main()
{
int a, attr, x = 0;
struct LIST_OF_OBJ *list_of_obj, *tmp_list_of_obj;
struct LIST_OF_OBJ_OPTIMIZED *list_of_obj_optimized,
*tmp_list_of_obj_optimized;
//TITLE
PRINT("= = = демонстраци¤ пагубного вли¤ни¤ страничной организации пам¤ти = = =\n");
PRINT_TITLE;
/* -----------------------------------------------------------------------
*
* обработка классического списка
*
------------------------------------------------------------------------ */
// выделение пам¤ти
list_of_obj = (struct LIST_OF_OBJ*)
_malloc32(N_ELEM*sizeof(struct LIST_OF_OBJ));
// иницилизаци¤ списка
for (a = 0; a < N_ELEM; a++)
list_of_obj[a].next = list_of_obj + a + 1;
list_of_obj[N_ELEM-1].next = 0;
// трассировка списка
AL_BEGIN;
CLEAR_L2_CACHE();
L_BEGIN(0)
tmp_list_of_obj = list_of_obj;
do{
for(attr = 0; attr < ATTR_SIZE; attr++)
x += tmp_list_of_obj[0].obj_attr[attr];
} while(tmp_list_of_obj = tmp_list_of_obj[0].next);
L_END(0)
AL_END;
L1_OUT("CLASSIC ");
/* ----------------------------------------------------------------------
*
* обработка оптимизированного раздельного списка
*
----------------------------------------------------------------------- */
// выделение пам¤ти
list_of_obj_optimized = (struct LIST_OF_OBJ_OPTIMIZED*)
_malloc32(N_ELEM*sizeof(struct LIST_OF_OBJ_OPTIMIZED));
// иницилизаци¤ списка
for (a = 0; a < N_ELEM ;a++)
{
list_of_obj_optimized[a].next = list_of_obj_optimized + a + 1;
#ifdef PESSIMIZE
list_of_obj_optimized[a].obj_attr = malloc(sizeof(int)*ATTR_SIZE);
// printf("*%X\n",list_of_obj_optimized[a].obj_attr);
#endif
list_of_obj_optimized[a].obj_body = malloc(sizeof(int)*BODY_SIZE);
} list_of_obj_optimized[N_ELEM-1].next = 0;
// трассировка списка
AL_BEGIN;
CLEAR_L2_CACHE();
L_BEGIN(1)
tmp_list_of_obj_optimized=list_of_obj_optimized;
do{
for(attr=0;attr<ATTR_SIZE;attr++)
x+=tmp_list_of_obj_optimized[0].obj_attr[attr];
} while(tmp_list_of_obj_optimized
= tmp_list_of_obj_optimized[0].next);
L_END(1)
AL_END;
Lx_OUT("OPTIMIZED ",Ax_GET(0),Ax_GET(1));
}
main()
{
int *p1,*p2;
float x = 0;
int a, b, per, z = 1;
// выделение пам¤ти
p1 = (int *) _malloc32(BLOCK_SIZE);
p2 = (int *) _malloc32(BLOCK_SIZE);
// TITLE
PRINT("= = = демонстраци¤ совмещени¤ вычислений с доступом к пам¤ти = = =\n");
PRINT_TITLE;
/* -----------------------------------------------------------------------
*
* cовмещение вычислений x+=cos(x) с чтением пам¤ти
*
------------------------------------------------------------------------ */
printf("### x+=cos(x) ###\n");
// 1:4
per=4; VVV;
A_BEGIN(0)
for(a = 0; a < BLOCK_SIZE; a += 4)
z += *(int*)((char *)p1+a);
for(a = 0; a < (BLOCK_SIZE/per); a++)
x+=cos(x);
A_END(0)
VVV;
A_BEGIN(1)
for(a = 0; a < BLOCK_SIZE; a += per)
{
z += *(int*)((char *)p1+a);
x+=cos(x);
}
A_END(1)
per=8;VVV;
A_BEGIN(2)
for(a = 0; a < BLOCK_SIZE; a+=4)
z += *(int*)((char *)p1+a);
for(a=0; a < (BLOCK_SIZE/per); a++)
x+=cos(x);
A_END(2)
VVV;
A_BEGIN(3)
for(a = 0; a < BLOCK_SIZE; a += per)
{
z += *(int*)((char *)p1 + a);
z += *(int*)((char *)p1 + a + 4);
x+=cos(x);
}
A_END(3)
per=16;VVV;
A_BEGIN(4)
for(a = 0; a < BLOCK_SIZE; a+=4)
z += *(int*)((char *)p1+a);
for(a=0; a < (BLOCK_SIZE/per); a++)
x+=cos(x);
A_END(4)
VVV;
A_BEGIN(5)
for(a = 0; a < BLOCK_SIZE; a += per)
{
z += *(int*)((char *)p1 + a);
z += *(int*)((char *)p1 + a + 4);
z += *(int*)((char *)p1 + a + 8);
z += *(int*)((char *)p1 + a + 12);
x+=cos(x);
}
A_END(5)
per=32;VVV;
A_BEGIN(6)
for(a = 0; a < BLOCK_SIZE; a+=4)
z += *(int*)((char *)p1+a);
for(a=0; a < (BLOCK_SIZE/per); a++)
x+=cos(x);
A_END(6)
VVV;
A_BEGIN(7)
for(a = 0; a < BLOCK_SIZE; a += per)
{
z += *(int*)((char *)p1 + a);
z += *(int*)((char *)p1 + a + 4);
z += *(int*)((char *)p1 + a + 8);
z += *(int*)((char *)p1 + a + 12);
z += *(int*)((char *)p1 + a + 16);
z += *(int*)((char *)p1 + a + 20);
z += *(int*)((char *)p1 + a + 24);
z += *(int*)((char *)p1 + a + 28);
x+=cos(x);
}
A_END(7)
per=64;VVV;
A_BEGIN(8)
for(a = 0; a < BLOCK_SIZE; a+=4)
z += *(int*)((char *)p1 + a);
for(a=0; a < (BLOCK_SIZE/per); a++)
x+=cos(x);
A_END(8)
VVV;
A_BEGIN(9)
for(a = 0; a < BLOCK_SIZE; a += per)
{
z += *(int*)((char *)p1 + a);
z += *(int*)((char *)p1 + a + 4);
z += *(int*)((char *)p1 + a + 8);
z += *(int*)((char *)p1 + a + 12);
z += *(int*)((char *)p1 + a + 16);
z += *(int*)((char *)p1 + a + 20);
z += *(int*)((char *)p1 + a + 24);
z += *(int*)((char *)p1 + a + 28);
z += *(int*)((char *)p1 + a + 32);
z += *(int*)((char *)p1 + a + 36);
z += *(int*)((char *)p1 + a + 40);
z += *(int*)((char *)p1 + a + 44);
z += *(int*)((char *)p1 + a + 48);
z += *(int*)((char *)p1 + a + 52);
z += *(int*)((char *)p1 + a + 56);
z += *(int*)((char *)p1 + a + 60);
x+=cos(x);
}
A_END(9)
Lx_OUT(" 1:4 ",Ax_GET(0),Ax_GET(1));
Lx_OUT(" 1:8 ",Ax_GET(2),Ax_GET(3));
Lx_OUT(" 1:16 ",Ax_GET(4),Ax_GET(5));
Lx_OUT(" 1:32 ",Ax_GET(6),Ax_GET(7));
Lx_OUT(" 1:64 ",Ax_GET(8),Ax_GET(9));
per=128;VVV;
A_BEGIN(0)
for(a = 0; a < BLOCK_SIZE; a+=4)
z += *(int*)((char *)p1 + a);
for(a=0; a < (BLOCK_SIZE/per); a++)
x+=cos(x);
A_END(0)
VVV;
A_BEGIN(1)
for(a = 0; a < BLOCK_SIZE; a += per)
{
z += *(int*)((char *)p1 + a);
z += *(int*)((char *)p1 + a + 4);
z += *(int*)((char *)p1 + a + 8);
z += *(int*)((char *)p1 + a + 12);
z += *(int*)((char *)p1 + a + 16);
z += *(int*)((char *)p1 + a + 20);
z += *(int*)((char *)p1 + a + 24);
z += *(int*)((char *)p1 + a + 28);
z += *(int*)((char *)p1 + a + 32);
z += *(int*)((char *)p1 + a + 36);
z += *(int*)((char *)p1 + a + 40);
z += *(int*)((char *)p1 + a + 44);
z += *(int*)((char *)p1 + a + 48);
z += *(int*)((char *)p1 + a + 52);
z += *(int*)((char *)p1 + a + 56);
z += *(int*)((char *)p1 + a + 60);
z += *(int*)((char *)p1 + a + 64);
z += *(int*)((char *)p1 + a + 68);
z += *(int*)((char *)p1 + a + 72);
z += *(int*)((char *)p1 + a + 76);
z += *(int*)((char *)p1 + a + 80);
z += *(int*)((char *)p1 + a + 84);
z += *(int*)((char *)p1 + a + 88);
z += *(int*)((char *)p1 + a + 92);
z += *(int*)((char *)p1 + a + 96);
z += *(int*)((char *)p1 + a + 100);
z += *(int*)((char *)p1 + a + 104);
z += *(int*)((char *)p1 + a + 108);
z += *(int*)((char *)p1 + a + 112);
z += *(int*)((char *)p1 + a + 116);
z += *(int*)((char *)p1 + a + 120);
z += *(int*)((char *)p1 + a + 124);
x+=cos(x);
}
A_END(1)
per=256;VVV;
A_BEGIN(2)
for(a = 0; a < BLOCK_SIZE; a+=4)
z += *(int*)((char *)p1 + a);
for(a=0; a < (BLOCK_SIZE/per); a++)
x+=cos(x);
A_END(2)
VVV;
A_BEGIN(3)
for(a = 0; a < BLOCK_SIZE; a += per)
{
z += *(int*)((char *)p1 + a);
z += *(int*)((char *)p1 + a + 4);
z += *(int*)((char *)p1 + a + 8);
z += *(int*)((char *)p1 + a + 12);
z += *(int*)((char *)p1 + a + 16);
z += *(int*)((char *)p1 + a + 20);
z += *(int*)((char *)p1 + a + 24);
z += *(int*)((char *)p1 + a + 28);
z += *(int*)((char *)p1 + a + 32);
z += *(int*)((char *)p1 + a + 36);
z += *(int*)((char *)p1 + a + 40);
z += *(int*)((char *)p1 + a + 44);
z += *(int*)((char *)p1 + a + 48);
z += *(int*)((char *)p1 + a + 52);
z += *(int*)((char *)p1 + a + 56);
z += *(int*)((char *)p1 + a + 60);
z += *(int*)((char *)p1 + a + 64);
z += *(int*)((char *)p1 + a + 68);
z += *(int*)((char *)p1 + a + 72);
z += *(int*)((char *)p1 + a + 76);
z += *(int*)((char *)p1 + a + 80);
z += *(int*)((char *)p1 + a + 84);
z += *(int*)((char *)p1 + a + 88);
z += *(int*)((char *)p1 + a + 92);
z += *(int*)((char *)p1 + a + 96);
z += *(int*)((char *)p1 + a + 100);
z += *(int*)((char *)p1 + a + 104);
z += *(int*)((char *)p1 + a + 108);
z += *(int*)((char *)p1 + a + 112);
z += *(int*)((char *)p1 + a + 116);
z += *(int*)((char *)p1 + a + 120);
z += *(int*)((char *)p1 + a + 124);
z += *(int*)((char *)p1 + a + 128);
z += *(int*)((char *)p1 + a + 132);
z += *(int*)((char *)p1 + a + 136);
z += *(int*)((char *)p1 + a + 140);
z += *(int*)((char *)p1 + a + 144);
z += *(int*)((char *)p1 + a + 148);
z += *(int*)((char *)p1 + a + 152);
z += *(int*)((char *)p1 + a + 156);
z += *(int*)((char *)p1 + a + 160);
z += *(int*)((char *)p1 + a + 164);
z += *(int*)((char *)p1 + a + 168);
z += *(int*)((char *)p1 + a + 172);
z += *(int*)((char *)p1 + a + 176);
z += *(int*)((char *)p1 + a + 180);
z += *(int*)((char *)p1 + a + 184);
z += *(int*)((char *)p1 + a + 188);
z += *(int*)((char *)p1 + a + 192);
z += *(int*)((char *)p1 + a + 196);
z += *(int*)((char *)p1 + a + 200);
z += *(int*)((char *)p1 + a + 204);
z += *(int*)((char *)p1 + a + 208);
z += *(int*)((char *)p1 + a + 212);
z += *(int*)((char *)p1 + a + 216);
z += *(int*)((char *)p1 + a + 220);
z += *(int*)((char *)p1 + a + 224);
z += *(int*)((char *)p1 + a + 228);
z += *(int*)((char *)p1 + a + 232);
z += *(int*)((char *)p1 + a + 236);
z += *(int*)((char *)p1 + a + 240);
z += *(int*)((char *)p1 + a + 244);
z += *(int*)((char *)p1 + a + 248);
z += *(int*)((char *)p1 + a + 252);
x+=cos(x);
}
A_END(3)
VVV;
A_BEGIN(4)
// комбинирование вычислений с доступом к пам¤ти
// без разворота цикла
for(a = 0; a < BLOCK_SIZE; a += per)
{
for(b = 0; b < per; b += 4)
z += *(int*)((char *)p1 + a + b);
x+=cos(x);
}
A_END(4)
// вывод результатов (окончание)
Lx_OUT(" 1:128",Ax_GET(0),Ax_GET(1));
Lx_OUT(" 1:256",Ax_GET(2),Ax_GET(3));
Lx_OUT("*1:256",Ax_GET(2),Ax_GET(3));
printf("##################\n\n");
return x+z;
}
int
__main (void *progxfer, void *cli_util, void *imghdr, void *image_file_desc,
unsigned int linkflag, unsigned int cliflag)
{
int argc;
int argv;
int envp;
int status;
char **argv64;
char **envp64;
unsigned int flags = (unsigned __int64)&__gcc_main_flags;
/* The argv and envp arrays are 32 bits pointers to 32 bits pointers. */
decc$main (progxfer, cli_util, imghdr, image_file_desc,
linkflag, cliflag, &argc, &argv, &envp);
if (flags & MAIN_FLAG_64BIT)
{
int i;
/* Reallocate argv and envp with 64 bit pointers. */
argv64 = (char **) _malloc32 (sizeof (char *) * (argc + 1));
for (i = 0; i < argc; i++)
argv64[i] = (char *) (__int64)(((int *) (__int64) argv)[i]);
argv64[argc] = NULL;
for (i = 0; ((int *) (__int64) envp)[i]; i++)
;
envp64 = (char **) _malloc32 (sizeof (char *) * (i + 1));
for (i = 0; ((int *) (__int64) envp)[i]; i++)
envp64[i] = (char *)(__int64)(((int *) (__int64) envp)[i]);
envp64[i] = NULL;
}
else
{
argv64 = (char **)(__int64)argv;
envp64 = (char **)(__int64)envp;
}
status = main (argc, argv64, envp64);
if (flags & MAIN_FLAG_POSIX)
{
/* Map into a range of 0 - 255. */
status &= 255;
if (status != 0)
{
int save_status = status;
status = (__int64) &C$_EXIT1 + ((status - 1) << STS$V_MSG_NO);
/* An exit failure status requires a "severe" error. All
status values are defined in errno with a successful (1)
severity but can be changed to an error (2) severity by
adding 1. In addition for compatibility with UNIX exit()
routines we inhibit a run-time error message from being
generated on exit(1). */
if (save_status == 1)
{
status++;
status |= STS$M_INHIB_MSG;
}
}
else
status = SS$_NORMAL;
}
return status;
}
