nb_kernel104_x86_64_sse2.s
来自「最著名最快的分子模拟软件」· S 代码 · 共 2,227 行 · 第 1/5 页
S
2,227 行
movapd nb104_half(%rsp),%xmm15 mulpd %xmm15,%xmm9 ## first iteration for rinvH1H1 mulpd %xmm15,%xmm10 ## first iteration for rinvH2H1 mulpd %xmm15,%xmm11 ## first iteration for rinvMH1 ## 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 nb104_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 nb104_half(%rsp),%xmm15 mulpd %xmm15,%xmm9 ## rinvH1H1 mulpd %xmm15,%xmm10 ## rinvH2H1 mulpd %xmm15,%xmm11 ## rinvMH1 ## H1 interactions movapd %xmm9,%xmm0 movapd %xmm10,%xmm1 movapd %xmm11,%xmm2 mulpd %xmm9,%xmm9 mulpd %xmm10,%xmm10 mulpd %xmm11,%xmm11 mulpd nb104_qqHH(%rsp),%xmm0 mulpd nb104_qqHH(%rsp),%xmm1 mulpd nb104_qqMH(%rsp),%xmm2 mulpd %xmm0,%xmm9 mulpd %xmm1,%xmm10 mulpd %xmm2,%xmm11 addpd nb104_vctot(%rsp),%xmm0 addpd %xmm2,%xmm1 addpd %xmm1,%xmm0 movapd %xmm0,nb104_vctot(%rsp) ## move j H1 forces to xmm0-xmm2 movq nb104_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 nb104_dxH1H1(%rsp),%xmm7 mulpd nb104_dyH1H1(%rsp),%xmm8 mulpd nb104_dzH1H1(%rsp),%xmm9 mulpd nb104_dxH2H1(%rsp),%xmm10 mulpd nb104_dyH2H1(%rsp),%xmm11 mulpd nb104_dzH2H1(%rsp),%xmm12 mulpd nb104_dxMH1(%rsp),%xmm13 mulpd nb104_dyMH1(%rsp),%xmm14 mulpd nb104_dzMH1(%rsp),%xmm15 addpd %xmm7,%xmm0 addpd %xmm8,%xmm1 addpd %xmm9,%xmm2 addpd nb104_fixH1(%rsp),%xmm7 addpd nb104_fiyH1(%rsp),%xmm8 addpd nb104_fizH1(%rsp),%xmm9 addpd %xmm10,%xmm0 addpd %xmm11,%xmm1 addpd %xmm12,%xmm2 addpd nb104_fixH2(%rsp),%xmm10 addpd nb104_fiyH2(%rsp),%xmm11 addpd nb104_fizH2(%rsp),%xmm12 addpd %xmm13,%xmm0 addpd %xmm14,%xmm1 addpd %xmm15,%xmm2 addpd nb104_fixM(%rsp),%xmm13 addpd nb104_fiyM(%rsp),%xmm14 addpd nb104_fizM(%rsp),%xmm15 movapd %xmm7,nb104_fixH1(%rsp) movapd %xmm8,nb104_fiyH1(%rsp) movapd %xmm9,nb104_fizH1(%rsp) movapd %xmm10,nb104_fixH2(%rsp) movapd %xmm11,nb104_fiyH2(%rsp) movapd %xmm12,nb104_fizH2(%rsp) movapd %xmm13,nb104_fixM(%rsp) movapd %xmm14,nb104_fiyM(%rsp) movapd %xmm15,nb104_fizM(%rsp) ## store back j H1 forces from xmm0-xmm2 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 nb104_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 movapd %xmm0,%xmm6 movapd %xmm1,%xmm7 movapd %xmm2,%xmm8 subpd nb104_ixH1(%rsp),%xmm0 subpd nb104_iyH1(%rsp),%xmm1 subpd nb104_izH1(%rsp),%xmm2 subpd nb104_ixH2(%rsp),%xmm3 subpd nb104_iyH2(%rsp),%xmm4 subpd nb104_izH2(%rsp),%xmm5 subpd nb104_ixM(%rsp),%xmm6 subpd nb104_iyM(%rsp),%xmm7 subpd nb104_izM(%rsp),%xmm8 movapd %xmm0,nb104_dxH1H2(%rsp) movapd %xmm1,nb104_dyH1H2(%rsp) movapd %xmm2,nb104_dzH1H2(%rsp) mulpd %xmm0,%xmm0 mulpd %xmm1,%xmm1 mulpd %xmm2,%xmm2 movapd %xmm3,nb104_dxH2H2(%rsp) movapd %xmm4,nb104_dyH2H2(%rsp) movapd %xmm5,nb104_dzH2H2(%rsp) mulpd %xmm3,%xmm3 mulpd %xmm4,%xmm4 mulpd %xmm5,%xmm5 movapd %xmm6,nb104_dxMH2(%rsp) movapd %xmm7,nb104_dyMH2(%rsp) movapd %xmm8,nb104_dzMH2(%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 jH2 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 nb104_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 nb104_half(%rsp),%xmm15 mulpd %xmm15,%xmm9 ## first iteration for rinvH1H2 mulpd %xmm15,%xmm10 ## first iteration for rinvH2H2 mulpd %xmm15,%xmm11 ## first iteration for rinvMH1H2 ## 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 nb104_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 nb104_half(%rsp),%xmm15 mulpd %xmm15,%xmm9 ## rinvH1H2 mulpd %xmm15,%xmm10 ## rinvH2H2 mulpd %xmm15,%xmm11 ## rinvMH2 ## H2 interactions movapd %xmm9,%xmm0 movapd %xmm10,%xmm1 movapd %xmm11,%xmm2 mulpd %xmm9,%xmm9 mulpd %xmm10,%xmm10 mulpd %xmm11,%xmm11 mulpd nb104_qqHH(%rsp),%xmm0 mulpd nb104_qqHH(%rsp),%xmm1 mulpd nb104_qqMH(%rsp),%xmm2 mulpd %xmm0,%xmm9 mulpd %xmm1,%xmm10 mulpd %xmm2,%xmm11 addpd nb104_vctot(%rsp),%xmm0 addpd %xmm2,%xmm1 addpd %xmm1,%xmm0 movapd %xmm0,nb104_vctot(%rsp) ## move j H2 forces to xmm0-xmm2 movq nb104_faction(%rbp),%rdi movlpd 48(%rdi,%rax,8),%xmm0 movlpd 56(%rdi,%rax,8),%xmm1 movlpd 64(%rdi,%rax,8),%xmm2 movhpd 48(%rdi,%rbx,8),%xmm0 movhpd 56(%rdi,%rbx,8),%xmm1 movhpd 64(%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 nb104_dxH1H2(%rsp),%xmm7 mulpd nb104_dyH1H2(%rsp),%xmm8 mulpd nb104_dzH1H2(%rsp),%xmm9 mulpd nb104_dxH2H2(%rsp),%xmm10 mulpd nb104_dyH2H2(%rsp),%xmm11 mulpd nb104_dzH2H2(%rsp),%xmm12 mulpd nb104_dxMH2(%rsp),%xmm13 mulpd nb104_dyMH2(%rsp),%xmm14 mulpd nb104_dzMH2(%rsp),%xmm15 addpd %xmm7,%xmm0 addpd %xmm8,%xmm1 addpd %xmm9,%xmm2 addpd nb104_fixH1(%rsp),%xmm7 addpd nb104_fiyH1(%rsp),%xmm8 addpd nb104_fizH1(%rsp),%xmm9 addpd %xmm10,%xmm0 addpd %xmm11,%xmm1 addpd %xmm12,%xmm2 addpd nb104_fixH2(%rsp),%xmm10 addpd nb104_fiyH2(%rsp),%xmm11 addpd nb104_fizH2(%rsp),%xmm12 addpd %xmm13,%xmm0 addpd %xmm14,%xmm1 addpd %xmm15,%xmm2 addpd nb104_fixM(%rsp),%xmm13 addpd nb104_fiyM(%rsp),%xmm14 addpd nb104_fizM(%rsp),%xmm15 movapd %xmm7,nb104_fixH1(%rsp) movapd %xmm8,nb104_fiyH1(%rsp) movapd %xmm9,nb104_fizH1(%rsp) movapd %xmm10,nb104_fixH2(%rsp) movapd %xmm11,nb104_fiyH2(%rsp) movapd %xmm12,nb104_fizH2(%rsp) movapd %xmm13,nb104_fixM(%rsp) movapd %xmm14,nb104_fiyM(%rsp) movapd %xmm15,nb104_fizM(%rsp) ## store back j H2 forces from xmm0-xmm2 movlpd %xmm0,48(%rdi,%rax,8) movlpd %xmm1,56(%rdi,%rax,8) movlpd %xmm2,64(%rdi,%rax,8) movhpd %xmm0,48(%rdi,%rbx,8) movhpd %xmm1,56(%rdi,%rbx,8) movhpd %xmm2,64(%rdi,%rbx,8) ## move j M coordinates to local temp variables movq nb104_pos(%rbp),%rsi movlpd 72(%rsi,%rax,8),%xmm0 movlpd 80(%rsi,%rax,8),%xmm1 movlpd 88(%rsi,%rax,8),%xmm2 movhpd 72(%rsi,%rbx,8),%xmm0 movhpd 80(%rsi,%rbx,8),%xmm1 movhpd 88(%rsi,%rbx,8),%xmm2 ## xmm0 = Mx ## xmm1 = My ## xmm2 = Mz movapd %xmm0,%xmm3 movapd %xmm1,%xmm4 movapd %xmm2,%xmm5 movapd %xmm0,%xmm6 movapd %xmm1,%xmm7 movapd %xmm2,%xmm8 subpd nb104_ixH1(%rsp),%xmm0 subpd nb104_iyH1(%rsp),%xmm1 subpd nb104_izH1(%rsp),%xmm2 subpd nb104_ixH2(%rsp),%xmm3 subpd nb104_iyH2(%rsp),%xmm4 subpd nb104_izH2(%rsp),%xmm5 subpd nb104_ixM(%rsp),%xmm6 subpd nb104_iyM(%rsp),%xmm7 subpd nb104_izM(%rsp),%xmm8 movapd %xmm0,nb104_dxH1M(%rsp) movapd %xmm1,nb104_dyH1M(%rsp) movapd %xmm2,nb104_dzH1M(%rsp) mulpd %xmm0,%xmm0 mulpd %xmm1,%xmm1 mulpd %xmm2,%xmm2 movapd %xmm3,nb104_dxH2M(%rsp) movapd %xmm4,nb104_dyH2M(%rsp) movapd %xmm5,nb104_dzH2M(%rsp) mulpd %xmm3,%xmm3 mulpd %xmm4,%xmm4 mulpd %xmm5,%xmm5 movapd %xmm6,nb104_dxMM(%rsp) movapd %xmm7,nb104_dyMM(%rsp) movapd %xmm8,nb104_dzMM(%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 jM 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 nb104_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 nb104_half(%rsp),%xmm15 mulpd %xmm15,%xmm9 ## first iteration for rinvH1M mulpd %xmm15,%xmm10 ## first iteration for rinvH2M mulpd %xmm15,%xmm11 ## first iteration for rinvMM ## 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 nb104_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⌨️ 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?