nb_kernel103_ia32_sse2.s
来自「最著名最快的分子模拟软件」· S 代码 · 共 1,641 行 · 第 1/4 页
S
1,641 行
#### $Id: nb_kernel103_ia32_sse2.s,v 1.1.2.2 2007/11/28 18:35:33 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_kernel103_ia32_sse2.globl _nb_kernel103_ia32_sse2nb_kernel103_ia32_sse2: _nb_kernel103_ia32_sse2: .set nb103_p_nri, 8.set nb103_iinr, 12.set nb103_jindex, 16.set nb103_jjnr, 20.set nb103_shift, 24.set nb103_shiftvec, 28.set nb103_fshift, 32.set nb103_gid, 36.set nb103_pos, 40.set nb103_faction, 44.set nb103_charge, 48.set nb103_p_facel, 52.set nb103_argkrf, 56.set nb103_argcrf, 60.set nb103_Vc, 64.set nb103_type, 68.set nb103_p_ntype, 72.set nb103_vdwparam, 76.set nb103_Vvdw, 80.set nb103_p_tabscale, 84.set nb103_VFtab, 88.set nb103_invsqrta, 92.set nb103_dvda, 96.set nb103_p_gbtabscale, 100.set nb103_GBtab, 104.set nb103_p_nthreads, 108.set nb103_count, 112.set nb103_mtx, 116.set nb103_outeriter, 120.set nb103_inneriter, 124.set nb103_work, 128 ## stack offsets for local variables ## bottom of stack is cache-aligned for sse2 use .set nb103_ixH1, 0.set nb103_iyH1, 16.set nb103_izH1, 32.set nb103_ixH2, 48.set nb103_iyH2, 64.set nb103_izH2, 80.set nb103_ixM, 96.set nb103_iyM, 112.set nb103_izM, 128.set nb103_iqM, 144.set nb103_iqH, 160.set nb103_dxH1, 176.set nb103_dyH1, 192.set nb103_dzH1, 208.set nb103_dxH2, 224.set nb103_dyH2, 240.set nb103_dzH2, 256.set nb103_dxM, 272.set nb103_dyM, 288.set nb103_dzM, 304.set nb103_qqM, 320.set nb103_qqH, 336.set nb103_vctot, 352.set nb103_fixM, 368.set nb103_fiyM, 384.set nb103_fizM, 400.set nb103_fixH1, 416.set nb103_fiyH1, 432.set nb103_fizH1, 448.set nb103_fixH2, 464.set nb103_fiyH2, 480.set nb103_fizH2, 496.set nb103_fjx, 512.set nb103_fjy, 528.set nb103_fjz, 544.set nb103_half, 560.set nb103_three, 576.set nb103_is3, 592.set nb103_ii3, 596.set nb103_innerjjnr, 600.set nb103_innerk, 604.set nb103_n, 608.set nb103_nn1, 612.set nb103_nri, 616.set nb103_nouter, 620.set nb103_ninner, 624.set nb103_salign, 628 pushl %ebp movl %esp,%ebp pushl %eax pushl %ebx pushl %ecx pushl %edx pushl %esi pushl %edi subl $632,%esp ## local stack space movl %esp,%eax andl $0xf,%eax subl %eax,%esp movl %eax,nb103_salign(%esp) emms ## Move args passed by reference to stack movl nb103_p_nri(%ebp),%ecx movl (%ecx),%ecx movl %ecx,nb103_nri(%esp) ## zero iteration counters movl $0,%eax movl %eax,nb103_nouter(%esp) movl %eax,nb103_ninner(%esp) ## create constant floating-point factors on stack movl $0x00000000,%eax ## lower half of double 0.5 IEEE (hex) movl $0x3fe00000,%ebx movl %eax,nb103_half(%esp) movl %ebx,nb103_half+4(%esp) movsd nb103_half(%esp),%xmm1 shufpd $0,%xmm1,%xmm1 ## splat to all elements movapd %xmm1,%xmm3 addpd %xmm3,%xmm3 ## 1.0 movapd %xmm3,%xmm2 addpd %xmm2,%xmm2 ## 2.0 addpd %xmm2,%xmm3 ## 3.0 movapd %xmm1,nb103_half(%esp) movapd %xmm3,nb103_three(%esp) ## assume we have at least one i particle - start directly movl nb103_iinr(%ebp),%ecx ## ecx = pointer into iinr[] movl (%ecx),%ebx ## ebx =ii movl nb103_charge(%ebp),%edx movsd 8(%edx,%ebx,8),%xmm3 movsd 24(%edx,%ebx,8),%xmm4 movl nb103_p_facel(%ebp),%esi movsd (%esi),%xmm5 mulsd %xmm5,%xmm3 mulsd %xmm5,%xmm4 shufpd $0,%xmm3,%xmm3 shufpd $0,%xmm4,%xmm4 movapd %xmm3,nb103_iqH(%esp) movapd %xmm4,nb103_iqM(%esp)_nb_kernel103_ia32_sse2.nb103_threadloop: movl nb103_count(%ebp),%esi ## pointer to sync counter movl (%esi),%eax_nb_kernel103_ia32_sse2.nb103_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_kernel103_ia32_sse2.nb103_spinlock ## if(nn1>nri) nn1=nri movl nb103_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,nb103_n(%esp) movl %ebx,nb103_nn1(%esp) subl %eax,%ebx ## calc number of outer lists movl %eax,%esi ## copy n to esi jg _nb_kernel103_ia32_sse2.nb103_outerstart jmp _nb_kernel103_ia32_sse2.nb103_end_nb_kernel103_ia32_sse2.nb103_outerstart: ## ebx contains number of outer iterations addl nb103_nouter(%esp),%ebx movl %ebx,nb103_nouter(%esp)_nb_kernel103_ia32_sse2.nb103_outer: movl nb103_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,nb103_is3(%esp) ## store is3 movl nb103_shiftvec(%ebp),%eax ## eax = base of shiftvec[] movsd (%eax,%ebx,8),%xmm0 movsd 8(%eax,%ebx,8),%xmm1 movsd 16(%eax,%ebx,8),%xmm2 movl nb103_iinr(%ebp),%ecx ## ecx = pointer into iinr[] movl (%ecx,%esi,4),%ebx ## ebx =ii movapd %xmm0,%xmm3 movapd %xmm1,%xmm4 movapd %xmm2,%xmm5 leal (%ebx,%ebx,2),%ebx ## ebx = 3*ii=ii3 movl nb103_pos(%ebp),%eax ## eax = base of pos[] movl %ebx,nb103_ii3(%esp) addsd 24(%eax,%ebx,8),%xmm3 addsd 32(%eax,%ebx,8),%xmm4 addsd 40(%eax,%ebx,8),%xmm5 shufpd $0,%xmm3,%xmm3 shufpd $0,%xmm4,%xmm4 shufpd $0,%xmm5,%xmm5 movapd %xmm3,nb103_ixH1(%esp) movapd %xmm4,nb103_iyH1(%esp) movapd %xmm5,nb103_izH1(%esp) movsd %xmm0,%xmm3 movsd %xmm1,%xmm4 movsd %xmm2,%xmm5 addsd 48(%eax,%ebx,8),%xmm0 addsd 56(%eax,%ebx,8),%xmm1 addsd 64(%eax,%ebx,8),%xmm2 addsd 72(%eax,%ebx,8),%xmm3 addsd 80(%eax,%ebx,8),%xmm4 addsd 88(%eax,%ebx,8),%xmm5 shufpd $0,%xmm0,%xmm0 shufpd $0,%xmm1,%xmm1 shufpd $0,%xmm2,%xmm2 shufpd $0,%xmm3,%xmm3 shufpd $0,%xmm4,%xmm4 shufpd $0,%xmm5,%xmm5 movapd %xmm0,nb103_ixH2(%esp) movapd %xmm1,nb103_iyH2(%esp) movapd %xmm2,nb103_izH2(%esp) movapd %xmm3,nb103_ixM(%esp) movapd %xmm4,nb103_iyM(%esp) movapd %xmm5,nb103_izM(%esp) ## clear vctot and i forces xorpd %xmm4,%xmm4 movapd %xmm4,nb103_vctot(%esp) movapd %xmm4,nb103_fixM(%esp) movapd %xmm4,nb103_fiyM(%esp) movapd %xmm4,nb103_fizM(%esp) movapd %xmm4,nb103_fixH1(%esp) movapd %xmm4,nb103_fiyH1(%esp) movapd %xmm4,nb103_fizH1(%esp) movapd %xmm4,nb103_fixH2(%esp) movapd %xmm4,nb103_fiyH2(%esp) movapd %xmm4,nb103_fizH2(%esp) movl nb103_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 nb103_pos(%ebp),%esi movl nb103_faction(%ebp),%edi movl nb103_jjnr(%ebp),%eax shll $2,%ecx addl %ecx,%eax movl %eax,nb103_innerjjnr(%esp) ## pointer to jjnr[nj0] movl %edx,%ecx subl $2,%edx addl nb103_ninner(%esp),%ecx movl %ecx,nb103_ninner(%esp) addl $0,%edx movl %edx,nb103_innerk(%esp) ## number of innerloop atoms jge _nb_kernel103_ia32_sse2.nb103_unroll_loop jmp _nb_kernel103_ia32_sse2.nb103_checksingle_nb_kernel103_ia32_sse2.nb103_unroll_loop: ## twice unrolled innerloop here movl nb103_innerjjnr(%esp),%edx ## pointer to jjnr[k] movl (%edx),%eax movl 4(%edx),%ebx addl $8,nb103_innerjjnr(%esp) ## advance pointer (unrolled 2) movl nb103_charge(%ebp),%esi ## base of charge[] movlpd (%esi,%eax,8),%xmm6 ## jq A movhpd (%esi,%ebx,8),%xmm6 ## jq B movapd nb103_iqM(%esp),%xmm3 movapd nb103_iqH(%esp),%xmm4 mulpd %xmm6,%xmm3 ## qqM mulpd %xmm6,%xmm4 ## qqH movapd %xmm3,nb103_qqM(%esp) movapd %xmm4,nb103_qqH(%esp) movl nb103_pos(%ebp),%esi ## base of pos[] leal (%eax,%eax,2),%eax ## replace jnr with j3 leal (%ebx,%ebx,2),%ebx ## move two coordinates to xmm0-xmm2 movlpd (%esi,%eax,8),%xmm0 movlpd 8(%esi,%eax,8),%xmm1 movlpd 16(%esi,%eax,8),%xmm2 movhpd (%esi,%ebx,8),%xmm0 movhpd 8(%esi,%ebx,8),%xmm1 movhpd 16(%esi,%ebx,8),%xmm2 ## move ixM-izM to xmm4-xmm6 movapd nb103_ixM(%esp),%xmm4 movapd nb103_iyM(%esp),%xmm5 movapd nb103_izM(%esp),%xmm6 ## calc dr subpd %xmm0,%xmm4 subpd %xmm1,%xmm5 subpd %xmm2,%xmm6 ## store dr movapd %xmm4,nb103_dxM(%esp) movapd %xmm5,nb103_dyM(%esp) movapd %xmm6,nb103_dzM(%esp) ## square it mulpd %xmm4,%xmm4 mulpd %xmm5,%xmm5 mulpd %xmm6,%xmm6 addpd %xmm5,%xmm4 addpd %xmm6,%xmm4 movapd %xmm4,%xmm7 ## rsqO in xmm7 ## move ixH1-izH1 to xmm4-xmm6 movapd nb103_ixH1(%esp),%xmm4 movapd nb103_iyH1(%esp),%xmm5 movapd nb103_izH1(%esp),%xmm6 ## calc dr subpd %xmm0,%xmm4 subpd %xmm1,%xmm5 subpd %xmm2,%xmm6 ## store dr movapd %xmm4,nb103_dxH1(%esp) movapd %xmm5,nb103_dyH1(%esp) movapd %xmm6,nb103_dzH1(%esp) ## square it mulpd %xmm4,%xmm4 mulpd %xmm5,%xmm5 mulpd %xmm6,%xmm6 addpd %xmm5,%xmm6 addpd %xmm4,%xmm6 ## rsqH1 in xmm6 ## move ixH2-izH2 to xmm3-xmm5 movapd nb103_ixH2(%esp),%xmm3 movapd nb103_iyH2(%esp),%xmm4 movapd nb103_izH2(%esp),%xmm5 ## calc dr subpd %xmm0,%xmm3 subpd %xmm1,%xmm4 subpd %xmm2,%xmm5 ## store dr movapd %xmm3,nb103_dxH2(%esp) movapd %xmm4,nb103_dyH2(%esp) movapd %xmm5,nb103_dzH2(%esp) ## square it mulpd %xmm3,%xmm3 mulpd %xmm4,%xmm4 mulpd %xmm5,%xmm5 addpd %xmm4,%xmm5 addpd %xmm3,%xmm5 ## rsqH2 in xmm5, rsqH1 in xmm6, rsqO in xmm7 ## start with rsqM - put seed in xmm2 cvtpd2ps %xmm7,%xmm2 rsqrtps %xmm2,%xmm2 cvtps2pd %xmm2,%xmm2 movapd %xmm2,%xmm3 mulpd %xmm2,%xmm2 movapd nb103_three(%esp),%xmm4 mulpd %xmm7,%xmm2 ## rsq*lu*lu subpd %xmm2,%xmm4 ## 30-rsq*lu*lu mulpd %xmm3,%xmm4 ## lu*(3-rsq*lu*lu) mulpd nb103_half(%esp),%xmm4 ## iter1 ( new lu) movapd %xmm4,%xmm3 mulpd %xmm4,%xmm4 ## lu*lu mulpd %xmm4,%xmm7 ## rsq*lu*lu movapd nb103_three(%esp),%xmm4 subpd %xmm7,%xmm4 ## 3-rsq*lu*lu mulpd %xmm3,%xmm4 ## lu*( 3-rsq*lu*lu) mulpd nb103_half(%esp),%xmm4 ## rinv movapd %xmm4,%xmm7 ## rinvM in xmm7 ## rsqH1 - seed in xmm2 cvtpd2ps %xmm6,%xmm2 rsqrtps %xmm2,%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?