nb_kernel314_x86_64_sse.s
来自「最著名最快的分子模拟软件」· S 代码 · 共 2,211 行 · 第 1/5 页
S
2,211 行
mulps %xmm2,%xmm2 addps %xmm1,%xmm0 addps %xmm2,%xmm0 ## have rsq in xmm0 mulps %xmm3,%xmm3 mulps %xmm4,%xmm4 mulps %xmm5,%xmm5 addps %xmm3,%xmm4 addps %xmm5,%xmm4 ## have rsq in xmm4 ## Save H1 data in H1H1 movaps %xmm4,nb314_rsqH1H1(%rsp) ## do 1/x for O and 1/sqrt(x) for H1 rcpss %xmm0,%xmm1 rsqrtps %xmm4,%xmm5 movss nb314_two(%rsp),%xmm2 movaps %xmm5,%xmm6 mulss %xmm1,%xmm0 mulps %xmm5,%xmm5 subss %xmm0,%xmm2 movaps nb314_three(%rsp),%xmm7 mulss %xmm1,%xmm2 ## 1/r2 mulps %xmm4,%xmm5 movss %xmm2,%xmm0 subps %xmm5,%xmm7 mulss %xmm2,%xmm2 mulps %xmm6,%xmm7 mulss %xmm0,%xmm2 ## 1/r6 mulps nb314_half(%rsp),%xmm7 ## rinv iH1 - j water movss %xmm2,%xmm1 movaps %xmm7,nb314_rinvH1H1(%rsp) mulss %xmm2,%xmm2 ## 1/r12 mulss nb314_c6(%rsp),%xmm1 mulss nb314_c12(%rsp),%xmm2 movss %xmm2,%xmm3 subss %xmm1,%xmm3 addss nb314_Vvdwtot(%rsp),%xmm3 movss %xmm3,nb314_Vvdwtot(%rsp) mulss nb314_six(%rsp),%xmm1 mulss nb314_twelve(%rsp),%xmm2 subss %xmm1,%xmm2 mulss %xmm2,%xmm0 ## fscal movss %xmm0,%xmm1 movss %xmm0,%xmm2 mulss nb314_dxOO(%rsp),%xmm0 mulss nb314_dyOO(%rsp),%xmm1 mulss nb314_dzOO(%rsp),%xmm2 xorps %xmm3,%xmm3 xorps %xmm4,%xmm4 xorps %xmm5,%xmm5 addss %xmm0,%xmm3 addss %xmm1,%xmm4 addss %xmm2,%xmm5 movaps %xmm3,nb314_fjxO(%rsp) movaps %xmm4,nb314_fjyO(%rsp) movaps %xmm5,nb314_fjzO(%rsp) addss nb314_fixO(%rsp),%xmm0 addss nb314_fiyO(%rsp),%xmm1 addss nb314_fizO(%rsp),%xmm2 movss %xmm0,nb314_fixO(%rsp) movss %xmm1,nb314_fiyO(%rsp) movss %xmm2,nb314_fizO(%rsp) ## do H1 coulomb interaction movaps nb314_rinvH1H1(%rsp),%xmm0 ## rinv movaps %xmm0,%xmm1 mulps nb314_rsqH1H1(%rsp),%xmm1 ## r mulps nb314_tsc(%rsp),%xmm1 movhlps %xmm1,%xmm2 cvttps2pi %xmm1,%mm6 cvttps2pi %xmm2,%mm7 ## mm6/mm7 contain lu indices cvtpi2ps %mm6,%xmm3 cvtpi2ps %mm7,%xmm2 movlhps %xmm2,%xmm3 subps %xmm3,%xmm1 ## xmm1=eps movaps %xmm1,%xmm2 mulps %xmm2,%xmm2 ## xmm2=eps2 pslld $2,%mm6 pslld $2,%mm7 psrlq $32,%mm6 movd %mm6,%ebx movd %mm7,%ecx psrlq $32,%mm7 movd %mm7,%edx ## table indices in ebx,ecx,edx movq nb314_VFtab(%rbp),%rsi movlps (%rsi,%rbx,4),%xmm4 movlps (%rsi,%rcx,4),%xmm3 movlps (%rsi,%rdx,4),%xmm7 movhps 8(%rsi,%rbx,4),%xmm4 movhps 8(%rsi,%rcx,4),%xmm3 movhps 8(%rsi,%rdx,4),%xmm7 movaps %xmm3,%xmm6 unpcklps %xmm7,%xmm6 unpckhps %xmm7,%xmm3 movaps %xmm4,%xmm5 movaps %xmm4,%xmm7 shufps $0x40,%xmm6,%xmm4 shufps $0xE4,%xmm6,%xmm5 movaps %xmm7,%xmm6 shufps $0x48,%xmm3,%xmm6 shufps $0xEC,%xmm3,%xmm7 ## coulomb table ready, in xmm4-xmm7 mulps %xmm1,%xmm6 ## xmm6=Geps mulps %xmm2,%xmm7 ## xmm7=Heps2 addps %xmm6,%xmm5 addps %xmm7,%xmm5 ## xmm5=Fp addps %xmm7,%xmm7 ## two*Heps2 ## fetch charges to xmm3 (temporary) xorps %xmm3,%xmm3 movss nb314_qqHH(%rsp),%xmm3 movhps nb314_qqMH(%rsp),%xmm3 shufps $193,%xmm3,%xmm3 ## 11000001 addps %xmm6,%xmm7 addps %xmm5,%xmm7 ## xmm7=FF mulps %xmm1,%xmm5 ## xmm5=eps*Fp addps %xmm4,%xmm5 ## xmm5=VV mulps %xmm3,%xmm5 ## vcoul=qq*VV mulps %xmm7,%xmm3 ## fijC=FF*qq ## at this point xmm5 contains vcoul and xmm3 fijC addps nb314_vctot(%rsp),%xmm5 movaps %xmm5,nb314_vctot(%rsp) mulps nb314_tsc(%rsp),%xmm3 xorps %xmm2,%xmm2 subps %xmm3,%xmm2 mulps %xmm2,%xmm0 movaps %xmm0,%xmm1 movaps %xmm0,%xmm2 mulps nb314_dxH1H1(%rsp),%xmm0 mulps nb314_dyH1H1(%rsp),%xmm1 mulps nb314_dzH1H1(%rsp),%xmm2 ## update forces H1 - j water movaps nb314_fjxO(%rsp),%xmm3 movaps nb314_fjyO(%rsp),%xmm4 movaps nb314_fjzO(%rsp),%xmm5 addps %xmm0,%xmm3 addps %xmm1,%xmm4 addps %xmm2,%xmm5 movaps %xmm3,nb314_fjxO(%rsp) movaps %xmm4,nb314_fjyO(%rsp) movaps %xmm5,nb314_fjzO(%rsp) addps nb314_fixH1(%rsp),%xmm0 addps nb314_fiyH1(%rsp),%xmm1 addps nb314_fizH1(%rsp),%xmm2 movaps %xmm0,nb314_fixH1(%rsp) movaps %xmm1,nb314_fiyH1(%rsp) movaps %xmm2,nb314_fizH1(%rsp) ## i H2 & M simultaneously first get i particle coords: movaps nb314_jxO(%rsp),%xmm0 movaps nb314_jyO(%rsp),%xmm1 movaps nb314_jzO(%rsp),%xmm2 movaps %xmm0,%xmm3 movaps %xmm1,%xmm4 movaps %xmm2,%xmm5 subps nb314_ixH2(%rsp),%xmm0 subps nb314_iyH2(%rsp),%xmm1 subps nb314_izH2(%rsp),%xmm2 subps nb314_ixM(%rsp),%xmm3 subps nb314_iyM(%rsp),%xmm4 subps nb314_izM(%rsp),%xmm5 movaps %xmm0,nb314_dxH2H2(%rsp) movaps %xmm1,nb314_dyH2H2(%rsp) movaps %xmm2,nb314_dzH2H2(%rsp) movaps %xmm3,nb314_dxMM(%rsp) movaps %xmm4,nb314_dyMM(%rsp) movaps %xmm5,nb314_dzMM(%rsp) mulps %xmm0,%xmm0 mulps %xmm1,%xmm1 mulps %xmm2,%xmm2 mulps %xmm3,%xmm3 mulps %xmm4,%xmm4 mulps %xmm5,%xmm5 addps %xmm1,%xmm0 addps %xmm3,%xmm4 addps %xmm2,%xmm0 ## have rsqH2 in xmm0 addps %xmm5,%xmm4 ## have rsqM in xmm4 ## start with H2, save data movaps %xmm0,nb314_rsqH2H2(%rsp) movaps %xmm4,nb314_rsqMM(%rsp) ## do invsqrt rsqrtps %xmm0,%xmm1 rsqrtps %xmm4,%xmm5 movaps %xmm1,%xmm2 movaps %xmm5,%xmm6 mulps %xmm1,%xmm1 mulps %xmm5,%xmm5 movaps nb314_three(%rsp),%xmm3 movaps %xmm3,%xmm7 mulps %xmm0,%xmm1 mulps %xmm4,%xmm5 subps %xmm1,%xmm3 subps %xmm5,%xmm7 mulps %xmm2,%xmm3 mulps %xmm6,%xmm7 mulps nb314_half(%rsp),%xmm3 ## rinv H2 - j water mulps nb314_half(%rsp),%xmm7 ## rinv M - j water movaps %xmm3,nb314_rinvH2H2(%rsp) movaps %xmm7,nb314_rinvMM(%rsp) movaps %xmm3,%xmm1 mulps nb314_rsqH2H2(%rsp),%xmm1 ## xmm1=r movaps %xmm3,%xmm0 ## xmm0=rinv mulps nb314_tsc(%rsp),%xmm1 movhlps %xmm1,%xmm2 cvttps2pi %xmm1,%mm6 cvttps2pi %xmm2,%mm7 ## mm6/mm7 contain lu indices cvtpi2ps %mm6,%xmm3 cvtpi2ps %mm7,%xmm2 movlhps %xmm2,%xmm3 subps %xmm3,%xmm1 ## xmm1=eps movaps %xmm1,%xmm2 mulps %xmm2,%xmm2 ## xmm2=eps2 pslld $2,%mm6 pslld $2,%mm7 psrlq $32,%mm6 movd %mm6,%ebx movd %mm7,%ecx psrlq $32,%mm7 movd %mm7,%edx ## table indices in ebx,ecx,edx movlps (%rsi,%rbx,4),%xmm4 movlps (%rsi,%rcx,4),%xmm3 movlps (%rsi,%rdx,4),%xmm7 movhps 8(%rsi,%rbx,4),%xmm4 movhps 8(%rsi,%rcx,4),%xmm3 movhps 8(%rsi,%rdx,4),%xmm7 movaps %xmm3,%xmm6 unpcklps %xmm7,%xmm6 unpckhps %xmm7,%xmm3 movaps %xmm4,%xmm5 movaps %xmm4,%xmm7 shufps $0x40,%xmm6,%xmm4 shufps $0xE4,%xmm6,%xmm5 movaps %xmm7,%xmm6 shufps $0x48,%xmm3,%xmm6 shufps $0xEC,%xmm3,%xmm7 ## coulomb table ready, in xmm4-xmm7 mulps %xmm1,%xmm6 ## xmm6=Geps mulps %xmm2,%xmm7 ## xmm7=Heps2 addps %xmm6,%xmm5 addps %xmm7,%xmm5 ## xmm5=Fp mulps nb314_two(%rsp),%xmm7 ## two*Heps2 xorps %xmm3,%xmm3 ## fetch charges to xmm3 (temporary) movss nb314_qqHH(%rsp),%xmm3 movhps nb314_qqMH(%rsp),%xmm3 shufps $193,%xmm3,%xmm3 ## 11000001 addps %xmm6,%xmm7 addps %xmm5,%xmm7 ## xmm7=FF mulps %xmm1,%xmm5 ## xmm5=eps*Fp addps %xmm4,%xmm5 ## xmm5=VV mulps %xmm3,%xmm5 ## vcoul=qq*VV mulps %xmm7,%xmm3 ## fijC=FF*qq ## at this point xmm5 contains vcoul and xmm3 fijC addps nb314_vctot(%rsp),%xmm5 movaps %xmm5,nb314_vctot(%rsp) xorps %xmm1,%xmm1 mulps nb314_tsc(%rsp),%xmm3 mulps %xmm0,%xmm3 subps %xmm3,%xmm1 movaps %xmm1,%xmm0 movaps %xmm1,%xmm2 mulps nb314_dxH2H2(%rsp),%xmm0 mulps nb314_dyH2H2(%rsp),%xmm1 mulps nb314_dzH2H2(%rsp),%xmm2 ## update forces H1 - j water movaps nb314_fjxO(%rsp),%xmm3 movaps nb314_fjyO(%rsp),%xmm4 movaps nb314_fjzO(%rsp),%xmm5 addps %xmm0,%xmm3 addps %xmm1,%xmm4 addps %xmm2,%xmm5 movaps %xmm3,nb314_fjxO(%rsp) movaps %xmm4,nb314_fjyO(%rsp) movaps %xmm5,nb314_fjzO(%rsp) addps nb314_fixH2(%rsp),%xmm0 addps nb314_fiyH2(%rsp),%xmm1 addps nb314_fizH2(%rsp),%xmm2 movaps %xmm0,nb314_fixH2(%rsp) movaps %xmm1,nb314_fiyH2(%rsp) movaps %xmm2,nb314_fizH2(%rsp) ## do table for i M - j water interaction movaps nb314_rinvMM(%rsp),%xmm0 movaps %xmm0,%xmm1 mulps nb314_rsqMM(%rsp),%xmm1 ## xmm0=rinv, xmm1=r mulps nb314_tsc(%rsp),%xmm1 movhlps %xmm1,%xmm2 cvttps2pi %xmm1,%mm6 cvttps2pi %xmm2,%mm7 ## mm6/mm7 contain lu indices cvtpi2ps %mm6,%xmm3 cvtpi2ps %mm7,%xmm2 movlhps %xmm2,%xmm3 subps %xmm3,%xmm1 ## xmm1=eps movaps %xmm1,%xmm2 mulps %xmm2,%xmm2 ## xmm2=eps2 pslld $2,%mm6 pslld $2,%mm7 psrlq $32,%mm6 movd %mm6,%ebx movd %mm7,%ecx psrlq $32,%mm7 movd %mm7,%edx ## table indices in ebx,ecx,edx movlps (%rsi,%rbx,4),%xmm4 movlps (%rsi,%rcx,4),%xmm3 movlps (%rsi,%rdx,4),%xmm7 movhps 8(%rsi,%rbx,4),%xmm4 movhps 8(%rsi,%rcx,4),%xmm3 movhps 8(%rsi,%rdx,4),%xmm7 movaps %xmm3,%xmm6 unpcklps %xmm7,%xmm6 unpckhps %xmm7,%xmm3 movaps %xmm4,%xmm5 movaps %xmm4,%xmm7 shufps $0x40,%xmm6,%xmm4 shufps $0xE4,%xmm6,%xmm5 movaps %xmm7,%xmm6 shufps $0x48,%xmm3,%xmm6 shufps $0xEC,%xmm3,%xmm7 ## coulomb table ready, in xmm4-xmm7 mulps %xmm1,%xmm6 ## xmm6=Geps mulps %xmm2,%xmm7 ## xmm7=Heps2 addps %xmm6,%xmm5 addps %xmm7,%xmm5 ## xmm5=Fp mulps nb314_two(%rsp),%xmm7 ## two*Heps2 xorps %xmm3,%xmm3 ## fetch charges to xmm3 (temporary) movss nb314_qqMH(%rsp),%xmm3 movhps nb314_qqMM(%rsp),%xmm3 shufps $193,%xmm3,%xmm3 ## 11000001 addps %xmm6,%xmm7 addps %xmm5,%xmm7 ## xmm7=FF mulps %xmm1,%xmm5 ## xmm5=eps*Fp addps %xmm4,%xmm5 ## xmm5=VV mulps %xmm3,%xmm5 ## vcoul=qq*VV mulps %xmm7,%xmm3 ## fijC=FF*qq ## at this point xmm5 contains vcoul and xmm3 fijC addps nb314_vctot(%rsp),%xmm5 movaps %xmm5,nb314_vctot(%rsp) xorps %xmm1,%xmm1 mulps nb314_tsc(%rsp),%xmm3 mulps %xmm0,%xmm3 subps %xmm3,%xmm1 movaps %xmm1,%xmm0 movaps %xmm1,%xmm2 mulps nb314_dxMM(%rsp),%xmm0 mulps nb314_dyMM(%rsp),%xmm1 mulps nb314_dzMM(%rsp),%xmm2 movaps nb314_fjxO(%rsp),%xmm3 movaps nb314_fjyO(%rsp),%xmm4 movaps nb314_fjzO(%rsp),%xmm5 addps %xmm0,%xmm3 addps %xmm1,%xmm4 addps %xmm2,%xmm5 movq nb314_faction(%rbp),%rsi movaps %xmm3,nb314_fjxO(%rsp) movaps %xmm4,nb314_fjyO(%rsp) movaps %xmm5,nb314_fjzO(%rsp) addps nb314_fixM(%rsp),%xmm0 addps nb314_fiyM(%rsp),%xmm1 addps nb314_fizM(%rsp),%xmm2 movaps %xmm0,nb314_fixM(%rsp) movaps %xmm1,nb314_fiyM(%rsp) movaps %xmm2,nb314_fizM(%rsp) ## update j water forces from local variables. ## transpose back first movaps nb314_fjxO(%rsp),%xmm0 ## Ox H1x H2x Mx movaps nb314_fjyO(%rsp),%xmm1 ## Oy H1y H2y My movaps nb314_fjzO(%rsp),%xmm2 ## Oz H1z H2z Mz movaps %xmm0,%xmm3 movaps %xmm0,%xmm4 unpcklps %xmm1,%xmm3 ## Ox Oy - - shufps $0x1,%xmm2,%xmm4 ## h1x - Oz - movaps %xmm1,%xmm5 movaps %xmm0,%xmm6 unpcklps %xmm2,%xmm5 ## - - H1y H1z unpckhps %xmm1,%xmm6 ## h2x h2y - - unpckhps %xmm2,%xmm1 ## - - My Mz shufps $0x32,%xmm0,%xmm2 ## (00110010) h2z - Mx - shufps $36,%xmm4,%xmm3 ## 00100100 ;# Ox Oy Oz H1x shufps $78,%xmm6,%xmm5 ## 01001110 ;# h1y h1z h2x h2y shufps $232,%xmm1,%xmm2 ## 11101000 ;# h2z mx my mz movlps (%rsi,%rax,4),%xmm0 movlps 16(%rsi,%rax,4),%xmm1 movlps 32(%rsi,%rax,4),%xmm4 movhps 8(%rsi,%rax,4),%xmm0 movhps 24(%rsi,%rax,4),%xmm1 movhps 40(%rsi,%rax,4),%xmm4 addps %xmm3,%xmm0 addps %xmm5,%xmm1 addps %xmm2,%xmm4 movlps %xmm0,(%rsi,%rax,4) movlps %xmm1,16(%rsi,%rax,4) movlps %xmm4,32(%rsi,%rax,4) movhps %xmm0,8(%rsi,%rax,4) movhps %xmm1,24(%rsi,%rax,4) movhps %xmm4,40(%rsi,%rax,4) decl nb314_innerk(%rsp) jz _nb_kernel314_x86_64_sse.nb314_updateouterdata jmp _nb_kernel314_x86_64_sse.nb314_single_loop_nb_kernel314_x86_64_sse.nb314_updateouterdata: movl nb314_ii3(%⌨️ 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?