nb_kernel132_x86_64_sse2.s
来自「最著名最快的分子模拟软件」· S 代码 · 共 2,287 行 · 第 1/5 页
S
2,287 行
movapd nb132_three(%rsp),%xmm9 movapd %xmm9,%xmm10 movapd %xmm9,%xmm11 mulpd %xmm0,%xmm1 ## rsq*lu*lu mulpd %xmm3,%xmm4 ## rsq*lu*lu mulpd %xmm6,%xmm7 ## rsq*lu*lu subpd %xmm1,%xmm9 subpd %xmm4,%xmm10 subpd %xmm7,%xmm11 ## 3-rsq*lu*lu mulpd %xmm2,%xmm9 mulpd %xmm5,%xmm10 mulpd %xmm8,%xmm11 ## lu*(3-rsq*lu*lu) movapd nb132_half(%rsp),%xmm15 mulpd %xmm15,%xmm9 ## first iteration for rinvOO mulpd %xmm15,%xmm10 ## first iteration for rinvH1O mulpd %xmm15,%xmm11 ## first iteration for rinvH2O ## second iteration step movapd %xmm9,%xmm2 movapd %xmm10,%xmm5 movapd %xmm11,%xmm8 mulpd %xmm2,%xmm2 ## lu*lu mulpd %xmm5,%xmm5 ## lu*lu mulpd %xmm8,%xmm8 ## lu*lu movapd nb132_three(%rsp),%xmm1 movapd %xmm1,%xmm4 movapd %xmm1,%xmm7 mulpd %xmm0,%xmm2 ## rsq*lu*lu mulpd %xmm3,%xmm5 ## rsq*lu*lu mulpd %xmm6,%xmm8 ## rsq*lu*lu subpd %xmm2,%xmm1 subpd %xmm5,%xmm4 subpd %xmm8,%xmm7 ## 3-rsq*lu*lu mulpd %xmm1,%xmm9 mulpd %xmm4,%xmm10 mulpd %xmm7,%xmm11 ## lu*(3-rsq*lu*lu) movapd nb132_half(%rsp),%xmm15 mulpd %xmm15,%xmm9 ## rinvOO mulpd %xmm15,%xmm10 ## rinvH1O mulpd %xmm15,%xmm11 ## rinvH2O ## O interactions ## rsq in xmm0,xmm3,xmm6 ## rinv in xmm9, xmm10, xmm11 movapd %xmm0,nb132_rsqOO(%rsp) movapd %xmm3,nb132_rsqOH1(%rsp) movapd %xmm6,nb132_rsqOH2(%rsp) movapd %xmm9,nb132_rinvOO(%rsp) movapd %xmm10,nb132_rinvOH1(%rsp) movapd %xmm11,nb132_rinvOH2(%rsp) ## table LJ interaction mulpd %xmm9,%xmm0 mulpd nb132_tsc(%rsp),%xmm0 ## rtab ## truncate and convert to integers cvttpd2dq %xmm0,%xmm1 ## convert back to float cvtdq2pd %xmm1,%xmm2 ## multiply by 8 pslld $3,%xmm1 ## move to integer registers pshufd $1,%xmm1,%xmm13 movd %xmm1,%r8d movd %xmm13,%r10d ## calculate eps subpd %xmm2,%xmm0 movq nb132_VFtab(%rbp),%rsi movlpd (%rsi,%r8,8),%xmm4 movlpd 8(%rsi,%r8,8),%xmm5 movlpd 16(%rsi,%r8,8),%xmm6 movlpd 24(%rsi,%r8,8),%xmm7 movlpd 32(%rsi,%r8,8),%xmm8 movlpd 40(%rsi,%r8,8),%xmm9 movlpd 48(%rsi,%r8,8),%xmm10 movlpd 56(%rsi,%r8,8),%xmm11 movhpd (%rsi,%r10,8),%xmm4 movhpd 8(%rsi,%r10,8),%xmm5 movhpd 16(%rsi,%r10,8),%xmm6 movhpd 24(%rsi,%r10,8),%xmm7 movhpd 32(%rsi,%r10,8),%xmm8 movhpd 40(%rsi,%r10,8),%xmm9 movhpd 48(%rsi,%r10,8),%xmm10 movhpd 56(%rsi,%r10,8),%xmm11 ## dispersion table in xmm4-xmm7, repulsion table in xmm8-xmm11 mulpd %xmm0,%xmm7 ## Heps mulpd %xmm0,%xmm11 mulpd %xmm0,%xmm6 ## Geps mulpd %xmm0,%xmm10 mulpd %xmm0,%xmm7 ## Heps2 mulpd %xmm0,%xmm11 addpd %xmm6,%xmm5 ## F+Geps addpd %xmm10,%xmm9 addpd %xmm7,%xmm5 ## F+Geps+Heps2 = Fp addpd %xmm11,%xmm9 addpd %xmm7,%xmm7 ## 2*Heps2 addpd %xmm11,%xmm11 addpd %xmm6,%xmm7 ## 2*Heps2+Geps addpd %xmm10,%xmm11 addpd %xmm5,%xmm7 ## FF = Fp + 2*Heps2 + Geps addpd %xmm9,%xmm11 mulpd %xmm0,%xmm5 ## eps*Fp mulpd %xmm0,%xmm9 movapd nb132_c6(%rsp),%xmm12 movapd nb132_c12(%rsp),%xmm13 addpd %xmm4,%xmm5 ## VV addpd %xmm8,%xmm9 mulpd %xmm12,%xmm5 ## VV*c6 = vnb6 mulpd %xmm13,%xmm9 ## VV*c12 = vnb12 addpd %xmm9,%xmm5 addpd nb132_Vvdwtot(%rsp),%xmm5 movapd %xmm5,nb132_Vvdwtot(%rsp) mulpd %xmm12,%xmm7 ## FF*c6 = fnb6 mulpd %xmm13,%xmm11 ## FF*c12 = fnb12 addpd %xmm11,%xmm7 mulpd nb132_tsc(%rsp),%xmm7 movapd nb132_rinvOO(%rsp),%xmm9 movapd nb132_rinvOH1(%rsp),%xmm10 movapd nb132_rinvOH2(%rsp),%xmm11 movapd %xmm9,%xmm0 movapd %xmm10,%xmm1 movapd %xmm11,%xmm2 mulpd %xmm10,%xmm10 mulpd %xmm11,%xmm11 mulpd nb132_qqOO(%rsp),%xmm0 mulpd nb132_qqOH(%rsp),%xmm1 mulpd nb132_qqOH(%rsp),%xmm2 mulpd %xmm0,%xmm9 mulpd %xmm1,%xmm10 mulpd %xmm2,%xmm11 subpd %xmm7,%xmm9 mulpd nb132_rinvOO(%rsp),%xmm9 addpd nb132_vctot(%rsp),%xmm0 addpd %xmm2,%xmm1 addpd %xmm1,%xmm0 movapd %xmm0,nb132_vctot(%rsp) ## move j O forces to xmm0-xmm2 movq nb132_faction(%rbp),%rdi movlpd (%rdi,%rax,8),%xmm0 movlpd 8(%rdi,%rax,8),%xmm1 movlpd 16(%rdi,%rax,8),%xmm2 movhpd (%rdi,%rbx,8),%xmm0 movhpd 8(%rdi,%rbx,8),%xmm1 movhpd 16(%rdi,%rbx,8),%xmm2 movapd %xmm9,%xmm7 movapd %xmm9,%xmm8 movapd %xmm11,%xmm13 movapd %xmm11,%xmm14 movapd %xmm11,%xmm15 movapd %xmm10,%xmm11 movapd %xmm10,%xmm12 mulpd nb132_dxOO(%rsp),%xmm7 mulpd nb132_dyOO(%rsp),%xmm8 mulpd nb132_dzOO(%rsp),%xmm9 mulpd nb132_dxH1O(%rsp),%xmm10 mulpd nb132_dyH1O(%rsp),%xmm11 mulpd nb132_dzH1O(%rsp),%xmm12 mulpd nb132_dxH2O(%rsp),%xmm13 mulpd nb132_dyH2O(%rsp),%xmm14 mulpd nb132_dzH2O(%rsp),%xmm15 addpd %xmm7,%xmm0 addpd %xmm8,%xmm1 addpd %xmm9,%xmm2 addpd nb132_fixO(%rsp),%xmm7 addpd nb132_fiyO(%rsp),%xmm8 addpd nb132_fizO(%rsp),%xmm9 addpd %xmm10,%xmm0 addpd %xmm11,%xmm1 addpd %xmm12,%xmm2 addpd nb132_fixH1(%rsp),%xmm10 addpd nb132_fiyH1(%rsp),%xmm11 addpd nb132_fizH1(%rsp),%xmm12 addpd %xmm13,%xmm0 addpd %xmm14,%xmm1 addpd %xmm15,%xmm2 addpd nb132_fixH2(%rsp),%xmm13 addpd nb132_fiyH2(%rsp),%xmm14 addpd nb132_fizH2(%rsp),%xmm15 movapd %xmm7,nb132_fixO(%rsp) movapd %xmm8,nb132_fiyO(%rsp) movapd %xmm9,nb132_fizO(%rsp) movapd %xmm10,nb132_fixH1(%rsp) movapd %xmm11,nb132_fiyH1(%rsp) movapd %xmm12,nb132_fizH1(%rsp) movapd %xmm13,nb132_fixH2(%rsp) movapd %xmm14,nb132_fiyH2(%rsp) movapd %xmm15,nb132_fizH2(%rsp) ## store back j O forces from xmm0-xmm2 movq nb132_faction(%rbp),%rdi movlpd %xmm0,(%rdi,%rax,8) movlpd %xmm1,8(%rdi,%rax,8) movlpd %xmm2,16(%rdi,%rax,8) movhpd %xmm0,(%rdi,%rbx,8) movhpd %xmm1,8(%rdi,%rbx,8) movhpd %xmm2,16(%rdi,%rbx,8) ## move j H1 coordinates to local temp variables movq nb132_pos(%rbp),%rsi movlpd 24(%rsi,%rax,8),%xmm0 movlpd 32(%rsi,%rax,8),%xmm1 movlpd 40(%rsi,%rax,8),%xmm2 movhpd 24(%rsi,%rbx,8),%xmm0 movhpd 32(%rsi,%rbx,8),%xmm1 movhpd 40(%rsi,%rbx,8),%xmm2 ## xmm0 = H1x ## xmm1 = H1y ## xmm2 = H1z movapd %xmm0,%xmm3 movapd %xmm1,%xmm4 movapd %xmm2,%xmm5 movapd %xmm0,%xmm6 movapd %xmm1,%xmm7 movapd %xmm2,%xmm8 subpd nb132_ixO(%rsp),%xmm0 subpd nb132_iyO(%rsp),%xmm1 subpd nb132_izO(%rsp),%xmm2 subpd nb132_ixH1(%rsp),%xmm3 subpd nb132_iyH1(%rsp),%xmm4 subpd nb132_izH1(%rsp),%xmm5 subpd nb132_ixH2(%rsp),%xmm6 subpd nb132_iyH2(%rsp),%xmm7 subpd nb132_izH2(%rsp),%xmm8 movapd %xmm0,nb132_dxOH1(%rsp) movapd %xmm1,nb132_dyOH1(%rsp) movapd %xmm2,nb132_dzOH1(%rsp) mulpd %xmm0,%xmm0 mulpd %xmm1,%xmm1 mulpd %xmm2,%xmm2 movapd %xmm3,nb132_dxH1H1(%rsp) movapd %xmm4,nb132_dyH1H1(%rsp) movapd %xmm5,nb132_dzH1H1(%rsp) mulpd %xmm3,%xmm3 mulpd %xmm4,%xmm4 mulpd %xmm5,%xmm5 movapd %xmm6,nb132_dxH2H1(%rsp) movapd %xmm7,nb132_dyH2H1(%rsp) movapd %xmm8,nb132_dzH2H1(%rsp) mulpd %xmm6,%xmm6 mulpd %xmm7,%xmm7 mulpd %xmm8,%xmm8 addpd %xmm1,%xmm0 addpd %xmm2,%xmm0 addpd %xmm4,%xmm3 addpd %xmm5,%xmm3 addpd %xmm7,%xmm6 addpd %xmm8,%xmm6 ## start doing invsqrt for jH1 atoms cvtpd2ps %xmm0,%xmm1 cvtpd2ps %xmm3,%xmm4 cvtpd2ps %xmm6,%xmm7 rsqrtps %xmm1,%xmm1 rsqrtps %xmm4,%xmm4 rsqrtps %xmm7,%xmm7 cvtps2pd %xmm1,%xmm1 cvtps2pd %xmm4,%xmm4 cvtps2pd %xmm7,%xmm7 movapd %xmm1,%xmm2 movapd %xmm4,%xmm5 movapd %xmm7,%xmm8 mulpd %xmm1,%xmm1 ## lu*lu mulpd %xmm4,%xmm4 ## lu*lu mulpd %xmm7,%xmm7 ## lu*lu movapd nb132_three(%rsp),%xmm9 movapd %xmm9,%xmm10 movapd %xmm9,%xmm11 mulpd %xmm0,%xmm1 ## rsq*lu*lu mulpd %xmm3,%xmm4 ## rsq*lu*lu mulpd %xmm6,%xmm7 ## rsq*lu*lu subpd %xmm1,%xmm9 subpd %xmm4,%xmm10 subpd %xmm7,%xmm11 ## 3-rsq*lu*lu mulpd %xmm2,%xmm9 mulpd %xmm5,%xmm10 mulpd %xmm8,%xmm11 ## lu*(3-rsq*lu*lu) movapd nb132_half(%rsp),%xmm15 mulpd %xmm15,%xmm9 ## first iteration for rinvOH1 mulpd %xmm15,%xmm10 ## first iteration for rinvH1H1 mulpd %xmm15,%xmm11 ## first iteration for rinvH2OH1 ## second iteration step movapd %xmm9,%xmm2 movapd %xmm10,%xmm5 movapd %xmm11,%xmm8 mulpd %xmm2,%xmm2 ## lu*lu mulpd %xmm5,%xmm5 ## lu*lu mulpd %xmm8,%xmm8 ## lu*lu movapd nb132_three(%rsp),%xmm1 movapd %xmm1,%xmm4 movapd %xmm1,%xmm7 mulpd %xmm0,%xmm2 ## rsq*lu*lu mulpd %xmm3,%xmm5 ## rsq*lu*lu mulpd %xmm6,%xmm8 ## rsq*lu*lu subpd %xmm2,%xmm1 subpd %xmm5,%xmm4 subpd %xmm8,%xmm7 ## 3-rsq*lu*lu mulpd %xmm1,%xmm9 mulpd %xmm4,%xmm10 mulpd %xmm7,%xmm11 ## lu*(3-rsq*lu*lu) movapd nb132_half(%rsp),%xmm15 mulpd %xmm15,%xmm9 ## rinvOH1 mulpd %xmm15,%xmm10 ## rinvH1H1 mulpd %xmm15,%xmm11 ## rinvH2H1 ## H1 interactions movapd %xmm9,%xmm0 movapd %xmm10,%xmm1 movapd %xmm11,%xmm2 mulpd %xmm9,%xmm9 mulpd %xmm10,%xmm10 mulpd %xmm11,%xmm11 mulpd nb132_qqOH(%rsp),%xmm0 mulpd nb132_qqHH(%rsp),%xmm1 mulpd nb132_qqHH(%rsp),%xmm2 mulpd %xmm0,%xmm9 mulpd %xmm1,%xmm10 mulpd %xmm2,%xmm11 addpd nb132_vctot(%rsp),%xmm0 addpd %xmm2,%xmm1 addpd %xmm1,%xmm0 movapd %xmm0,nb132_vctot(%rsp) ## move j H1 forces to xmm0-xmm2 movq nb132_faction(%rbp),%rdi movlpd 24(%rdi,%rax,8),%xmm0 movlpd 32(%rdi,%rax,8),%xmm1 movlpd 40(%rdi,%rax,8),%xmm2 movhpd 24(%rdi,%rbx,8),%xmm0 movhpd 32(%rdi,%rbx,8),%xmm1 movhpd 40(%rdi,%rbx,8),%xmm2 movapd %xmm9,%xmm7 movapd %xmm9,%xmm8 movapd %xmm11,%xmm13 movapd %xmm11,%xmm14 movapd %xmm11,%xmm15 movapd %xmm10,%xmm11 movapd %xmm10,%xmm12 mulpd nb132_dxOH1(%rsp),%xmm7 mulpd nb132_dyOH1(%rsp),%xmm8 mulpd nb132_dzOH1(%rsp),%xmm9 mulpd nb132_dxH1H1(%rsp),%xmm10 mulpd nb132_dyH1H1(%rsp),%xmm11 mulpd nb132_dzH1H1(%rsp),%xmm12 mulpd nb132_dxH2H1(%rsp),%xmm13 mulpd nb132_dyH2H1(%rsp),%xmm14 mulpd nb132_dzH2H1(%rsp),%xmm15 addpd %xmm7,%xmm0 addpd %xmm8,%xmm1 addpd %xmm9,%xmm2 addpd nb132_fixO(%rsp),%xmm7 addpd nb132_fiyO(%rsp),%xmm8 addpd nb132_fizO(%rsp),%xmm9 addpd %xmm10,%xmm0 addpd %xmm11,%xmm1 addpd %xmm12,%xmm2 addpd nb132_fixH1(%rsp),%xmm10 addpd nb132_fiyH1(%rsp),%xmm11 addpd nb132_fizH1(%rsp),%xmm12 addpd %xmm13,%xmm0 addpd %xmm14,%xmm1 addpd %xmm15,%xmm2 addpd nb132_fixH2(%rsp),%xmm13 addpd nb132_fiyH2(%rsp),%xmm14 addpd nb132_fizH2(%rsp),%xmm15 movapd %xmm7,nb132_fixO(%rsp) movapd %xmm8,nb132_fiyO(%rsp) movapd %xmm9,nb132_fizO(%rsp) movapd %xmm10,nb132_fixH1(%rsp) movapd %xmm11,nb132_fiyH1(%rsp) movapd %xmm12,nb132_fizH1(%rsp) movapd %xmm13,nb132_fixH2(%rsp) movapd %xmm14,nb132_fiyH2(%rsp) movapd %xmm15,nb132_fizH2(%rsp) ## store back j H1 forces from xmm0-xmm2 movq nb132_faction(%rbp),%rdi movlpd %xmm0,24(%rdi,%rax,8) movlpd %xmm1,32(%rdi,%rax,8) movlpd %xmm2,40(%rdi,%rax,8) movhpd %xmm0,24(%rdi,%rbx,8) movhpd %xmm1,32(%rdi,%rbx,8) movhpd %xmm2,40(%rdi,%rbx,8) ## move j H2 coordinates to local temp variables movq nb132_pos(%rbp),%rsi movlpd 48(%rsi,%rax,8),%xmm0 movlpd 56(%rsi,%rax,8),%xmm1 movlpd 64(%rsi,%rax,8),%xmm2 movhpd 48(%rsi,%rbx,8),%xmm0 movhpd 56(%rsi,%rbx,8),%xmm1 movhpd 64(%rsi,%rbx,8),%xmm2 ## xmm0 = H2x ## xmm1 = H2y ## xmm2 = H2z movapd %xmm0,%xmm3 movapd %xmm1,%xmm4 movapd %xmm2,%xmm5⌨️ 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?