nb_kernel114_x86_64_sse.intel_syntax.s
来自「最著名最快的分子模拟软件」· S 代码 · 共 2,204 行 · 第 1/5 页
S
2,204 行
add qword ptr [rsp + nb114_innerjjnr], 16 ;# advance pointer (unroll 4) mov rsi, [rbp + nb114_pos] ;# base of pos[] lea rax, [rax + rax*2] ;# replace jnr with j3 lea rbx, [rbx + rbx*2] lea rcx, [rcx + rcx*2] ;# replace jnr with j3 lea rdx, [rdx + rdx*2] ;# load j O coordinates movlps xmm0, [rsi + rax*4] ;# jxOa jyOa - - movlps xmm1, [rsi + rcx*4] ;# jxOc jyOc - - movhps xmm0, [rsi + rbx*4] ;# jxOa jyOa jxOb jyOb movhps xmm1, [rsi + rdx*4] ;# jxOc jyOc jxOd jyOd movss xmm2, [rsi + rax*4 + 8] ;# jzOa - - - movss xmm3, [rsi + rcx*4 + 8] ;# jzOc - - - movhps xmm2, [rsi + rbx*4 + 8] ;# jzOa - jzOb - movhps xmm3, [rsi + rdx*4 + 8] ;# jzOc - jzOd - movaps xmm4, xmm0 unpcklps xmm0, xmm1 ;# jxOa jxOc jyOa jyOc unpckhps xmm4, xmm1 ;# jxOb jxOd jyOb jyOd movaps xmm1, xmm0 unpcklps xmm0, xmm4 ;# x unpckhps xmm1, xmm4 ;# y shufps xmm2, xmm3, 136 ;# 10001000 => jzOa jzOb jzOc jzOd ;# xmm0 = Ox ;# xmm1 = Oy ;# xmm2 = Oz subps xmm0, [rsp + nb114_ixO] subps xmm1, [rsp + nb114_iyO] subps xmm2, [rsp + nb114_izO] movaps xmm4, xmm0 movaps xmm5, xmm1 movaps xmm6, xmm2 ;# square it mulps xmm0, xmm0 mulps xmm1, xmm1 mulps xmm2, xmm2 addps xmm1, xmm0 addps xmm1, xmm2 ;# rsq in xmm1 ;# move j O forces to local temp variables movlps xmm10, [rdi + rax*4] ;# jxOa jyOa - - movlps xmm11, [rdi + rcx*4] ;# jxOc jyOc - - movhps xmm10, [rdi + rbx*4] ;# jxOa jyOa jxOb jyOb movhps xmm11, [rdi + rdx*4] ;# jxOc jyOc jxOd jyOd movss xmm12, [rdi + rax*4 + 8] ;# jzOa - - - movss xmm13, [rdi + rcx*4 + 8] ;# jzOc - - - movhps xmm12, [rdi + rbx*4 + 8] ;# jzOa - jzOb - movhps xmm13, [rdi + rdx*4 + 8] ;# jzOc - jzOd - shufps xmm12, xmm13, 136 ;# 10001000 => jzOa jzOb jzOc jzOd ;# xmm10: jxOa jyOa jxOb jyOb ;# xmm11: jxOc jyOc jxOd jyOd ;# xmm12: jzOa jzOb jzOc jzOd ;# calc rinvsq=1/rsq rcpps xmm2, xmm1 movaps xmm0, [rsp + nb114_two] mulps xmm1, xmm2 subps xmm0, xmm1 mulps xmm0, xmm2 ;# xmm0=rinvsq movaps xmm1, xmm0 ;# rinvsq mulps xmm0, xmm0 ;# rinv4 mulps xmm0, xmm1 ;# rinv6 movaps xmm2, xmm0 mulps xmm2, xmm2 ;# xmm2=rinv12 mulps xmm0, [rsp + nb114_c6] mulps xmm2, [rsp + nb114_c12] movaps xmm8, xmm2 subps xmm2, xmm0 ;# Vvdw=Vvdw12-Vvdw6 mulps xmm0, [rsp + nb114_six] mulps xmm8, [rsp + nb114_twelve] subps xmm8, xmm0 mulps xmm8, xmm1 ;# xmm8=total fscal ;# add potential to Vvdwtot addps xmm2, [rsp + nb114_Vvdwtot] movaps [rsp + nb114_Vvdwtot], xmm2 ;# calculate scalar force by multiplying dx/dy/dz with fscal mulps xmm4, xmm8 mulps xmm5, xmm8 mulps xmm6, xmm8 ;# increment i force movaps xmm0, [rsp + nb114_fixO] movaps xmm1, [rsp + nb114_fiyO] movaps xmm2, [rsp + nb114_fizO] addps xmm0, xmm4 addps xmm1, xmm5 addps xmm2, xmm6 movaps [rsp + nb114_fixO], xmm0 movaps [rsp + nb114_fiyO], xmm1 movaps [rsp + nb114_fizO], xmm2 ;# update O forces ;# xmm3 = fH1x , xmm4 = fH1y movaps xmm3, xmm4 unpcklps xmm4, xmm5 ;# fjx1 fjx1 fjy1 fjy2 unpckhps xmm3, xmm5 ;# fjx3 fjx4 fjy3 fjy4 addps xmm10, xmm4 addps xmm11, xmm3 addps xmm12, xmm6 movhlps xmm13, xmm12 ;# fH1zc fH1zd movlps [rdi + rax*4], xmm10 movhps [rdi + rbx*4], xmm10 movlps [rdi + rcx*4], xmm11 movhps [rdi + rdx*4], xmm11 movss [rdi + rax*4 + 8], xmm12 movss [rdi + rcx*4 + 8], xmm13 shufps xmm12, xmm12, 1 shufps xmm13, xmm13, 1 movss [rdi + rbx*4 + 8], xmm12 movss [rdi + rdx*4 + 8], xmm13 ;# done with OO interaction. ;# move j H1 coordinates to local temp variables movlps xmm0, [rsi + rax*4 + 12] ;# jxH1a jyH1a - - movlps xmm1, [rsi + rcx*4 + 12] ;# jxH1c jyH1c - - movhps xmm0, [rsi + rbx*4 + 12] ;# jxH1a jyH1a jxH1b jyH1b movhps xmm1, [rsi + rdx*4 + 12] ;# jxH1c jyH1c jxH1d jyH1d movss xmm2, [rsi + rax*4 + 20] ;# jzH1a - - - movss xmm3, [rsi + rcx*4 + 20] ;# jzH1c - - - movhps xmm2, [rsi + rbx*4 + 20] ;# jzH1a - jzH1b - movhps xmm3, [rsi + rdx*4 + 20] ;# jzH1c - jzH1d - movaps xmm4, xmm0 unpcklps xmm0, xmm1 ;# jxH1a jxH1c jyH1a jyH1c unpckhps xmm4, xmm1 ;# jxH1b jxH1d jyH1b jyH1d movaps xmm1, xmm0 unpcklps xmm0, xmm4 ;# x unpckhps xmm1, xmm4 ;# y shufps xmm2, xmm3, 136 ;# 10001000 => jzH1a jzH1b jzH1c jzH1d ;# xmm0 = H1x ;# xmm1 = H1y ;# xmm2 = H1z movaps xmm3, xmm0 movaps xmm4, xmm1 movaps xmm5, xmm2 movaps xmm6, xmm0 movaps xmm7, xmm1 movaps xmm8, xmm2 subps xmm0, [rsp + nb114_ixH1] subps xmm1, [rsp + nb114_iyH1] subps xmm2, [rsp + nb114_izH1] subps xmm3, [rsp + nb114_ixH2] subps xmm4, [rsp + nb114_iyH2] subps xmm5, [rsp + nb114_izH2] subps xmm6, [rsp + nb114_ixM] subps xmm7, [rsp + nb114_iyM] subps xmm8, [rsp + nb114_izM] movaps [rsp + nb114_dxH1H1], xmm0 movaps [rsp + nb114_dyH1H1], xmm1 movaps [rsp + nb114_dzH1H1], xmm2 mulps xmm0, xmm0 mulps xmm1, xmm1 mulps xmm2, xmm2 movaps [rsp + nb114_dxH2H1], xmm3 movaps [rsp + nb114_dyH2H1], xmm4 movaps [rsp + nb114_dzH2H1], xmm5 mulps xmm3, xmm3 mulps xmm4, xmm4 mulps xmm5, xmm5 movaps [rsp + nb114_dxMH1], xmm6 movaps [rsp + nb114_dyMH1], xmm7 movaps [rsp + nb114_dzMH1], 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 jH1 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 + nb114_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 + nb114_half] mulps xmm9, xmm0 ;# rinvH1H1 mulps xmm10, xmm0 ;# rinvH2H1 mulps xmm11, xmm0 ;# rinvMH1 ;# H1 interactions movaps xmm0, xmm9 movaps xmm1, xmm10 movaps xmm2, xmm11 mulps xmm9, xmm9 mulps xmm10, xmm10 mulps xmm11, xmm11 mulps xmm0, [rsp + nb114_qqHH] mulps xmm1, [rsp + nb114_qqHH] mulps xmm2, [rsp + nb114_qqMH] mulps xmm9, xmm0 mulps xmm10, xmm1 mulps xmm11, xmm2 addps xmm0, [rsp + nb114_vctot] addps xmm1, xmm2 addps xmm0, xmm1 movaps [rsp + nb114_vctot], xmm0 ;# move j H1 forces to local temp variables movlps xmm0, [rdi + rax*4 + 12] ;# jxH1a jyH1a - - movlps xmm1, [rdi + rcx*4 + 12] ;# jxH1c jyH1c - - movhps xmm0, [rdi + rbx*4 + 12] ;# jxH1a jyH1a jxH1b jyH1b movhps xmm1, [rdi + rdx*4 + 12] ;# jxH1c jyH1c jxH1d jyH1d movss xmm2, [rdi + rax*4 + 20] ;# jzH1a - - - movss xmm3, [rdi + rcx*4 + 20] ;# jzH1c - - - movhps xmm2, [rdi + rbx*4 + 20] ;# jzH1a - jzH1b - movhps xmm3, [rdi + rdx*4 + 20] ;# jzH1c - jzH1d - shufps xmm2, xmm3, 136 ;# 10001000 => jzH1a jzH1b jzH1c jzH1d ;# xmm0: jxH1a jyH1a jxH1b jyH1b ;# xmm1: jxH1c jyH1c jxH1d jyH1d ;# xmm2: jzH1a jzH1b jzH1c jzH1d movaps xmm7, xmm9 movaps xmm8, xmm9 movaps xmm13, xmm11 movaps xmm14, xmm11 movaps xmm15, xmm11 movaps xmm11, xmm10 movaps xmm12, xmm10 mulps xmm7, [rsp + nb114_dxH1H1] mulps xmm8, [rsp + nb114_dyH1H1] mulps xmm9, [rsp + nb114_dzH1H1] mulps xmm10, [rsp + nb114_dxH2H1] mulps xmm11, [rsp + nb114_dyH2H1] mulps xmm12, [rsp + nb114_dzH2H1] mulps xmm13, [rsp + nb114_dxMH1] mulps xmm14, [rsp + nb114_dyMH1] mulps xmm15, [rsp + nb114_dzMH1] movaps xmm3, xmm7 movaps xmm4, xmm8 addps xmm2, xmm9 addps xmm7, [rsp + nb114_fixH1] addps xmm8, [rsp + nb114_fiyH1] addps xmm9, [rsp + nb114_fizH1] addps xmm3, xmm10 addps xmm4, xmm11 addps xmm2, xmm12 addps xmm10, [rsp + nb114_fixH2] addps xmm11, [rsp + nb114_fiyH2] addps xmm12, [rsp + nb114_fizH2] addps xmm3, xmm13 addps xmm4, xmm14 addps xmm2, xmm15 addps xmm13, [rsp + nb114_fixM] addps xmm14, [rsp + nb114_fiyM] addps xmm15, [rsp + nb114_fizM] movaps [rsp + nb114_fixH1], xmm7 movaps [rsp + nb114_fiyH1], xmm8 movaps [rsp + nb114_fizH1], xmm9 movaps [rsp + nb114_fixH2], xmm10 movaps [rsp + nb114_fiyH2], xmm11 movaps [rsp + nb114_fizH2], xmm12 movaps [rsp + nb114_fixM], xmm13 movaps [rsp + nb114_fiyM], xmm14 movaps [rsp + nb114_fizM], xmm15 ;# xmm3 = fH1x , xmm4 = fH1y 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 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 - - - movhps xmm2, [rsi + rbx*4 + 32] ;# jzH2a - jzH2b - movhps xmm3, [rsi + rdx*4 + 32] ;# 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 + nb114_ixH1] subps xmm1, [rsp + nb114_iyH1] subps xmm2, [rsp + nb114_izH1] subps xmm3, [rsp + nb114_ixH2] subps xmm4, [rsp + nb114_iyH2] subps xmm5, [rsp + nb114_izH2] subps xmm6, [rsp + nb114_ixM] subps xmm7, [rsp + nb114_iyM] subps xmm8, [rsp + nb114_izM] movaps [rsp + nb114_dxH1H2], xmm0 movaps [rsp + nb114_dyH1H2], xmm1 movaps [rsp + nb114_dzH1H2], xmm2 mulps xmm0, xmm0 mulps xmm1, xmm1 mulps xmm2, xmm2 movaps [rsp + nb114_dxH2H2], xmm3 movaps [rsp + nb114_dyH2H2], xmm4 movaps [rsp + nb114_dzH2H2], xmm5 mulps xmm3, xmm3 mulps xmm4, xmm4 mulps xmm5, xmm5 movaps [rsp + nb114_dxMH2], xmm6 movaps [rsp + nb114_dyMH2], xmm7 movaps [rsp + nb114_dzMH2], 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 + nb114_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 + nb114_half] mulps xmm9, xmm0 ;# rinvH1H2 mulps xmm10, xmm0 ;# rinvH2H2⌨️ 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?