nb_kernel134_x86_64_sse.intel_syntax.s
来自「最著名最快的分子模拟软件」· S 代码 · 共 2,215 行 · 第 1/5 页
S
2,215 行
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 + nb134_ixH1] subps xmm1, [rsp + nb134_iyH1] subps xmm2, [rsp + nb134_izH1] subps xmm3, [rsp + nb134_ixH2] subps xmm4, [rsp + nb134_iyH2] subps xmm5, [rsp + nb134_izH2] subps xmm6, [rsp + nb134_ixM] subps xmm7, [rsp + nb134_iyM] subps xmm8, [rsp + nb134_izM] movaps [rsp + nb134_dxH1H2], xmm0 movaps [rsp + nb134_dyH1H2], xmm1 movaps [rsp + nb134_dzH1H2], xmm2 mulps xmm0, xmm0 mulps xmm1, xmm1 mulps xmm2, xmm2 movaps [rsp + nb134_dxH2H2], xmm3 movaps [rsp + nb134_dyH2H2], xmm4 movaps [rsp + nb134_dzH2H2], xmm5 mulps xmm3, xmm3 mulps xmm4, xmm4 mulps xmm5, xmm5 movaps [rsp + nb134_dxMH2], xmm6 movaps [rsp + nb134_dyMH2], xmm7 movaps [rsp + nb134_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 + nb134_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 + nb134_half] mulps xmm9, xmm0 ;# rinvH1H2 mulps xmm10, xmm0 ;# rinvH2H2 mulps xmm11, xmm0 ;# rinvMH2 ;# H2 interactions movaps xmm0, xmm9 movaps xmm1, xmm10 movaps xmm2, xmm11 mulps xmm9, xmm9 mulps xmm10, xmm10 mulps xmm11, xmm11 mulps xmm0, [rsp + nb134_qqHH] mulps xmm1, [rsp + nb134_qqHH] mulps xmm2, [rsp + nb134_qqMH] mulps xmm9, xmm0 mulps xmm10, xmm1 mulps xmm11, xmm2 addps xmm0, [rsp + nb134_vctot] addps xmm1, xmm2 addps xmm0, xmm1 movaps [rsp + nb134_vctot], xmm0 ;# move j H2 forces to local temp variables mov rdi, [rbp + nb134_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 - - - movhps xmm2, [rdi + rbx*4 + 32] ;# jzH2a - jzH2b - movhps xmm3, [rdi + rdx*4 + 32] ;# 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 + nb134_dxH1H2] mulps xmm8, [rsp + nb134_dyH1H2] mulps xmm9, [rsp + nb134_dzH1H2] mulps xmm10, [rsp + nb134_dxH2H2] mulps xmm11, [rsp + nb134_dyH2H2] mulps xmm12, [rsp + nb134_dzH2H2] mulps xmm13, [rsp + nb134_dxMH2] mulps xmm14, [rsp + nb134_dyMH2] mulps xmm15, [rsp + nb134_dzMH2] movaps xmm3, xmm7 movaps xmm4, xmm8 addps xmm2, xmm9 addps xmm7, [rsp + nb134_fixH1] addps xmm8, [rsp + nb134_fiyH1] addps xmm9, [rsp + nb134_fizH1] addps xmm3, xmm10 addps xmm4, xmm11 addps xmm2, xmm12 addps xmm10, [rsp + nb134_fixH2] addps xmm11, [rsp + nb134_fiyH2] addps xmm12, [rsp + nb134_fizH2] addps xmm3, xmm13 addps xmm4, xmm14 addps xmm2, xmm15 addps xmm13, [rsp + nb134_fixM] addps xmm14, [rsp + nb134_fiyM] addps xmm15, [rsp + nb134_fizM] movaps [rsp + nb134_fixH1], xmm7 movaps [rsp + nb134_fiyH1], xmm8 movaps [rsp + nb134_fizH1], xmm9 movaps [rsp + nb134_fixH2], xmm10 movaps [rsp + nb134_fiyH2], xmm11 movaps [rsp + nb134_fizH2], xmm12 movaps [rsp + nb134_fixM], xmm13 movaps [rsp + nb134_fiyM], xmm14 movaps [rsp + nb134_fizM], xmm15 ;# xmm3 = fH2x , xmm4 = fH2y 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 ;# move j M coordinates to local temp variables mov rsi, [rbp + nb134_pos] ;# base of pos[] movlps xmm0, [rsi + rax*4 + 36] ;# jxMa jyMa - - movlps xmm1, [rsi + rcx*4 + 36] ;# jxMc jyMc - - movhps xmm0, [rsi + rbx*4 + 36] ;# jxMa jyMa jxMb jyMb movhps xmm1, [rsi + rdx*4 + 36] ;# jxMc jyMc jxMd jyMd movss xmm2, [rsi + rax*4 + 44] ;# jzMa - - - movss xmm3, [rsi + rcx*4 + 44] ;# jzMc - - - movss xmm5, [rsi + rbx*4 + 44] ;# jzMb - - - movss xmm6, [rsi + rdx*4 + 44] ;# jzMd - - - movlhps xmm2, xmm5 ;# jzMa - jzMb - movlhps xmm3, xmm6 ;# jzMc - jzMd - movaps xmm4, xmm0 unpcklps xmm0, xmm1 ;# jxMa jxMc jyMa jyMc unpckhps xmm4, xmm1 ;# jxMb jxMd jyMb jyMd movaps xmm1, xmm0 unpcklps xmm0, xmm4 ;# x unpckhps xmm1, xmm4 ;# y shufps xmm2, xmm3, 136 ;# 10001000 => jzMa jzMb jzMc jzMd ;# xmm0 = Mx ;# xmm1 = My ;# xmm2 = Mz movaps xmm3, xmm0 movaps xmm4, xmm1 movaps xmm5, xmm2 movaps xmm6, xmm0 movaps xmm7, xmm1 movaps xmm8, xmm2 subps xmm0, [rsp + nb134_ixH1] subps xmm1, [rsp + nb134_iyH1] subps xmm2, [rsp + nb134_izH1] subps xmm3, [rsp + nb134_ixH2] subps xmm4, [rsp + nb134_iyH2] subps xmm5, [rsp + nb134_izH2] subps xmm6, [rsp + nb134_ixM] subps xmm7, [rsp + nb134_iyM] subps xmm8, [rsp + nb134_izM] movaps [rsp + nb134_dxH1M], xmm0 movaps [rsp + nb134_dyH1M], xmm1 movaps [rsp + nb134_dzH1M], xmm2 mulps xmm0, xmm0 mulps xmm1, xmm1 mulps xmm2, xmm2 movaps [rsp + nb134_dxH2M], xmm3 movaps [rsp + nb134_dyH2M], xmm4 movaps [rsp + nb134_dzH2M], xmm5 mulps xmm3, xmm3 mulps xmm4, xmm4 mulps xmm5, xmm5 movaps [rsp + nb134_dxMM], xmm6 movaps [rsp + nb134_dyMM], xmm7 movaps [rsp + nb134_dzMM], 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 jM 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 + nb134_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 + nb134_half] mulps xmm9, xmm0 ;# rinvH1M mulps xmm10, xmm0 ;# rinvH2M mulps xmm11, xmm0 ;# rinvMM ;# M interactions movaps xmm0, xmm9 movaps xmm1, xmm10 movaps xmm2, xmm11 mulps xmm9, xmm9 mulps xmm10, xmm10 mulps xmm11, xmm11 mulps xmm0, [rsp + nb134_qqMH] mulps xmm1, [rsp + nb134_qqMH] mulps xmm2, [rsp + nb134_qqMM] mulps xmm9, xmm0 mulps xmm10, xmm1 mulps xmm11, xmm2 addps xmm0, [rsp + nb134_vctot] addps xmm1, xmm2 addps xmm0, xmm1 movaps [rsp + nb134_vctot], xmm0 ;# move j M forces to local temp variables mov rdi, [rbp + nb134_faction] movlps xmm0, [rdi + rax*4 + 36] ;# jxMa jyMa - - movlps xmm1, [rdi + rcx*4 + 36] ;# jxMc jyMc - - movhps xmm0, [rdi + rbx*4 + 36] ;# jxMa jyMa jxMb jyMb movhps xmm1, [rdi + rdx*4 + 36] ;# jxMc jyMc jxMd jyMd movss xmm2, [rdi + rax*4 + 44] ;# jzMa - - - movss xmm3, [rdi + rcx*4 + 44] ;# jzMc - - - movss xmm7, [rdi + rbx*4 + 44] ;# jzMb - - - movss xmm8, [rdi + rdx*4 + 44] ;# jzMd - - - movlhps xmm2, xmm7 ;# jzMa - jzMb - movlhps xmm3, xmm8 ;# jzMc - jzMd - shufps xmm2, xmm3, 136 ;# 10001000 => jzMa jzMb jzMc jzMd ;# xmm0: jxMa jyMa jxMb jyMb ;# xmm1: jxMc jyMc jxMd jyMd ;# xmm2: jzMa jzMb jzMc jzMd movaps xmm7, xmm9 movaps xmm8, xmm9 movaps xmm13, xmm11 movaps xmm14, xmm11 movaps xmm15, xmm11 movaps xmm11, xmm10 movaps xmm12, xmm10 mulps xmm7, [rsp + nb134_dxH1M] mulps xmm8, [rsp + nb134_dyH1M] mulps xmm9, [rsp + nb134_dzH1M] mulps xmm10, [rsp + nb134_dxH2M] mulps xmm11, [rsp + nb134_dyH2M] mulps xmm12, [rsp + nb134_dzH2M] mulps xmm13, [rsp + nb134_dxMM] mulps xmm14, [rsp + nb134_dyMM] mulps xmm15, [rsp + nb134_dzMM] movaps xmm3, xmm7 movaps xmm4, xmm8 addps xmm2, xmm9 addps xmm7, [rsp + nb134_fixH1] addps xmm8, [rsp + nb134_fiyH1] addps xmm9, [rsp + nb134_fizH1] addps xmm3, xmm10 addps xmm4, xmm11 addps xmm2, xmm12 addps xmm10, [rsp + nb134_fixH2] addps xmm11, [rsp + nb134_fiyH2] addps xmm12, [rsp + nb134_fizH2] addps xmm3, xmm13 addps xmm4, xmm14 addps xmm2, xmm15 addps xmm13, [rsp + nb134_fixM] addps xmm14, [rsp + nb134_fiyM] addps xmm15, [rsp + nb134_fizM] movaps [rsp + nb134_fixH1], xmm7 movaps [rsp + nb134_fiyH1], xmm8 movaps [rsp + nb134_fizH1], xmm9 movaps [rsp + nb134_fixH2], xmm10 movaps [rsp + nb134_fiyH2], xmm11 movaps [rsp + nb134_fizH2], xmm12 movaps [rsp + nb134_fixM], xmm13 movaps [rsp + nb134_fiyM], xmm14 movaps [rsp + nb134_fizM], xmm15 ;# xmm3 = fMx , xmm4 = fMy movaps xmm5, xmm3 unpcklps xmm3, xmm4 unpckhps xmm5, xmm4 addps xmm0, xmm3 addps xmm1, xmm5 movhlps xmm3, xmm2 ;# fMzc fMzd movlps [rdi + rax*4 + 36], xmm0 movhps [rdi + rbx*4 + 36], xmm0 movlps [rdi + rcx*4 + 36], xmm1 movhps [rdi + rdx*4 + 36], xmm1 movss [rdi + rax*4 + 44], xmm2 movss [rdi + rcx*4 + 44], xmm3 shufps xmm2, xmm2, 1 shufps xmm3, xmm3, 1 movss [rdi + rbx*4 + 44], xmm2 movss [rdi + rdx*4 + 44], xmm3 ;# should we do one more iteration? sub dword ptr [rsp + nb134_innerk], 4 jl .nb134_single_check jmp .nb134_unroll_loop.nb134_single_check: add dword ptr [rsp + nb134_innerk], 4 jnz .nb134_single_loop jmp .nb134_updateouterdata.nb134_single_loop: mov rdx, [rsp + nb134_innerjjnr] ;# pointer to jjnr[k] mov eax, [rdx] add qword ptr [rsp + nb134_innerjjnr], 4 mov rsi, [rbp + nb134_pos] lea rax, [rax + rax*2] ;# fetch j coordinates movlps xmm3, [rsi + rax*4] ;# Ox Oy movlps xmm4, [rsi + rax*4 + 16] ;# H1y H1z movlps xmm5, [rsi + rax*4 + 32] ;# H2z Mx movhps xmm3, [rsi + rax*4 + 8] ;# Ox Oy Oz H1x movhps xmm4, [rsi + rax*4 + 24] ;# H1y H1z H2x H2y movhps xmm5, [rsi + rax*4 + 40] ;# H2z Mx My Mz ;# transpose movaps xmm0, xmm4 movaps xmm1, xmm3 movaps xmm2, xmm4 movaps xmm6, xmm3 shufps xmm4, xmm5, 18 ;# (00010010) h2x - Mx - shufps xmm3, xmm0, 193 ;# (11000001) Oy - H1y - shufps xmm2, xmm5, 35 ;# (00100011) H2y - My - shufps xmm1, xmm0, 18 ;# (00010010) Oz - H1z - ;# xmm6: Ox - - H1x xmm5: H2z - - Mz shufps xmm6, xmm4, 140 ;# (10001100) Ox H1x H2x Mx ⌨️ 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?