nb_kernel132_x86_64_sse.intel_syntax.s
来自「最著名最快的分子模拟软件」· S 代码 · 共 2,232 行 · 第 1/5 页
S
2,232 行
mulps xmm10, [rsp + nb132_dxH1H1] mulps xmm11, [rsp + nb132_dyH1H1] mulps xmm12, [rsp + nb132_dzH1H1] mulps xmm13, [rsp + nb132_dxH2H1] mulps xmm14, [rsp + nb132_dyH2H1] mulps xmm15, [rsp + nb132_dzH2H1] movaps xmm3, xmm7 movaps xmm4, xmm8 addps xmm2, xmm9 addps xmm7, [rsp + nb132_fixO] addps xmm8, [rsp + nb132_fiyO] addps xmm9, [rsp + nb132_fizO] addps xmm3, xmm10 addps xmm4, xmm11 addps xmm2, xmm12 addps xmm10, [rsp + nb132_fixH1] addps xmm11, [rsp + nb132_fiyH1] addps xmm12, [rsp + nb132_fizH1] addps xmm3, xmm13 addps xmm4, xmm14 addps xmm2, xmm15 addps xmm13, [rsp + nb132_fixH2] addps xmm14, [rsp + nb132_fiyH2] addps xmm15, [rsp + nb132_fizH2] movaps [rsp + nb132_fixO], xmm7 movaps [rsp + nb132_fiyO], xmm8 movaps [rsp + nb132_fizO], xmm9 movaps [rsp + nb132_fixH1], xmm10 movaps [rsp + nb132_fiyH1], xmm11 movaps [rsp + nb132_fizH1], xmm12 movaps [rsp + nb132_fixH2], xmm13 movaps [rsp + nb132_fiyH2], xmm14 movaps [rsp + nb132_fizH2], xmm15 ;# xmm0 = fH1x ;# xmm1 = fH1y ;# xmm2 = fH1z movaps xmm5, xmm3 unpcklps xmm3, xmm4 unpckhps xmm5, xmm4 addps xmm0, xmm3 addps xmm1, xmm5 movhlps xmm3, xmm2 ;# fH1zc fH1zd movlps [rdi + rax*4 + 12], xmm0 movhps [rdi + rbx*4 + 12], xmm0 movlps [rdi + rcx*4 + 12], xmm1 movhps [rdi + rdx*4 + 12], xmm1 movss [rdi + rax*4 + 20], xmm2 movss [rdi + rcx*4 + 20], xmm3 shufps xmm2, xmm2, 1 shufps xmm3, xmm3, 1 movss [rdi + rbx*4 + 20], xmm2 movss [rdi + rdx*4 + 20], xmm3 ;# move j H2 coordinates to local temp variables mov rsi, [rbp + nb132_pos] ;# base of pos[] movlps xmm0, [rsi + rax*4 + 24] ;# jxH2a jyH2a - - movlps xmm1, [rsi + rcx*4 + 24] ;# jxH2c jyH2c - - movhps xmm0, [rsi + rbx*4 + 24] ;# jxH2a jyH2a jxH2b jyH2b movhps xmm1, [rsi + rdx*4 + 24] ;# jxH2c jyH2c jxH2d jyH2d movss xmm2, [rsi + rax*4 + 32] ;# jzH2a - - - movss xmm3, [rsi + rcx*4 + 32] ;# jzH2c - - - movss xmm5, [rsi + rbx*4 + 32] ;# jzH2b - - - movss xmm6, [rsi + rdx*4 + 32] ;# jzH2d - - - movlhps xmm2, xmm5 ;# jzH2a - jzH2b - movlhps xmm3, xmm6 ;# jzH2c - jzH2d - movaps xmm4, xmm0 unpcklps xmm0, xmm1 ;# jxH2a jxH2c jyH2a jyH2c unpckhps xmm4, xmm1 ;# jxH2b jxH2d jyH2b jyH2d movaps xmm1, xmm0 unpcklps xmm0, xmm4 ;# x unpckhps xmm1, xmm4 ;# y shufps xmm2, xmm3, 136 ;# 10001000 => jzH2a jzH2b jzH2c jzH2d ;# xmm0 = H2x ;# xmm1 = H2y ;# xmm2 = H2z movaps xmm3, xmm0 movaps xmm4, xmm1 movaps xmm5, xmm2 movaps xmm6, xmm0 movaps xmm7, xmm1 movaps xmm8, xmm2 subps xmm0, [rsp + nb132_ixO] subps xmm1, [rsp + nb132_iyO] subps xmm2, [rsp + nb132_izO] subps xmm3, [rsp + nb132_ixH1] subps xmm4, [rsp + nb132_iyH1] subps xmm5, [rsp + nb132_izH1] subps xmm6, [rsp + nb132_ixH2] subps xmm7, [rsp + nb132_iyH2] subps xmm8, [rsp + nb132_izH2] movaps [rsp + nb132_dxOH2], xmm0 movaps [rsp + nb132_dyOH2], xmm1 movaps [rsp + nb132_dzOH2], xmm2 mulps xmm0, xmm0 mulps xmm1, xmm1 mulps xmm2, xmm2 movaps [rsp + nb132_dxH1H2], xmm3 movaps [rsp + nb132_dyH1H2], xmm4 movaps [rsp + nb132_dzH1H2], xmm5 mulps xmm3, xmm3 mulps xmm4, xmm4 mulps xmm5, xmm5 movaps [rsp + nb132_dxH2H2], xmm6 movaps [rsp + nb132_dyH2H2], xmm7 movaps [rsp + nb132_dzH2H2], xmm8 mulps xmm6, xmm6 mulps xmm7, xmm7 mulps xmm8, xmm8 addps xmm0, xmm1 addps xmm0, xmm2 addps xmm3, xmm4 addps xmm3, xmm5 addps xmm6, xmm7 addps xmm6, xmm8 ;# start doing invsqrt for jH2 atoms rsqrtps xmm1, xmm0 rsqrtps xmm4, xmm3 rsqrtps xmm7, xmm6 movaps xmm2, xmm1 movaps xmm5, xmm4 movaps xmm8, xmm7 mulps xmm1, xmm1 ;# lu*lu mulps xmm4, xmm4 ;# lu*lu mulps xmm7, xmm7 ;# lu*lu movaps xmm9, [rsp + nb132_three] movaps xmm10, xmm9 movaps xmm11, xmm9 mulps xmm1, xmm0 ;# rsq*lu*lu mulps xmm4, xmm3 ;# rsq*lu*lu mulps xmm7, xmm6 ;# rsq*lu*lu subps xmm9, xmm1 subps xmm10, xmm4 subps xmm11, xmm7 ;# 3-rsq*lu*lu mulps xmm9, xmm2 mulps xmm10, xmm5 mulps xmm11, xmm8 ;# lu*(3-rsq*lu*lu) movaps xmm0, [rsp + nb132_half] mulps xmm9, xmm0 ;# rinvOH2 mulps xmm10, xmm0 ;# rinvH1H2 mulps xmm11, xmm0 ;# rinvH2H2 ;# H2 interactions movaps xmm0, xmm9 movaps xmm1, xmm10 movaps xmm2, xmm11 mulps xmm9, xmm9 mulps xmm10, xmm10 mulps xmm11, xmm11 mulps xmm0, [rsp + nb132_qqOH] mulps xmm1, [rsp + nb132_qqHH] mulps xmm2, [rsp + nb132_qqHH] mulps xmm9, xmm0 mulps xmm10, xmm1 mulps xmm11, xmm2 addps xmm0, [rsp + nb132_vctot] addps xmm1, xmm2 addps xmm0, xmm1 movaps [rsp + nb132_vctot], xmm0 ;# move j H2 forces to local temp variables mov rdi, [rbp + nb132_faction] ;# base of faction[] movlps xmm0, [rdi + rax*4 + 24] ;# jxH2a jyH2a - - movlps xmm1, [rdi + rcx*4 + 24] ;# jxH2c jyH2c - - movhps xmm0, [rdi + rbx*4 + 24] ;# jxH2a jyH2a jxH2b jyH2b movhps xmm1, [rdi + rdx*4 + 24] ;# jxH2c jyH2c jxH2d jyH2d movss xmm2, [rdi + rax*4 + 32] ;# jzH2a - - - movss xmm3, [rdi + rcx*4 + 32] ;# jzH2c - - - movss xmm7, [rdi + rbx*4 + 32] ;# jzH2b - - - movss xmm8, [rdi + rdx*4 + 32] ;# jzH2d - - - movlhps xmm2, xmm7 ;# jzH2a - jzH2b - movlhps xmm3, xmm8 ;# jzH2c - jzH2d - shufps xmm2, xmm3, 136 ;# 10001000 => jzH2a jzH2b jzH2c jzH2d ;# xmm0: jxH2a jyH2a jxH2b jyH2b ;# xmm1: jxH2c jyH2c jxH2d jyH2d ;# xmm2: jzH2a jzH2b jzH2c jzH2d movaps xmm7, xmm9 movaps xmm8, xmm9 movaps xmm13, xmm11 movaps xmm14, xmm11 movaps xmm15, xmm11 movaps xmm11, xmm10 movaps xmm12, xmm10 mulps xmm7, [rsp + nb132_dxOH2] mulps xmm8, [rsp + nb132_dyOH2] mulps xmm9, [rsp + nb132_dzOH2] mulps xmm10, [rsp + nb132_dxH1H2] mulps xmm11, [rsp + nb132_dyH1H2] mulps xmm12, [rsp + nb132_dzH1H2] mulps xmm13, [rsp + nb132_dxH2H2] mulps xmm14, [rsp + nb132_dyH2H2] mulps xmm15, [rsp + nb132_dzH2H2] movaps xmm3, xmm7 movaps xmm4, xmm8 addps xmm2, xmm9 addps xmm7, [rsp + nb132_fixO] addps xmm8, [rsp + nb132_fiyO] addps xmm9, [rsp + nb132_fizO] addps xmm3, xmm10 addps xmm4, xmm11 addps xmm2, xmm12 addps xmm10, [rsp + nb132_fixH1] addps xmm11, [rsp + nb132_fiyH1] addps xmm12, [rsp + nb132_fizH1] addps xmm3, xmm13 addps xmm4, xmm14 addps xmm2, xmm15 addps xmm13, [rsp + nb132_fixH2] addps xmm14, [rsp + nb132_fiyH2] addps xmm15, [rsp + nb132_fizH2] movaps [rsp + nb132_fixO], xmm7 movaps [rsp + nb132_fiyO], xmm8 movaps [rsp + nb132_fizO], xmm9 movaps [rsp + nb132_fixH1], xmm10 movaps [rsp + nb132_fiyH1], xmm11 movaps [rsp + nb132_fizH1], xmm12 movaps [rsp + nb132_fixH2], xmm13 movaps [rsp + nb132_fiyH2], xmm14 movaps [rsp + nb132_fizH2], xmm15 ;# xmm0 = fH2x ;# xmm1 = fH2y ;# xmm2 = fH2z movaps xmm5, xmm3 unpcklps xmm3, xmm4 unpckhps xmm5, xmm4 addps xmm0, xmm3 addps xmm1, xmm5 movhlps xmm3, xmm2 ;# fH2zc fH2zd movlps [rdi + rax*4 + 24], xmm0 movhps [rdi + rbx*4 + 24], xmm0 movlps [rdi + rcx*4 + 24], xmm1 movhps [rdi + rdx*4 + 24], xmm1 movss [rdi + rax*4 + 32], xmm2 movss [rdi + rcx*4 + 32], xmm3 shufps xmm2, xmm2, 1 shufps xmm3, xmm3, 1 movss [rdi + rbx*4 + 32], xmm2 movss [rdi + rdx*4 + 32], xmm3 ;# should we do one more iteration? sub dword ptr [rsp + nb132_innerk], 4 jl .nb132_single_check jmp .nb132_unroll_loop.nb132_single_check: add dword ptr [rsp + nb132_innerk], 4 jnz .nb132_single_loop jmp .nb132_updateouterdata.nb132_single_loop: mov rdx, [rsp + nb132_innerjjnr] ;# pointer to jjnr[k] mov eax, [rdx] add qword ptr [rsp + nb132_innerjjnr], 4 mov rsi, [rbp + nb132_pos] lea rax, [rax + rax*2] ;# fetch j coordinates xorps xmm0, xmm0 xorps xmm1, xmm1 xorps xmm2, xmm2 movss xmm0, [rsi + rax*4] ;# jxO - - - movss xmm1, [rsi + rax*4 + 4] ;# jyO - - - movss xmm2, [rsi + rax*4 + 8] ;# jzO - - - movlps xmm6, [rsi + rax*4 + 12] ;# xmm6 = jxH1 jyH1 - - movss xmm7, [rsi + rax*4 + 20] ;# xmm7 = jzH1 - - - movhps xmm6, [rsi + rax*4 + 24] ;# xmm6 = jxH1 jyH1 jxH2 jyH2 movss xmm5, [rsi + rax*4 + 32] ;# xmm5 = jzH2 - - - ;# have all coords, time for some shuffling. shufps xmm6, xmm6, 216 ;# 11011000 ;# xmm6 = jxH1 jxH2 jyH1 jyH2 unpcklps xmm7, xmm5 ;# xmm7 = jzH1 jzH2 - - movlhps xmm0, xmm6 ;# xmm0 = jxO 0 jxH1 jxH2 shufps xmm1, xmm6, 228 ;# 11100100 ;# xmm1 = jyO 0 jyH1 jyH2 shufps xmm2, xmm7, 68 ;# 01000100 ;# xmm2 = jzO 0 jzH1 jzH2 ;# store all j coordinates in jO movaps [rsp + nb132_jxO], xmm0 movaps [rsp + nb132_jyO], xmm1 movaps [rsp + nb132_jzO], xmm2 subps xmm0, [rsp + nb132_ixO] subps xmm1, [rsp + nb132_iyO] subps xmm2, [rsp + nb132_izO] movaps [rsp + nb132_dxOO], xmm0 movaps [rsp + nb132_dyOO], xmm1 movaps [rsp + nb132_dzOO], xmm2 mulps xmm0, xmm0 mulps xmm1, xmm1 mulps xmm2, xmm2 addps xmm0, xmm1 addps xmm0, xmm2 ;# have rsq in xmm0 movaps [rsp + nb132_rsqOO], xmm0 ;# do invsqrt rsqrtps xmm1, xmm0 movaps xmm2, xmm1 mulps xmm1, xmm1 movaps xmm3, [rsp + nb132_three] mulps xmm1, xmm0 subps xmm3, xmm1 mulps xmm3, xmm2 mulps xmm3, [rsp + nb132_half] ;# rinv iO - j water in xmm3 movaps [rsp + nb132_rinvOO], xmm3 movaps xmm0, xmm3 ;# rinv xorps xmm4, xmm4 ;# fetch charges to xmm4 (temporary) movss xmm4, [rsp + nb132_qqOO] movhps xmm4, [rsp + nb132_qqOH] mulps xmm3, xmm4 ;# vcoul movaps xmm6, xmm3 mulps xmm3, xmm0 movaps [rsp + nb132_fstmp], xmm3 ;# save it addps xmm6, [rsp + nb132_vctot] movaps [rsp + nb132_vctot], xmm6 movss xmm1, xmm0 ;# rinv mulss xmm1, [rsp + nb132_rsqOO] ;# xmm1=r mulss xmm1, [rsp + nb132_tsc] cvttps2pi mm6, xmm1 cvtpi2ps xmm3, mm6 subss xmm1, xmm3 ;# xmm1=eps movss xmm2, xmm1 mulss xmm2, xmm2 ;# xmm2=eps2 pslld mm6, 3 mov rsi, [rbp + nb132_VFtab] movd r8d, mm6 ;# dispersion movlps xmm5, [rsi + r8*4] movaps xmm4, xmm5 shufps xmm4, xmm7, 136 ;# constant 10001000 shufps xmm5, xmm7, 221 ;# constant 11011101 movlps xmm7, [rsi + r8*4 + 8] movaps xmm6, xmm7 shufps xmm6, xmm3, 136 ;# constant 10001000 shufps xmm7, xmm3, 221 ;# constant 11011101 ;# dispersion table ready, in xmm4-xmm7 mulss xmm6, xmm1 ;# xmm6=Geps mulss xmm7, xmm2 ;# xmm7=Heps2 addss xmm5, xmm6 addss xmm5, xmm7 ;# xmm5=Fp addss xmm7, xmm7 ;# two*Heps2 addss xmm7, xmm6 addss xmm7, xmm5 ;# xmm7=FF mulss xmm5, xmm1 ;# xmm5=eps*Fp addss xmm5, xmm4 ;# xmm5=VV movss xmm4, [rsp + nb132_c6] mulss xmm7, xmm4 ;# fijD mulss xmm5, xmm4 ;# Vvdw6 movss xmm3, [rsp + nb132_fstmp] mulps xmm7, [rsp + nb132_tsc] subss xmm3, xmm7 movss [rsp + nb132_fstmp], xmm3 addss xmm5, [rsp + nb132_Vvdwtot] movss [rsp + nb132_Vvdwtot], xmm5 ;# repulsion movlps xmm5, [rsi + r8*4 + 16] movaps xmm4, xmm5 shufps xmm4, xmm7, 136 ;# constant 10001000 shufps xmm5, xmm7, 221 ;# constant 11011101 movlps xmm7, [rsi + r8*4 + 24] movaps xmm6, xmm7 shufps xmm6, xmm3, 136 ;# constant 10001000 shufps xmm7, xmm3, 221 ;# constant 11011101 ;# table ready, in xmm4-xmm7 mulss xmm6, xmm1 ;# xmm6=Geps mulss xmm7, xmm2 ;# xmm7=Heps2 addss xmm5, xmm6 addss xmm5, xmm7 ;# xmm5=Fp addss xmm7, xmm7 ;# two*Heps2 addss xmm7, xmm6 addss xmm7, xmm5 ;# xmm7=FF mulss xmm5, xmm1 ;# xmm5=eps*Fp addss xmm5, xmm4 ;# xmm5=VV movss xmm4, [rsp + nb132_c12] mulss xmm7, xmm4 ;# fijR mulss xmm5, xmm4 ;# Vvdw12 movaps xmm3, [rsp + nb132_fstmp] mulss xmm7, [rsp + nb132_tsc] subss xmm3, xmm7 addss xmm5, [rsp + nb132_Vvdwtot] movss [rsp + nb132_Vvdwtot], xmm5 mulps xmm0, xmm3 movaps xmm1, xmm0 movaps xmm2, xmm0 mulps xmm0, [rsp + nb132_dxOO] mulps xmm1, [rsp + nb132_dyOO] mulps xmm2, [rsp + nb132_dzOO] ;# initial update for j forces movaps xmm3, xmm0 movaps xmm4, xmm1 movaps xmm5, xmm2⌨️ 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?