm4,%xmm5 addps %xmm7,%xmm8 movaps nb202_crf(%rsp),%xmm14 subps %xmm14,%xmm2 ## rinv+krsq-crf subps %xmm14,%xmm5 subps %xmm14,%xmm8 movaps nb202_qqOO(%rsp),%xmm12 movaps nb202_qqOH(%rsp),%xmm13 mulps %xmm12,%xmm2 ## voul=qq*(rinv+ krsq-crf) mulps %xmm13,%xmm5 ## voul=qq*(rinv+ krsq-crf) mulps %xmm13,%xmm8 ## voul=qq*(rinv+ krsq-crf) addps %xmm0,%xmm0 ## 2*krsq addps %xmm3,%xmm3 addps %xmm6,%xmm6 subps %xmm0,%xmm1 ## rinv-2*krsq subps %xmm3,%xmm4 subps %xmm6,%xmm7 mulps %xmm12,%xmm1 ## (rinv-2*krsq)*qq mulps %xmm13,%xmm4 mulps %xmm13,%xmm7 addps nb202_vctot(%rsp),%xmm2 addps %xmm8,%xmm5 addps %xmm5,%xmm2 movaps %xmm2,%xmm15 mulps %xmm9,%xmm1 ## fscal mulps %xmm10,%xmm4 mulps %xmm11,%xmm7 ## move j O forces to local temp variables movlps (%rdi,%rax,4),%xmm9 ## jxOa jyOa - - movlps (%rdi,%rcx,4),%xmm10 ## jxOc jyOc - - movhps (%rdi,%rbx,4),%xmm9 ## jxOa jyOa jxOb jyOb movhps (%rdi,%rdx,4),%xmm10 ## jxOc jyOc jxOd jyOd movss 8(%rdi,%rax,4),%xmm11 ## jzOa - - - movss 8(%rdi,%rcx,4),%xmm12 ## jzOc - - - movhps 8(%rdi,%rbx,4),%xmm11 ## jzOa - jzOb - movhps 8(%rdi,%rdx,4),%xmm12 ## jzOc - jzOd - shufps $136,%xmm12,%xmm11 ## 10001000 => jzOa jzOb jzOc jzOd ## xmm9: jxOa jyOa jxOb jyOb ## xmm10: jxOc jyOc jxOd jyOd ## xmm11: jzOa jzOb jzOc jzOd movaps %xmm1,%xmm0 movaps %xmm1,%xmm2 movaps %xmm4,%xmm3 movaps %xmm4,%xmm5 movaps %xmm7,%xmm6 movaps %xmm7,%xmm8 mulps nb202_dxOO(%rsp),%xmm0 mulps nb202_dyOO(%rsp),%xmm1 mulps nb202_dzOO(%rsp),%xmm2 mulps nb202_dxH1O(%rsp),%xmm3 mulps nb202_dyH1O(%rsp),%xmm4 mulps nb202_dzH1O(%rsp),%xmm5 mulps nb202_dxH2O(%rsp),%xmm6 mulps nb202_dyH2O(%rsp),%xmm7 mulps nb202_dzH2O(%rsp),%xmm8 movaps %xmm0,%xmm13 movaps %xmm1,%xmm14 addps %xmm2,%xmm11 addps nb202_fixO(%rsp),%xmm0 addps nb202_fiyO(%rsp),%xmm1 addps nb202_fizO(%rsp),%xmm2 addps %xmm3,%xmm13 addps %xmm4,%xmm14 addps %xmm5,%xmm11 addps nb202_fixH1(%rsp),%xmm3 addps nb202_fiyH1(%rsp),%xmm4 addps nb202_fizH1(%rsp),%xmm5 addps %xmm6,%xmm13 addps %xmm7,%xmm14 addps %xmm8,%xmm11 addps nb202_fixH2(%rsp),%xmm6 addps nb202_fiyH2(%rsp),%xmm7 addps nb202_fizH2(%rsp),%xmm8 movaps %xmm0,nb202_fixO(%rsp) movaps %xmm1,nb202_fiyO(%rsp) movaps %xmm2,nb202_fizO(%rsp) movaps %xmm3,nb202_fixH1(%rsp) movaps %xmm4,nb202_fiyH1(%rsp) movaps %xmm5,nb202_fizH1(%rsp) movaps %xmm6,nb202_fixH2(%rsp) movaps %xmm7,nb202_fiyH2(%rsp) movaps %xmm8,nb202_fizH2(%rsp) ## xmm9 = fOx ## xmm10 = fOy ## xmm11 = fOz movaps %xmm13,%xmm0 unpcklps %xmm14,%xmm13 unpckhps %xmm14,%xmm0 addps %xmm13,%xmm9 addps %xmm0,%xmm10 movhlps %xmm11,%xmm12 ## fOzc fOzd movlps %xmm9,(%rdi,%rax,4) movhps %xmm9,(%rdi,%rbx,4) movlps %xmm10,(%rdi,%rcx,4) movhps %xmm10,(%rdi,%rdx,4) movss %xmm11,8(%rdi,%rax,4) movss %xmm12,8(%rdi,%rcx,4) shufps $1,%xmm11,%xmm11 shufps $1,%xmm12,%xmm12 movss %xmm11,8(%rdi,%rbx,4) movss %xmm12,8(%rdi,%rdx,4) ## move j H1 coordinates to local temp variables movlps 12(%rsi,%rax,4),%xmm0 ## jxH1a jyH1a - - movlps 12(%rsi,%rcx,4),%xmm1 ## jxH1c jyH1c - - movhps 12(%rsi,%rbx,4),%xmm0 ## jxH1a jyH1a jxH1b jyH1b movhps 12(%rsi,%rdx,4),%xmm1 ## jxH1c jyH1c jxH1d jyH1d movss 20(%rsi,%rax,4),%xmm2 ## jzH1a - - - movss 20(%rsi,%rcx,4),%xmm3 ## jzH1c - - - movhps 20(%rsi,%rbx,4),%xmm2 ## jzH1a - jzH1b - movhps 20(%rsi,%rdx,4),%xmm3 ## jzH1c - jzH1d - movaps %xmm0,%xmm4 unpcklps %xmm1,%xmm0 ## jxH1a jxH1c jyH1a jyH1c unpckhps %xmm1,%xmm4 ## jxH1b jxH1d jyH1b jyH1d movaps %xmm0,%xmm1 unpcklps %xmm4,%xmm0 ## x unpckhps %xmm4,%xmm1 ## y shufps $136,%xmm3,%xmm2 ## 10001000 => jzH1a jzH1b jzH1c jzH1d ## xmm0 = H1x ## xmm1 = H1y ## xmm2 = H1z movaps %xmm0,%xmm3 movaps %xmm1,%xmm4 movaps %xmm2,%xmm5 movaps %xmm0,%xmm6 movaps %xmm1,%xmm7 movaps %xmm2,%xmm8 subps nb202_ixO(%rsp),%xmm0 subps nb202_iyO(%rsp),%xmm1 subps nb202_izO(%rsp),%xmm2 subps nb202_ixH1(%rsp),%xmm3 subps nb202_iyH1(%rsp),%xmm4 subps nb202_izH1(%rsp),%xmm5 subps nb202_ixH2(%rsp),%xmm6 subps nb202_iyH2(%rsp),%xmm7 subps nb202_izH2(%rsp),%xmm8 movaps %xmm0,nb202_dxOH1(%rsp) movaps %xmm1,nb202_dyOH1(%rsp) movaps %xmm2,nb202_dzOH1(%rsp) mulps %xmm0,%xmm0 mulps %xmm1,%xmm1 mulps %xmm2,%xmm2 movaps %xmm3,nb202_dxH1H1(%rsp) movaps %xmm4,nb202_dyH1H1(%rsp) movaps %xmm5,nb202_dzH1H1(%rsp) mulps %xmm3,%xmm3 mulps %xmm4,%xmm4 mulps %xmm5,%xmm5 movaps %xmm6,nb202_dxH2H1(%rsp) movaps %xmm7,nb202_dyH2H1(%rsp) movaps %xmm8,nb202_dzH2H1(%rsp) mulps %xmm6,%xmm6 mulps %xmm7,%xmm7 mulps %xmm8,%xmm8 addps %xmm1,%xmm0 addps %xmm2,%xmm0 addps %xmm4,%xmm3 addps %xmm5,%xmm3 addps %xmm7,%xmm6 addps %xmm8,%xmm6 ## start doing invsqrt for jH1 atoms rsqrtps %xmm0,%xmm1 rsqrtps %xmm3,%xmm4 rsqrtps %xmm6,%xmm7 movaps %xmm1,%xmm2 movaps %xmm4,%xmm5 movaps %xmm7,%xmm8 mulps %xmm1,%xmm1 ## lu*lu mulps %xmm4,%xmm4 ## lu*lu mulps %xmm7,%xmm7 ## lu*lu movaps nb202_three(%rsp),%xmm9 movaps %xmm9,%xmm10 movaps %xmm9,%xmm11 mulps %xmm0,%xmm1 ## rsq*lu*lu mulps %xmm3,%xmm4 ## rsq*lu*lu mulps %xmm6,%xmm7 ## rsq*lu*lu subps %xmm1,%xmm9 subps %xmm4,%xmm10 subps %xmm7,%xmm11 ## 3-rsq*lu*lu mulps %xmm2,%xmm9 mulps %xmm5,%xmm10 mulps %xmm8,%xmm11 ## lu*(3-rsq*lu*lu) movaps nb202_half(%rsp),%xmm4 mulps %xmm4,%xmm9 ## rinvOH1 mulps %xmm4,%xmm10 ## rinvH1H1 mulps %xmm4,%xmm11 ## rinvH2H1 ## H1 interactions ## rsq in xmm0,xmm3,xmm6 ## rinv in xmm9, xmm10, xmm11 movaps %xmm9,%xmm1 ## copy of rinv movaps %xmm10,%xmm4 movaps %xmm11,%xmm7 movaps nb202_krf(%rsp),%xmm2 mulps %xmm9,%xmm9 ## rinvsq mulps %xmm10,%xmm10 mulps %xmm11,%xmm11 mulps %xmm2,%xmm0 ## k*rsq mulps %xmm2,%xmm3 mulps %xmm2,%xmm6 movaps %xmm0,%xmm2 ## copy of k*rsq movaps %xmm3,%xmm5 movaps %xmm6,%xmm8 addps %xmm1,%xmm2 ## rinv+krsq addps %xmm4,%xmm5 addps %xmm7,%xmm8 movaps nb202_crf(%rsp),%xmm14 subps %xmm14,%xmm2 ## rinv+krsq-crf subps %xmm14,%xmm5 subps %xmm14,%xmm8 movaps nb202_qqOH(%rsp),%xmm12 movaps nb202_qqHH(%rsp),%xmm13 mulps %xmm12,%xmm2 ## xmm6=voul=qq*(rinv+ krsq-crf) mulps %xmm13,%xmm5 ## xmm6=voul=qq*(rinv+ krsq-crf) mulps %xmm13,%xmm8 ## xmm6=voul=qq*(rinv+ krsq-crf) addps %xmm0,%xmm0 ## 2*krsq addps %xmm3,%xmm3 addps %xmm6,%xmm6 subps %xmm0,%xmm1 ## rinv-2*krsq subps %xmm3,%xmm4 subps %xmm6,%xmm7 mulps %xmm12,%xmm1 ## (rinv-2*krsq)*qq mulps %xmm13,%xmm4 mulps %xmm13,%xmm7 addps %xmm2,%xmm15 addps %xmm8,%xmm5 addps %xmm5,%xmm15 mulps %xmm9,%xmm1 ## fscal mulps %xmm10,%xmm4 mulps %xmm11,%xmm7 ## move j H1 forces to local temp variables movlps 12(%rdi,%rax,4),%xmm9 ## jxH1a jyH1a - - movlps 12(%rdi,%rcx,4),%xmm10 ## jxH1c jyH1c - - movhps 12(%rdi,%rbx,4),%xmm9 ## jxH1a jyH1a jxH1b jyH1b movhps 12(%rdi,%rdx,4),%xmm10 ## jxH1c jyH1c jxH1d jyH1d movss 20(%rdi,%rax,4),%xmm11 ## jzH1a - - - movss 20(%rdi,%rcx,4),%xmm12 ## jzH1c - - - movhps 20(%rdi,%rbx,4),%xmm11 ## jzH1a - jzH1b - movhps 20(%rdi,%rdx,4),%xmm12 ## jzH1c - jzH1d - shufps $136,%xmm12,%xmm11 ## 10001000 => jzH1a jzH1b jzH1c jzH1d ## xmm9: jxH1a jyH1a jxH1b jyH1b ## xmm10: jxH1c jyH1c jxH1d jyH1d ## xmm11: jzH1a jzH1b jzH1c jzH1d movaps %xmm1,%xmm0 movaps %xmm1,%xmm2 movaps %xmm4,%xmm3 movaps %xmm4,%xmm5 movaps %xmm7,%xmm6 movaps %xmm7,%xmm8 mulps nb202_dxOH1(%rsp),%xmm0 mulps nb202_dyOH1(%rsp),%xmm1 mulps nb202_dzOH1(%rsp),%xmm2 mulps nb202_dxH1H1(%rsp),%xmm3 mulps nb202_dyH1H1(%rsp),%xmm4 mulps nb202_dzH1H1(%rsp),%xmm5 mulps nb202_dxH2H1(%rsp),%xmm6 mulps nb202_dyH2H1(%rsp),%xmm7 mulps nb202_dzH2H1(%rsp),%xmm8 movaps %xmm0,%xmm13 movaps %xmm1,%xmm14 addps %xmm2,%xmm11 addps nb202_fixO(%rsp),%xmm0 addps nb202_fiyO(%rsp),%xmm1 addps nb202_fizO(%rsp),%xmm2 addps %xmm3,%xmm13 addps %xmm4,%xmm14 addps %xmm5,%xmm11 addps nb202_fixH1(%rsp),%xmm3 addps nb202_fiyH1(%rsp),%xmm4 addps nb202_fizH1(%rsp),%xmm5 addps %xmm6,%xmm13 addps %xmm7,%xmm14 addps %xmm8,%xmm11 addps nb202_fixH2(%rsp),%xmm6 addps nb202_fiyH2(%rsp),%xmm7 addps nb202_fizH2(%rsp),%xmm8 movaps %xmm0,nb202_fixO(%rsp) movaps %xmm1,nb202_fiyO(%rsp) movaps %xmm2,nb202_fizO(%rsp) movaps %xmm3,nb202_fixH1(%rsp) movaps %xmm4,nb202_fiyH1(%rsp) movaps %xmm5,nb202_fizH1(%rsp) movaps %xmm6,nb202_fixH2(%rsp) movaps %xmm7,nb202_fiyH2(%rsp) movaps %xmm8,nb202_fizH2(%rsp) ## xmm9 = fH1x ## xmm10 = fH1y ## xmm11 = fH1z movaps %xmm13,%xmm0 unpcklps %xmm14,%xmm13 unpckhps %xmm14,%xmm0 addps %xmm13,%xmm9 addps %xmm0,%xmm10 movhlps %xmm11,%xmm12 ## fH1zc fH1zd movlps %xmm9,12(%rdi,%rax,4) movhps %xmm9,12(%rdi,%rbx,4) movlps %xmm10,12(%rdi,%rcx,4) movhps %xmm10,12(%rdi,%rdx,4) movss %xmm11,20(%rdi,%rax,4) movss %xmm12,20(%rdi,%rcx,4) shufps $1,%xmm11,%xmm11 shufps $1,%xmm12,%xmm12 movss %xmm11,20(%rdi,%rbx,4) movss %xmm12,20(%rdi,%rdx,4) ## move j H2 coordinates to local temp variables movlps 24(%rsi,%rax,4),%xmm0 ## jxH2a jyH2a - -⌨️ 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?