nb_kernel210_x86_64_sse2.s
来自「最著名最快的分子模拟软件」· S 代码 · 共 1,159 行 · 第 1/3 页
S
1,159 行
#### $Id: nb_kernel210_x86_64_sse2.s,v 1.4.2.3 2006/09/22 08:40:34 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##.globl nb_kernel210_x86_64_sse2.globl _nb_kernel210_x86_64_sse2nb_kernel210_x86_64_sse2: _nb_kernel210_x86_64_sse2: ## Room for return address and rbp (16 bytes).set nb210_fshift, 16.set nb210_gid, 24.set nb210_pos, 32.set nb210_faction, 40.set nb210_charge, 48.set nb210_p_facel, 56.set nb210_argkrf, 64.set nb210_argcrf, 72.set nb210_Vc, 80.set nb210_type, 88.set nb210_p_ntype, 96.set nb210_vdwparam, 104.set nb210_Vvdw, 112.set nb210_p_tabscale, 120.set nb210_VFtab, 128.set nb210_invsqrta, 136.set nb210_dvda, 144.set nb210_p_gbtabscale, 152.set nb210_GBtab, 160.set nb210_p_nthreads, 168.set nb210_count, 176.set nb210_mtx, 184.set nb210_outeriter, 192.set nb210_inneriter, 200.set nb210_work, 208 ## stack offsets for local variables ## bottom of stack is cache-aligned for sse2 use .set nb210_ix, 0.set nb210_iy, 16.set nb210_iz, 32.set nb210_iq, 48.set nb210_dx, 64.set nb210_dy, 80.set nb210_dz, 96.set nb210_c6, 112.set nb210_c12, 128.set nb210_six, 144.set nb210_twelve, 160.set nb210_vctot, 176.set nb210_Vvdwtot, 192.set nb210_fix, 208.set nb210_fiy, 224.set nb210_fiz, 240.set nb210_half, 256.set nb210_three, 272.set nb210_two, 288.set nb210_krf, 304.set nb210_crf, 320.set nb210_nri, 336.set nb210_iinr, 344.set nb210_jindex, 352.set nb210_jjnr, 360.set nb210_shift, 368.set nb210_shiftvec, 376.set nb210_facel, 384.set nb210_innerjjnr, 392.set nb210_is3, 400.set nb210_ii3, 404.set nb210_ntia, 408.set nb210_innerk, 412.set nb210_n, 416.set nb210_nn1, 420.set nb210_ntype, 424.set nb210_nouter, 428.set nb210_ninner, 432 push %rbp movq %rsp,%rbp push %rbx emms push %r12 push %r13 push %r14 push %r15 subq $456,%rsp ## local variable stack space (n*16+8) ## zero 32-bit iteration counters movl $0,%eax movl %eax,nb210_nouter(%rsp) movl %eax,nb210_ninner(%rsp) movl (%rdi),%edi movl %edi,nb210_nri(%rsp) movq %rsi,nb210_iinr(%rsp) movq %rdx,nb210_jindex(%rsp) movq %rcx,nb210_jjnr(%rsp) movq %r8,nb210_shift(%rsp) movq %r9,nb210_shiftvec(%rsp) movq nb210_p_ntype(%rbp),%rdi movl (%rdi),%edi movl %edi,nb210_ntype(%rsp) movq nb210_p_facel(%rbp),%rsi movsd (%rsi),%xmm0 movsd %xmm0,nb210_facel(%rsp) movq nb210_argkrf(%rbp),%rsi movq nb210_argcrf(%rbp),%rdi movsd (%rsi),%xmm1 movsd (%rdi),%xmm2 shufpd $0,%xmm1,%xmm1 shufpd $0,%xmm2,%xmm2 movapd %xmm1,nb210_krf(%rsp) movapd %xmm2,nb210_crf(%rsp) ## create constant floating-point factors on stack movl $0x00000000,%eax ## lower half of double half IEEE (hex) movl $0x3fe00000,%ebx movl %eax,nb210_half(%rsp) movl %ebx,nb210_half+4(%rsp) movsd nb210_half(%rsp),%xmm1 shufpd $0,%xmm1,%xmm1 ## splat to all elements movapd %xmm1,%xmm3 addpd %xmm3,%xmm3 ## one movapd %xmm3,%xmm2 addpd %xmm2,%xmm2 ## two addpd %xmm2,%xmm3 ## three movapd %xmm3,%xmm4 addpd %xmm4,%xmm4 ## six movapd %xmm4,%xmm5 addpd %xmm5,%xmm5 ## twelve movapd %xmm1,nb210_half(%rsp) movapd %xmm2,nb210_two(%rsp) movapd %xmm3,nb210_three(%rsp) movapd %xmm4,nb210_six(%rsp) movapd %xmm5,nb210_twelve(%rsp)_nb_kernel210_x86_64_sse2.nb210_threadloop: movq nb210_count(%rbp),%rsi ## pointer to sync counter movl (%rsi),%eax_nb_kernel210_x86_64_sse2.nb210_spinlock: movl %eax,%ebx ## ebx=*count=nn0 addl $1,%ebx ## ebx=nn1=nn0+10 lock cmpxchgl %ebx,(%rsi) ## write nn1 to *counter, ## if it hasnt changed. ## or reread *counter to eax. pause ## -> better p4 performance jnz _nb_kernel210_x86_64_sse2.nb210_spinlock ## if(nn1>nri) nn1=nri movl nb210_nri(%rsp),%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,nb210_n(%rsp) movl %ebx,nb210_nn1(%rsp) subl %eax,%ebx ## calc number of outer lists movl %eax,%esi ## copy n to esi jg _nb_kernel210_x86_64_sse2.nb210_outerstart jmp _nb_kernel210_x86_64_sse2.nb210_end_nb_kernel210_x86_64_sse2.nb210_outerstart: ## ebx contains number of outer iterations addl nb210_nouter(%rsp),%ebx movl %ebx,nb210_nouter(%rsp)_nb_kernel210_x86_64_sse2.nb210_outer: movq nb210_shift(%rsp),%rax ## rax = pointer into shift[] movl (%rax,%rsi,4),%ebx ## rbx=shift[n] lea (%rbx,%rbx,2),%rbx ## rbx=3*is movl %ebx,nb210_is3(%rsp) ## store is3 movq nb210_shiftvec(%rsp),%rax ## rax = base of shiftvec[] movsd (%rax,%rbx,8),%xmm0 movsd 8(%rax,%rbx,8),%xmm1 movsd 16(%rax,%rbx,8),%xmm2 movq nb210_iinr(%rsp),%rcx ## rcx = pointer into iinr[] movl (%rcx,%rsi,4),%ebx ## ebx =ii movq nb210_charge(%rbp),%rdx movsd (%rdx,%rbx,8),%xmm3 mulsd nb210_facel(%rsp),%xmm3 shufpd $0,%xmm3,%xmm3 movq nb210_type(%rbp),%rdx movl (%rdx,%rbx,4),%edx imull nb210_ntype(%rsp),%edx shll %edx movl %edx,nb210_ntia(%rsp) lea (%rbx,%rbx,2),%rbx ## rbx = 3*ii=ii3 movq nb210_pos(%rbp),%rax ## rax = base of pos[] addsd (%rax,%rbx,8),%xmm0 addsd 8(%rax,%rbx,8),%xmm1 addsd 16(%rax,%rbx,8),%xmm2 movapd %xmm3,nb210_iq(%rsp) shufpd $0,%xmm0,%xmm0 shufpd $0,%xmm1,%xmm1 shufpd $0,%xmm2,%xmm2 movapd %xmm0,nb210_ix(%rsp) movapd %xmm1,nb210_iy(%rsp) movapd %xmm2,nb210_iz(%rsp) movl %ebx,nb210_ii3(%rsp) ## clear vctot and i forces xorpd %xmm12,%xmm12 movapd %xmm12,%xmm13 movapd %xmm12,%xmm14 movapd %xmm12,%xmm15 movapd %xmm12,nb210_vctot(%rsp) movapd %xmm12,nb210_Vvdwtot(%rsp) movq nb210_jindex(%rsp),%rax movl (%rax,%rsi,4),%ecx ## jindex[n] movl 4(%rax,%rsi,4),%edx ## jindex[n+1] subl %ecx,%edx ## number of innerloop atoms movq nb210_pos(%rbp),%rsi movq nb210_faction(%rbp),%rdi movq nb210_jjnr(%rsp),%rax shll $2,%ecx addq %rcx,%rax movq %rax,nb210_innerjjnr(%rsp) ## pointer to jjnr[nj0] movl %edx,%ecx subl $2,%edx addl nb210_ninner(%rsp),%ecx movl %ecx,nb210_ninner(%rsp) addl $0,%edx movl %edx,nb210_innerk(%rsp) ## number of innerloop atoms jge _nb_kernel210_x86_64_sse2.nb210_unroll_loop jmp _nb_kernel210_x86_64_sse2.nb210_checksingle_nb_kernel210_x86_64_sse2.nb210_unroll_loop: ## twice unrolled innerloop here movq nb210_innerjjnr(%rsp),%rdx ## pointer to jjnr[k] movl (%rdx),%r10d movl 4(%rdx),%r11d addq $8,nb210_innerjjnr(%rsp) ## advance pointer (unrolled 2) movq nb210_pos(%rbp),%rsi ## base of pos[] lea (%r10,%r10,2),%rax ## replace jnr with j3 lea (%r11,%r11,2),%rbx ## move two coordinates to xmm4-xmm6 movlpd (%rsi,%rax,8),%xmm4 movlpd 8(%rsi,%rax,8),%xmm5 movlpd 16(%rsi,%rax,8),%xmm6 movhpd (%rsi,%rbx,8),%xmm4 movhpd 8(%rsi,%rbx,8),%xmm5 movhpd 16(%rsi,%rbx,8),%xmm6 ## calc dr subpd nb210_ix(%rsp),%xmm4 subpd nb210_iy(%rsp),%xmm5 subpd nb210_iz(%rsp),%xmm6 ## store dr movapd %xmm4,%xmm9 movapd %xmm5,%xmm10 movapd %xmm6,%xmm11 movq nb210_charge(%rbp),%rsi ## base of charge[] ## square it mulpd %xmm4,%xmm4 mulpd %xmm5,%xmm5 mulpd %xmm6,%xmm6 addpd %xmm5,%xmm4 addpd %xmm6,%xmm4 ## rsq in xmm4 movq nb210_type(%rbp),%rdi movlpd (%rsi,%r10,8),%xmm3 cvtpd2ps %xmm4,%xmm5 rsqrtps %xmm5,%xmm5 cvtps2pd %xmm5,%xmm2 ## lu in low xmm2 movhpd (%rsi,%r11,8),%xmm3 movl (%rdi,%r10,4),%r8d movl (%rdi,%r11,4),%r9d movapd nb210_krf(%rsp),%xmm7 ## lookup seed in xmm2 movapd %xmm2,%xmm5 ## copy of lu mulpd %xmm2,%xmm2 ## lu*lu movapd nb210_three(%rsp),%xmm1 mulpd %xmm4,%xmm7 ## krsq mulpd %xmm4,%xmm2 ## rsq*lu*lu shll %r8d shll %r9d movl nb210_ntia(%rsp),%edi movapd nb210_half(%rsp),%xmm0 movq nb210_vdwparam(%rbp),%rsi subpd %xmm2,%xmm1 ## 30-rsq*lu*lu mulpd %xmm5,%xmm1 mulpd %xmm0,%xmm1 ## xmm0=iter1 of rinv (new lu) mulpd nb210_iq(%rsp),%xmm3 ## qq addl %edi,%r8d addl %edi,%r9d movapd %xmm1,%xmm5 ## copy of lu mulpd %xmm1,%xmm1 ## lu*lu movapd nb210_three(%rsp),%xmm2 mulpd %xmm4,%xmm1 ## rsq*lu*lu movapd nb210_half(%rsp),%xmm0 movlpd (%rsi,%r8,8),%xmm4 movlpd 8(%rsi,%r8,8),%xmm6 movhpd (%rsi,%r9,8),%xmm4 movhpd 8(%rsi,%r9,8),%xmm6 movapd %xmm4,nb210_c6(%rsp) movapd %xmm6,nb210_c12(%rsp) subpd %xmm1,%xmm2 ## 30-rsq*lu*lu mulpd %xmm5,%xmm2 mulpd %xmm2,%xmm0 ## xmm0=rinv movapd %xmm0,%xmm4 mulpd %xmm4,%xmm4 ## xmm4=rinvsq movapd %xmm0,%xmm6 addpd %xmm7,%xmm6 ## xmm6=rinv+ krsq movapd %xmm4,%xmm1 subpd nb210_crf(%rsp),%xmm6 mulpd %xmm4,%xmm1 mulpd %xmm4,%xmm1 ## xmm1=rinvsix movapd %xmm1,%xmm2 mulpd %xmm2,%xmm2 ## xmm2=rinvtwelve mulpd %xmm3,%xmm6 ## xmm6=vcoul=qq*(rinv+ krsq) addpd %xmm7,%xmm7 mulpd nb210_c6(%rsp),%xmm1 mulpd nb210_c12(%rsp),%xmm2 movapd %xmm2,%xmm5 subpd %xmm1,%xmm5 ## Vvdw=Vvdw12-Vvdw6 addpd nb210_Vvdwtot(%rsp),%xmm5 mulpd nb210_six(%rsp),%xmm1 mulpd nb210_twelve(%rsp),%xmm2 subpd %xmm1,%xmm2 subpd %xmm7,%xmm0 mulpd %xmm0,%xmm3 addpd %xmm3,%xmm2 mulpd %xmm2,%xmm4 ## xmm4=total fscal addpd nb210_vctot(%rsp),%xmm6 ## the fj's - start by accumulating forces from memory movq nb210_faction(%rbp),%rdi movlpd (%rdi,%rax,8),%xmm0 movlpd 8(%rdi,%rax,8),%xmm1 movlpd 16(%rdi,%rax,8),%xmm2 movhpd (%rdi,%rbx,8),%xmm0 movhpd 8(%rdi,%rbx,8),%xmm1 movhpd 16(%rdi,%rbx,8),%xmm2 movapd %xmm6,nb210_vctot(%rsp) movapd %xmm5,nb210_Vvdwtot(%rsp) mulpd %xmm4,%xmm9 mulpd %xmm4,%xmm10⌨️ 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?