nb_kernel112_x86_64_sse.s
来自「最著名最快的分子模拟软件」· S 代码 · 共 2,103 行 · 第 1/5 页
S
2,103 行
subps nb112_ixO(%rsp),%xmm0 subps nb112_iyO(%rsp),%xmm1 subps nb112_izO(%rsp),%xmm2 subps nb112_ixH1(%rsp),%xmm3 subps nb112_iyH1(%rsp),%xmm4 subps nb112_izH1(%rsp),%xmm5 subps nb112_ixH2(%rsp),%xmm6 subps nb112_iyH2(%rsp),%xmm7 subps nb112_izH2(%rsp),%xmm8 movaps %xmm0,nb112_dxOO(%rsp) movaps %xmm1,nb112_dyOO(%rsp) movaps %xmm2,nb112_dzOO(%rsp) mulps %xmm0,%xmm0 mulps %xmm1,%xmm1 mulps %xmm2,%xmm2 movaps %xmm3,nb112_dxH1O(%rsp) movaps %xmm4,nb112_dyH1O(%rsp) movaps %xmm5,nb112_dzH1O(%rsp) mulps %xmm3,%xmm3 mulps %xmm4,%xmm4 mulps %xmm5,%xmm5 movaps %xmm6,nb112_dxH2O(%rsp) movaps %xmm7,nb112_dyH2O(%rsp) movaps %xmm8,nb112_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 nb112_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 nb112_half(%rsp),%xmm0 mulps %xmm0,%xmm9 ## rinvOO mulps %xmm0,%xmm10 ## rinvH1O mulps %xmm0,%xmm11 ## rinvH2O ## O interactions movaps %xmm9,%xmm0 movaps %xmm10,%xmm1 movaps %xmm11,%xmm2 mulps %xmm9,%xmm9 ## rinvsq mulps %xmm10,%xmm10 mulps %xmm11,%xmm11 movaps %xmm9,%xmm12 mulps %xmm12,%xmm12 ## rinv4 mulps %xmm9,%xmm12 ## rinv6 mulps nb112_qqOO(%rsp),%xmm0 mulps nb112_qqOH(%rsp),%xmm1 mulps nb112_qqOH(%rsp),%xmm2 movaps %xmm12,%xmm13 ## rinv6 mulps %xmm12,%xmm12 ## rinv12 mulps nb112_c6(%rsp),%xmm13 mulps nb112_c12(%rsp),%xmm12 movaps %xmm12,%xmm14 subps %xmm13,%xmm14 addps nb112_Vvdwtot(%rsp),%xmm14 mulps nb112_six(%rsp),%xmm13 mulps nb112_twelve(%rsp),%xmm12 movaps %xmm14,nb112_Vvdwtot(%rsp) subps %xmm13,%xmm12 ## LJ fscal addps %xmm0,%xmm12 mulps %xmm12,%xmm9 mulps %xmm1,%xmm10 mulps %xmm2,%xmm11 addps nb112_vctot(%rsp),%xmm0 addps %xmm2,%xmm1 addps %xmm1,%xmm0 movaps %xmm0,nb112_vctot(%rsp) ## move j O forces to local temp variables movlps (%rdi,%rax,4),%xmm0 ## jxOa jyOa - - movlps (%rdi,%rcx,4),%xmm1 ## jxOc jyOc - - movhps (%rdi,%rbx,4),%xmm0 ## jxOa jyOa jxOb jyOb movhps (%rdi,%rdx,4),%xmm1 ## jxOc jyOc jxOd jyOd movss 8(%rdi,%rax,4),%xmm2 ## jzOa - - - movss 8(%rdi,%rcx,4),%xmm3 ## jzOc - - - movhps 8(%rdi,%rbx,4),%xmm2 ## jzOa - jzOb - movhps 8(%rdi,%rdx,4),%xmm3 ## jzOc - jzOd - shufps $136,%xmm3,%xmm2 ## 10001000 => jzOa jzOb jzOc jzOd ## xmm0: jxOa jyOa jxOb jyOb ## xmm1: jxOc jyOc jxOd jyOd ## xmm2: jzOa jzOb jzOc jzOd movaps %xmm9,%xmm7 movaps %xmm9,%xmm8 movaps %xmm11,%xmm13 movaps %xmm11,%xmm14 movaps %xmm11,%xmm15 movaps %xmm10,%xmm11 movaps %xmm10,%xmm12 mulps nb112_dxOO(%rsp),%xmm7 mulps nb112_dyOO(%rsp),%xmm8 mulps nb112_dzOO(%rsp),%xmm9 mulps nb112_dxH1O(%rsp),%xmm10 mulps nb112_dyH1O(%rsp),%xmm11 mulps nb112_dzH1O(%rsp),%xmm12 mulps nb112_dxH2O(%rsp),%xmm13 mulps nb112_dyH2O(%rsp),%xmm14 mulps nb112_dzH2O(%rsp),%xmm15 movaps %xmm7,%xmm3 movaps %xmm8,%xmm4 addps %xmm9,%xmm2 addps nb112_fixO(%rsp),%xmm7 addps nb112_fiyO(%rsp),%xmm8 addps nb112_fizO(%rsp),%xmm9 addps %xmm10,%xmm3 addps %xmm11,%xmm4 addps %xmm12,%xmm2 addps nb112_fixH1(%rsp),%xmm10 addps nb112_fiyH1(%rsp),%xmm11 addps nb112_fizH1(%rsp),%xmm12 addps %xmm13,%xmm3 addps %xmm14,%xmm4 addps %xmm15,%xmm2 addps nb112_fixH2(%rsp),%xmm13 addps nb112_fiyH2(%rsp),%xmm14 addps nb112_fizH2(%rsp),%xmm15 movaps %xmm7,nb112_fixO(%rsp) movaps %xmm8,nb112_fiyO(%rsp) movaps %xmm9,nb112_fizO(%rsp) movaps %xmm10,nb112_fixH1(%rsp) movaps %xmm11,nb112_fiyH1(%rsp) movaps %xmm12,nb112_fizH1(%rsp) movaps %xmm13,nb112_fixH2(%rsp) movaps %xmm14,nb112_fiyH2(%rsp) movaps %xmm15,nb112_fizH2(%rsp) ## xmm3 = fOx , xmm4 = fOy, xmm5=fOz movaps %xmm3,%xmm5 unpcklps %xmm4,%xmm3 unpckhps %xmm4,%xmm5 addps %xmm3,%xmm0 addps %xmm5,%xmm1 movhlps %xmm2,%xmm3 ## fOzc fOzd movlps %xmm0,(%rdi,%rax,4) movhps %xmm0,(%rdi,%rbx,4) movlps %xmm1,(%rdi,%rcx,4) movhps %xmm1,(%rdi,%rdx,4) movss %xmm2,8(%rdi,%rax,4) movss %xmm3,8(%rdi,%rcx,4) shufps $1,%xmm2,%xmm2 shufps $1,%xmm3,%xmm3 movss %xmm2,8(%rdi,%rbx,4) movss %xmm3,8(%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 nb112_ixO(%rsp),%xmm0 subps nb112_iyO(%rsp),%xmm1 subps nb112_izO(%rsp),%xmm2 subps nb112_ixH1(%rsp),%xmm3 subps nb112_iyH1(%rsp),%xmm4 subps nb112_izH1(%rsp),%xmm5 subps nb112_ixH2(%rsp),%xmm6 subps nb112_iyH2(%rsp),%xmm7 subps nb112_izH2(%rsp),%xmm8 movaps %xmm0,nb112_dxOH1(%rsp) movaps %xmm1,nb112_dyOH1(%rsp) movaps %xmm2,nb112_dzOH1(%rsp) mulps %xmm0,%xmm0 mulps %xmm1,%xmm1 mulps %xmm2,%xmm2 movaps %xmm3,nb112_dxH1H1(%rsp) movaps %xmm4,nb112_dyH1H1(%rsp) movaps %xmm5,nb112_dzH1H1(%rsp) mulps %xmm3,%xmm3 mulps %xmm4,%xmm4 mulps %xmm5,%xmm5 movaps %xmm6,nb112_dxH2H1(%rsp) movaps %xmm7,nb112_dyH2H1(%rsp) movaps %xmm8,nb112_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 nb112_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 nb112_half(%rsp),%xmm0 mulps %xmm0,%xmm9 ## rinvOH1 mulps %xmm0,%xmm10 ## rinvH1H1 mulps %xmm0,%xmm11 ## rinvH2H1 ## H1 interactions movaps %xmm9,%xmm0 movaps %xmm10,%xmm1 movaps %xmm11,%xmm2 mulps %xmm9,%xmm9 mulps %xmm10,%xmm10 mulps %xmm11,%xmm11 mulps nb112_qqOH(%rsp),%xmm0 mulps nb112_qqHH(%rsp),%xmm1 mulps nb112_qqHH(%rsp),%xmm2 mulps %xmm0,%xmm9 mulps %xmm1,%xmm10 mulps %xmm2,%xmm11 addps nb112_vctot(%rsp),%xmm0 addps %xmm2,%xmm1 addps %xmm1,%xmm0 movaps %xmm0,nb112_vctot(%rsp) ## move j H1 forces to local temp variables movlps 12(%rdi,%rax,4),%xmm0 ## jxH1a jyH1a - - movlps 12(%rdi,%rcx,4),%xmm1 ## jxH1c jyH1c - - movhps 12(%rdi,%rbx,4),%xmm0 ## jxH1a jyH1a jxH1b jyH1b movhps 12(%rdi,%rdx,4),%xmm1 ## jxH1c jyH1c jxH1d jyH1d movss 20(%rdi,%rax,4),%xmm2 ## jzH1a - - - movss 20(%rdi,%rcx,4),%xmm3 ## jzH1c - - - movhps 20(%rdi,%rbx,4),%xmm2 ## jzH1a - jzH1b - movhps 20(%rdi,%rdx,4),%xmm3 ## jzH1c - jzH1d - shufps $136,%xmm3,%xmm2 ## 10001000 => jzH1a jzH1b jzH1c jzH1d ## xmm0: jxH1a jyH1a jxH1b jyH1b ## xmm1: jxH1c jyH1c jxH1d jyH1d ## xmm2: jzH1a jzH1b jzH1c jzH1d movaps %xmm9,%xmm7 movaps %xmm9,%xmm8 movaps %xmm11,%xmm13 movaps %xmm11,%xmm14 movaps %xmm11,%xmm15 movaps %xmm10,%xmm11 movaps %xmm10,%xmm12 mulps nb112_dxOH1(%rsp),%xmm7 mulps nb112_dyOH1(%rsp),%xmm8 mulps nb112_dzOH1(%rsp),%xmm9 mulps nb112_dxH1H1(%rsp),%xmm10 mulps nb112_dyH1H1(%rsp),%xmm11 mulps nb112_dzH1H1(%rsp),%xmm12 mulps nb112_dxH2H1(%rsp),%xmm13 mulps nb112_dyH2H1(%rsp),%xmm14 mulps nb112_dzH2H1(%rsp),%xmm15 movaps %xmm7,%xmm3 movaps %xmm8,%xmm4 addps %xmm9,%xmm2 addps nb112_fixO(%rsp),%xmm7 addps nb112_fiyO(%rsp),%xmm8 addps nb112_fizO(%rsp),%xmm9 addps %xmm10,%xmm3 addps %xmm11,%xmm4 addps %xmm12,%xmm2 addps nb112_fixH1(%rsp),%xmm10 addps nb112_fiyH1(%rsp),%xmm11 addps nb112_fizH1(%rsp),%xmm12 addps %xmm13,%xmm3 addps %xmm14,%xmm4 addps %xmm15,%xmm2 addps nb112_fixH2(%rsp),%xmm13 addps nb112_fiyH2(%rsp),%xmm14 addps nb112_fizH2(%rsp),%xmm15 movaps %xmm7,nb112_fixO(%rsp) movaps %xmm8,nb112_fiyO(%rsp) movaps %xmm9,nb112_fizO(%rsp) movaps %xmm10,nb112_fixH1(%rsp) movaps %xmm11,nb112_fiyH1(%rsp) movaps %xmm12,nb112_fizH1(%rsp) movaps %xmm13,nb112_fixH2(%rsp) movaps %xmm14,nb112_fiyH2(%rsp) movaps %xmm15,nb112_fizH2(%rsp) ## xmm0 = fH1x ## xmm1 = fH1y ## xmm2 = fH1z movaps %xmm3,%xmm5 unpcklps %xmm4,%xmm3 unpckhps %xmm4,%xmm5 addps %xmm3,%xmm0 addps %xmm5,%xmm1 movhlps %xmm2,%xmm3 ## fH1zc fH1zd movlps %xmm0,12(%rdi,%rax,4) movhps %xmm0,12(%rdi,%rbx,4) movlps %xmm1,12(%rdi,%rcx,4) movhps %xmm1,12(%rdi,%rdx,4) movss %xmm2,20(%rdi,%rax,4) movss %xmm3,20(%rdi,%rcx,4) shufps $1,%xmm2,%xmm2 shufps $1,%xmm3,%xmm3 movss %xmm2,20(%rdi,%rbx,4) movss %xmm3,20(%rdi,%rdx,4) ## move j H2 coordinates to local temp variables movlps 24(%rsi,%rax,4),%xmm0 ## jxH2a jyH2a - - movlps 24(%rsi,%rcx,4),%xmm1 ## jxH2c jyH2c - - movhps 24(%rsi,%rbx,4),%xmm0 ## jxH2a jyH2a jxH2b jyH2b movhps 24(%rsi,%rdx,4),%xmm1 ## jxH2c jyH2c jxH2d jyH2d movss 32(%rsi,%rax,4),%xmm2 ## jzH2a - - - movss 32(%rsi,%rcx,4),%xmm3 ## jzH2c - - - movss 32(%rsi,%rbx,4),%xmm5 ## jzH2b - - - movss 32(%rsi,%rdx,4),%xmm6 ## jzH2d - - - movlhps %xmm5,%xmm2 ## jzH2a - jzH2b - movlhps %xmm6,%xmm3 ## jzH2c - jzH2d - movaps %xmm0,%xmm4 unpcklps %xmm1,%xmm0 ## jxH2a jxH2c jyH2a jyH2c unpckhps %xmm1,%xmm4 ## jxH2b jxH2d jyH2b jyH2d movaps %xmm0,%xmm1⌨️ 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?