nb_kernel030_ia32_3dnow.intel_syntax.s
来自「最著名最快的分子模拟软件」· S 代码 · 共 1,107 行 · 第 1/3 页
S
1,107 行
;#;# $Id: nb_kernel030_ia32_3dnow.intel_syntax.s,v 1.1.2.1 2006/03/01 15:18:28 lindahl 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;#;# These files require GNU binutils 2.10 or later, since we;# use intel syntax for portability, or a recent version ;# of NASM that understands Extended 3DNow and SSE2 instructions.;# (NASM is normally only used with MS Visual C++).;# Since NASM and gnu as disagree on some definitions and use ;# completely different preprocessing options I have to introduce a;# trick: NASM uses ';' for comments, while gnu as uses '#' on x86.;# Gnu as treats ';' as a line break, i.e. ignores it. This is the;# reason why all comments need both symbols...;# The source is written for GNU as, with intel syntax. When you use;# NASM we redefine a couple of things. The false if-statement around ;# the following code is seen by GNU as, but NASM doesn't see it, so ;# the code inside is read by NASM but not gcc.; .if 0 # block below only read by NASM%define .section section%define .long dd%define .align align%define .globl global;# NASM only wants 'dword', not 'dword ptr'.%define ptr%macro .equiv 2 %1 equ %2%endmacro; .endif # End of NASM-specific block; .intel_syntax noprefix # Line only read by gnu as.globl nb_kernel030_ia32_3dnow.globl _nb_kernel030_ia32_3dnownb_kernel030_ia32_3dnow: _nb_kernel030_ia32_3dnow: .equiv nb030_p_nri, 8.equiv nb030_iinr, 12.equiv nb030_jindex, 16.equiv nb030_jjnr, 20.equiv nb030_shift, 24.equiv nb030_shiftvec, 28.equiv nb030_fshift, 32.equiv nb030_gid, 36.equiv nb030_pos, 40 .equiv nb030_faction, 44.equiv nb030_charge, 48.equiv nb030_p_facel, 52.equiv nb030_p_krf, 56 .equiv nb030_p_crf, 60 .equiv nb030_Vc, 64 .equiv nb030_type, 68.equiv nb030_p_ntype, 72.equiv nb030_vdwparam, 76 .equiv nb030_Vvdw, 80 .equiv nb030_p_tabscale, 84 .equiv nb030_VFtab, 88.equiv nb030_invsqrta, 92 .equiv nb030_dvda, 96.equiv nb030_p_gbtabscale, 100.equiv nb030_GBtab, 104.equiv nb030_p_nthreads, 108.equiv nb030_count, 112.equiv nb030_mtx, 116.equiv nb030_outeriter, 120.equiv nb030_inneriter, 124.equiv nb030_work, 128 ;# stack offsets for local variables .equiv nb030_is3, 0.equiv nb030_ii3, 4.equiv nb030_ix, 8.equiv nb030_iy, 12.equiv nb030_iz, 16.equiv nb030_Vvdwtot, 20 .equiv nb030_c6, 28 .equiv nb030_c12, 36 .equiv nb030_two, 44 .equiv nb030_n1, 52 .equiv nb030_tsc, 60 .equiv nb030_ntia, 68.equiv nb030_innerjjnr, 72.equiv nb030_innerk, 76 .equiv nb030_fix, 80.equiv nb030_fiy, 84.equiv nb030_fiz, 88.equiv nb030_dx1, 92.equiv nb030_dy1, 96.equiv nb030_dz1, 100.equiv nb030_dx2, 104.equiv nb030_dy2, 108.equiv nb030_dz2, 112 .equiv nb030_n, 116 ;# idx for outer loop.equiv nb030_nn1, 120 ;# number of outer iterations.equiv nb030_nri, 124.equiv nb030_ntype, 128.equiv nb030_nouter, 132.equiv nb030_ninner, 136 push ebp mov ebp,esp push eax push ebx push ecx push edx push esi push edi sub esp, 140 ;# local stack space femms ;# move data to local stack mov eax, 0x40000000 mov [esp + nb030_two], eax mov [esp + nb030_two+4], eax mov ecx, [ebp + nb030_p_nri] mov edx, [ebp + nb030_p_ntype] mov esi, [ebp + nb030_p_tabscale] mov ecx, [ecx] mov edx, [edx] mov [esp + nb030_nri], ecx mov [esp + nb030_ntype], edx movd mm3, [esi] punpckldq mm3,mm3 movq [esp + nb030_tsc], mm3 ;# zero iteration counters mov eax, 0 mov [esp + nb030_nouter], eax mov [esp + nb030_ninner], eax.nb030_threadloop: mov esi, [ebp + nb030_count] ;# pointer to sync counter mov eax, [esi].nb030_spinlock: mov ebx, eax ;# ebx=*count=nn0 add ebx, 1 ;# ebx=nn1=nn0+10 lock cmpxchg [esi], ebx ;# write nn1 to *counter, ;# if it hasnt changed. ;# or reread *counter to eax. pause ;# -> better p4 performance jnz .nb030_spinlock ;# if(nn1>nri) nn1=nri mov ecx, [esp + nb030_nri] mov edx, ecx sub ecx, ebx cmovle ebx, edx ;# if(nn1>nri) nn1=nri ;# Cleared the spinlock if we got here. ;# eax contains nn0, ebx contains nn1. mov [esp + nb030_n], eax mov [esp + nb030_nn1], ebx sub ebx, eax ;# calc number of outer lists mov esi, eax ;# copy n to esi jg .nb030_outerstart jmp .nb030_end.nb030_outerstart: ;# ebx contains number of outer iterations add ebx, [esp + nb030_nouter] mov [esp + nb030_nouter], ebx .nb030_outer: mov eax, [ebp + nb030_shift] ;# eax = pointer into shift[] mov ebx, [eax + esi*4] ;# ebx=shift[n] lea ebx, [ebx + ebx*2] ;# ebx=3*is mov [esp + nb030_is3],ebx ;# store is3 mov eax, [ebp + nb030_shiftvec] ;# eax = base of shiftvec[] movq mm0, [eax + ebx*4] ;# move shX/shY to mm0 and shZ to mm1 movd mm1, [eax + ebx*4 + 8] mov ecx, [ebp + nb030_iinr] ;# ecx = pointer into iinr[] mov ebx, [ecx+esi*4] ;# ebx=ii mov edx, [ebp + nb030_type] mov edx, [edx + ebx*4] imul edx, [esp + nb030_ntype] shl edx, 1 mov [esp + nb030_ntia], edx lea ebx, [ebx + ebx*2] ;# ebx = 3*ii=ii3 mov eax, [ebp + nb030_pos] ;# eax = base of pos[] pfadd mm0, [eax + ebx*4] ;# ix = shX + posX (and iy too) movd mm3, [eax + ebx*4 + 8] ;# cant use direct memory add for 4 bytes (iz) mov [esp + nb030_ii3], ebx pfadd mm1, mm3 movq [esp + nb030_ix], mm0 movd [esp + nb030_iz], mm1 ;# clear total potential and i forces pxor mm7,mm7 movq [esp + nb030_Vvdwtot], mm7 movq [esp + nb030_fix], mm7 movd [esp + nb030_fiz], mm7 mov eax, [ebp + nb030_jindex] mov ecx, [eax + esi*4] ;# jindex[n] mov edx, [eax + esi*4 + 4] ;# jindex[n+1] sub edx, ecx ;# number of innerloop atoms mov esi, [ebp + nb030_pos] mov edi, [ebp + nb030_faction] mov eax, [ebp + nb030_jjnr] shl ecx, 2 add eax, ecx mov [esp + nb030_innerjjnr], eax ;# pointer to jjnr[nj0] mov ecx, edx sub edx, 2 add ecx, [esp + nb030_ninner] mov [esp + nb030_ninner], ecx mov [esp + nb030_innerk], edx ;# number of innerloop atoms add edx, 0 jge .nb030_unroll_loop jmp .nb030_finish_inner.nb030_unroll_loop: ;# paired innerloop starts here mov ecx, [esp + nb030_innerjjnr] ;# pointer to jjnr[k] mov eax, [ecx] mov ebx, [ecx + 4] ;# eax/ebx=jnr add dword ptr [esp + nb030_innerjjnr], 8 ;# advance pointer (unrolled 2) prefetch [ecx + 16] ;# prefetch data - trial and error says 16 is best mov ecx, [ebp + nb030_type] mov edx, [ecx + eax*4] ;# type [jnr1] mov ecx, [ecx + ebx*4] ;# type [jnr2] mov esi, [ebp + nb030_vdwparam] ;# base of vdwparam shl edx, 1 shl ecx, 1 add edx, [esp + nb030_ntia] ;# tja = ntia + 2*type add ecx, [esp + nb030_ntia] movq mm5, [esi + edx*4] ;# mm5 = 1st c6 / c12 movq mm7, [esi + ecx*4] ;# mm7 = 2nd c6 / c12 movq mm6,mm5 punpckldq mm5,mm7 ;# mm5 = 1st c6 / 2nd c6 punpckhdq mm6,mm7 ;# mm6 = 1st c12 / 2nd c12 movq [esp + nb030_c6], mm5 movq [esp + nb030_c12], mm6 lea eax, [eax + eax*2] ;# replace jnr with j3 lea ebx, [ebx + ebx*2] mov esi, [ebp + nb030_pos] movq mm0, [esp + nb030_ix] movd mm1, [esp + nb030_iz] movq mm4, [esi + eax*4] ;# fetch first j coordinates movd mm5, [esi + eax*4 + 8] pfsubr mm4,mm0 ;# dr = ir - jr pfsubr mm5,mm1 movq [esp + nb030_dx1], mm4 ;# store dr movd [esp + nb030_dz1], mm5 pfmul mm4,mm4 ;# square dx,dy,dz pfmul mm5,mm5 pfacc mm4, mm5 ;# accumulate to get dx*dx+ dy*dy+ dz*dz pfacc mm4, mm5 ;# first rsq in lower mm4 movq mm6, [esi + ebx*4] ;# fetch second j coordinates movd mm7, [esi + ebx*4 + 8] pfsubr mm6,mm0 ;# dr = ir - jr pfsubr mm7,mm1 movq [esp + nb030_dx2], mm6 ;# store dr movd [esp + nb030_dz2], mm7 pfmul mm6,mm6 ;# square dx,dy,dz pfmul mm7,mm7 pfacc mm6, mm7 ;# accumulate to get dx*dx+ dy*dy+ dz*dz pfacc mm6, mm7 ;# second rsq in lower mm6 pfrsqrt mm0, mm4 ;# lookup inverse square root seed pfrsqrt mm1, mm6 punpckldq mm0,mm1 punpckldq mm4,mm6 ;# now 4 has rsq and 0 the seed for both pairs movq mm2,mm0 ;# amd 3dnow N-R iteration to get full precision pfmul mm0,mm0 pfrsqit1 mm0,mm4 pfrcpit2 mm0,mm2 pfmul mm4, mm0 movq mm1, mm4 ;# mm0 is invsqrt, and mm1 r ;# do potential and fscal pfmul mm1, [esp + nb030_tsc] ;# mm1=rt pf2iw mm4,mm1 movq [esp + nb030_n1], mm4 pi2fd mm4,mm4 pfsub mm1, mm4 ;# now mm1 is eps and mm4 is n0 movq mm2,mm1 pfmul mm2,mm2 ;# mm1 is eps, mm2 is eps2 mov edx, [ebp + nb030_VFtab] ;# dispersion table mov ecx, [esp + nb030_n1] shl ecx, 3 ;# load all the table values we need movd mm4, [edx + ecx*4] movd mm5, [edx + ecx*4 + 4] movd mm6, [edx + ecx*4 + 8] movd mm7, [edx + ecx*4 + 12] mov ecx, [esp + nb030_n1+4] shl ecx, 3 punpckldq mm4, [edx + ecx*4] punpckldq mm5, [edx + ecx*4 + 4] punpckldq mm6, [edx + ecx*4 + 8] punpckldq mm7, [edx + ecx*4 + 12] pfmul mm6, mm1 ;# mm6 = Geps pfmul mm7, mm2 ;# mm7 = Heps2 pfadd mm5, mm6 pfadd mm5, mm7 ;# mm5 = Fp pfmul mm7, [esp + nb030_two] ;# two*Heps2 pfadd mm7, mm6 pfadd mm7, mm5 ;# mm7=FF pfmul mm5, mm1 ;# mm5=eps*Fp pfadd mm5, mm4 ;# mm5= VV movq mm4, [esp + nb030_c6] pfmul mm7, mm4 ;# fijD pfmul mm5, mm4 ;# Vvdw6 movq mm3, mm7 ;# add to fscal ;# update Vvdwtot to release mm5! pfadd mm5, [esp + nb030_Vvdwtot] ;# add the earlier value movq [esp + nb030_Vvdwtot], mm5 ;# store the sum ;# repulsion table mov ecx, [esp + nb030_n1] shl ecx, 3 ;# load all the table values we need movd mm4, [edx + ecx*4 + 16] movd mm5, [edx + ecx*4 + 20] movd mm6, [edx + ecx*4 + 24] movd mm7, [edx + ecx*4 + 28] mov ecx, [esp + nb030_n1+4] shl ecx, 3 punpckldq mm4, [edx + ecx*4 + 16] punpckldq mm5, [edx + ecx*4 + 20] punpckldq mm6, [edx + ecx*4 + 24] punpckldq mm7, [edx + ecx*4 + 28] pfmul mm6, mm1 ;# mm6 = Geps pfmul mm7, mm2 ;# mm7 = Heps2 pfadd mm5, mm6 pfadd mm5, mm7 ;# mm5 = Fp pfmul mm7, [esp + nb030_two] ;# two*Heps2 pfadd mm7, mm6
⌨️ 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?