nb_kernel214_x86_64_sse.intel_syntax.s
来自「最著名最快的分子模拟软件」· S 代码 · 共 2,201 行 · 第 1/5 页
S
2,201 行
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 + nb214_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 xmm4, [rsp + nb214_half] mulps xmm9, xmm4 ;# rinvH1H2 mulps xmm10, xmm4 ;# rinvH2H2 mulps xmm11, xmm4 ;# rinvMH2 ;# H2 interactions ;# rsq in xmm0,xmm3,xmm6 ;# rinv in xmm9, xmm10, xmm11 movaps xmm1, xmm9 ;# copy of rinv movaps xmm4, xmm10 movaps xmm7, xmm11 movaps xmm2, [rsp + nb214_krf] mulps xmm9, xmm9 ;# rinvsq mulps xmm10, xmm10 mulps xmm11, xmm11 mulps xmm0, xmm2 ;# k*rsq mulps xmm3, xmm2 mulps xmm6, xmm2 movaps xmm2, xmm0 ;# copy of k*rsq movaps xmm5, xmm3 movaps xmm8, xmm6 addps xmm2, xmm1 ;# rinv+krsq addps xmm5, xmm4 addps xmm8, xmm7 movaps xmm14, [rsp + nb214_crf] subps xmm2, xmm14 ;# rinv+krsq-crf subps xmm5, xmm14 subps xmm8, xmm14 movaps xmm12, [rsp + nb214_qqHH] movaps xmm13, [rsp + nb214_qqMH] mulps xmm2, xmm12 ;# xmm6=voul=qq*(rinv+ krsq-crf) mulps xmm5, xmm12 ;# xmm6=voul=qq*(rinv+ krsq-crf) mulps xmm8, xmm13 ;# xmm6=voul=qq*(rinv+ krsq-crf) addps xmm0, xmm0 ;# 2*krsq addps xmm3, xmm3 addps xmm6, xmm6 subps xmm1, xmm0 ;# rinv-2*krsq subps xmm4, xmm3 subps xmm7, xmm6 mulps xmm1, xmm12 ;# (rinv-2*krsq)*qq mulps xmm4, xmm12 mulps xmm7, xmm13 addps xmm15, xmm2 addps xmm5, xmm8 addps xmm15, xmm5 mulps xmm1, xmm9 ;# fscal mulps xmm4, xmm10 mulps xmm7, xmm11 ;# move j H2 forces to local temp variables movlps xmm9, [rdi + rax*4 + 24] ;# jxH2a jyH2a - - movlps xmm10, [rdi + rcx*4 + 24] ;# jxH2c jyH2c - - movhps xmm9, [rdi + rbx*4 + 24] ;# jxH2a jyH2a jxH2b jyH2b movhps xmm10, [rdi + rdx*4 + 24] ;# jxH2c jyH2c jxH2d jyH2d movss xmm11, [rdi + rax*4 + 32] ;# jzH2a - - - movss xmm12, [rdi + rcx*4 + 32] ;# jzH2c - - - movhps xmm11, [rdi + rbx*4 + 32] ;# jzH2a - jzH2b - movhps xmm12, [rdi + rdx*4 + 32] ;# jzH2c - jzH2d - shufps xmm11, xmm12, 136 ;# 10001000 => jzH2a jzH2b jzH2c jzH2d ;# xmm9: jxH2a jyH2a jxH2b jyH2b ;# xmm10: jxH2c jyH2c jxH2d jyH2d ;# xmm11: jzH2a jzH2b jzH2c jzH2d movaps xmm0, xmm1 movaps xmm2, xmm1 movaps xmm3, xmm4 movaps xmm5, xmm4 movaps xmm6, xmm7 movaps xmm8, xmm7 mulps xmm0, [rsp + nb214_dxH1H2] mulps xmm1, [rsp + nb214_dyH1H2] mulps xmm2, [rsp + nb214_dzH1H2] mulps xmm3, [rsp + nb214_dxH2H2] mulps xmm4, [rsp + nb214_dyH2H2] mulps xmm5, [rsp + nb214_dzH2H2] mulps xmm6, [rsp + nb214_dxMH2] mulps xmm7, [rsp + nb214_dyMH2] mulps xmm8, [rsp + nb214_dzMH2] movaps xmm13, xmm0 movaps xmm14, xmm1 addps xmm11, xmm2 addps xmm0, [rsp + nb214_fixH1] addps xmm1, [rsp + nb214_fiyH1] addps xmm2, [rsp + nb214_fizH1] addps xmm13, xmm3 addps xmm14, xmm4 addps xmm11, xmm5 addps xmm3, [rsp + nb214_fixH2] addps xmm4, [rsp + nb214_fiyH2] addps xmm5, [rsp + nb214_fizH2] addps xmm13, xmm6 addps xmm14, xmm7 addps xmm11, xmm8 addps xmm6, [rsp + nb214_fixM] addps xmm7, [rsp + nb214_fiyM] addps xmm8, [rsp + nb214_fizM] movaps [rsp + nb214_fixH1], xmm0 movaps [rsp + nb214_fiyH1], xmm1 movaps [rsp + nb214_fizH1], xmm2 movaps [rsp + nb214_fixH2], xmm3 movaps [rsp + nb214_fiyH2], xmm4 movaps [rsp + nb214_fizH2], xmm5 movaps [rsp + nb214_fixM], xmm6 movaps [rsp + nb214_fiyM], xmm7 movaps [rsp + nb214_fizM], xmm8 ;# xmm9 = fH2x ;# xmm10 = fH2y ;# xmm11 = fH2z movaps xmm0, xmm13 unpcklps xmm13, xmm14 unpckhps xmm0, xmm14 addps xmm9, xmm13 addps xmm10, xmm0 movhlps xmm12, xmm11 ;# fH2zc fH2zd movlps [rdi + rax*4 + 24], xmm9 movhps [rdi + rbx*4 + 24], xmm9 movlps [rdi + rcx*4 + 24], xmm10 movhps [rdi + rdx*4 + 24], xmm10 movss [rdi + rax*4 + 32], xmm11 movss [rdi + rcx*4 + 32], xmm12 shufps xmm11, xmm11, 1 shufps xmm12, xmm12, 1 movss [rdi + rbx*4 + 32], xmm11 movss [rdi + rdx*4 + 32], xmm12 ;# move j M coordinates to local temp variables 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 + nb214_ixH1] subps xmm1, [rsp + nb214_iyH1] subps xmm2, [rsp + nb214_izH1] subps xmm3, [rsp + nb214_ixH2] subps xmm4, [rsp + nb214_iyH2] subps xmm5, [rsp + nb214_izH2] subps xmm6, [rsp + nb214_ixM] subps xmm7, [rsp + nb214_iyM] subps xmm8, [rsp + nb214_izM] movaps [rsp + nb214_dxH1M], xmm0 movaps [rsp + nb214_dyH1M], xmm1 movaps [rsp + nb214_dzH1M], xmm2 mulps xmm0, xmm0 mulps xmm1, xmm1 mulps xmm2, xmm2 movaps [rsp + nb214_dxH2M], xmm3 movaps [rsp + nb214_dyH2M], xmm4 movaps [rsp + nb214_dzH2M], xmm5 mulps xmm3, xmm3 mulps xmm4, xmm4 mulps xmm5, xmm5 movaps [rsp + nb214_dxMM], xmm6 movaps [rsp + nb214_dyMM], xmm7 movaps [rsp + nb214_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 + nb214_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 xmm4, [rsp + nb214_half] mulps xmm9, xmm4 ;# rinvH1M mulps xmm10, xmm4 ;# rinvH2M mulps xmm11, xmm4 ;# rinvMM ;# M interactions ;# rsq in xmm0,xmm3,xmm6 ;# rinv in xmm9, xmm10, xmm11 movaps xmm1, xmm9 ;# copy of rinv movaps xmm4, xmm10 movaps xmm7, xmm11 movaps xmm2, [rsp + nb214_krf] mulps xmm9, xmm9 ;# rinvsq mulps xmm10, xmm10 mulps xmm11, xmm11 mulps xmm0, xmm2 ;# k*rsq mulps xmm3, xmm2 mulps xmm6, xmm2 movaps xmm2, xmm0 ;# copy of k*rsq movaps xmm5, xmm3 movaps xmm8, xmm6 addps xmm2, xmm1 ;# rinv+krsq addps xmm5, xmm4 addps xmm8, xmm7 movaps xmm14, [rsp + nb214_crf] subps xmm2, xmm14 ;# rinv+krsq-crf subps xmm5, xmm14 subps xmm8, xmm14 movaps xmm12, [rsp + nb214_qqMH] movaps xmm13, [rsp + nb214_qqMM] mulps xmm2, xmm12 ;# xmm6=voul=qq*(rinv+ krsq-crf) mulps xmm5, xmm12 ;# xmm6=voul=qq*(rinv+ krsq-crf) mulps xmm8, xmm13 ;# xmm6=voul=qq*(rinv+ krsq-crf) addps xmm0, xmm0 ;# 2*krsq addps xmm3, xmm3 addps xmm6, xmm6 subps xmm1, xmm0 ;# rinv-2*krsq subps xmm4, xmm3 subps xmm7, xmm6 mulps xmm1, xmm12 ;# (rinv-2*krsq)*qq mulps xmm4, xmm12 mulps xmm7, xmm13 addps xmm5, xmm8 addps xmm2, xmm15 addps xmm2, xmm5 movaps [rsp + nb214_vctot], xmm2 mulps xmm1, xmm9 ;# fscal mulps xmm4, xmm10 mulps xmm7, xmm11 ;# move j M forces to local temp variables movlps xmm9, [rdi + rax*4 + 36] ;# jxMa jyMa - - movlps xmm10, [rdi + rcx*4 + 36] ;# jxMc jyMc - - movhps xmm9, [rdi + rbx*4 + 36] ;# jxMa jyMa jxMb jyMb movhps xmm10, [rdi + rdx*4 + 36] ;# jxMc jyMc jxMd jyMd movss xmm11, [rdi + rax*4 + 44] ;# jzMa - - - movss xmm12, [rdi + rcx*4 + 44] ;# jzMc - - - movss xmm2, [rdi + rbx*4 + 44] ;# jzMb - - - movss xmm3, [rdi + rdx*4 + 44] ;# jzMd - - - movlhps xmm11, xmm2 ;# jzMa - jzMb - movlhps xmm12, xmm3 ;# jzMc - jzMd - shufps xmm11, xmm12, 136 ;# 10001000 => jzMa jzMb jzMc jzMd ;# xmm9: jxMa jyMa jxMb jyMb ;# xmm10: jxMc jyMc jxMd jyMd ;# xmm11: jzMa jzMb jzMc jzMd movaps xmm0, xmm1 movaps xmm2, xmm1 movaps xmm3, xmm4 movaps xmm5, xmm4 movaps xmm6, xmm7 movaps xmm8, xmm7 mulps xmm0, [rsp + nb214_dxH1M] mulps xmm1, [rsp + nb214_dyH1M] mulps xmm2, [rsp + nb214_dzH1M] mulps xmm3, [rsp + nb214_dxH2M] mulps xmm4, [rsp + nb214_dyH2M] mulps xmm5, [rsp + nb214_dzH2M] mulps xmm6, [rsp + nb214_dxMM] mulps xmm7, [rsp + nb214_dyMM] mulps xmm8, [rsp + nb214_dzMM] movaps xmm13, xmm0 movaps xmm14, xmm1 addps xmm11, xmm2 addps xmm0, [rsp + nb214_fixH1] addps xmm1, [rsp + nb214_fiyH1] addps xmm2, [rsp + nb214_fizH1] addps xmm13, xmm3 addps xmm14, xmm4 addps xmm11, xmm5 addps xmm3, [rsp + nb214_fixH2] addps xmm4, [rsp + nb214_fiyH2] addps xmm5, [rsp + nb214_fizH2] addps xmm13, xmm6 addps xmm14, xmm7 addps xmm11, xmm8 addps xmm6, [rsp + nb214_fixM] addps xmm7, [rsp + nb214_fiyM] addps xmm8, [rsp + nb214_fizM] movaps [rsp + nb214_fixH1], xmm0 movaps [rsp + nb214_fiyH1], xmm1 movaps [rsp + nb214_fizH1], xmm2 movaps [rsp + nb214_fixH2], xmm3 movaps [rsp + nb214_fiyH2], xmm4 movaps [rsp + nb214_fizH2], xmm5 movaps [rsp + nb214_fixM], xmm6 movaps [rsp + nb214_fiyM], xmm7 movaps [rsp + nb214_fizM], xmm8 ;# xmm0 = fMx ;# xmm1 = fMy ;# xmm2 = fMz movaps xmm0, xmm13 unpcklps xmm13, xmm14 unpckhps xmm0, xmm14 addps xmm9, xmm13 addps xmm10, xmm0 movhlps xmm12, xmm11 ;# fMzc fMzd movlps [rdi + rax*4 + 36], xmm9 movhps [rdi + rbx*4 + 36], xmm9 movlps [rdi + rcx*4 + 36], xmm10 movhps [rdi + rdx*4 + 36], xmm10 movss [rdi + rax*4 + 44], xmm11 movss [rdi + rcx*4 + 44], xmm12 shufps xmm11, xmm11, 1 shufps xmm12, xmm12, 1 movss [rdi + rbx*4 + 44], xmm11 movss [rdi + rdx*4 + 44], xmm12 ;# should we do one more iteration? sub dword ptr [rsp + nb214_innerk], 4 jl .nb214_single_check jmp .nb214_unroll_loop.nb214_single_check: add dword ptr [rsp + nb214_innerk], 4 jnz .nb214_single_loop jmp .nb214_updateouterdata.nb214_single_loop: mov rdx, [rsp + nb214_innerjjnr] ;# pointer to jjnr[k] mov eax, [rdx] add qword ptr [rsp + nb214_innerjjnr], 4 mov rsi, [rbp + nb214_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⌨️ 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?