nb_kernel112_ppc_altivec.c

来自「最著名最快的分子模拟软件」· C语言 代码 · 共 1,557 行 · 第 1/5 页

				v10             = vec_madd(v10,v1,v0);

				vec_st(v24,240,(float *)stackdata); /* store Vvdwtot */
				/* calculate vectorial forces and accumulate fj. 
				 * v10-v18 has fs11-fs33 now. 
				 * First load iO-* dx,dy,dz vectors to v1-v9 
				 * and load iO forces to v28,v29,v30 
				 * use v19-v27 to accumulate j 3atoms forces 
				 */
				v28             = vec_ld(256, (float *) stackdata);
				v29             = vec_ld(272, (float *) stackdata);
				v30             = vec_ld(288, (float *) stackdata);

				v1              = vec_ld(400, (float *) stackdata);
				v2              = vec_ld(416, (float *) stackdata);
				v23             = vec_add(v23,v9); /* incr. vctot */
				v3              = vec_ld(432, (float *) stackdata);
				v4              = vec_ld(448, (float *) stackdata);
				v5              = vec_ld(464, (float *) stackdata);
				v6              = vec_ld(480, (float *) stackdata);
				vec_st(v23,224,(float *)stackdata); /* store vctot to stack */
				v7              = vec_ld(496, (float *) stackdata);
				v8              = vec_ld(512, (float *) stackdata);
				v9              = vec_ld(528, (float *) stackdata);

				v28             = vec_madd(v10,v1,v28);
				v19             = vec_nmsub(v10,v1,v0);
				v29             = vec_madd(v10,v2,v29);
				v20             = vec_nmsub(v10,v2,v0);
				v30             = vec_madd(v10,v3,v30);
				v21             = vec_nmsub(v10,v3,v0);

				v28             = vec_madd(v11,v4,v28);
				v22             = vec_nmsub(v11,v4,v0);
				v29             = vec_madd(v11,v5,v29);
				v23             = vec_nmsub(v11,v5,v0);
				v30             = vec_madd(v11,v6,v30);
				v24             = vec_nmsub(v11,v6,v0);

				v28             = vec_madd(v12,v7,v28);
				v25             = vec_nmsub(v12,v7,v0);
				v29             = vec_madd(v12,v8,v29);
				v26             = vec_nmsub(v12,v8,v0);
				v30             = vec_madd(v12,v9,v30);
				v27             = vec_nmsub(v12,v9,v0);

				/* store these i forces, and repeat the 
				 * procedure for the iH1-* force 
				 */
				vec_st(v28,256,(float *)stackdata);
				vec_st(v29,272,(float *)stackdata);
				vec_st(v30,288,(float *)stackdata);

				v28             = vec_ld(304,(float *) stackdata);
				v29             = vec_ld(320,(float *) stackdata);
				v30             = vec_ld(336,(float *) stackdata);
				/* load new vectorial distances */
				v1              = vec_ld(544, (float *) stackdata);
				v2              = vec_ld(560, (float *) stackdata);
				v3              = vec_ld(576, (float *) stackdata);
				v4              = vec_ld(592, (float *) stackdata);
				v5              = vec_ld(608, (float *) stackdata);
				v6              = vec_ld(624, (float *) stackdata);
				v7              = vec_ld(640, (float *) stackdata);
				v8              = vec_ld(656, (float *) stackdata);
				v9              = vec_ld(672, (float *) stackdata);
      
				v28             = vec_madd(v13,v1,v28);
				v19             = vec_nmsub(v13,v1,v19);
				v29             = vec_madd(v13,v2,v29);
				v20             = vec_nmsub(v13,v2,v20);
				v30             = vec_madd(v13,v3,v30);
				v21             = vec_nmsub(v13,v3,v21);

				v28             = vec_madd(v14,v4,v28);
				v22             = vec_nmsub(v14,v4,v22);
				v29             = vec_madd(v14,v5,v29);
				v23             = vec_nmsub(v14,v5,v23);
				v30             = vec_madd(v14,v6,v30);
				v24             = vec_nmsub(v14,v6,v24);

				v28             = vec_madd(v15,v7,v28);
				v25             = vec_nmsub(v15,v7,v25);
				v29             = vec_madd(v15,v8,v29);
				v26             = vec_nmsub(v15,v8,v26);
				v30             = vec_madd(v15,v9,v30);
				v27             = vec_nmsub(v15,v9,v27);

				/* store these i forces, and repeat the
				 * procedure for the iH2-* force 
				 */
				vec_st(v28,304,(float *)stackdata);
				vec_st(v29,320,(float *)stackdata);
				vec_st(v30,336,(float *)stackdata);
				v28             = vec_ld(352,(float *) stackdata);
				v29             = vec_ld(368,(float *) stackdata);
				v30             = vec_ld(384,(float *) stackdata);
				/* load new vectorial distances */
				v1              = vec_ld(688, (float *) stackdata);
				v2              = vec_ld(704, (float *) stackdata);
				v3              = vec_ld(720, (float *) stackdata);
				v4              = vec_ld(736, (float *) stackdata);
				v5              = vec_ld(752, (float *) stackdata);
				v6              = vec_ld(768, (float *) stackdata);
				v7              = vec_ld(784, (float *) stackdata);
				v8              = vec_ld(800, (float *) stackdata);
				v9              = vec_ld(816, (float *) stackdata);
      
				v28             = vec_madd(v16,v1,v28);
				v19             = vec_nmsub(v16,v1,v19);
				v29             = vec_madd(v16,v2,v29);
				v20             = vec_nmsub(v16,v2,v20);
				v30             = vec_madd(v16,v3,v30);
				v21             = vec_nmsub(v16,v3,v21);

				v28             = vec_madd(v17,v4,v28);
				v22             = vec_nmsub(v17,v4,v22);
				v29             = vec_madd(v17,v5,v29);
				v23             = vec_nmsub(v17,v5,v23);
				v30             = vec_madd(v17,v6,v30);
				v24             = vec_nmsub(v17,v6,v24);

				v28             = vec_madd(v18,v7,v28);
				v25             = vec_nmsub(v18,v7,v25);
				v29             = vec_madd(v18,v8,v29);
				v26             = vec_nmsub(v18,v8,v26);
				v30             = vec_madd(v18,v9,v30);
				v27             = vec_nmsub(v18,v9,v27);

				/* store these i forces */
				vec_st(v28,352,(float *)stackdata);
				vec_st(v29,368,(float *)stackdata);
				vec_st(v30,384,(float *)stackdata);

				/* j forces present in v19-v27 */    
				
				/*  Oxa  Oza  Oxb  Ozb */
				v1              = vec_mergeh(v19,v21); 
				/*  Oxc  Ozc  Oxd  Ozd */
				v19             = vec_mergel(v19,v21); 
				/*  Oya H1xa  Oyb H1xb */
				v21             = vec_mergeh(v20,v22); 
				/*  Oyc H1xc  Oyd H1xd */
				v20             = vec_mergel(v20,v22); 
				/* H1ya H2xa H1yb H2xb */
				v22             = vec_mergeh(v23,v25); 
				/* H1yc H2xc H1yd H2xd */
				v23             = vec_mergel(v23,v25); 
				/* H1za H2ya H1zb H2yb */
				v25             = vec_mergeh(v24,v26); 
				/* H1zc H2yc H1zd H2yd */
				v24             = vec_mergel(v24,v26); 

				/* H2za   0  H2zb   0  */
				v26             = vec_mergeh(v27,v0);   
				/* H2zc   0  H2zd   0  */
				v27             = vec_mergel(v27,v0);   
      
				/*  Oxa  Oya  Oza H1xa */
				v2              = vec_mergeh(v1,v21);   
				/*  Oxb  Oyb  Ozb H1xb */
				v21             = vec_mergel(v1,v21);   
				/*  Oxc  Oyc  Ozc H1xc */
				v1              = vec_mergeh(v19,v20);    
				/*  Oxd  Oyd  Ozd H1xd */
				v19             = vec_mergel(v19,v20);    
				/* H1ya H1za H2xa H2ya */
				v20             = vec_mergeh(v22,v25);  
				/* H1yb H1zb H2xb H2yb */
				v22             = vec_mergel(v22,v25);  
				/* H1yc H1zc H2xc H2yc */
				v25             = vec_mergeh(v23,v24);  
				/* H1yd H1zd H2xd H2yd */
				v23             = vec_mergel(v23,v24);  
				/* H2za   0    0    0  */
				v24             = vec_mergeh(v26,v0);   
				/* H2zb   0    0    0  */
				v26             = vec_mergel(v26,v0);   
				/* H2zc   0    0    0  */
				v3              = vec_mergeh(v27,v0);   
				/* H2zd   0    0    0  */
				v27             = vec_mergel(v27,v0);   
 
				v29             = (vector float)vec_splat_s32(-1);

				/* move into position, load and add */  
				v30            = (vector float)vec_lvsr(0,(int *)faction+j3a); 
				v4              = vec_ld(  0, faction+j3a);
				v6              = vec_ld( 16, faction+j3a);
				v8              = vec_ld( 32, faction+j3a);
				v10             = vec_perm(v0,v29,(vector unsigned char)v30);
				v12             = vec_perm(v0,v2,(vector unsigned char)v30);
				v12             = vec_add(v12,v4);
				v14             = vec_perm(v2,v20,(vector unsigned char)v30);
				v2              = vec_add(v14,v6);
				v16             = vec_perm(v20,v24,(vector unsigned char)v30);
				v20             = vec_add(v16,v8);
				v12             = vec_sel(v4,v12,(vector unsigned int)v10);
				vec_st(v12,  0, faction+j3a);
				v10             = vec_sld(v0,v10,12);
				vec_st(v2, 16, faction+j3a);
				v12             = vec_sel(v20,v8,(vector unsigned int)v10);
				vec_st(v12, 32, faction+j3a);


				/* Finished 1, now do 2  */
				v30            = (vector float)vec_lvsr(0,(int *)faction+j3b); 
				v4              = vec_ld(  0, faction+j3b);
				v6              = vec_ld( 16, faction+j3b);
				v8              = vec_ld( 32, faction+j3b);
				v10             = vec_perm(v0,v29,(vector unsigned char)v30);
				v12             = vec_perm(v0,v21,(vector unsigned char)v30);
				v24             = vec_add(v12,v4);
				v12             = vec_perm(v21,v22,(vector unsigned char)v30);
				v21             = vec_add(v12,v6);
				v12             = vec_perm(v22,v26,(vector unsigned char)v30);
				v22             = vec_add(v12,v8);
				v12             = vec_sel(v4,v24,(vector unsigned int)v10);
				vec_st(v12,  0, faction+j3b);
				v10             = vec_sld(v0,v10,12);
				vec_st(v21, 16, faction+j3b);
				v12             = vec_sel(v22,v8,(vector unsigned int)v10);
				vec_st(v12, 32, faction+j3b);

				/* water 3 */
				v31            = (vector float)vec_lvsr(0,(int *)faction+j3c);
				v5              = vec_ld(  0, faction+j3c);
				v7              = vec_ld( 16, faction+j3c);
				v9              = vec_ld( 32, faction+j3c);
				v11             = vec_perm(v0,v29,(vector unsigned char)v31);
				v13             = vec_perm(v0,v1,(vector unsigned char)v31);
				v13             = vec_add(v13,v5);
				v15             = vec_perm(v1,v25,(vector unsigned char)v31);
				v1              = vec_add(v15,v7);      
				v17             = vec_perm(v25,v3,(vector unsigned char)v31);
				v25             = vec_add(v17,v9);
				v13             = vec_sel(v5,v13,(vector unsigned int)v11);
				vec_st(v13,  0, faction+j3c);
				v11             = vec_sld(v0,v11,12);
				vec_st(v1, 16, faction+j3c);
				v13             = vec_sel(v25,v9,(vector unsigned int)v11);
				vec_st(v13, 32, faction+j3c);

				/* water 4 */
				v31            = (vector float)vec_lvsr(0,(int *)faction+j3d);    
				v5              = vec_ld(  0, faction+j3d);
				v7              = vec_ld( 16, faction+j3d);
				v9              = vec_ld( 32, faction+j3d);
				v11             = vec_perm(v0,v29,(vector unsigned char)v31);      
				v13             = vec_perm(v0,v19,(vector unsigned char)v31);
				v25             = vec_add(v13,v5);
				v13             = vec_perm(v19,v23,(vector unsigned char)v31);
				v19             = vec_add(v13,v7);
				v13             = vec_perm(v23,v27,(vector unsigned char)v31);
				v23             = vec_add(v13,v9);
				v13             = vec_sel(v5,v25,(vector unsigned int)v11);
				vec_st(v13,  0, faction+j3d);
				v11             = vec_sld(v0,v11,12);
				vec_st(v19, 16, faction+j3d);
				v13             = vec_sel(v23,v9,(vector unsigned int)v11);
				vec_st(v13, 32, faction+j3d);
			} 
			if(k<(nj1-2)) {
				jnra            = jjnr[k];
				jnrb            = jjnr[k+1];
				jnrc            = jjnr[k+2];
				j3a             = 3*jnra;
				j3b             = 3*jnrb;
				j3c             = 3*jnrc;

				v1              = (vector float)vec_lvsl(0, pos+j3a);
				v8              = (vector float)vec_lvsl(0, pos+j3b);
				v15             = (vector float)vec_lvsl(0, pos+j3c);

				v2              = vec_ld(0, pos+j3a);
				v9              = vec_ld(0, pos+j3b);
				v16             = vec_ld(0, pos+j3c);
				v3              = vec_ld(16, pos+j3a);
				v10             = vec_ld(16, pos+j3b);
				v17             = vec_ld(16, pos+j3c);
				v4              = vec_ld(32, pos+j3a);
				v11             = vec_ld(32, pos+j3b);
				v18             = vec_ld(32, pos+j3c);
				/*  Oxa  Oya  Oza H1xa */
				v5              = vec_perm(v2,v3,(vector unsigned char)v1); 
				 /*  Oxb  Oyb  Ozb H1xb */
				v12             = vec_perm(v9,v10,(vector unsigned char)v8); 
				 /*  Oxc  Oyc  Ozc H1xc */
				v19             = vec_perm(v16,v17,(vector unsigned char)v15);

				 /* H1ya H1za H2xa H2ya */
				v6              = vec_perm(v3,v4,(vector unsigned char)v1);
				 /* H1yb H1zb H2xb H2yb */
				v13             = vec_perm(v10,v11,(vector unsigned char)v8);
				 /* H1yc H1zc H2xc H2yc */
				v20             = vec_perm(v17,v18,(vector unsigned char)v15);
				/* H2za   -   -   - */    
				v7              = vec_perm(v4,v4,(vector unsigned char)v1); 
				 /* H2zb   -   -   - */
				v14             = vec_perm(v11,v11,(vector unsigned char)v8);
				/* H2zc   -   -   - */
				v21             = vec_perm(v18,v18,(vector unsigned char)v15); 
      
				/* permute 3atoms coordinates */
				/*  Oxa  Oxc  Oya  Oyc */
				v3              = vec_mergeh(v5,v19);  
				v5              = vec_mergel(v5,v19); /*  Oza  Ozc H1xa H1xc */
				v19             = vec_mergeh(v12,v0); /*  Oxb   -   Oyb   -  */
				v12             = vec_mergel(v12,v0); /*  Ozb   -  H1xb   -  */
      
				v26             = vec_mergeh(v6,v20); /* H1ya H1yc H1za H1zc */
				v16              = vec_mergel(v6,v20);/* H2xa H2xc H2ya H2yc */