nb_kernel100_ia32_sse.s
来自「最著名最快的分子模拟软件」· S 代码 · 共 1,206 行 · 第 1/3 页
S
1,206 行
#### $Id: nb_kernel100_ia32_sse.s,v 1.1.2.2 2007/11/28 09:36:32 spoel Exp $#### Gromacs 4.0 Copyright (c) 1991-2003 ## David van der Spoel, Erik Lindahl#### This program is free software; you can redistribute it and/or## modify it under the terms of the GNU General Public License## as published by the Free Software Foundation; either version 2## of the License, or (at your option) any later version.#### To help us fund GROMACS development, we humbly ask that you cite## the research papers on the package. Check out http://www.gromacs.org## ## And Hey:## Gnomes, ROck Monsters And Chili Sauce##.globl nb_kernel100_ia32_sse.globl _nb_kernel100_ia32_ssenb_kernel100_ia32_sse: _nb_kernel100_ia32_sse: .set nb100_p_nri, 8.set nb100_iinr, 12.set nb100_jindex, 16.set nb100_jjnr, 20.set nb100_shift, 24.set nb100_shiftvec, 28.set nb100_fshift, 32.set nb100_gid, 36.set nb100_pos, 40.set nb100_faction, 44.set nb100_charge, 48.set nb100_p_facel, 52.set nb100_p_krf, 56.set nb100_p_crf, 60.set nb100_Vc, 64.set nb100_type, 68.set nb100_p_ntype, 72.set nb100_vdwparam, 76.set nb100_Vvdw, 80.set nb100_p_tabscale, 84.set nb100_VFtab, 88.set nb100_invsqrta, 92.set nb100_dvda, 96.set nb100_p_gbtabscale, 100.set nb100_GBtab, 104.set nb100_p_nthreads, 108.set nb100_count, 112.set nb100_mtx, 116.set nb100_outeriter, 120.set nb100_inneriter, 124.set nb100_work, 128 ## stack offsets for local variables ## bottom of stack is cache-aligned for sse use .set nb100_ix, 0.set nb100_iy, 16.set nb100_iz, 32.set nb100_iq, 48.set nb100_dx, 64.set nb100_dy, 80.set nb100_dz, 96.set nb100_vctot, 112.set nb100_fix, 128.set nb100_fiy, 144.set nb100_fiz, 160.set nb100_half, 176.set nb100_three, 192.set nb100_is3, 208.set nb100_ii3, 212.set nb100_innerjjnr, 216.set nb100_innerk, 220.set nb100_n, 224.set nb100_nn1, 228.set nb100_nri, 232.set nb100_facel, 236.set nb100_nouter, 240.set nb100_ninner, 244.set nb100_salign, 248 pushl %ebp movl %esp,%ebp pushl %eax pushl %ebx pushl %ecx pushl %edx pushl %esi pushl %edi subl $252,%esp ## local stack space movl %esp,%eax andl $0xf,%eax subl %eax,%esp movl %eax,nb100_salign(%esp) emms ## Move args passed by reference to stack movl nb100_p_nri(%ebp),%ecx movl nb100_p_facel(%ebp),%esi movl (%ecx),%ecx movl (%esi),%esi movl %ecx,nb100_nri(%esp) movl %esi,nb100_facel(%esp) ## zero iteration counters movl $0,%eax movl %eax,nb100_nouter(%esp) movl %eax,nb100_ninner(%esp) ## create constant floating-point factors on stack movl $0x3f000000,%eax ## half in IEEE (hex) movl %eax,nb100_half(%esp) movss nb100_half(%esp),%xmm1 shufps $0,%xmm1,%xmm1 ## splat to all elements movaps %xmm1,%xmm2 addps %xmm2,%xmm2 ## one movaps %xmm2,%xmm3 addps %xmm2,%xmm2 ## two addps %xmm2,%xmm3 ## three movaps %xmm1,nb100_half(%esp) movaps %xmm3,nb100_three(%esp)_nb_kernel100_ia32_sse.nb100_threadloop: movl nb100_count(%ebp),%esi ## pointer to sync counter movl (%esi),%eax_nb_kernel100_ia32_sse.nb100_spinlock: movl %eax,%ebx ## ebx=*count=nn0 addl $1,%ebx ## ebx=nn1=nn0+10 lock cmpxchgl %ebx,(%esi) ## write nn1 to *counter, ## if it hasnt changed. ## or reread *counter to eax. pause ## -> better p4 performance jnz _nb_kernel100_ia32_sse.nb100_spinlock ## if(nn1>nri) nn1=nri movl nb100_nri(%esp),%ecx movl %ecx,%edx subl %ebx,%ecx cmovlel %edx,%ebx ## if(nn1>nri) nn1=nri ## Cleared the spinlock if we got here. ## eax contains nn0, ebx contains nn1. movl %eax,nb100_n(%esp) movl %ebx,nb100_nn1(%esp) subl %eax,%ebx ## calc number of outer lists movl %eax,%esi ## copy n to esi jg _nb_kernel100_ia32_sse.nb100_outerstart jmp _nb_kernel100_ia32_sse.nb100_end_nb_kernel100_ia32_sse.nb100_outerstart: ## ebx contains number of outer iterations addl nb100_nouter(%esp),%ebx movl %ebx,nb100_nouter(%esp)_nb_kernel100_ia32_sse.nb100_outer: movl nb100_shift(%ebp),%eax ## eax = pointer into shift[] movl (%eax,%esi,4),%ebx ## ebx=shift[n] leal (%ebx,%ebx,2),%ebx ## ebx=3*is movl %ebx,nb100_is3(%esp) ## store is3 movl nb100_shiftvec(%ebp),%eax ## eax = base of shiftvec[] movss (%eax,%ebx,4),%xmm0 movss 4(%eax,%ebx,4),%xmm1 movss 8(%eax,%ebx,4),%xmm2 movl nb100_iinr(%ebp),%ecx ## ecx = pointer into iinr[] movl (%ecx,%esi,4),%ebx ## ebx =ii movl nb100_charge(%ebp),%edx movss (%edx,%ebx,4),%xmm3 mulss nb100_facel(%esp),%xmm3 shufps $0,%xmm3,%xmm3 leal (%ebx,%ebx,2),%ebx ## ebx = 3*ii=ii3 movl nb100_pos(%ebp),%eax ## eax = base of pos[] addss (%eax,%ebx,4),%xmm0 addss 4(%eax,%ebx,4),%xmm1 addss 8(%eax,%ebx,4),%xmm2 movaps %xmm3,nb100_iq(%esp) shufps $0,%xmm0,%xmm0 shufps $0,%xmm1,%xmm1 shufps $0,%xmm2,%xmm2 movaps %xmm0,nb100_ix(%esp) movaps %xmm1,nb100_iy(%esp) movaps %xmm2,nb100_iz(%esp) movl %ebx,nb100_ii3(%esp) ## clear vctot and i forces xorps %xmm4,%xmm4 movaps %xmm4,nb100_vctot(%esp) movaps %xmm4,nb100_fix(%esp) movaps %xmm4,nb100_fiy(%esp) movaps %xmm4,nb100_fiz(%esp) movl nb100_jindex(%ebp),%eax movl (%eax,%esi,4),%ecx ## jindex[n] movl 4(%eax,%esi,4),%edx ## jindex[n+1] subl %ecx,%edx ## number of innerloop atoms movl nb100_pos(%ebp),%esi movl nb100_faction(%ebp),%edi movl nb100_jjnr(%ebp),%eax shll $2,%ecx addl %ecx,%eax movl %eax,nb100_innerjjnr(%esp) ## pointer to jjnr[nj0] movl %edx,%ecx subl $4,%edx addl nb100_ninner(%esp),%ecx movl %ecx,nb100_ninner(%esp) addl $0,%edx movl %edx,nb100_innerk(%esp) ## number of innerloop atoms jge _nb_kernel100_ia32_sse.nb100_unroll_loop jmp _nb_kernel100_ia32_sse.nb100_finish_inner_nb_kernel100_ia32_sse.nb100_unroll_loop: ## quad-unrolled innerloop here movl nb100_innerjjnr(%esp),%edx ## pointer to jjnr[k] movl (%edx),%eax movl 4(%edx),%ebx movl 8(%edx),%ecx movl 12(%edx),%edx ## eax-edx=jnr1-4 addl $16,nb100_innerjjnr(%esp) ## advance pointer (unrolled 4) movl nb100_charge(%ebp),%esi ## base of charge[] movss (%esi,%eax,4),%xmm3 movss (%esi,%ecx,4),%xmm4 movss (%esi,%ebx,4),%xmm6 movss (%esi,%edx,4),%xmm7 movaps nb100_iq(%esp),%xmm5 shufps $0,%xmm6,%xmm3 shufps $0,%xmm7,%xmm4 shufps $136,%xmm4,%xmm3 ## constant 10001000 movl nb100_pos(%ebp),%esi ## base of pos[] leal (%eax,%eax,2),%eax ## replace jnr with j3 leal (%ebx,%ebx,2),%ebx mulps %xmm5,%xmm3 leal (%ecx,%ecx,2),%ecx ## replace jnr with j3 leal (%edx,%edx,2),%edx ## move four coordinates to xmm0-xmm2 movlps (%esi,%eax,4),%xmm4 ## x1 y1 - - movlps (%esi,%ecx,4),%xmm5 ## x3 y3 - - movss 8(%esi,%eax,4),%xmm2 ## z1 - - - movss 8(%esi,%ecx,4),%xmm6 ## z3 - - - movhps (%esi,%ebx,4),%xmm4 ## x1 y1 x2 y2 movhps (%esi,%edx,4),%xmm5 ## x3 y3 x4 y4 movss 8(%esi,%ebx,4),%xmm0 ## z2 - - - movss 8(%esi,%edx,4),%xmm1 ## z4 - - - shufps $0,%xmm0,%xmm2 ## z1 z1 z2 z2 shufps $0,%xmm1,%xmm6 ## z3 z3 z4 z4 movaps %xmm4,%xmm0 ## x1 y1 x2 y2 movaps %xmm4,%xmm1 ## x1 y1 x2 y2 shufps $136,%xmm6,%xmm2 ## constant 10001000 ;# z1 z2 z3 z4 shufps $136,%xmm5,%xmm0 ## constant 10001000 ;# x1 x2 x3 x4 shufps $221,%xmm5,%xmm1 ## constant 11011101 ;# y1 y2 y3 y4 movl nb100_faction(%ebp),%edi ## move nb100_ix-iz to xmm4-xmm6 movaps nb100_ix(%esp),%xmm4 movaps nb100_iy(%esp),%xmm5 movaps nb100_iz(%esp),%xmm6 ## calc dr subps %xmm0,%xmm4 subps %xmm1,%xmm5 subps %xmm2,%xmm6 ## store dr movaps %xmm4,nb100_dx(%esp) movaps %xmm5,nb100_dy(%esp) movaps %xmm6,nb100_dz(%esp) ## square it mulps %xmm4,%xmm4 mulps %xmm5,%xmm5 mulps %xmm6,%xmm6 addps %xmm5,%xmm4 addps %xmm6,%xmm4 ## rsq in xmm4 rsqrtps %xmm4,%xmm5 ## lookup seed in xmm5 movaps %xmm5,%xmm2 mulps %xmm5,%xmm5 movaps nb100_three(%esp),%xmm1 mulps %xmm4,%xmm5 ## rsq*lu*lu movaps nb100_half(%esp),%xmm0 subps %xmm5,%xmm1 ## constant 30-rsq*lu*lu mulps %xmm2,%xmm1 mulps %xmm1,%xmm0 ## xmm0=rinv movaps %xmm0,%xmm4 mulps %xmm4,%xmm4 ## xmm4=rinvsq movaps nb100_vctot(%esp),%xmm5 mulps %xmm0,%xmm3 ## xmm3=vcoul mulps %xmm3,%xmm4 ## xmm4=fscal addps %xmm3,%xmm5 movaps nb100_dx(%esp),%xmm0 movaps nb100_dy(%esp),%xmm1 movaps nb100_dz(%esp),%xmm2 movaps %xmm5,nb100_vctot(%esp) mulps %xmm4,%xmm0 mulps %xmm4,%xmm1 mulps %xmm4,%xmm2 ## xmm0-xmm2 contains tx-tz (partial force) ## now update f_i movaps nb100_fix(%esp),%xmm3 movaps nb100_fiy(%esp),%xmm4 movaps nb100_fiz(%esp),%xmm5 addps %xmm0,%xmm3 addps %xmm1,%xmm4 addps %xmm2,%xmm5 movaps %xmm3,nb100_fix(%esp) movaps %xmm4,nb100_fiy(%esp) movaps %xmm5,nb100_fiz(%esp) ## the fj's - start by accumulating x & y forces from memory movlps (%edi,%eax,4),%xmm4 movlps (%edi,%ecx,4),%xmm6 movhps (%edi,%ebx,4),%xmm4 movhps (%edi,%edx,4),%xmm6 movaps %xmm4,%xmm3 shufps $136,%xmm6,%xmm3 ## constant 10001000 shufps $221,%xmm6,%xmm4 ## constant 11011101 ## now xmm3-xmm5 contains fjx, fjy, fjz subps %xmm0,%xmm3 subps %xmm1,%xmm4 ## unpack them back so we can store them - first x & y in xmm3/xmm4 movaps %xmm3,%xmm6 unpcklps %xmm4,%xmm6 unpckhps %xmm4,%xmm3 ## xmm6(l)=x & y for j1, (h) for j2 ## xmm3(l)=x & y for j3, (h) for j4 movlps %xmm6,(%edi,%eax,4) movlps %xmm3,(%edi,%ecx,4) movhps %xmm6,(%edi,%ebx,4) movhps %xmm3,(%edi,%edx,4) ## and the z forces movss 8(%edi,%eax,4),%xmm4 movss 8(%edi,%ebx,4),%xmm5 movss 8(%edi,%ecx,4),%xmm6 movss 8(%edi,%edx,4),%xmm7 subss %xmm2,%xmm4 shufps $229,%xmm2,%xmm2 ## constant 11100101 subss %xmm2,%xmm5 shufps $234,%xmm2,%xmm2 ## constant 11101010 subss %xmm2,%xmm6 shufps $255,%xmm2,%xmm2 ## constant 11111111 subss %xmm2,%xmm7 movss %xmm4,8(%edi,%eax,4) movss %xmm5,8(%edi,%ebx,4) movss %xmm6,8(%edi,%ecx,4) movss %xmm7,8(%edi,%edx,4) ## should we do one more iteration? subl $4,nb100_innerk(%esp) jl _nb_kernel100_ia32_sse.nb100_finish_inner jmp _nb_kernel100_ia32_sse.nb100_unroll_loop_nb_kernel100_ia32_sse.nb100_finish_inner: ## check if at least two particles remain addl $4,nb100_innerk(%esp) movl nb100_innerk(%esp),%edx andl $2,%edx jnz _nb_kernel100_ia32_sse.nb100_dopair jmp _nb_kernel100_ia32_sse.nb100_checksingle_nb_kernel100_ia32_sse.nb100_dopair: movl nb100_charge(%ebp),%esi movl nb100_pos(%ebp),%edi movl nb100_innerjjnr(%esp),%ecx movl (%ecx),%eax movl 4(%ecx),%ebx addl $8,nb100_innerjjnr(%esp)⌨️ 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?